Citation
Geographic significance of delivery zones in an e-commerce grocery delivery strategy

Material Information

Title:
Geographic significance of delivery zones in an e-commerce grocery delivery strategy
Creator:
Herrel, Keith
Publication Date:
Language:
English
Physical Description:
viii, 215 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Customers ( jstor )
Databases ( jstor )
Geographic information systems ( jstor )
Geographic regions ( jstor )
Geography ( jstor )
Grocers ( jstor )
Logistics ( jstor )
Simulations ( jstor )
Supply chain management ( jstor )
Trucks ( jstor )
Dissertations, Academic -- Geography -- UF ( lcsh )
Geography thesis, Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 2006.
Bibliography:
Includes bibliographical references.
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Keith Herrel.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright [name of dissertation author]. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
3594259 ( ALEPH )
778802067 ( OCLC )

Downloads

This item has the following downloads:


Full Text











GEOGRAPHIC SIGNIFICANCE OF DELIVERY ZONES IN AN E-
COMMERCE ENABLED GROCERY DELIVERY STRATEGY










By

KEITH CARL HERREL


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA


2006















ACKNOWLEDGE E ENTS

The completion of this dissertation would not have been possible if it were not for

the loving devotion of my wife, Tsuneko Herrel (also known as "Jaato") and the

enduring love and support of my mother Jean Herrel, who passed away before she

could see me graduate with my Ph.D. I know, however, that she will see me graduate

as she looks down from heaven. My three darling children, Schawn (also known as

"Chub"), Lina Lee (also know as "Screaming Meemie" or "Angel Princess"), and

Aaren (also known as "BoupToYouBoy"), have given me the will, perseverance,

fortitude, and just plain heartwarming desire to achieve in life, both academically and

professionally.









TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS............................................................ ........1..

LIST OF FIGURES........................ ....... ........................................v............. v

ABSTRACT .................................................. viii

CONCEPTS AND METHODOLOGY.......................................................

Purpose................................................. ...... ..................
C concepts ................... ........... ................... ...................3
Methodology................................ ...... .... ................4

LITERATURE REVIEW.....................................................................8

Online Grocers' Failures, Difficulties, and Possible Solutions...................9
Spatial Decision Support Systems................................................. 18
Online Retailer Geographic Issues ..................................................25
Grocer Web Site Design Issues......................................................29
A artificial Intelligence........... ........................................ ................ 30
Online Grocers' Store Locations.....................................................45
Geographic Information Systems for Traffic Flow Simulation...................48
Geographic Information Systems for Vehicle Routing and Logistics.............51
Networks ............................................ ........ .............56
Data Visualization and Data Representation........................................59
Geographic Data and Surface Representation...................................66
Geographic Information Systems in Business......................................86
Enabling Geographic Information Systems Technologies.........................92
Spatiotemporality......................................................................101
Other Supporting Literature........................................................ 107

SIMULATION ELUCIDATION........................................................... 122

THREE SIMULATION SEQUENCES THAT RESULTED IN EXCEPTIONAL,
AVERAGE, AND MEDIUM PROFIT...................................................... 178

Exceptional Profit.................................................................... 178
Low Profit ................................ ... ...................................... 179
Medium Profit .................................................................... 179

C O N C LU SIO N ............................................................................. .. 186

SOURCES CITED...........................................................................199









BIOGRAPHICAL SKETCH................................................................ .. 215















LIST OF FIGURES

Figure

1 Hexagonal layout......................... ...........................................

2 Duration of delivery in days.................................................... 146

3 Profit options settings............................................................. 146

4 Minimum dollar delivery.......................................................147

5 Mandatory profit.................................................................. 147

6 Delivery charge per concentric region........................................ 148

7 Delivery charge per region interface........................................... 148

8 Delivery charge per region pop up form.......................................149

9 Simulation options form......................................................... 149

10 Number of orders error catching.............................................. 150

11 Number of deliveries drop-down list........................................150

12 Duration of deliveries error catching ..........................................151

13 Duration of deliveries drop-down list........................................ 151

14 Region and legend color coordination..................................... 152

15 Report dashboard for the simulation run on April 13......................152

16 Simulation main map result with profit settings..............................153

17 Pause function.................................................................. 153

18 Profit options form........................................................... 154

19 Allow dispatch alerts radio button............................................. 154









Figure

20 Truck dispatch notification.................................................. 155

21 Simulation main map result showing truck usage............................155

22 Report dashboard for the simulation run on April 15......................156

23 Population report ............................................................ 156

24 Profitability chart................................................................ 157

25 Total mandatory profit per simulation report.................................158

26 Residence report.......... .......................................... ........ 159

27 Settings report.................................................................... 160

28 D delivery report .................................................................. 161

29 Simulation options form.......................... ....... ................... 161

30 Simulation main map result showing three simulations.................... 162

31 Simulation charts report....................................................... 163

32 Simulation charts report showing second and third simulation ...........164

33 Settings report showing three simulations......................................165

34 Total mandatory profit per simulation report showing three simulations.. 165

35 Residence report showing orders from the second and third simulations.. 166

36 Mandatory profit per concentric region on profit options form........... 166

37 Mandatory profit per concentric region on settings report.................. 167

38 Submit Delivery Charge button.................................... ........... 168

39 Delivery charge per region interface........................................ 168

40 Delivery charge per region form............................................169

41 Region A3 showing the delivery charge....................................... 169









Figure

42 Profitability chart showing profit for region A3............................170

43 Mandatory profit per simulation report showing profit for region A3.... 170

44 Settings used with Intra-Regional Drive Time function...................171

45 Intra-regional drive time interface.......................................... 171

46 Intra-regional drive time pop-up form........................................172

47 Settings report showing 15 minute intra-regional drive time............ 172

48 Delivery report showing increased intra-regional drive times.............173

49 Inter-regional drive time interface................... ........ ......... 173

50 Settings report showing increased inter-regional drive time..............174

51 Delivery report showing increased inter-regional drive times........... 174

52 Overall average population per region.....................................175

53 Population per region interface................................................ 175

54 Main map showing various amounts of orders placed....................... 176

55 Regions showing higher and zero populations............................176

56 Report dashboard showing various populations...........................177

57 Population report showing various populations........................... 177

58 Profit and simulation options for high-profit scenario....................... 180

59 Profitability quotient and profit gauge for high-profit scenario .........181

60 Profit and simulation options for low-profit scenario......................182

61 Profitability quotient and profit gauge for low-profit scenario............. 183

62 Profit and simulation options for medium-profit scenario................ 184

63 Profitability quotient and profit gauge for medium-profit scenario........ 185















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

GEOGRAPHIC SIGNIFICANCE OF DELIVERY ZONES IN AN E-
COMMERCE ENABLED GROCERY DELIVERY STRATEGY

By

Keith Carl Herrel

May 2006

Chair: Grant Ian Thrall
Major Department: Geography

This work contains a geographic and logistic tool that will allow grocers who

are contemplating to deliver groceries via e-commerce to better decide whether they

should embark on that initiative. As web-based orders become more commonplace,

this computer application will have the potential to assist both large and small grocers

in the critical decision of starting a delivery service. The causes of failures of grocers

are traced. An operational computer program is demonstrated, with embedded

algorithms that make for a realistic simulation. The computer simulation together with

the user-friendly interface "front end" provide a practical and also scholarly solution to

a complex geographic problem.















CONCEPTS AND METHODOLOGY

Purpose

The potentially lucrative grocery home-delivery market largely remains untapped

because of the financial risk and operational difficulty of running such a service. Also

recent history is fraught with unsuccessful attempts at delivering groceries. The rapid

bankruptcy of the high-profile Webvan company is possibly the worst debacle of all the

grocery delivery attempts.

This dissertation contains a logistical simulation application to solve one of the most

plaguing problems that some grocers have, which is the problem of deciding to initiate a

home delivery service or not. The primary purpose of the simulation is to reduce the

financial risk involved with rolling out an e-grocery initiative.

This means that grocers can take a quantum leap forward in satisfying the annual

multi-billion dollar United States grocery market by using my application. This is also

true for other industrialized countries. How is this possible? First, the great number of

grocer home delivery failures is no secret. This could be resulting in many grocers

hesitating to start a delivery initiative. Second, there could be a notable benefit to the

people in much of the industrialized world if grocers can successfully deliver to time-

strapped customers. Traffic jams and fuel consumption could decrease if people do not

use their cars to shop after work. Third, a grocer can decrease the number of employees,

such as cash register attendants and stock persons needed in the brick-and-mortar store

because customers who would otherwise go to the store will not do so if the groceries are









home delivered. Fourth, the infirm and elderly will benefit because the competition to

satisfy people who accept home deliveries could very well drive down delivery prices

that otherwise would be out of the economic means of these people. This list of benefits

is not exclusive, but should give justification to the claim that grocery home delivery is a

benefit to not only the grocer who could make a greater profit or acquire more customers,

but also to society in general.

No computer simulation application as presented here has ever been presented in a

public forum, including academic journal articles or news media. Grocers who are

contemplating deliveries will find this profit estimating simulation application easy to use,

powerful in its functionality, and versatile to allow its use in the actual delivery

operational operations Actual customer addresses can be input into the application to

generate simulated deliveries to loyal and prospective customers. Actual times that are

required to deliver to each region can be input before the simulation begins. The output of

the delivery data is shown is easy-to-read automatically generated reports. Round trip

truck times that include the average time required to deliver within each region can be

input separately into the application to give an accurate report of how long it takes a truck

to leave the store, deliver to the households, and return to the store. Also, the number of

trucks required to deliver to the population of all the regions can either be specified or

calculated.

Because rarely is it possible to give a clear-cut yes or no answer about whether a

business venture should be undertaken or not, the conditions where a online e-grocer (or

a grocer who is taking orders by telephone) will reach a predetermined amount of profit

can be determined by running this application. In other words, questions like, "by what









pathways can profitability be achieved?" and "what values do you need in combination to

reach a predetermined profit?" must be answered before a grocer should begin delivering

groceries.

Because profitability is naturally the goal of any grocer, this application gives an

indication to the grocer about what clusters or sets of attributes that will work to achieve

the required level of profit. Each simulation that is run can portray a different scenario of

what is needed to achieve profitability.

Concepts

In terms of this logistic application the term "Mandatory Profit" is used. The

Mandatory Profit is the profit that must be attained on each run to a region. A region,

from a grocer's standpoint, is an area that has customers with a certain profile, or takes a

certain amount of time of travel. A grocer must determine if deliveries to a region will be

profitable or not. If the grocer decides that each dispatch of a delivery truck (or "run") to

a region must result that revenues must exceed operating costs by $X, then all the grocer

needs to do is input the $X profit figure into the Mandatory Profit dialogue box of this

simulation. The dialogue box is a data input box on the logistic computerized interface

that I have developed. Additionally, the Mandatory Profit can be specified according to

different regions, allowing myriad possibilities for different profitability attainment

scenarios.

The expected profit margin can be input into the application. This percentage is

calculated in conjunction with the Mandatory Profit and the Delivery Charge to ascertain

whether the initiative can be profitable. The Profit Margin can be varied from 0.1% to 3%

of the value of the groceries delivered.









The Delivery Charge is another variable that can be input. The grocer can specify

any amount of delivery charge, from zero to as high as he or she would like. This

delivery charge is added to the calculation of each order generated and is used to

calculate the Mandatory Profit.

Another variable that can be added to each simulation scenario is the Minimum

Dollar Amount. If a grocer feels that it is not worth his while to deliver anything less than

$20 of groceries, the merchant could input that option into the simulation.

Succinctly stated, an objective of the simulation is the determination of the

conditions where an online e-grocer (or a grocer who is taking orders by a telephone) will

reach a predetermined amount of profitability.

When a grocer uses the logistical application, he or she will be in a significantly

better position to decide whether to initiate a home delivery service, or at least make a

more calculated decision about whether more costly analyses are justified or not.

Although the logistical simulation application contains sophisticated algorithms, and

performs computational tasks, a main purpose of mine is to create an application that is

intuitive and easy to use. This I have done.

Methodology

This presents a spatial decision support system for the delivery of groceries from

fixed "brick and mortar" locations to customers' homes. The spatial decision support

system (SDSS) will be demonstrated with a simulation. The simulation will perform the

following functions.

* Determine the mix of variables that are necessary for the online grocer to generate a
profit while delivering groceries.









* Tell the total profit that can be generated through deliveries, given a certain profit
margin.

* Determine the amount of trucks necessary to deliver to the customers.

* Show what regions normally reach a profitability margin, and at what times they do
so.

* Provide very good, medium, and poor profitability scenarios using the simulation
application I developed.

The simulation includes identification of the algorithms required to implement the

SDSS. When combined with reasonable data inputs, the information flow and algorithms

will allow for the identification of which geographic settings the SDSS can be

successfully deployed, and which are unlikely to be met with success. Even if no

geographic settings are indicated as being likely candidates for an actual SDSS of this

type, the simulation will be a success as it will indicate to investors that profit margins in

the contemporary grocery industry are insufficient to take advantage of the technology

and online ordering and store distribution of groceries to customers.

It is assumed that the grocery store has surveyed its current customer demand base to

determine the amount of groceries the customers will order from home each week. In

other words, the grocery store management is assumed to have reasonable expectations of

market penetration by neighborhood under various pricing schemes.

The geography of the delivery market will be show as follows.

* An urban region will be depicted on a map.

* A pattern that consists of honeycomb regions will be overlain on the map (Figure 1).
* The home store will be located at the center of the honeycomb area. In Figure 1
diagram, the store is located in the center blue hexagon.

* The hexagonal regions will be identified by capital letters. The home store in the
center region will be in region "H" meaning "home".






6


* Each customer will be associated with a hexagonal region.

* A random number generator will select customers for use in the simulation.

* The simulation will compress time.

* The contiguous hexagonal pattern will continue outward from the store, in a radial
manner (Figure 1).






7
















Figure 1. Hexagonal layout















LITERATURE REVIEW

Although the main purpose of this dissertation is to emphasize the geographic

significance of delivery zones in an e-commerce enabled grocery delivery strategy, the

work necessarily encompasses more than geography and GIS (geographic information

systems), and this multi-disciplinary approach is reflected in the literature review. An

important part of this section is "Grocer web Site Design Issues." This is because,

although the logistic application that I developed is run through Microsoft ExcelTM, the

application can be used as a web service, which basically means it can be used by

computers at remote sites for a fee. The section about artificial intelligence is included to

better acquaint the reader with potentially important grocer decision-making tools such as

intelligent agents that could traverse networks and read mobile computers in delivery

trucks. The Literature Review would be incomplete without the incorporation of the two

sections that include GIS for traffic flow simulation, vehicle routing, and logistics. These

concepts are the important technologies for grocers who want to optimize their trucks'

routes, which can result in more satisfied customers because of timely deliveries, and less

fuel and manpower expenditures. The manner which polygons, points, and lines are

represented on a GIS along with the way data is depicted near these geometric figures

should be important to grocers who deliver through a GIS interface so as not to include

superfluous data, or omit important data to delivery, operational, and managerial

personnel. A review of the literature about these concepts is also included in this section.









Online Grocer's Failures, Difficulties and Possible Solutions

There is a lot of supporting and relevant literature that, when combined, can provide

the building blocks for the utilization of delivery zones in geographic enterprise

optimization systems (GEOS) for online grocers. After reviewing the below-described

literature, the solution proposed in this dissertation for online grocer deliveries is not only

necessary, but also feasible.

In 2000, at the minimum seven pure online grocers were operating in the United

States. By the end of that year the only two that remained were Webvan and Peapod. In

an ominous statement, Ring and Tigert (2001) said, "no company has ever made a profit

in this (pure online grocery) business."

Perhaps the most visible of all grocery delivery companies' demises is that of

Webvan, the multi-billion dollar venture that started optimistically, but ended soon after

it began. Much literature exists concerning the quick bankruptcy of this and other online

grocery delivery companies (Partch, 1999; Saccomano, 1999b; Tweney, 1999; Briody,

2000; Bubny, 2000; Sanborn, 2000; Evans, 2001; Heun, 2001; Partch, 2001; Ring, 2001).

The founder of Borders Books, Louis Borders, created Webvan in 1999 to deliver

groceries and other household items to homes within 30 minutes of receiving the order

(Tweney, 1999). George Shaheen, a managing partner of Anderson Consulting, whose

father was in the grocery business, was the CEO in charge when Webvan went bankrupt

(Partch, 2001). There was much hope and skepticism when $11 billion was raised in

Webvan's 1999 initial public offering (Evans, 2001). The hope was that Webvan,

optimistically called the spot where "the internet meets your doorstep," could capture a

portion of the annual $650 billion United States online grocery, prepared food, and









drugstore market (Sanborn, 2000). Of this amount, $450 billion of food is purchased

annually in the United States (Ring, 2001).

It was a grand scheme that depended on a network of large and expensive automated

warehouse distribution centers as springboards for the delivery vans. Partch (1999) wrote

in "Home Delivery, Another Problem Solved," that a single distribution center would

have the capacity to dispatch as many groceries as 20 supermarkets. The distribution

centers were approximately 330,000 square feet each. The first distribution center was

located in Oakland, California; and the company planned to build another 25 similar

centers in other cities (Ring, 2001).

A portion of Webvan's plan did have a customer allure to it. The fledgling company

offered 24-hour free delivery (with orders more than $50) with prices up to 10% cheaper

than brick-and-mortar grocers (Partch, 1999).

Besides the money raised from the IPO, deep-pocketed investors like Knight-Ridder,

CBS, and Softbank sank money into the company (Saccomano, 1999b). But, contrary to

stockholder and other investor optimism, literature shows that many people were also

pessimistic about Webvan's viability. Dan Rabinowitz of Peapod Inc., another online

grocer, called Webvan's business model "a lot of smoke and mirrors" (Saccomano,

1999b). In the same article, grocery consultant Cristopher Hoyt shows his doubt by

saying that Webvan would not be able to solve the perishable food distribution problem,

and that even if they can deliver the dry foods efficiently, the grocery market for dry

foods is being supplanted by cheaper super centers.

Webvan's investors and managers would have benefited from Michael Brown's

commentary on automated grocery distribution centers. Brown said in 1993, "the best









strategy at present may be to postpone investment in a centralized warehouse with high

levels of automation, correspondingly high capital costs, and a lengthy payback period"

(Browne, 1993).

Miles Cook, vice president of consulting firm Bain and Company, said, "online

grocery is a murderously difficult market," (Heun, 2001). Evidently Cook, Rabinowitz,

Hoyt, and other skeptics were right because only 20 months after its inception, Webvan

fired 2,000 employees and filed for bankruptcy (Evans, 2001).

It seems that the less-than-optimal business model employed by Webvan is not the

only factor to blame for that company's demise. Paul Bubny (2000) explains that

customer sentiment regarding online purchases is an inhibiting factor. His

PricewaterhouseCoopers' study revealed that 21% of internet users would never purchase

groceries online. This says nothing of the percentage of current non-internet users that

will not buy groceries online in the future when they begin to use the internet. Bubny

(2000) calls grocery shopping a "care-giving function," and says consumers are not ready

to allow a "detached process such as online shopping" to supply their food. Further, a

survey conducted by NPD group revealed that a mere 2% of consumers would buy

groceries exclusively from an online site.

The same article, however, states that, with proper incentives, customers can be

convinced to buy groceries online. Respondents to the PricewaterhouseCoopers' survey

stated that price is their main concern when buying online groceries, which includes free

delivery. This implies that if other factors such as freshness and rapid delivery are

fulfilled, consumers may be persuaded to buy groceries online if the price is right.









Mainstream grocers are not turning a blind eye to online grocers. In fact, the threat

posed to large grocers' markets by smaller online grocers is real. Alan Mitchell says that

even a 5 to 10% rise in home grocery shopping could "severely dent" a superstore's

underlying profitability (Mitchell, 1999). This statement can be especially portentous to

brick-and-mortar grocers if they consider that overall online purchases are estimated to

keep on increasing (Killgren, 1999). Evidently with this in mind, large United States

grocers like Albertsons and Safeway are initiating brick and click grocery business

models (Briody, 2000). Safeway and Royal Ahold, a Dutch conglomerate, both formed

joint alliances with pure online grocers. By joining with foundering Peabody, Royal

Ahold saved Peabody from probable bankruptcy. Safeway invested $30 million in

Groceryworks.com. Not long after Groceryworks.com built two fulfillment centers in the

south-central United States, the company closed them down. Safeway subsequently

decided that filling online orders from its store shelves was a better model and began

doing so (Heun, 2001).

Albertsons is moving forward with a purely internal brick-and-click delivery

initiative. Possibly as a result of Webvan's closure, Albertsons' online sales in the Seattle

area increased 300% after Webvan stopped operating. Like Safeway, Albertsons fills

orders from its stores' shelves. Safeway is also implementing a fleet management system

to better route their trucks and more closely monitor driver and vehicle performance

(Barnes, 2002).

Ring and Tigert (Tigert, 2001) enumerate six groups of dissuasive factors

contributing to customer skepticism toward online grocery purchases. According to them,

place, the first factor, is important because customers like to be in a place where they can









touch and smell the merchandise. The next group, product, says that online grocers may

not have enough stock keeping units (SKU) that the customer desires. An SKU is

essentially an individual type of product. Third, value, is "more bang for the buck,"

which is often sacrificed by relatively high delivery fees. Service comes next. When a

customer visits a store, there are hopefully more knowledgeable and courteous employees

present than a single delivery truck driver. Communication, the fifth factor, comes in the

form of advertisement or notification to the customer. Ring and Tigert (2001) do not

expound on how this group is an inhibiting factor for online grocers. The sixth group,

technical problems, appears in the form of wrong orders or late deliveries, which can ruin

online grocers.

Another problem, as told by McKinnon and Tallam (2003), is the security of

delivering to homes, especially when the home is unoccupied. Groceries can be delivered

four ways to unoccupied residences. 1. The driver can enter the house. 2. A drop box is

available. 3. Another customer-designated location is available. 4. The goods are left at a

separate local agency, which delivers the groceries when the customer is home.

The article does say that overall, unattended delivery of groceries is relatively

secure when compared to the unattended delivery of more valuable items. Of course, the

worth of the goods matter little if the cause of the security breach is the delivery driver

having access to the interior of the home. The article does say that although the

unattended groceries may not be stolen, the fact that they are left visible implies that the

homeowner is not present which can invite burglary into the premises.

McKinnon and Tallams' (2003) security implications are relevant because a

significant amount of the prospective home grocery delivery market may not by online









because of the security implications. In the same article, the authors list different ways

groceries could be left at an unattended home. Those ways are a fixed interior box with

three compartments of various temperatures with a keypad-lockable door; a fixed

external box connected to an external wall; a mobile box delivered to the customer and

connected to a cable to the house; and communal collection boxes. The communal boxes

can be of various types. Primarily they are in a central place like a popular parking lot or

near a train station. Communal boxes can have numerous compartments, each with its

own lock and changeable code to open the lock (McKinnon and Tallam, 2003). When

deciding on a grocery delivery model, consideration should be given to each of these

drop-off strategies to decide which would probably work best given the catchment

neighborhood, type of customer lifestyle segment profiles (LSP) catered to, and logistical

strategy used by the online grocer. Of the LSP databases used for this type of analysis,

MOSAIC, ACORN, and NDL's Lifestyle database are three of the most highly regarded

(O'Malley, 1997).

The type of drop-off method is not unrelated to the "skill" the drivers have in

delivering the groceries. Skill here relates to the ability of the drivers to learn routes on

the fly. If an online grocer delivers to individual residences, especially when guaranteeing

delivery within an hour, it becomes highly unlikely that any two routes taken within, say,

one or two days, will be exactly the same. Michael Haughton expounds upon the

relationship of uncertain daily deliveries with the drivers' ability to learn new routes on

the fly (Haughton, 2002).

Haughton (2002) states that fluctuating demand can result in inefficient delivery and

dissatisfied customers. He refers to a "learning burden," and proposes a correlation metric









to assess the familiarity of a driver with certain rectilinear delivery routes. Because my

research is focused on solving the problem of individual residence delivery, which

presupposes daily fluctuating demands, Haughton's (2002) literature is relevant to this

dissertation.

Cross and Neil (2000) call online grocery shopping a "booming global phenomenon."

Obviously this statement does not seem to include the plights of the failed United States

online grocers. Evidently, these authors are referring to online grocers in Great Britain,

Scandinavia, and Iceland that are not doing as badly as those in the United States.

Optimistically, however, Cross and Neil tell how convenience is a driving factor toward

the success of online grocers. They mention how technologically savvy customers are the

most profitable ones because they are more inclined to shop online. The article also says

how employers can use online grocers as a fringe benefit to their employees. The

employer can pay for the delivery service to the office where the employees pick up their

groceries when they leave work.

Another factor that could provide inertia toward successful online grocer businesses

is UCCNet. This is a subsidiary of the Uniform Code Council, which is the organization

that developed the universal product code. UCCNet allows e-commerce companies to

register their supply-chain related data. UCCNet is an internet trading network that could

benefit buyers and sellers of food, beverages, and other items (Violono, 1999). Because

the network aids collaborative planning, forecasting, and replenishment, smaller online

grocers who use UCCNet should be better able to procure needed stock in times of

unexpected spikes in online purchases. More literature pertaining to UCCNet is reviewed

below.









Catering to the growing United States Hispanic market is an online grocer called

LatinGrocer. Saccomano (1999a) states that by 2009, 40% of the United States

population will be Hispanic, creating a very large grocery market. LatinGrocer's unique

point is that its services are also offered in Puerto Rico.

An effort to identify the success factors in e-grocery home delivery was made by

Punakivi and Saranen (2001) using simulation. The article supports the intent of this

dissertation by explaining that the lack of a proper logistical home delivery infrastructure

is the main deterrent to a successful online grocer.

Using Finland as an example, Punakivi and Saranen (2001) concluded that home

delivery can be as much as 43% cheaper for customers when compared to the customer

driving to the supermarket. Punakivi and Saranen (2001) write that the cost to deliver

groceries will be less when the delivery time window is larger and customers live in

densely populated areas. These are intuitive factors. What might not be so intuitive,

however, is that the number of needed delivery vehicles increases by 250% when a

manned delivery model (when the customer must be home) is used instead of an

unmanned model which uses drop boxes. Tanskanen and Punakivi (2002) write that the

only factor more costly than the deliveries in a manned home delivery model is the actual

picking and packing of the groceries One reason for the higher manned delivery costs is

that up to 60% of the packages might be undeliverable because of people not being home.

In Great Britain, some main grocers are using unattended delivery methods. In Finland,

grocers are using shared multi-temperature drop boxes for unattended delivery. A factor

that affects the costs and profitability of the shared reception box model is the number of

separate compartments in each box (Punakivi and Tanskanen, 2002).









It is important to note that online grocers might have to operate somewhat contrarily

to flexible production techniques. This is because although flexible production uses

strategies such as just in time deliveries (Browne, 1993), online grocers who precisely

forecast a significant proportion of their deliveries should consider pre-packing those

groceries to save time during busier order packing times. Shorter delivery time windows

notwithstanding, the ability for customers to choose what time the groceries are delivered

is another factor that increases delivery costs. In other words, given a 40-minute delivery

time window, the model that allows customers to choose what delivery time window is

the more expensive delivery method (as opposed to the grocer deciding what delivery

time window to dispatch the truck) (Punakivi and Tanskanen, 2002).

Speaking of delivery costs, online grocers must consider the tradeoff of shipping

from fewer sites, which increase delivery costs (Browne, 1993), or dispatching the

groceries from more outlets, which could increase the packing costs. Browne writes that

products with relatively low value densities should be distributed by local delivery. This

would apply to most groceries.

Gorr, Johnson, and Roehrig (2001) talk about the difficulty of home delivery

services and how GIS can help with delivery solutions. They describe the constraints

imposed upon delivering hot meals to residents home. Their research is applicable to this

dissertation because online grocers may offer hot delicatessen foods delivered to homes.

Indeed, it might be this hot (and cold) food service that helps the online grocer to gain

market share. But, as this particular article states, the time window that exists to deliver

these foods is greatly decreased because of the necessity to maintain very hot or very cold









temperatures. Gorr, Johnson, and Roehrig (2001) write that 45 minutes is the maximum

allowable delivery time for these deliveries.

Gorr, Johnson, and Roehrig (2001) also discuss the feasibility of using a spatial

decision support system (SDSS) for delivering hot meals. The combination of GIS,

algorithms, and managerial judgment are parts of SDSS. A particular problem the authors

address is the situation created when customers change addresses. The SDSS algorithms

should be robust enough to handle this type of change (Gorr, Johnson, and Roehrig,

2001).

One approach that online grocers might pursue when considering their overall

strategic plans to deliver groceries is to look at the composite customer equity that the

delivery regions encompass. Customer equity, which is often used as an augmentation to

brand equity, is the sum of the lifetime values of all of the customers a business has

(Niraj, 2001). If, when contemplating a delivery plan, the customer equity calculation

does not look as if it could be increased or even sustainable, it might be advisable to

rethink the grocery delivery plan. Some aspects to look at when determining the

sustainability of customers are complexity factors and efficiency factors of those

customers (Niraj, 2001). Succinctly stated, these two factors are antithetical forces.

Whereas complexity factors, such as difficulty fulfilling a grocery order, decrease a

customer's lifetime value, efficiency factors, which make a delivery to a particular

address easier, can help increase the lifetime value of the customer.

Spatial Decision Support Systems

Research about using GIS as an integral part of SDSS began years before the above-

noted article by Gorr, Johnson, and Roehrig (2001). Crossland et al. (2001) studied the









effectiveness of using GIS in SDSS to see if the time used to solve problems could be

significantly decreased. The article reinforces the important point that most business data

have "one or more spatial components." Three of these components, as stated in the

article, are relevant to online grocer business models customer addresses, delivery

vehicle locations, and site selection. Calling GIS a "highly evolved technical toolbox,"

the article, says that GIS should not only be used to find solutions to problems, but also to

guide the users to problems that might have been unforeseen before the adoption of a

GIS-enabled SDSS (Crossland, 1995). The worth of decision support systems (DSS) in

general is stated by (Vlachopoulou et al., 2001). They say that DSS "couples the

intellectual resources of individuals with the capabilities of the computer to improve the

quality of decisions and to support managerial judgment" (Vlachopoulou et al., 2001).

The value of spatial aids to managerial judgment should not be underestimated. Thrall

(1995) endorses the value of using GIS to optimize judgment in managerial decision

making. Thrall writes that, "...using GIS to improve judgment is a crucial stage in

reasoning with GIS because it links GIS to the market economy." Thrall expounds on his

GIS-facilitated judgment maxim by describing the value of business geography. Thrall

said, "Business geography integrates geographic analysis, reasoning, and technology for

the improvement of the business judgmental decision. Without the demonstrated ability

to improve the business decision, there is no business geography. This differentiates

business geography from the traditional descriptive or explanatory objective of economic

and urban geography" (Thrall, 2002).

Gorr et al. (1995) tell how a GIS-enabled SDSS can allow decision makers, or their

assistants, to use databases in real time. This is of course important for online grocers that









employ metamorphic delivery zones because, as this dissertation will show, it is when a

delivery zone reaches a certain profitability threshold that the truck should be dispatched

to that zone. A decision maker, or his or her assistant, must be able to make the real-time

decision to send the truck. Gorr et al. (1995) show that spatially related decisions can be

made faster and with less error when SDSS are used. Supporting the assertion that

deliveries should be made only after a certain profitability level has been reached is Niraj

et al.'s (2001) research that says a company should "eliminate transactions that do not

add value." If value to a company (considered here as profitability) is not created by a

delivery to a zone, then perhaps delivery to that zone should be deferred to a later time

when the value of that zone increases. Care should be taken that this strategy does not

result in dropping customers who might be unprofitable in the short run, but might

become valuable customers in the foreseeable future (Niraj et al., 2001).

A SDSS incorporates GIS with a database management system (DBMS) and

peripheral software tools that allow interaction with the analysts and other users

(Tarantilis et al., 2002). Tarantilis et al. call SDSS a "new scientific area of information

systems applications." Possibly, the relative newness of SDSS is in part the reason that

little if any literature exists about the utilization of this technology to successfully solve

the problem of delivering groceries that are purchased online. I will demonstrate in this

dissertation that spatial decision support systems can be the difference between the

success and failure for any online grocer of significant size.

The importance of routing information systems is underscored by Browne as he says

that the end of the era of "cheap freight" transport is here, where reliability used to be

attained without modern information systems (Browne, 1993).









Tarantilis and Kiranoudis (2002) use SDSS to solve a vehicle routing problem (VRP)

using the backtracking adaptive threshold accepting (BATA) model. The constraints

posed in these authors' research mirror those that would be inherent in a typical grocery

delivery situation. They are minimize the distance traveled, do not exceed the capacity of

a vehicle, and, one vehicle delivers each customer's order. Although the article does not

specifically mention web services, it does say that reusable software components are

important parts of a SDSS. Because, more recently, the importance and viability of web

services (for example, services created by the .Net platform) have gained mainstream

exposure, the plausibility of creating a component-based SDSS is greater than it is

without using web services.

Tarantilis and Kiranoudis (2002) give a detailed and graphical explanation of all the

components of a SDSS. ArcView is the GIS used. BATA, in conjunction with the

Dijkstra algorithm, determines which vehicle should deliver to which customers and the

order that it should do so. BATA is a "metaheuristic" method, meaning that it relies on

more than one heuristic to arrive at a solution. The article states that the SDSS is used by

Athens, Greece taxi companies, newspaper companies, and for meal delivery. Benefits

attained by using the SDSS include the ability to make semi-structured or unstructured

vehicle routing decisions, transport cost reduction, and better control of distribution

(Tarantilis and Kiranoudis, 2002). Also, when routing decisions are made using a SDSS,

the decisions made might be more intuitive if some sort of multimedia capabilities exist

within the system. Cartwright and Hunter explain how decision support systems can be

enhanced by the incorporation of multimedia. These authors say, "users of geographic

information need far more than the printed map, and they are likely to need rapidly-









produced and well designed spatial displays combining a number of different types of

data" (Cartwright, 2001).

As pertains to this dissertation, the key words expressed by Cartwright and Hunter

are "rapidly-produced.. spatial displays." Another notion proposed by these authors is

that although the incorporation of multimedia into GIS is very valuable, care must be

taken to not "overwhelm" the users. Ogao and Kraak (2002) agree with this by saying

that animation can distract the users if it is overused. These cautions should be adhered to

by online grocers who incorporate delivery zones into their geographic enterprise

optimization systems.

Ralston, Tharakan, and Liu (1994) espoused the virtues of SDSS before GIS was

considered a mainstream technology. The authors said that GIS must be merged with

spatial analysis tools for GIS to be used to its fullest potential. Agreeably, GIS alone,

without being augmented with tools such as simulation or artificial intelligence, is a

powerful technology but remains a mere shell of what it could be when used with these

tools. In Ralston et al.'s article it is said that transportation networks are comprised of

physical and logical links. Physical links are the actual roads, railways, or inland water

routes. The Logical links are points along the physical links for deliveries, pick ups, and

inter-modal transfers. Physical links of the same type can be divided into classes

depending on factors such as cost, capacity, and speed. Paths between a source and

destination always have both physical and logical links (Ralston, 1994).

Nasirin and Birks expound upon the types of SDSS used by retailers in Great Britain.

Knowing customer proximity to the store and those customers' purchasing behaviors are

two common functions performed when SDSS is used by retailers. The literature states









that one reason retailers are adopting GIS-enabled decision support systems is because of

the increasing availability and portability of spatial data which often exists on compact

disks. Despite this growing adoption of GIS by retailers, there has been little research

about the methodologies that retailers employ when using GIS to make decisions, even

though successful usage of GIS by retailers involves "enormous numbers" of factors

(Nasirin, 2003). An important point made in this article is that GIS should be used by

retailers to determine who the competition is and where they are located.

Keenan (1998) says that GIS is a technology that can "take advantage of other

technologies," when used to make routing decisions. Keenan also states that straight-line

distance is insufficient for routing calculations, and that time traveled is a better indicator

of the route. Using a decision support system (DSS) is important when ascertaining the

best route because sometimes the best route may actually look circuitous (Keenan, 1998).

According to Keenan (1998), this is where the "soft" constraints, such as decision-

makers' judgment, apply. The importance of the decision-makers' input is why solutions

to routing problems require the use of DSS more and expert systems less (Keenan, 1998).

Many times data and information applicable to the routing problem may already be in the

corporate management information system (MIS). Often information such as customer

demand and vehicle size is business-specific, thereby requiring a customized link to the

data from the GIS (Keenan, 1998). Because spatial data is getting less inexpensive to

purchase, the availability of comprehensive highway databases will become more readily

available Also, according to Keenan, highway data within a GIS when used in with

sophisticated management science routing algorithms can result in useful SDSS.









Genetic algorithms, although not strictly a spatial decision support system

component, are intriguing problem solvers that can help with difficult and time-

consuming geographic problems. Clarke (1998) believes that Genetic algorithms (GA)

utilization will soon become more prominent in GIS. Genetic algorithms use a

"Darwinian" method to find feasible solutions for problems by forcing solutions to

"evolve" by allowing only the "strongest" genre of variables to "survive" repeated

iterations through the problem. Chromosomes, consisting of genes, which themselves

contain alleles, "evolves the population of solutions by repeated selection and mating"

(Van Dijk et al., 2002). Van Dijk et al. demonstrate the use of GA to solve a map labeling

problem. Some of the dynamics of the GA problem solving method are Selection, which

means better parents produce fit children. By using tournament selection iteratively

choosing parents until n sets of two parents are chosen a set of parents is only chosen if

their children are fit. Reproduction, which means each set of parents produce two

children. Crossover refers to a certain amount of genes from each parent which are

imbued into the children. Replacement means some less fit children are exterminated to

make room for other stronger offspring. Finally Termination occurs when, at some point,

the population "converges," or become as close to optimal as possible.

Obviously, this is a very abstract explanation of GA. As a concrete example, Van

Dijk et al. show how map labels are optimally placed using GA while specifying the

direction of the label in respect to the city, degree of congestion of labels, and how large

the labels are in proportion to city population. Online grocers should investigate the

utilization of GA in their routing and customer selection optimization strategies. This is

because if delivery time is considered a scarce resource, the profitability of customers









becomes a paramount issue when deciding routes. Customers can be chosen using GA by

"natural selection" using profitability as one of the gene traits, as described above.

Online Retailer Geographic Issues

"Microscale" competitor analysis is possible with GIS (Sadahiro, 2001). Sadahiro

(2001) uses this approach as he looks at retail location in Japan and discovers that stores

are usually clustered around railroad stations.

Clustering, although often used to denote a sort of condensation of entities within a

region, is also a group of techniques use to analyze complex geographic data to ascertain

if spatial relations exist or not. Pattern spotting and data mining are two ways to

determine the presence or absence of clustering (Murray, 1999). While looking at two

types of clustering problems, the median clustering problem and the central point

clustering problem, Murray discovered that the median clustering problem requires less

computational power to solve than the central point clustering problem.

An aspect of retail clustering that pure online grocers (those without a brick-and-

mortar store) should consider is that they might not be able to capitalize on the important

aspect of being within customers' comparison shopping region. Sadahiro (2001)

recognizes this important factor and develops a probability density function, which

quantifiably tells the degree of agglomeration of retail stores of the same classification.

One reason stores in the same classification locate near each other is to provide customers

with the opportunity to shop comparatively (Sadahiro, 2001). Although this is not the

only force that results in retail agglomeration, it should be recognized as a force that

might take customers away from online grocery sites. Guy describes comparison

shopping as one factor to consider when investigating the grouping of retail stores. He









says that comparison shopping has an "enjoyable" aspect to it (Guy, 1998). Can

comparing prices of competing online grocers online be "enjoyable?" This question is

worth considering by online grocers.

Baker addresses another problem that is encountered by online retailers. He says that

online retailers do not adequately consider central places when selling online. This is

because the fast speed of most internet communications lends itself to the

underestimation of the importance of spatiality in cyberspace. He calls the failure of

retailers to comprehend distance minimization strategies such as the gravity model "a

major barrier to successful marketing and profitability for internet retailing" (Baker,

2001).

Recently, consumer activity patterns have been given greater scrutiny by retailers.

Consumer activity patterns could be relevant to online grocers for various reasons. For

instance, activity patterns could help online grocers decide whether to deliver to the home

or use communal drop boxes. Decision tree induction, which is used by artificial

intelligence and statistics, has recently been said to be useful to help model activity-based

models (Arentze, 2003). Albatross models model travel demand using a rule-based

decision tree. Arentze (2003) says that decision tree induction can be useful for modeling

spatiotemporal behavior, such as the behavior exhibited in activity patterns. Activity-

based travel behavior (ABTB) is the term coined for this type of activity monitoring

(Frihida et al., 2002). The concept of "space-time paths" is covered by Frihida et al.

(2002) in their article about ABTB. According to these authors, space-time paths can help

explain, predict, and plan for past and future activity patterns of commuters.









The reason for the attention given to activity patterns is that previously, retailers

looked at discrete trips when calculating the amount purchased by consumers. The

current consumer trend, however, is to engage in multi-stop "tours." The itinerary of

these tours is influenced by retailer policies, travel costs, and other services and

entertainment consumed on the tour. These "hybrid" multi-stop tours are blurring the

distinction between retail shopping trips and other types of trips (Roy et al., 2001).

Online grocers might be interested in the efforts to classify hybrid trips. Some

classifications enumerated by Roy et al. are tours that a) start and end at home, b) start

and end at home with a work stop, c) start and end at work, and d) tele-orders, which are

orders placed over the telephone.

Of course online grocers would be interested in the fourth hybrid trip category tele-

orders. But there should be ample reason for online grocers to concern themselves with

the other three categories as well. This is because, according to Roy et al. (2001),

consumers consider aggregate properties of tours. This means that if groceries ordered

online can become part of the tour (by picking up the groceries at communal drop boxes),

or might not become part of the tour (thereby saving the consumer time by having the

groceries delivered at home).

The duration and sequence of activities undertaken during tours is important; so

much so that various models of activity duration during trips have been developed. Two

of the models are the unconditional and conditional risk models. Some of the many

determinations made by Popkowski et al. (2002) when using these models are shopping

after work decreases with age; single people shop often after work; poorer people are

more likely to indulge in leisure after work; after leisure people are likely to conduct









personal business, then shop. Online grocers could look at these determinations when

deciding web page content, delivery routes, and more.

Task allocation is similar to activity patterns. Task allocation is the allocation of

subsets of tasks by household members while considering context variables such as

constraints and imperatives relevant to those household members such as car availability,

and urgency (Borgers et al., 2001). Other issues to consider when attempting to predict if

a household member will perform a certain task are whether the task is mandatory or

discretionary; or is shared or performed alone. Borgers et al. (2001) assume that task

allocation can be correctly simulated. Although this assumption is debatable, it may be

worthwhile for online grocers to consider situations when household members must

perform a mandatory task, such as shop for groceries, while at the same time be

constrained by only having one vehicle because a spouse or child is using the sole car

when the shopping trip must be performed.

The logistical requirements of online grocers are different than that of brick-and-

mortar supermarkets. Although there is an overabundance of literature about retail

supermarket logistics, there is little that applies solely to the logistic requirements of

online grocers.

Third party logistic firms (3PL) can supply all or a portion of a companies'

warehousing and distribution needs (Balakrishnan, 2000). This could be a novel idea for

online grocers who do not want to pack and ship their own groceries. It might

also be an innovative service provided by a new kind of 3PL that caters exclusively to

grocery delivery. A difficulty in using 3PLs for grocery delivery could be in the

compensation scheme. There are many variables, such at type of goods (frozen or hot),









weight, changing demand, and expedited foods, to name just a few. According to

Balakrishnan (2000), achieving fairness in compensating multiple distributors can be

challenging. Fee tables usually are made to calculate delivery compensation. Two

primary ways to determine fee values are tariff and cost-based approaches. The former

gauges the delivery compensation by geographic regions, while the latter estimates the

delivery costs using a function of distance and weight of the goods (Balakrishnan, 2000).

Microsoft's Solver contains linear programming capabilities that can be useful to

even smaller online grocers that are developing fee schedules or in-house delivery driver

compensation.

Grocer Web Site Design Issues

Wang and Gerchak (2001) support the importance of stock keeping units (SKU) in

the decision of a customer to shop at the store. Of course, customers prefer stores with

more SKUs. Therefore, a tradeoff has to be reached when allowing greater "virtual" shelf

space for more profitable items or showing more SKUs per web page.

Wang and Gerchak's (2001) research discusses the relationship between demand and

the allocation of retail shelf space. Because, according to the article, demand increases

when more shelf space is allocated to a product, an online grocer that more prominently

displays the higher profit items on its web site might have lower delivery costs relative to

the cost of the groceries delivered. This issue deserves more research. Interestingly,

Wang and Gerchak (2001) say that for any two identical retailers, the total profit made

between them depends only on their total inventory level, not on how the total inventory

is allotted between them. This could be significant when two online grocers, or an online

grocer and a brick and mortar grocer compete in the same market area. The word









"identical" can be ambiguous. But, if identical is defined as the same amount of SKU's,

then, if this proposition is correct, an online grocer that has the same amount of SKUs as

a brick and mortar grocer should have a reasonably good chance of selling as much as the

brick and mortar store. In "Designing an Effective Cyber Store Interface," Kim and Eom

(2002) talk about the importance of clearly showing on a web site that risk-free and on

time delivery is offered. They propose four elements of customer satisfaction. The

elements are product and/or service, support, bad experience recovery, and extraordinary

service. The factors of risk-free and on time delivery could be classified under any or all

of Kim and Eom's (2002) elements of customer satisfaction. Throughout their article it is

said that an e-commerce site should allow customers to comparison shop. Taken in

conjunction with showing the greatest possible amount of SKUs on a web site, if a site

allows, or even encourages customers to compare the prices of the host grocer's SKUs

with the prices of other grocers products (online or not), customer satisfaction and sales

might increase.

Artificial Intelligence

The study of neural networks is a branch of artificial intelligence that pertains to

roughly mimicking the brain to make calculated decisions. Literature exists that shows

the attempts to include the decision-making power of artificial neural networks (ANN) in

GIS. Although I have not located any literature that specifically uses ANN to determine

delivery routes for online grocers, ANN has been used with GIS for other types of

applications, which shows the feasibility of using these technologies together.

The predictive power of the combination of GIS and ANN to forecast land use

changes in Michigan is documented by Pijanowski et al. (2002). The article says the use









of GIS and ANN can "aid in the complex process of land use change." Also, "ANNs are

powerful tools that use a machine learning approach to quantify and model complex

behavior and patterns." This is precisely the type of functionality online grocers need in

order to predict demand and create tentative delivery zones to better prepare for that

demand. Pijanowski et al. state that because of the spatial nature of the variables, the

incorporation of ANN with GIS is "essential." The same should hold true for an online

grocer's logistical dynamic routing application.

Black (1995) writes that using ANN with the gravity model. He uses a gravity

artificial neural network (GANN) to show commodity flows between nine United States

census regions. Using as input values regional flow of commodities, regional flow

attraction, and interregional distance, he discovered that the accuracy of the gravity

model improved when he moved from the unconstrained (conventional) gravity model to

the GANN. Black concludes that flow modeling can be "revolutionized" by using the

GANN model. What Black's research contributes to the purpose of this dissertation is the

clarification that ANNs increase the flexibility and usefulness of not only GIS, but also

spatial modeling in general.

Black (1993) states that an industrial firm must make three decisions after it has

decided to operate. 1. Where to locate, 2. What technology to use, and 3. What the scale

of production will be. Although Black was mainly referring to manufacturing firms, his

three points can also apply to an online grocery initiative. The technology inferred by

Black can include GIS, computer simulation, SDSS, intelligent traffic systems and more.

Scale could refer to the scope of the business model as referring to delivery methods and

delivery zones, while firm location is of course where the grocery store and its









distribution centers) are located in comparison to the customer base. These three factors

should be considered within the entire online grocery delivery model.

It should be noted that a concern of both Pijanowski and Black in the above-

mentioned research about ANN and spatial technology is the problem of spatial

autocorrelation. Research exists addressing this problem. One example is from Duckham

(2000) who discusses "error sensitive GIS." Relevantly, this literature also touches on the

realm of artificial intelligence because it uses induction, an established artificial

intelligence technique. According to this article, "an inductive learning algorithm should

be able to automatically deduce rules that embody the patterns in that data..." According

to Gahegan (2000), inductive machine learning is gaining popularity in the geographic

community.

Interestingly, even though, according to Duckham et al. (2000), the inductive

algorithm deduces; it does not induce, this literature is noteworthy because it endorses an

artificial intelligence (AI) solution to spatial problems such as autocorrelation that could

be otherwise intensified by the incorporation of another AI solution, namely artificial

neural networks (ANN). Additionally, because of the low profitability and labor

intensiveness of ascertaining the amount of error in geographic data, the article says that

there probably will not be many companies strictly concerned with quality issues in

geographic data. Duckham et al.'s (2000) research therefore recommends the inductive

learning algorithm because it is a relatively low cost way of ascertaining spatial data

quality.

Neural spatial interaction models are somewhat related to gravity models. Three

gravity model input variables relate to measures of "origin propulsiveness, destination









attractiveness and spatial separation" (Fischer, 2002). Neural spatial interaction models

relate to gravity models not only because they expedite mathematical modeling, but

because they are adaptive enough to "deal with incomplete, inaccurate, distorted, missing,

noisy, and confusing data" (Fischer, 2002).

Although the field of artificial intelligence includes technology such as expert

systems, neural networks, and fuzzy logic, intelligent agents possess a particularly

promising niche in the field of online grocery supply chain optimization and visualization.

Because of their ability to "automatically" scan intranets, extranets, and the internet,

intelligent software agents (or just "intelligent agents"), or bots, can be programmed to

periodically monitor databases, data marts, and directories. The gleaned data could then

be made available in a more informative format, such as in GEOS-enabled online grocery

logistic solutions.

Although using bots to collect data can be difficult (Fontana, 2002), companies, such

as SAP, are aggressively pursuing the utilization of bots in ERP (enterprise resource

management) and SCM (supply chain management) solutions. Steve Ranger tells how

SAP is using bots to gather supply chain partner data and make the resulting information

available through portals (Ranger, 2001).

One of the forerunners in intelligent agent research is the Carnegie Mellon

University. A look at their intelligent software agent website gives an inkling of the

branches of this field of artificial intelligence (Carnegie Mellon, 2001).

The Carnegie Mellon site shows a few bot technologies that could be of particular

value to online grocery logistic plans. Local area discovery and multi-agent learning are









two such technologies. In multi-agent learning the bots learn in "dynamic environments"

such as that which exist between different networks of companies interior supply chains.

Business rules markup language (BRML) is an intelligent-agent facilitated

technology that could help heterogeneous e-commerce-related applications to exchange

"rules." It is referred to as an "interlingua" technology that was developed by IBM while

the company was working on the Business Rules for E-Commerce project. What this has

to do with intelligent agents is that bots have been relying on the knowledge interchange

format (KIF) to transfer knowledge. BRML is supposed to add "prioritized conflict

handling" capability to KIF, which is important for maintaining business rules and when

transferring information between systems such as those between different internal supply

chain partners.

West and Hess (2002) say that software agents (intelligent agents) should be

employed to help users with interactivity and difficult spatially relevant jobs. This also

pertains to the user friendly, or procedural aspect of GIS knowledge management, even

when the procedures to use the specific GIS package are contained within the instruction

manual that is geared for GIS programmers and analysts (West and Hess, 2002).

Intelligent agents can facilitate comparison shopping. Kim and Eom (2002) suggest

using intelligent agents to not only allow customers to compare product prices, but also to

compare products on multi attributes. For online grocers the various attributes could be

product freshness, nutritional information, average delivery time per product, and more.

The article states that just as important as allowing your customers to do comparison

shopping through agents, is the need to not block competitors' intelligent agents from

entering your site to perform comparisons (Kim and Eom, 2002). This is because if a









store does not appear on an online shopper's online comparison list because that store

blocked the incoming agents, that shopper might think that particular store does not carry

the item that the shopper is looking for.

Intelligent agents are discussed by Tsou and Buttenfield (2002) in their literature

about GIServices (GISServices are explained below). Agents can be used on a distributed

GIS-enabled computer architecture to find and bind spatial data objects across networks.

According to these authors, three intelligent agents perform the distributed GIS system

find and bind functionality. They are filer agents, interpreter agents, and decision agents.

Collectively, these agents find the requested information, filter out unnecessary data,

bridge heterogeneous data environments, and autonomously make decisions such as what

server to use to send the requested information to the GIS user. Sometimes spatially

enabled artificial agents communicate between themselves using the knowledge query

and manipulation language (KQML) (Tsou and Buttenfield, 2002). Online grocers with

more than one distribution center should consider using this type of technology when

adopting GIS-enabled delivery solutions.

Although not strictly necessary for the development of delivery zones, swarm

intelligence and complexity are scientific fields that have recently come to the fore in

some business strategies. A primary reason that this literature is reviewed is because of

the intuitive value that these two concepts can have for the construction of delivery

regions in a geographic enterprise optimization system (GEOS).

Literature exists that explains the combination of GIS with "swarm intelligence." A

project undertaken by the Department of Geography at Southwest Texas State University









incorporates swarm intelligence with GIS to "study and simulate" various entities in

supply chain situations (Zhang, undated).

Swarm intelligence can loosely be defined as a neural network of interacting

intelligent agents. Whereas a neural network is commonly a closed network, meaning that

inputs and outputs are controlled, swarm intelligence can denote the utilization of

intelligent agents moving over disparate systems to detect unforeseen patterns (Payman,

undated).

Eric Bonebau is one of the pioneers of applying the concept of swarm intelligence to

managing business. He is a coauthor of "Swarm Intelligence: From Natural to Artificial

Systems, 1999. According to Fredman, ants set up supply chains to accomplish

sophisticated tasks, and the emulation of these processes can help companies deal with

complex environments. Her literature explains how logistics is a "natural application of

these ant-foraging algorithms."

Conventional supply chains use more or less centralized methods to forecast demand

and fulfillment. According to Roy (1998) centralized planning methods can be disrupted

by fluctuations in customer demand. More suitable than centralized planning is

conceptualizing the supply chain as a supply web, and incorporating swarm intelligence

to traverse the web to better calculate and compensate for demand fluctuations (Roy,

1998). One particular statement by Roy (1998) summarizes a main point of my emphasis

on profitability thresholds as driving forces behind the shape of delivery zones. This is

"any changes in customer demand...can easily affect how feasible or profitable a given

plan may be. These unplanned events will always happen, so it is desirable to create a

management system that adjusts to them more gracefully" (Roy, 1998). He uses a vehicle









analogy is used to demonstrate this point. Simply stated, instead of having a "centralized"

vehicle and adding parts to it, it can be better to conceptualize having thousands of parts

that "flock together" to become a vehicle (Roy, 1998). This concept may be very useful

when parceling groceries together for individual orders or routes.

According to Roy (1998), if swarming intelligent agents learn about region-specific

pricing dynamics they could help "develop new strategies to take advantage of changing

market conditions. Roy (1998) says that each swarming agent can individually generate

"internal make" suborders. It is not difficult to visualize how this could facilitate

packaging and shipping of groceries purchased online. After some time the agents can

attain the ability to "create new forecasting techniques and learn which of the techniques

are most accurate" (Roy, 1998). When it comes down to it, accurate forecasting is the

linchpin that online grocers depend on. At an extreme, if forecasting could be 100%

accurate, no wasted efforts in packing or shipping would be incurred, which would of

course result in significant competitive advantage.

Enterprise Resource Planning (ERP) is an established technological methodology

that is used to facilitate and assimilate information flows between and within departments

in an enterprise. ERP uses centralized databases, common user interfaces, and

sophisticated middleware to optimize the information flow within a company, or between

supply chain members, thereby increasing the accuracy and timeliness of inter-corporate

and intra-corporate decision making. ERP vendors are increasingly incorporating

forecasting functionality into integrated applications. J.D. Edwards sells the Demand

Planning 4.0 collaborative application module to enable partners to better make forecasts

throughout the supply chain (Ferguson, 2002a). According to Ferguson, the software uses









"market intelligence" and demand information to predict product demand. Although the

article does not state from where the market intelligence originates, it would be desirable

if the information comes from current industry-specific marketing research data. Demand

Planning 4.0 also accounts for holidays and promotions when forecasting demand. A key

word in the description of this supply-chain centered product is "collaboration." This

effort by J.D. Edwards shows that there is importance in having forecasting information

readily available when collaborating throughout the supply chain. More collaboration-

related literature is reviewed below.

Southwest Airlines is saving $10,000,000 annually by using swarm intelligence to

help route cargo (Fredman, 2003). Problems were arising from ground crews using their

judgment about what flights they should forward luggage and other cargo to. Naturally,

the cargo handlers would wait for the next flight that had space in the cargo hold that was

going in the same direction as the luggage. This procedure, however, was not optimal, as

discovered after using swarm intelligence. This technology revealed that many times it is

better to leave the luggage on a plane than to use ground crew manpower to change the

luggage in mid route. Southwest's freight transfer rates have reduced by up to 80%, the

necessity for storage facilities has been decreased, and payroll has been reduced because

of the decrease in necessary manpower (Fredman, 2003). Other noteworthy companies

that are using swarm intelligence-related solutions effectively are Unilever and Capital

One. As stated by Fredman (2003), the use of swarm intelligence in complex

organizations helps these organizations harness their tacit knowledge. Tacit knowledge is

knowledge that is difficult to codify in databases, but has high value because of the

exclusivity it has to the company that is using it. In other words, tacit knowledge can give









a company its competitive edge. This is precisely what is needed for businesses to

succeed in the low profitability margin online grocery industry.

As expected, "swarming" intelligent agents perform tasks that mimic life forms such

as ants, bees, or birds which exhibit swarming behavior (Resnick, 1998). Foraging for

data (like ants foraging for food) is the primary task of swarming agents. It is not hard to

imagine how swarming intelligence could help create delivery zones. Swarming agents

could collect data from tens of thousands of entities in delivery regions such as PDAs,

trucks, databases, and directories, and present this information in real time to strategic

decision support systems.

Deborah Gordon (2002) published an interesting article about task allocation within

an ant colony. Amazingly, although essential tasks such as caring for the young ants, nest

construction and maintenance, and foraging are accomplished in a very efficient manner,

nothing is in charge of "managing" these operations (Gordon, 2002). According to the

literature, it seems like some of the task allocation within a nest depends upon the number

of ants that are already involved in performing that particular task. For example, if an ant

leaves the nest to forage for food and does not return, no other ants leave the nest that day.

But if it does return, successively increasing amounts of ants will leave to find food and

perform other tasks like carrying any dead ants back to the nest.

Communication between ants is performed when the insects touch their antenna.

Also, ants emit an odor that is specific to the task that they are performing. The ants do

not tell the other ants what to do; instead it is the interaction pattern that determines the

probability that an ant will perform a job (Gordon, 2002). Another relevant fact is that

because ants live only a year, there is no handing down of knowledge from one









generation to another. The work that gets done is due to indigenous knowledge that the

nest collectively possesses.

Gordon's (2002) research alludes to ongoing work that uses intelligent agents to

simulate the collective jobs that are performed by social insects such as ants and wasps. It

should be relatively easy to imagine the benefit that the incorporation of these types of

agents into an information system that creates delivery zones. The potential decision-

making accuracy, resilience, forecasting ability, and longevity of such a system are some

of the benefits that make the utilization of swarm intelligence a technology that is

worthwhile of study by online grocers.

A relevant project undertaken at the department of Zoology at Michigan State

University is the Multi-Agent Based Economic Landscape (MABEL). The project

simulates the group dynamics of individuals in institutional and organizational

environments by using swarm intelligence with GIS (Mabel, 2002). Although the

incorporation of intelligent agents into GIS-augmented online grocery delivery routing

applications has yet to develop, many of the technological tools already exist to make this

combination possible.

Ants and bees have been studied to determine how these insects efficiently work

together to accomplish sophisticated tasks such as procuring nutrition or building

tectonically durable nests and hives. These types of studies are not only the

underpinnings of swarm intelligence theories and practices, but also integral aspects of a

field called complexity science.

Van Uden et al. (2001) call complexity science a body of knowledge that is "not

trivial" for business, and the "contender for the top spot in the next era of management









science." Carol Kennedy (2000) says that complexity science has "exciting possibilities"

for various fields including logistics. The Southwest Airlines example above was an

instance of a company using artificial intelligence to solve a "chaotic" situation.

Complexity science recognizes that systems, whether those systems are biological,

technical, or organizational, exist in a chaotic world where forces are constantly

impinging upon the system. Regions, such as delivery regions serviced by online grocers,

can clearly be considered examples of systems existing in a chaotic and complex

environment. Whenever online orders are coming into the grocer's system, the regions

that need delivery should change shape or composition. Various complexity issues are

inherent when considering the delivery, not least of which is the profitability threshold

that encompasses myriad variables such as packing time, crime rate along the route, fuel

costs, driver familiarity with the area, customers) lifetime value, competitive pricing,

traffic en route, position of other delivery drivers, customer payment history and payment

method, customer profile, and much more. The boundaries of the region are a constantly

changing factor because the shape and geographic area of the region's boundaries can

allow, or disallow the region to meet a profitability threshold.

Many properties make a system complex, such as "incompressibility" which is the

inability to explain a system in a level that is less complex than the complexity level of

the system itself, without losing some of the explanatory aspects of the system. Another

notion that is perhaps more relevant to online grocery delivery methods is that a complex

system is both deterministically chaotic and anti-chaotic. Deterministically chaotic means

being "incredibly sensitive to small disturbances," and anti-chaotic means being

"incredibly insensitive to large disturbances" (Van Uden et al., 2001). It is probably best









for a delivery zone delineation system to be resilient enough to withstand large,

unprecedented spikes in orders, while being selectively reactionary to small patterns or

states that could make a region profitable or not.

Van Uden et al. (2001) sum up complexity science by saying that it is the body of

knowledge that claims that "everything is connected to everything else." This author's

literature also tells how unwise it is to concentrate on select parts of a system while

ignoring the interactions between these parts. If we consider delivery zones as parts of a

logistics system, we might find that interactions between these zones are important in

determining if the zones should be combined into one deliver zone to meet the

profitability threshold, or remain separated.

Complexity science tells us that permanent boundaries never exist (Van Uden et al.,

2001). Here again is an argument in favor of changing the boundaries of delivery zones in

real time, depending on the overall logistical, operational, and profit "habitat" at hand.

"In complexity science, all boundaries are emergent and temporary" (Van Uden et al.,

2001).

Another complexity science maxim that should be followed by online grocer

delivery zones is the property of self organization, which is synonymous with anti-chaos.

As was explained in the section about swarming insects, complex systems have the trait

that they can self organize. Concisely stated, when there are a large amount of elements

in a large state space, when the conditions are random, these elements "tend to converge

into small areas of this space" (Van Uden et al., 2001). The delivery zones should do

exactly this. Instead of considering each and every customer as an individual profit

creating entity, the system should converge the customers into profitability zones,









regardless of the geographic proximity of the customers with each other and with the

store (up to a certain range).

Interestingly enough, this literature says that a trait of self-organization is that if the

starting conditions are similar, "a quantitatively similar pattern will always emerge" (Van

Uden et al., 2001). This can be somewhat reassuring for online grocers because there will

of course be times when a certain proportion of the deliveries will be similar to previous

deliveries. This phenomenon should significantly aid in forecasting and pre-packing the

groceries.

Stephan Toffler's book "Adaptive Corporation" is based on many complexity

science maxims. He advises the usage of "complex adaptive systems" to cope with the

"unpredictability of the modern marketplace" (Lloyd, 2000). Similar to the self-

organizing nature of complex systems, Toffler's "Sense-and-Respond" system allows the

realization of benefits such as mass customization and the ability to capitalize upon wild

marketplace swings. This book shows the growing awareness of the viability of running

organizations as complex systems, instead of operating the organizations hierarchically.

The ability to mass customize can be very important for online grocers, especially if we

consider packaging as an operational entity that also can be mass customized which

indeed it is because the contents of each package is different than each other package.

Therefore, the concepts offered in Toffler's book can apply to online grocers. Both the

meaning of "complexity science" and "complex" can vary depending on who is defining

the terms) (Van Uden et al., 2001). Christoph Adamnil (2002) said, "Nobody knows

precisely what is meant by the word 'complexity.'" John Casti (2001) attempts to clarify

what the meaning of complex is. By using the examples of infinitely continuing repeating









decimals as opposed to finite decimals, he says that of course the infinite non-repeating

decimals are more complex. He goes on to say that randomness is also measured by

degrees, while Adamil (2002) asserts that randomness is the opposite of periodical

occurrences. The relevance of this literature is that an organization might want to

determine the degree of complexity or randomness of occurrences in an environment

before deciding on what type of system to implement. If one agrees with Adamil's (2002)

assertion that "randomness does not give rise to organisms," then before a company

establishes a system to cope with complexity, the randomness of the environment should

be determined in order to predict if the system will be viable that is, if any type of

analogy between an information system and an organism exists. Casti gives a general

outline about how to determine complexity and randomness in various environments.

Adamil (2002) views complexity as the amount of information that an organism stores

about its environment.

Global companies are beginning to become aware of the value of complexity science

in their business operations. General Motors, Capital One, and Ford are three such

companies (Wujciak, 2003). Perhaps the example most germane to this dissertation is that

of Ford. Ford allowed a large array of automobile configurations to be ordered by

distributors, which created difficult to manage manufacturing schedules and inaccurate

demand forecasts. Ford solved the configuration problem by using methods to reduce

complexity.

One possibly useful way of thinking about "populating" a routing information

system that creates delivery zones on the fly is a concept offered by Adamil (2002). He

says that entropy is "potential knowledge" and that "sequence entropy" is a length of a









tape, and the marks on the tape is the information. Measurement, which in the case of

online grocers would be the geographic measurements of the delivery zones, populates

the tape. In other words, the geographic measurements of the delivery zones turn the

online grocer's entropy into information. Most relevantly, Adamil (2002) states, "the

information-filled tape allows you to make predictions about the state of the system that

the sequence is information about." In this case the state would be the state of online

orders at any particular time, and the predictions would be forecasts of those orders.

Intuitively, Adamil (2002) says that a "complexity catastrophe" is a rapidly changing

environment. We could view all of the above-stated online grocer failures as resulting

from complexity catastrophes. That is if we consider "rapidly changing" to be the state

caused by the reception and response to online grocery orders. "If the changes are fast

and extreme, not only will the organism be maladapted to this new environment, but also

its measurable physical complexity will have decreased commensurately" (Adamil, 2002).

This is exactly the situation Webvan found itself in, which of course led to its demise.

Therefore, any online grocer that begins a delivery initiative must keep its physical

complexity at the utmost minimum that is necessary to fulfill the orders. In other words,

multimillion dollar distribution centers, such as those created by Webvan, may not be

able to cope with the dynamic environment created by online grocery sales.

Online Grocers' Store Locations

If an online grocer has not yet decided on a site to open the brick-and-click store, the

site selection process becomes highly relevant to the businesses' overall strategy. Thrall

(2000) asserts that investment in retail location is one of the most important investments a

business can make. It can also be the most costly.









Moreover, when one considers that the central store is the epicenter of grocery

dispatches, which in turn can contribute to the success or failure of the company because

of the high delivery expense and the low profit margin, the site selection process takes on

a high magnitude of importance. There is much literature that extols the value of GIS for

the retail industry (Grimshaw, 1989; Brown, 2000; Murayama, 2001a; Thrall, 2002a;

Hakala, 2003). Although I have found no literature that deals with the use of GIS to

create metamorphic delivery regions for online grocers, an abundance of literature exists

telling how important GIS is for retail location, delivery, and retail marketing.

Online grocers should use heuristics and techniques that are different from brick and

mortar grocers when deciding on the location of their store or distribution center.

Although the term "online grocer" is used throughout this dissertation, I suggest that any

store selling online should be of the brick-and-click kind. This paper will refer to

exclusively online grocers as "pure online grocers."

Vlachopoulou et al. (2001) enumerate transportation type(s), customer type(s),

competitor locationss, and sales levels as some of the factors to consider when choosing

a site. The authors recommend using GIS to facilitate the site selection process because

of the data and spatial modeling complexity involved. Site selection tools fall in various

categories. These categories include analog models, rules of thumb, linear programming,

simulation, checklists, gravity models, and linear programming. The "parasitic" approach

is mostly used by smaller stores that copy the location decisions of larger retailers (Clarke,

1998).

A broad rule of thumb for online grocers to consider is that goods that are consumed

frequently should be distributed from a dense network of locations (Borchert, 1998).









Groceries certainly fit this category, so an online grocer will have to reach a tradeoff in

the site selection decision as to have more distribution centers, or stores, and less delivery

trucks, or vice versa.

Hernandez et al. (1998) say that because retailers have notoriously been "cavalier" in

their approach to store location. Techniques for deciding on the location of a store range

from using checklists to sophisticated artificial neural network (ANN) systems. Moreover,

some location decisions are in response to location decisions made by the competition.

To better strategize retail location, some companies have made semi-autonomous

property divisions within the company to manage store location (Hernandez et al., 1998).

This means that smaller scale online grocers might have to have some type of "equalizer"

because these stores will probably not have the resources to create an entire department

devoted to store location or even marketing. Therefore, online grocers could use

Hernandez et al.'s (1998) tactical, or local marketing, strata to help compete. Hernandez

et al. divides retail location techniques into three strata strategic, monadic (location mix,

such as relocation, re-fascia, and re-merchandising), and tactical (local marketing). One

tactical technique that online grocers should be able to efficiently capitalize upon is

changing online food prices in response to competitors' prices. Although brick-and-

mortar stores can do this by using the UPC (universal product code) and point-of-sale

technology, these stores must rely on periodic advertisement to inform customers of

prices. Online grocers who have developed a significant customer base can inform these

customers of price changes and coupons by e-mail or as soon as the customers log on to

the website.









The maxim that those who do not know history are doomed to repeat it can be

applicable to online grocers' site selection strategies. Thrall gives a detailed account of

the family "club," showing how the amount of people per dwelling has decreased though

time in the United States (Thrall, 2002a). Borchert (1998) supports this fact, while also

talking about decreasing birth rates of families. The ramification of these and other

demographic trends should be looked at closely by online grocers when deciding where

to locate.

Borchert (1998) writes about how lower level stores that were located in city centers

are losing their place in established retail hierarchies. It might be possible, however, for

online grocers to stem the tide of location displacement by virtue of their delivery

strategy. This conjecture remains to be seen, however. Nevertheless, the reinvigoration of

many downtown areas, plus the governmental dissuasion of commuters to use cars in

downtown areas (primarily in Europe) (Borchert, 1998), can also have an influence on

online grocers location strategy.

Clarke (1998) takes the utilization of GIS for retail location planning and divides it

into three eras, before approximately 1985, when the technology was not used at all for

retail location; from that time to the late 1990's, when GIS gained a foothold in many

retail organizations; and the present time when data mining and optimization techniques

are used with GIS for retail site selection. He supports his threefold division by showing

that retail location has become much more complicated recently, necessitating the usage

of artificial intelligence and optimization techniques.









Geographic Information Systems for Traffic-Flow Simulation

Recently efforts have been made to use GIS for simulating traffic flows for various

reasons. Considering the great amount of money to be made or lost with an online

grocery, simulating the time to delivery should be an integral aspect of any strategic plan.

Above, it is described how Punakivi and Saranen (2001) used simulation to

determine the delivery times to houses using various criteria such as attended and

unattended deliveries. Simulation is also used to determine traffic flows within cities'

main commerce areas and between different city centers (Medda, 2003). Traffic

simulation can be divided into macroscopic, mesoscopic and microscopic traffic flow

simulation. Mesoscopic simulation shows traffic flows at a scale between macroscopic

and microscopic. Nokel et al. (2002) have published research about mesoscopic traffic

simulation, while Wahle, Chrobok, Pottmeier, and Schreckenberg researched microscopic

traffic flow simulation. The research about mesoscopic traffic simulation states that there

are many times when the area simulated must be wide enough to be considered

macroscopic, but the finer level of roadway detail necessitates a granularity that the

authors classify as mesoscopic (Nokel et al., 2002). On the other hand, Wahle et al.

(2002) contend that because of the increasing processing power of computers,

microscopic traffic simulation is becoming more feasible. The article also says that

microscopic traffic simulation is valuable as a transportation planning tool which helps

vehicles navigate the roadway easier.

If a comprehensive grocery delivery plan is to be implemented, simulation using all

three scopes of simulation (macro-, meso-, and microscopic) should be considered.

Wahle et al. (2002) say that the output, such as travel times, from traffic simulation









systems can be used as input for other intelligent systems. This is precisely what is

necessary for a dynamic (real time) grocery delivery schedule information system. The

system tested by Wahle et al. (2002) supposedly functions in real time.

Simulation can be used to overcome the current analytical and dynamic modeling

inadequacies of GIS applications (Wu, 1999). He recognizes the importance of modeling

space-time spatiotemporall) processes through simulation with GIS, and is supported by

literature by Bernard and Kruger (2000). Integration of GIS with spatiotemporal models

will allow the "forecasting of space-time patterns in a reasonable, repeatable and

consistent way," precisely what would be required of an application that shows delivery

trucks' performance based on profitability thresholds. While noting that GIS-based

simulation is quite complex, Wu (1999) says that three contributions can be made to

decision makers by this type of simulation description, prediction, and prescription.

Relating to the macro, meso, and microscopic simulation scopes stated by the authors

mentioned above, Wu says that GIS-enabled simulation may reveal relationships between

micro and macroscopic behaviors. As applied to grocery delivery, the profitability

ramifications related to small deviations in order times might not be discernable in a

microscopic simulation, but might be noticeable if the simulation was run on a

macroscopic scale. As stated by Wu (1999), the "purpose of simulation is to see how a

global structure is evolving from uncoordinated individual behaviors."

Another benefit of GIS-enabled simulation noted by Wu is that the simulation results

can be stored in GIS "scenario libraries" for use in later simulations (Wu, 1999). Bernard

et al. discuss the integration of GIS with simulation in some detail. The difference

between coupling and integration of GIS applications is explained. Coupling is









transferring data between two GIS, while integration is a "monolithic" model where GIS

and simulation tools are implemented on top of a common data and method base

(Bernard, 2000). The literature goes on to explain common ground between initiatives

undertaken by both the OpenGIS Consortium (OGC) and the Simulation Interoperability

Standards Organization (SISO). The OGC's mission is to facilitate the worldwide

proliferation of "geoinformation services." This is to be accomplished in part by the

utilization of two models and a specific architecture, called the Open Geodata Model, the

Information Community Model, and the Service Architecture. Much of this initiative is

fashioned after the web services paradigm. Of similar importance is the High Level

Architecture which was developed by the US Department of Defense, and is supposed to

be a standard for interoperable simulation components (Bernard, 2000). According to this

literature, these initiatives are necessary because there is a "gap between the offered

interfaces and services in the GIS area and the requirements of the simulation area."

The integration of simulation and GIS can be highly relevant for online grocers who

deliver. This integration should allow the use of "what if" analyses to ascertain if delivery

regions will meet their profitability threshold or not, which, in turn, will allow managers

to better decide when to pre-pack the groceries that have the most probability of being

purchased by customers in those regions.

Geographic Information Systems for Vehicle Routing and Logistics

Pain Stubing of Ernst and Young said, "Execution is the most important thing for

any retailer" (Reardon, 2000). This can not be overstated when considering online retailer

delivery dispatches. GIS, and the role the technology plays in routing, can help online

grocers dispatch delivery trucks in the most cost-effective manner.









Closely related to traffic simulations are routing scenarios. There are significant

amounts of literature that pertains to the role GIS plays in vehicle routing that can apply

to online grocery delivery. Sometimes the GIS can be a component in a spatial decision

support system that helps determine the best routes.

Campbell et al. (2001) tell how GIS has helped create new logistical solutions. Their

paper talks about a hybrid distance approximation solution to assign routes to snow

removal trucks in Montreal. Their hybrid model strives to "combine accuracy of shortest

paths ... with the simplicity of approximations." One aspect tying this research to my

dissertation is that the shortest path algorithm alone is perhaps insufficient to determine

the best route for delivering groceries. There are other factors that might make a longer-

distance path preferable. Essentially, these authors' hybrid method speeds up the routing

process by creating a reduced road network that eliminates superfluous roads. According

to the authors, the calculation of routes in very large networks can be cost and time

prohibitive (Campbell et al., 2001). A problem with many routing tools is that they view

the network and traffic costs as "static" entities, or entities that have attained an

"equilibrium" state (Wu, 2001). This is clearly unacceptable for fast changing online

grocery delivery schedules. The article reinforces this idea by saying "... a static model

of congestion is an oxymoron." The need for forecasting traffic flows and congestion is

addressed in the article. The acronym GIS dynamic traffic assignment (GIS-DTA) is

coined by the authors. Aided by a GIS, analysts can change delivery routes according to

changes in predicted traffic flows. This simulation was performed using Arc/Info GIS.

The importance of coordinating delivery schedules with traffic congestion is underscored









by Browne as he says that delivery reliability decreases as congestion increases (Browne,

1993).

A notable amount of recent literature exists about using simulation in conjunction

with spatial decision support systems (SDSS) to solve the vehicle routing problem.

While explaining their Hierarchical Path View Routing (HPVR) model, Huang

breaks down vehicle routing systems into two broad types. One is where individual

vehicles perform their own routing calculations using CD ROM based maps and on-board

computers. The other method uses a centralized path (route) discovery model. The

centralized path discovery model was demonstrated by Huang (1997) to be less expensive

to implement

The HPVR model is relevant to this dissertation. The model uses only a subset of all

the roads possible to suggest a route for a vehicle. They call this method "fragmentation"

of arteries. The fragments are classified by road type (highway, side road, etc). This

method makes routing computational costs cheaper because not all the roads are used to

calculate any particular path from origin to destination. Because of constantly changing

road conditions, the status of routes must be recalculated frequently. If HPVR is used,

this frequent recalculation of routes can be performed more quickly because fewer roads

are used within each hierarchy.

It is worth mentioning here that the hierarchical set that includes highways would be

used to calculate paths between regions, which may not be necessary for local grocery

deliveries. However, this model has merit because if it is employed by local grocers,

unnecessary or superfluous roads may be omitted from local route calculations, thereby

speeding up the route-optimization process. Alternately, if the online grocer wishes to









branch out (for example, on weekends) to other customer regions, it could incorporate the

main artery hierarchy into the analysis. Also, although I don't propose using a

hierarchical method per say to form delivery routes, the hierarchical segregation of roads

for logistics and transportation planning has intriguing potential to be used for further

grocery delivery route optimization efforts.

GIS can also assist with determining accessibility measures for roads depicted with

ITS (intelligent transportation systems). Miller et al. have developed a method of ranking

the accessibility of arteries using ArcInfo (Miller, 2000b). Their space-time accessibility

model (STAM) calculates the accessibility of nodes in a network based upon drivers'

activity schedules. Activity schedules are essentially the times that a specific set of users

will use the roads. Accessibility to roads will be greater or less depending upon the time

of day. Based upon "travel diaries," drivers' activity schedules are divided into

mandatory and discretionary travel times. A similar concept might be used to group

online grocery customers into mandatory or discretionary delivery time windows.

Another consideration online grocers might have when developing a logistics

strategy is the type of traffic jam that occurs during a certain time of day. Mahnke and

Kaupuzs (2001) classify traffic flow into three "regimes." The first is a free flow of small

densities of vehicles, which is obviously the preferable time to embark on a delivery

route. Next is the "coexisting phase," where traffic jam clusters coexist within free

flowing traffic. And the "viscous overcrowded situation," where a high density of cars

move at low velocities." These broad classifications could be assigned as variables to the

delivery zones to better ascertain if the groceries can be delivered on time. Quite possibly,

for example, easternmost delivery zones could be experiencing the first category, while









northwest sections of town may be undergoing the viscous overcrowding condition. With

close inspection, it could be discovered that this pattern repeats itself daily. This

information should help to enable the scheduling of grocery deliveries with close regard

to factors such as congestion patterns.

An inseparable aspect of online grocers' delivery solutions should be the

incorporation of GIS into their logistic plans. The entire concept of delivery zones is

merely a hollow conception if not executed with a GIS.

A company called IVU Traffic Technologies uses GIS as an integral part of its e-

logistics solution (IVU, 2000). This company shows foresight by creating a solution that

allows visualization of logistical spatial information to allow informed management

decisions. The company asserts that GIS functionality is important for e-logistics

solutions because of the importance of location visualization for management decisions.

IVU claims, as this paper also asserts, that data visualization, as facilitated by GIS, is

especially important for e-businesses that utilize e-logistics solutions. An example of

advances made in the GIS visualization field is 3D Analyst. This software application is

an Environmental Science Research Institute (ESRI) product that allows realistic

animations such being able to "fly through" a region (Thrall, 2002b).

A Korean governmental organization that incorporates GIS into an e-logistics

solution is the Electronics and Telecommunications Research Institute (ETRI) which

develops GIS / e-logistics postal solutions in their Postal Technology Research Center

(Electronics and Telecommunications Research Institute, 2003).

Kipling Holding AB, a German mobile internet technology company also develops

e-logistics solutions using GIS. Kipling's niche is the utilization of global systems for









mobile communication (GSM) in their e-logistic solutions. GSM is a prominent

technology used for second and 2.5 generation wireless telecommunication services

(GEO Community, 2001).

Global systems for mobile communications is showing itself to be a valuable

technology in the advancement of e-logistics solutions as evinced by an Indian e-logistics

solution provider called, most appropriately, e-Logistics Ltd. According to the owner of

e-Logistics Ltd., Mr. V. Sanjeevi, the usage of GPS to monitor trucks is more expensive

than using GSM in the e-Logistics solutions (GIS Development, 2001). Sanjeevi also

asserts that en-route delay of delivery trucks can be reduced significantly if e-logistics

solutions are utilized.

According to IVU Traffic Technologies, "about 80% of all data in a company is

location driven" (IVU Traffic Technologies, 2001). The company's FilialWeb technology

enables the selective presentation of spatial information. The word selective is important

here because even if companies are able to garner and store real-time spatial information,

it is another matter to be able to present it selectively to the operational employees that

must make a decision about whether a grocery delivery route is profitable or not.

Networks

LeHeron et al. (2001) explain how supply chains should be networks of tacit and

codified knowledge; and that within the last decade there has been a large-scale change in

the networks of food consumption. If localized food consumption, especially the

consumption that is fulfilled through online grocers, is thought about as being fulfilled

through a network of supply chain knowledge, then more innovative models of delivery

may be contrived. Using this mode of thought, online grocers might be able to devise









ways of having their suppliers coordinate with the delivery drivers to possibly deliver

more bulky items to the drivers as they are en route to or from the customer premises.

According to Balakrishnan et al. (2000), there has been "increasing emphasis in

recent years on customer orientation providing the right product at the right price, time,

and place." And this emphasis has "propagated up the supply chain."

Industrial networks have been defined as the "spatial pattern of sales linkages"

(Benenson, 1998). According to this literature, dense and complicated networks can exist

even when small sets of linkages (such as those established by small to medium sized

online grocers with their suppliers and customers) exist. These types of networks are not

unlike matrices, where the dimensions are equal to the number of interplaying objects.

Benenson et al.'s (1998) comparison of networks to multidimensional matrices lends

itself to the thought that some problems encountered by online grocers can possibly be

solved by viewing these problems with the assistance of multidimensional databases and

multidimensional visualization methods such as virtual reality. These thoughts should be

discussed in further research about online grocery delivery solutions. Indeed, albeit in an

abstract sense, Benenson et al. (1998) talk about discovering "visually meaningful

patterns" in multidimensional information about business networks. They give an

example of sales patterns that, they say, could be too complicated to interpret by visual

means if depicted on a one or two dimensional map. Moreover, this research cautions to

be aware of sub-networks that can exist within networks.

More in line with the intelligent transportation definition of networks, Zhou et al.

(2000) talks about modeling networks with objects. Under this model, objects that

embody both spatial and thematic information are used to show spatial entities on the GIS.









This literature says that most GIS transportation (GIS-T) applications use a feature-based

model to show roads and other pertinent entities. Features are like classes in object-

oriented programming and modeling. Therefore, GIS-T applications should go beyond

strictly representing features, and use the objects that result from the features. This is

similar to the widely accepted object oriented modeling and programming paradigm. By

using this object oriented approach, transportation networks can be more holisticallyy"

depicted because each entity such as roads, points, intersections, etc, can be objects that

contain the data and functions that work upon the data (Zhou et al., 2000).

Whereas Zhou et al. (2000) recommend the use of "virtual networks" to better depict

multi-modal traffic network analyses, virtual multi-mode networks exist by sharing nodes

and sharing points in a network that does not necessarily reflect the exact shape of the

real multi-mode transportation network. The virtual network abstractly shows the

topological relations of the multi-mode network. Evidently, this is a more efficient way

of representing multi-mode networks, as the representation of multi-mode networks, but

according to Southworth et al. (2000), this is very difficult. Actually, Southworth et al.

(2000), while attempting to model intermodal (similar to multi-modal) freight networks

discovered that it was easier to model these networks without using GIS! Online grocers

may find that a multi-modal delivery system consisting of suppliers trucks, company

trucks (or cars), and even possibly bicycles or employees that walk the groceries to the

nearest customers. A delivery model where these "walkers" and/or bicycle delivery

employees might even meet delivery trucks at predetermined locations to more rapidly

deliver to the "last mile" of customers may be depicted using multi-modal classifications.

Although I do not necessarily suggest this type of break-of-bulk delivery strategy in this









dissertation, the possibility of modeling multi-mode logistics might make it more feasible

to use simulation to better ascertain the cost effectiveness of such last mile break-of-bulk

points.

Southworth et al. (2000) claim that because most GIS packages depict networks

geometrically, the use of these packages to show multi-modal logistics models is

practically useless. This literature says logical intermodal route representations are more

appropriate.

Data Visualization and Data Representation

A scientific discipline exists called "visualization in scientific computing" that

contains collections of methods that are used to perform high-definition simulation

(Bernard, 1998). The existence of this discipline demonstrates the current emphasis on

the importance of data visualization. Gahegan (2000) restates the definition of

visualization as "the process of creating and viewing graphical images of data...with the

aim of increasing human understanding." He also lists visualization as an important

means to achieve collaboration by experts (Gahegan, 2000). Spatial data visualization,

especially when adopted by companies that use integrally spatial strategies, such as

routing and delivering groceries, should put significant emphasis on adopting

visualization technologies to make operational, managerial, and strategic decisions.

The value of clear inter- and intra-corporate visibility is articulated well by John

Fontana (2002), He said, "Companies must learn to read their supply chain's

performance...as a gauge of the health of their business." He goes on to tell how a

reflexive phenomenon occurs when something goes wrong in the supply chain (Fontana,

2002). This occurrence, often called the bullwhip effect (Turban, 2002) should be









detected as soon as possible to preempt undesirable ripple effects throughout an online

grocer's interior supply chain. Some undesirable outcomes from the bullwhip effect could

be overstocking or stock-outs in peak periods. The bullwhip effect can cause operations

to stock unnecessary inventory to hedge against unforeseen online spikes in demand.

Robert Malone (2002), executive editor of Inbound Logistics magazine, said,

"companies need greater visibility along as much of the supply chain as possible." Clyde

Witt (2002) supports the notion of the value of supply chain visibility. His article about

Ford's six-sigma quality control program tells how visibility along the supply chain is

important in achieving six sigma logistical operations. If a supplier fails to deliver on

time, or throughput on the factory floor drops below a certain amount, the proper people

can be notified, creating a clear picture of the events that occur throughout the supply

chain.

Jan Bowland, managing director of KPMG Consulting, agrees with the notion of

overall supply chain visibility as an important aspect of logistics. She asserts that

shipment visibility is rapidly gaining recognition as an important priority for shippers,

particularly inter-modal carriers (Kuhel, 2002). Bowland adds that the value of supply

chain visibility goes further than just the awareness of what product is at what location at

what time. Full spatiotemporal knowledge should help to alleviate the necessity for safety

stocks which are kept in case of emergencies. This is because emergencies are less likely

to occur if full visibility into the supply chain is attained.

Despite the relative awareness of the importance of visibility into supply chain

operations, the amount of executives that are actually using sophisticated spatiotemporal









monitoring models or spatiotemporal monitoring technologies is surprisingly low.

According to John Zipperer (2002), less than 10% of senior executives surveyed properly

track supply chain performance, less than 7% collect correct or relevant information, and

less than one third of the executives track supply chain performance beyond the home

corporation.

According to Zipperer (2002), telephones and fax machines are still the preferred

method for many buyers to send orders to manufacturers, which creates problems such as

unnecessary errors. The article goes on to mention some outstanding problems with

supply chain monitoring, while noting that the problem of lack of visibility throughout

the chain is the most important one. Because the visualization of the internal supply chain

of online grocers is of primary importance to pare unnecessary efforts, the lack of

awareness of visualization solutions must be addressed when incorporating a grocery

delivery plan.

Cartographic animation is valuable when it results in more intuitive judgments.

Using cartographic animation, users can see "geospatial transitions as they happen in

time, as opposed to simply viewing the end states" (Ogao, 2002). According to these

authors, cartographic animation can allow spatial phenomenon to be viewed holistically

instead of disparate "instances of time." The increase of bandwidth because of optical

fiber networks can facilitate the rendition of three-dimensional rendering technologies

such as virtual reality markup language across networks (Shiode, 2001).

Goodchild (2002) states that transportation research requires many models but

comparatively few types of data. The data that is used, however, is chiefly geographic.

Defining the combination of GIS and transportation science as "GIS-T," Goodchild









(2002) elucidates the three viewpoints of GIS-T, the map view, navigation view, and

behavioral view. Bergougnoux (2000) supports this standpoint and adds that these three

viewpoints can help researchers gain a perspective on the development of GIS-

Transportation (GIS-T) solutions. He also refers to GIS-T as "a significant area for

research and development." Goodchild (2002) uses nodes and links to describe the map

view. This is similar to Keenan's physical links and logical links. But Goodchild (2002)

notes some flaws with the link/node view such as features that overshoot or undershoot

the endpoints. Also, because streets are one-dimensional lines, information pertaining to

the respective sides of the streets is lost. Interoperability between spatial databases is

another problem. These undesirables are characteristics of the map view.

The navigation view "requires a massive extension of attributes provided in the map

view" (Goodchild, 2000). These attributes include factors that show hindrances to the

flow of traffic such as turning restrictions (for example, illegal U turns) and one-way

streets. An inhibitor to successful representation of streets in the navigation view is that

collecting data about the flow of traffic and representing that data in lane views is more

expensive than just showing single line, one dimensional streets.

The behavioral view adds the time dimension and takes into consideration how

discrete transport objects such as vehicles, boats, trains or people move on the network.

Goodchild (2002) says that all the different methods of representation necessary in the

behavioral view are not yet included in any one GIS package. Whether the logistics

model used by an online grocer needs the navigation view or the behavioral view can be a

topic for further research.









Research performed by Winter (2002) underscores the point made by Goodchild that

the representation of turning restrictions are an important aspect of the navigation view.

According to Winter (2002), every turn has a cost to it such as time used to decelerate

and then accelerate, along with waiting time; and these costs should be incorporated in

any representation of a transportation network used for logistical purposes. However,

incorporating turn costs in a network is expensive because they greatly increase the data

necessary to represent the graph, and makes it necessary to create more sophisticated

routing algorithms (Winter, 2002). One particular point made by Winter (2002) could

especially apply to online grocers delivery plans. He states that a circular tour is when an

entity leaves a point, makes a trip, and then returns to that point. This is of course what

the grocery delivery truck will do, so estimating the cost, amount, and location of turns

that will facilitate, and not hinder, the delivery truck making a circular tour back to the

main distribution center should be important factors in the overall logistical plan for

grocery delivery.

Data modeling features of GIS should be improved to allow better user interfaces

such as "exploratory" data analyses that allow users to progress through non-

predetermined paths (Spaccapietra, 2001). This functionality would obviously be useful

for grocery delivery.

West and Hess (2002) explain that metadata pertains to both the technical and

business aspects of a business, and that the business metadata is especially useful to the

end users. Much of what West and Hess (2002) say about the importance of intelligent

agents and metadata for end users is reflected by Tsou and Buttenfield (2002) as they

discuss both of these aspects in the context of distributed geographic information services.









According to Tsou and Buttenfield (2002) two type of metadata are required to

implement geographic information services (literature pertaining to GIServices is below)

- operation metadata and connectivity metadata. Operation metadata facilitates the

representation of cartographic information, such as coordinate projection and spatial

footprints, across heterogeneous networks. Data connectivity metadata specifies data

connection protocols such as java database connectivity (JDBC) and open database

connectivity (ODBC) that are used to access spatial data across the networkss.

According to this literature, the use of both operation and data connectivity metadata

enables geographic objects to be "more accessible, self-describing, and self-managing.

Dashboards are tools used to quickly and concisely visualize trends in corporate

performance by employees that have responsibility to monitor those trends. Often

referred to as digital dashboards (Ricadela, 1999a), these tools send important operational

indicators to the computers of corporate decision makers. Some of the uses of dashboards,

as stated by Whiting (2002a), are mitigating threats, highlighting business opportunities,

and depicting performance figures. The underlying technology used in dashboards are

data warehouses that amalgamate information from disparate data stores. According to

Whiting, GE is a prominent user of dashboard technology. The company incorporates

uses a corporate-wide dashboard called the "cockpit" (Whiting, 2002a).

Dashboards are intra-network or internet-enabled business intelligence tools. A

difference between dashboards and portals is that dashboards seem to be more task-

exclusive, whereas corporate portals can include links to a diverse amount of information

sources. Nevertheless, dashboards are becoming popular and are worthwhile

investigating as an integral part of GEOS-enabled online grocer logistic plans. Also, the









term "dashboard" is appealing because it connotes control mechanisms in a functional

layout. The incorporation of spatiotemporal "instruments" into dashboards such as the

depiction of delivery zones could be useful for online grocer logistics strategies and profit

maximization.

PeopleSoft and Microsoft are two companies that have begun developing dashboard

solutions for businesses (Ricadela, 1999a; Ewalt, 2002). These applications, often

classified as business intelligence solutions, show potential to be incorporated into online

grocer logistic plans. According to Ewalt (2002), dashboards can give front-line

employees better decision making power. An integral part of a GEOS should be some

type of dashboard that shows the color-coded regions to be delivered to changing

dynamically. At times, operations managers might have to incorporate their judgment

about what route a dispatch should go on, especially if the dashboard-enabled GEOS

shows that two routes exist, but only one driver is available at the moment. An instance

of managers using their GIS-assisted judgment to dispatch delivery trucks is a reflection

of Thrall's assertion that, "GIS is one part of a larger information technology that may be

drawn upon to improve our judgment" (Thrall, 1995b).

Some similarities can exist between dashboards that contain GIS and geographic

information portals. Indeed, some literature refers to portals and dashboards

synonymously (Ricadela, 1999b). A differentiating factor between the two technologies

could be that geographic information portals include smart maps. Many aspects of smart

maps are similar to the functions that I said should be contained in GEOS in my

geography master's thesis entitled "The Conceptual Integration of Geographic

Information Systems into Enterprise Resource Planning." Those functions essentially









facilitate the retrieval of information pertinent to an area selected on a map. Called

"center pieces of graphical information portals," smart maps can show information not

usually included in a GIS such as real time information and interactive regions. As an

example, when a user places a cursor over a region of interest on a smart map, an

associated "subject tree" or pop up menu appears, providing the user with choices

germane to the region (Peinel and Rose, 2001). If an information source contained in the

tree is selected, a file will be downloaded from a web browser. According to Peinel and

Rose (2001), geographic information portals can help non-GIS professionals use relevant

spatial information because of the intuitive, visual nature of the information retrieval

aspects of the information trees. Their literature explains that logistics is an application

area where geographic information portals can be utilized. The real-time information

depiction qualities of smart maps might be suitable for the representation of online grocer

delivery zones.

Geographic Data and Surface Representation

Pienarr and Van Brakel (1999) summarize the value of data contained within GIS by

saying, "the value of the GIS is dependent on the quality of the data contained within the

system, undoubtedly making it the most important component of a GIS."

It could be financially prohibitive for online grocers to generate their own data about

roads and customer profiles at least during the first few years of operations. This is why

online grocers should be aware of the various types of geographic data available to them.

Pienarr (1999) divide online geographic data sources into five categories. They are,

educational, commercial, information providers, interactive mapping, and online data

searching. Educational resources include universities and other schools; businesses and









organizations sometimes provide GIS data, often on a subscription basis; information

providers include online GIS journals. Generator services, map browser services, and

real-time maps comprise the interactive mapping geographic data source of data. Online

data searching providers can have search capabilities using metadata, and permit transfer

of data by FTP (Pienarr and Van Brakel, 1999).

Declarative knowledge exists within metadata. When end users have to choose

between hundreds of map themes, declarative knowledge can help with their choices

(West, 2002). This could be especially important if the end users are not GIS

professionals, which is often the case when the average manager or end user is not

trained in cartographic principals, which are necessary to create professional GIS

coverages. Literature by Miller underscores this point as he discusses the potential for

improved geographic representation in spatial analysis through the use of GIScience

(Miller, 2000a). Spaccapietra (2001) says the type of end GIS user is changing. Most end

users are not GIS specialists, and GIS has not become interactive enough for

inexperienced GIS users to use quickly (Spaccapietra, 2001).

Swink and Speier (1999) also highlight the fact that the effectiveness of SDSS is

dependent upon the cartographical representation of the data. Swink and Speier (1999) go

on to tell that because of the increasing complexity of business data that must be

incorporated into maps (such as the data that pertains to logistics and grocery delivery)

the level of cartographic detail increases greatly. Many times it is impossible to show the

complete level of detail that is available, so GIS must be used judiciously to show data

patterns that help users make better decisions. Essentially, as the number of data points to

be considered increases, the complexity of the decision at hand increases Swink and









Speier, 1999). Despite this, Swink and Speier (1999) have discovered that better

transportation-related decisions are made when more customer zones are included in the

analysis instead of less. Results vary, however, if the decision makers have varying

degrees of "spatial orientation skills."

Also worthy of consideration is literature by Bittner and Frank (1999) who argue that

although most of the spatial representations rendered by GIS are currently based upon

analytical geometry, other ways of representing geographic space might be considered.

Most relevant is their proposition that "constraint-based" GIS representation be

considered. They also say that language can be a way to represent formal models on a

GIS. This too is noteworthy because the results of the calculations for delivery zones can

be expressed in words superimposed upon color-coded zones.

Somewhat supporting Bittner and Frank's (1999) research is research by Marble

(2000) who states that GIS inadequately uses the spatial tools available to it. Calling

current GIS-related computational approaches "myopic," he says that the dimension of

time has often been omitted in spatial analyses (Marble, 2000). Granted, since the year

2000, more literature has surfaced that pertains to spatiotemporality; but Bittner predicts

that GIS and spatial analysis are on the "brink of a major revolution." In a sense I agree

with him because much of the utilization of GIS to determine optimal (profit

maximizing) routes for dynamic online businesses like grocers has not yet begun to be

realized. This is an important point that is emphasized in my dissertation. Miller (Miller,

2000a) further expresses the inadequacies of current geographic representation through

GIS by saying that the inception and rise of GIScience might herald in a "comprehensive

re-examination of geographic representation in spatial analysis" (Miller, 2000b). Much of









his assertions are based upon the premise that because computational power is increasing,

so should the power of decision makers who use this spatial information. An example of

the utilization of computerization to help make spatial decisions is the SpaceStat and

DynESDA extensions for ArcView, which perform sophisticated statistical analyses, and

depict those analyses graphically (Anselin, 2000). Anselin's (2000) research also says

that much of the spatial analytical capacity of applications, especially if those

applications can be employed by mainstream GIS packages such as ArcView, MapInfo,

or ArcInfo, will allow non GIS experts to better perform these spatial analyses. If this is

true, decision makers in industries such as online grocery will benefit from the increased

ability to perform complex statistical spatial calculations to ascertain profitable deliveries

and more. Also, possibly as a way to make end users of grocery delivery spatial

applications acclimated to these types of applications, the users and decision makers

could be trained first on more simplistic location-allocation programs such as NEWLAP,

that have intuitive menu-driven interfaces (Lindquist, 2002). Because data (such as the

data in relational databases) is tabular, an important aspect of GIS is that it can combine

tabular data with cartographic data (on maps). When one considers that much customer

transaction data is initially stored in tables, this functionality apparently becomes

valuable. The value of GIS is increased further when combined with lifestyle data about

customers in retailers' catchment areas. In a United Kingdom survey most respondents

preferred the MOSAIC lifestyle database, with ACORN the second-most preferred

(O'Malley, 1997). Further, when extensions to GIS networking software, such as those

that have location-allocation functions, are implemented, catchment areas can be more

accurately determined. Geographic location-allocation software assigns demand areas to









supply centers while simultaneously maximizing supplier coverage and minimizing the

travel time for customers (Figueroa, 2000).

O'Malley et al. (1997) state that the most important data for retailers are the data that

are generated internally. Online grocers can have one important advantage pertaining to

internally generated data, which is the generation of internet-generated data. All data

relating to purchases made online can and should be added to corporate databases to

allow customer analyses. Moreover, many of these analyses can be performed in real

time using procedures like online analytical processing. Whereas brick-and-mortar

grocers use analytics (more on analytics-related literature is below) to generate coupons

on the back of sales receipts to be used on the next purchase; online grocers can generate

incentives such as coupons to be used on the current purchase depending on what is in the

customers virtual online shopping cart at the time. O'Malley et al. (1997) support this

notion, saying that "the future is likely to see greater use of real-time data for

analyzing..."

Geographic data is inherently multidimensional, and recent developments in

constraint databases are allowing better representation of multidimensional data.

Succinctly stated, constraint databases attempt to show infinite collections of points in a

finite number of dimensional spaces (Grumbach, 2001). An example would be to show

three-dimensional polygons as conjunctions of linear inequalities. Depending on the

stipulated constraints, query languages can perform operations on the points that

constitute the polygons.

Assessing the exact extent of catchment areas has always been problematic. The

reason for this is that the outer boundaries of geographic "sales cones" have fuzzy









boundaries that can only be estimated (Loffler, 1998). Sales cones are areas that radiate

away from the store, with the vertex of the cone representing the store (Christaller, 1933).

The outer range represents the farthest point at which a customer will travel to a store or a

store can deliver to a customer (King, 1984). This outer range is the boundary of a region,

where the geographic extent of the region is dictated by the distance decay influence on

the customers of a particular retail store that is located within the region. Thrall (2002b)

said, "The rate at which demand declines with distance to the retail center is known as the

distance decay of store patronage."

I claim that the shape of customer sales cones for online grocers will be different

than for strict brick-and-mortar grocers at that same location. Further, the sales cones for

an exclusively brick-and-mortar grocer will change if that grocer begins an online sales

and delivery initiative. Additionally, because of the data gleaned from georeferenced

online sales, the boundaries of the sales cones will become less fuzzy. Also, speaking of

borders of sales cones, it is relevant to note that a "zone of indifference" exists between

market areas where the attraction of neither market area is greater for a particular

customer (Akwawua, 2001).

Also, it is important for online grocers to calculate the decrease of the customers'

friction of distance that occurs because of the zero drive times, and how those non-

existent customer drive times affect the gravity model as pertaining to the "law of retail

trade gravitation," which is, the ratio of shares of turnover of two shopping centers or

central places is proportional to the ratio of their attractiveness and vice versa (Loffler,

1998; Thrall, 2000b).









Customer drive times should not be overlooked when developing an online grocery

delivery strategy. This is because brick-and mortar grocers' first criteria used when

ascertaining a stores prospective customers' trade areas is the customers' drive times to

the store (O'Malley, 1997). Therefore, if, through GIS, an online grocer can discover

areas where the residents' drive times to the nearest brick-and-mortar grocer are

comparatively high, the online grocer could use this information to cater to those

residents because they would benefit most by home-delivered groceries bought online in

comparison to people living closer to the store.

Loffler (1998) expects the relative cost of shopping trips to increase in the future

because of increased drive times to stores combined with greater fuel consumption. This

means that the value of purchasing groceries online should increase proportionately with

the increases in fuel prices and fuel consumption. Luoma et al. (1993) supply a somewhat

antithetical theory by saying that because of the increasing wealth of the customers, the

customers will drive even farther, regardless of the size of a shopping center that he or

she is going to (Luoma et al., 1993). Still another perspective is given by Borchert (1998),

who says that despite any improved customer spending ability, the amount spent for retail

goods increases only marginally. Although all of these postulates may be debatable,

online grocers should investigate and test these assumptions in their regions to better

decide upon marketing and delivery strategies.

Buor (2002) demonstrates how distance decay is relevant for medical services in

Ghana. It should also be mentioned that literature has been published that attempts to

refine some aspects of the gravity model. For instance, Hu and Pooler (2002) propose a

competing destinations model that tries to remove some bias from distance decay









parameters in conventional gravity models. The underlying proposition in the competing

destinations model is that an alternative location can "compensate for the bad features of

another alternative". Said another way, "destinations compete for the attention of

decision makers," (Akwawua, 2001). Luoma et al. (1993) support this by saying that

some models do not consider competing destinations that lie within a customers

"personal threshold."

The theory of intervening opportunities takes the competing destinations model a

step further. According to this theory, "the number of persons going a given distance is

directly proportional to the number of opportunities at that distance and inversely

proportional to the number of intervening opportunities" (Akwawua et al., 2001).

Akwawua et al. (2001) go on to say that some people use the internet to prioritize their

search before embarking on a route. Although there are competing destinations and

intervening opportunities within the regions) that a person is willing to travel, a person

will limit the destinations to which he or she will travel to those that are "perceived

clearly." Online grocers should capitalize on this assumption by having their web sites

outstanding in clarity and functionality.

Chen and Jiang (2000), while emphasizing the importance of integrating disparate

spatial data across networked systems, tell how an "event driven" model of

spatiotemporal database changes can be an effective way to utilize spatial decision

support systems (SDSS). The idea of using events as the basis for computer applications

is not new. Visual basic for applications (VBA) is an event driven programming language.

This means that an event, such as a mouse click, pressing of a command button, etc,

triggers the underlying code to run. Chen and Jiangs' (2000) literature, however, is an









attempt to use the event driven model to represent spatiotemporal data. The authors say

that representation of events in databases is a "hot research topic," and "this event-driven

approach ...provides a new way for simulating system workflow." Although their

research pertains to events within a land parceling SDSS, it is not difficult to extrapolate

from this literature how valuable an event based database schema could be for online

grocery delivery businesses. If a sudden or gradual increase or decrease of online

purchases occurs within two groups of geographically distinct customers, it would be

important to determine the causes) of this change in ordering patterns. Possibly, the

increase of online orders is occurring because certain sections of two adjacent regions are

serviced by two or more delivery trucks on certain days, decreasing delivery times and

increasing customer satisfaction. The event here would be the superimposition of parts of

two or more delivery drivers' routes (a union of area). The end "state" of the region

served by two routes simultaneously would be increased orders (possibly including new

customers). Therefore, the connection between the event and the end state, as shown by

the event-driven database schema, could show to management the importance of

periodically doubling up drivers in certain areas (possibly in areas of high-value

customers). Alternately, the correlation of events and end states, in this scenario, might

show management, after a cost / benefit analysis is performed, that the amount of

increase in customer orders is not sufficient enough to pay for the added expense of

doubling up of delivery trucks for that region or time period. For online grocers interested

in possibly initiating an event-driven database schema, Chen and Jiang (2000) list useful

operators (in augmentation to conventional SQL operators such as AND, OR, etc). Three

of these operators are the sequence operator (;) showing the sequence of events, the









periodic operator (P) designating an event that occurs periodically, the FIRST operator

- which designates that an event occurred before another event (Chen and Jiang, 2000).

Roddick et al. (2001) address other schema issues while talking about the

spatiotemporal receptivity (or non-receptivity) of databases. By using "schema

versioning," temporality can be added to a spatial database. This literature makes a

distinction between schema versioning and schema evolution by saying that schema

versioning is a special type of schema evolution. Schema evolution is the modification of

the database schema without any loss of data, while schema versioning is the ability to

query all data within a database "both retrospectively and prospectively." For dynamic

delivery zone related businesses, such as online grocers, knowledge of schema versioning

and evolution might be of some value. However, as Roddick et al. (2001) state, schema

versioning has not been incorporated well into spatiotemporal databases, but also say,

"spatial schema versioning would be a useful adjunct to many systems."

Spatially extended structured query languages (SQL) that use operators such as

overlap, direction, contains, and distance, are better at representing data, rather than

analyzing the data (Huang, 1999). Two more recent SQL efforts attempt to allow more

spatial analytic capabilities. They are SQL3 MultiMedia Specification (SQL/MM), and

Open GIS Simple Features Specification for SQL (OpenGIS SQL). SQL/Temporal is an

international standard for spatiotemporal data modeling (Peuquet, 2001). Many of the

functions contained in SQL/Spatial are similar to those in ArcView. Some operations are

VORONI, CONVEXHULL, DISJOINT, WITHIN, and DIFFERENCE. According to

Huang et al., the functions that create new types of spatial features are the most difficult

to define in SQL. SQL/Spatial can be run on a client / server architecture, with an









interactive front end for making queries and a SQL/Spatial server on the back end (Huang,

1999).

While considering employing any database schema technique, online grocers should

not forget that not all street databases are directly compatible with each other.

Interpretation of features (ontology) and level of detail are two aspects where

cartographic data can differ (Noronha, 2000). Objects such as turn lanes and traffic

circles might not be represented on all maps. Also, Noronha (2000) says that most

databases store addresses as ranges rather than individual address points. This clearly

could be a problem for online grocers. Another consideration is that if delivery drivers

will be navigating while using mobile cartographic devices (which they should be doing

in order to respond in real time to online grocery order changes), is that maps take much

longer to download than other types of representations. Bertolotto and Egenhofer (2001)

propose a more expeditious way to download maps, which allows initially only a partial

representation of the map that is being downloaded. This literature explains the

difficulties of segmenting portions of a vector map to be downloaded. These difficulties

do not necessarily apply to raster maps. One difficulty is that when maps are overlaid in a

GIS, the process of sending those maps over a network is computationally expensive.

Map generalization, which is decreasing a vector map's detail, is a "complex and time-

consuming process," (Bertolotto and Egenhofer, 2001). Nevertheless, this type of

problem must be solved by online grocers who want to send maps in real time over the

network to delivery drivers.

Surface representation of elevation also may have relevance to routing solutions. The

point at which a delivery zone meets a profitability threshold can be calculated by









incorporating some principles of fluid dynamics that use interpolated depressions and

relief of a map using a GIS. Atkinson says that measures of elevation are relatively

simple to apply to maps, and that this measurement "is of fundamental importance for a

range of applications." Moreover, supporting the assertion that profitability can more

intuitively, and perhaps more accurately, be depicted and calculated over a region is

literature by Chang and Harrington (2000) say that profit can form a "landscape." Their

literature goes on to discuss how different degrees of "fitness" can affect the profitability

landscape. If Chang and Harrington's (2000) concept of "consumer preferences" can be

considered analogous to changing consumer online grocery purchases, we can see how

the spatial representation of profitability might be worthwhile endeavor.

Etzelmuller (2000) discusses different ways of showing changes in surfaces by using

grid-based digital elevation models (DEM). By using DEM, it is possible to quantify

surfaces to compare the surfaces for aspects such as roughness, changes to the surfacess,

and "noisier" or less noisy surfaces (Etzelmuller, 2000). Some of this calculation is

performed by assigning wavelengths to surfaces, and having the higher amplitudes

representing relief, and wavelength valleys representing depressions. The DEM

calculations result in a coefficient that when equal to 1, represents small topographical

changes, and 0 or a negative number means larger topographical changes. This type of

allocation of coefficients to topography could have relevance to the "terrain" transcended

by grocery delivery vehicles. Therefore, this research deserves further consideration

when developing logistical applications for businesses.

Additionally, besides just quantifying surfaces in DEM using Etzelmuller's (2000)

wavelength technique, data pertaining to various features can be overlaid on digital









elevation models. This might be useful in a trough and relief method for calculating

profitability thresholds.

I propose that "conventional" representation of polygons on GIS are not suitable for

many business decisions, such as those decisions that must be made by online grocers

about most profitable routes. Cressie et al. (2000) reinforce this idea by stating,

"polygons are typically counties, health districts, or states, which are politically chosen

entities that often have nothing to do with the etiology of the phenomenon." In many

ways, using predetermined polygons for routing profitability analysis is too constraining,

and cannot lead to accurate determinations of the value of any particular route. Also,

attributes related to polygons change all the time. Stagnant cartographic polygons may

not be able to adequately show these changes. Indeed, Cressie et al. (2000) say that

causation of events can be better determined when data relating to polygons are modeled

according to time.

The literature pertaining to fuzzy points, fuzzy polygons, and spatial uncertainty is

provided because of the possibility of applying these types of uncertainty measurement to

online grocery delivery systems.

Besides the actual representation of polygons, the relationship between the polygons

and the points (customer nodes, delivery trucks, etc) must be considered. When some sort

of non-definitiveness or imprecision exists about the nodes within a polygon we can say

that the points exist with some sort of uncertainty (Leung, 1997; Morris, 2003; Robinson,

2003). In essence, fuzzy sets add a dimension to Boolean algebra that proposes every

proposition is either true or false. Fuzzy sets add a "middle," subjective possibility called

a membership function (Robinson, 2003). These membership functions can be shown on









an x and y graph as triangular, trapezoidal, or Gaussian. Fuzzy sets can provide an

infinite set of values that apply to variables (Morris, 2003). In other words, the degree of

membership of a variable to a class is expressed rather than the probability of

membership of a variable to a class (Peuquet, 2001). Robinson (2003) predicts that as the

sophistication of GIS increases, the need for human input for the definition of the fuzzy

set function should decrease.

We could consider grocery orders from customers within a region that come in to the

online grocer with no determinable pattern (completely stochastic) as random orders.

This could lead to decisions to be made under uncertainty. Some of these decisions would

be to have a buffer amount of inventory, or have extra drivers on standby in case

forecasted orders actually come in. Decisions made under uncertainty relate to fuzzy set

concepts, which have been attempted with GIS (Leung, 1997). A point or polygon

represented on a GIS can be considered fuzzy if either object's shape or location is

imprecisely recognized. Basically stated, a point or polygon can have degrees of

belongingness to sets of properties. The body of literature pertaining to using GIS to

make decisions under uncertainty is growing, however, the literature pertaining to

allotting degrees of fuzziness to points and polygons is limited. Robinson (Robinson,

2003) stated that although the use of fuzzy sets in GIS has grown in the last decade,

commercially available GIS that support fuzzy information processing is rare. The above-

cited authors' research can possibly be applied to grocery deliver zones, especially if the

shape of those zones are considered random or fuzzy. Moreover, because fuzzy sets allow

the meanings of natural sentences (that contain indefinite words like near, where, how,

large, and small) to better be projected on a map, the advancement of fuzzy sets in GIS









seems inevitable, and the online grocers that become early adopters of the technology

may gain valuable first mover advantage in this highly competitive market.

Cheng et al. (1999) say that most natural phenomena are bounded by fuzzy transition

zones. Can the changes in customer catchment region boundaries be considered natural

phenomena? If not, Cheng et al.'s (1999) assertion could also apply to "unnatural"

phenomena. Nevertheless, since online grocers' delivery areas may change, sometimes

hourly, the conception of fuzzy transition zones could apply to these regions. However, it

is difficult to measure the changes in transition zones, especially since fuzzy transition

zones may overlap in contiguous regions (Cheng et al., 1999). According to these authors,

regions change shape during "epochs;" and the changes in the regions are measured by

comparing their size at different epochs, during which time they might have shrunk,

shifted, or expanded. This is called state transition. Peuquet (2001) supports this view by

stating that regional boundaries, along with boundary categories can be fuzzy. This mode

of thought can be of use to online grocers who want to calculate to different degrees

about if a region is profitable or not before dispatching a truck.

Allan and Lowell (2002) propose another way of dealing with spatial uncertainty by

using "abjects." They say that all points, lines, and polygons on a map can belong to

several map classes, which denotes that the prominent map class to which these entities

belong is not always certain. Polygons whose classes are indeterminate are abjects.

Abjects that are surrounded by objects (which definitively belong to a particular class)

can, under certain circumstances be joined or merged with the class, thereby becoming

part of the object itself (Allan and Lowell, 2002).









Funamoto (2000) endorses the notion of multi-level "clumps" of points that exist

within geographic polygons. According to him, geographic clumps develop because of

various phenomenon, such as concentrations of cancer victims within a region. Defined, a

clump is a cluster of points, with each point having a circle of a certain radius emanating

from it. The "clump radius" is the collective radii of the circles. What this research seems

to be is another way of depicting geographic concentration, which could be of value to

logistic models. Funamoto's (2000) research is noted here because of the potential of

incorporating his theories into online grocery delivery models.

Clumps should not be confused with clusters, however. Clusters are the much-

debated macroeconomic theory by Michael Porter about how related businesses "cluster"

around each other (Martin, 2003).

When geographic data is represented in layers, as is commonly done by a GIS, the

resulting spatial entities (points, lines, and polygons) are quantified to arrive at certain

conclusions. Once the data is quantified, it can be studied by order, topology, or algebraic

methods (Yongli, 2000). An online grocer might transform visual cartographic

information into quantitative information by layering a coverage containing the distance

from the nearest distribution center with a coverage containing the amount of purchases

per week to derive a new coverage that can tell if closer customers order more often.

One way to show the quantification of map layers is through map algebra, which is a

way to sample continuous space, and show those samples on a template that contains

rows and columns. Interestingly, algebra, which is most often associated exclusively with

numbers, is actually a way to "facilitate integration and cross-fertilization of abstract

models," (Pullar, 2001). In this case, the abstract model is a map.









MapScript implements map algebra using C++ class libraries which include

functionality for arithmetic, trigonometric, exponential, and relational operations on

templates, where templates refer to arbitrary neighborhoods. Groups of cells within a

template are also neighborhoods. According to Pullar (2001), by using the cell / template

representation of neighborhood values, a "large class of high level statistical and

mathematical operations" can be performed, including time-dependent analyses. This

form of modeling can be used to examine changes across space and time. Pullar (2001)

says, "travel time is related to the cost, or work effort, to move a certain distance." My

addition to this postulate is that since time essentially equals money, if the work effort

results in enough profit, then the cost of the travel will decrease or even become non-

frictional. The relationship of map algebra to the method I am suggesting is that instead

of conceptualizing neighborhoods using two-dimensional templates, such as those used

by map algebra, we can conceptualize the neighborhoods (delivery zones) as three

dimensional entities, incorporating the profitability variable (or other variables) as the

third dimension. This could be called map calculus, but since the online grocer delivery

zones could change, often in real time, fluid dynamics may be the more appropriate way

to conceptualize this regional quantification. As mentioned by Bernard and Kruger,

adaptive geographic grids can be used to simulate the movement of fluid. Also, if

cartographic representations are three dimensional, hydrodynamic movement may be

simulated (Bernard, 2000). Therefore, if innovative thinking is used, modeling methods

such as map algebra and fluid dynamics may be very important steps for quantifying

delivery regions for online grocers. It should be noted, however, that because three-

dimensional data is not readily available on the commercial marketplace (Koninger and









Bartel, 1998) online grocers might have to pay for expensive data collection if this option

is employed. Also, according to Koninger and Bartel (1998), there exists a way to

represent solids called constructive solid geometry, which uses Boolean operations to

make more complex solid objects from simpler three dimensional objects. Pullar (2001)

explains that "mathematical morphology" within map algebra is the theory for analyzing

spatial structures. Two morphological operators, dilation and erosion, change the region

under study by either increasing or decreasing the size of the region. Also, cells within a

neighborhood template can either die, be born, or continue to live, depending on the

properties of the nearest cells. It seems feasible that the dilation and erosion operators, or

some semblance of them, can be used to create delivery regions.

When incorporating map algebra, or virtually any spatial quantification method,

some thought should be given to linear referencing, which can render the necessity of

locating points on the earth's surface unnecessary when performing spatial

quantifications (Scarponcini, 2002). This is done by using anchor points, such as

benchmarks, which are georeferenced points used as initial reference points. Then, from

the anchor points arbitrarily spaced points are placed. The points are one-dimensional

entities used to specify locations. Once points are arranged in a route or network topology

from an anchor point or between anchor points, a linear data topology exists. One trait

inherent in a linear topology is that if an anchor point changes because of, for example,

the change in the location of a traffic intersection, the linear topology does not

necessarily have to change. The origin of the topology merely begins at a different anchor

point. A section bastioned by an anchor point is called an anchor section. Scarponcini

(2002) says that anchor sections are more stable than routes because routes become









unusable as the entire linear datum topology changes. This is not the case in linear

referenced topologies because in linear referencing, measurements relative to the anchor

points) are made when depicting the datum topology, as opposed to absolute

measurements such as measurements made to place mile markers at permanent intervals

from a state or county line, for example. When a route comprised of absolute

measurements changes, each measured point on the route must change. This is not so

with linear referenced topologies (Scarponcini, 2002).

Just as absolutely measured distances can cause problems when route changes occur,

spatial scale poses its own problems when using GIS to decide on best routes, such as

delivery routes used by online grocers. Even the definition of "scale" can be ambiguous,

meaning both "spatial extent" and "amount of detail" (Atkinson, 2000). An aspect of

scale that is important to online grocers is that entities shown at one scale can acquire

considerably different relevance when viewed at another scale. As an example, areas not

serviced might not be visible when viewed on a GIS in small scale, whereas these areas

might become highly noticeable when viewing a subsection of the original area at a larger

scale. According to Atkinson, heterogeneity and regularity of cartographic entities can

change when the scale is adjusted. The same grocery delivery routes that look elliptical at

a small scale might look polygonal at a larger scale. An exception to this phenomenon is

fractals. Fractals are patterns that do not significantly deviate when viewed at different

scales. Fractal patterns exist in both physical and human geography (Atkinson, 2000).

One way to decide on a scale appropriate for the job at hand, such as viewing the

most grocery delivery zones, while seeing the closest points any two delivery trucks are

at the same time, is to scale the map down, then scale it up until "robust" parameters are









discovered. Robust parameters are those parameters that are visible and relevant at all

scales within the study area. The area studied is also known as the "support" (Atkinson,

2000).

Adjustments in scale can also affect the way spatial dependence between regions is

viewed. The fact that households of certain LSPs (lifestyle segmentation profiles) live

closely together within a region can be lost if the regions are viewed at too small a scale.

Some of this information loss can be mitigated by kriging, or "smoothing" of information

from neighboring regions to adjacent regions.

Moreover, because scale can be divided into spatial and temporal realms (Maruca,

2002) the problems associated with scale become that much more complex.

Somewhat related to the above-stated micro-, macro- and meso-scopic traffic flow

simulation scales is micro-, macro-, and mesoscale scale representations (Bernard, 1998).

Mesoscale representations extend from a few to several hundred kilometers. Microscale

and macroscale representations are, of course, respectively less than, or greater than this

range.

According to Haining et al. (2000), quantitative measures that are incorporated into

GIS packages are not sufficiently inclusive or sophisticated. Nevertheless, it is important

that online grocers become aware of some of the existing quantitative functionality.

There is no shortage of literature pertaining to this topic, although no writings could be

found that are exclusively focused toward online grocer decision-making using GIS-

enabled statistical processes. Although considerable breadth could be allotted to this

subject, an account of just a few of the more pertinent spatial statistical packages is

included in this literature review.









Spatial analysis in a GIS environment (SAGE) is a statistical package that can be

integrated with Arc/Info. SAGE works with polygon coverages and includes data

management, graphical drawing, querying, and classification tools (Haining et al., 2000).

Polygons input into SAGE can be "dissolved" to create coverages for sets of regions.

Linear regression, a common statistical method used by geographers and other scientists,

is enabled by SAGE. Ordinary linear regression, linear regression with spatially

correlated errors, and linear regression with a spatially lagged response variable are the

three types of linear regression that can be performed by using SAGE. Dynamic brushing,

the operation of quantifying the movement of a shape over a region, is not handled very

well by SAGE, however (Haining, 2000).

a spatial data analysis package that can be integrated with ArcView is called S-PLUS.

Besides being a quantitative system with more than 2000 functions, S-PLUS is also an

object-oriented language. By creating customized ArcView Avenue scripts, users can

increase the analytical capabilities of S-PLUS. Data transfer between ArcView and S-

PLUS is supposed to be relatively seamless (Bao, 2000). S-PLUS graphs include

boxplots, histograms, variograms, and multiple three-dimensional layouts.

Geographic Information Systems in Business

The literature cited leaves little room for doubt that GIS in combination with

decision-making tools and artificial intelligence will continue to be useful as a facilitative

system to allow optimal decisions in many business settings, including the online grocery

business. Business geography and "managerial geography" are gradually being

recognized as a means to a competitive edge (Risto, 1998). Dr. Grant Thrall (2000) has

recently published an important book about business geography, which gives greater









impetus to the recognition of the field of business geography as a bona fide aspect of

geography and an important aspect of business location strategy. Following is a review of

other literature that asserts the usefulness of GIS as a managerial tool.

Grimshaw (2001) says that the primary business benefits attained through scheduling

and routing systems is gained by linking geographic knowledge about customers to these

systems. Although there have been great efforts by small and medium enterprises to

integrate corporate-wide data, such as through ERP systems, most businesses have

ignored this "key dimension" of data. Businesses concentrate on the "what, how, and

why" of strategy but to a much less extent consider the "where" dimension (Heinritz,

1998; Grimshaw, 2001). One reason for this non-integration of geographic data is that

smaller businesses, such as many online grocers, do not have the technical or intellectual

resources to do this. Despite this, O'Malley et al. (1997) have found that 48% of survey

respondents (United Kingdom retailers) develop in-house spatial databases. Yet, these

authors have stated that retailers are "data rich...but... information poor" (Hernandez,

1998). Nevertheless, as Grimshaw (2001) says, because "benefits from improved routing

and scheduling go hand in hand with benefits from improved customer relations," the

geographic aspect of data should not be overlooked, especially by the online grocery

industry, which has low profit margins.

Perhaps one of the most important points Grimshaw (2001) makes is that geographic

knowledge about who the most profitable customers are. In order to better calculate the

profitability of each delivery route, georeferenced customer profitability variables should

be incorporated into the equation. Niraj et al. (2001) bolster the notion of the different

value of customers. Their literature also explains that the lifetime value of a customer









should be calculated. Although their research concentrates on the lifetime value of supply

chain partners, some of their findings can be applicable to the lifetime value of online

grocery customers. One thing they say is that the cost of serving each customer can be

different, thereby affecting the lifetime value of the customer (Niraj et al., 2001). This

becomes clear in an online grocery delivery context when variables like time to walk to

the back door, presence of mean dogs, and high crime neighborhoods are added into the

lifetime value calculation for individual customers.

Of the three categories of GIS stated by Brown (2000), the category of GIS as an

analytical system best fits the needs of online grocery delivery initiatives. As an

analytical system, the use of GIS has various benefits for business decision makers.

Hickman goes as far as to say that by converting large amounts of geographic and tabular

information into a more useful form, GIS "enhances every aspect of the business"

(Hickman, 1999). Most pertinent to this dissertation is the article's mention of the

usefulness of GIS to help track routes in real time.

As "early as" 1989 the value of GIS for business was predicted by some literature.

While reinforcing the fact that much industrial data have geographic dimensions,

Grimshaw (1989) said that retailers can "revolutionize" their marketing plans by

georeferencing customer data. Two of the four factors that he said would influence the

adoption of GIS in business are particularly true for the online grocery industry. Those

factors are better techniques for handling spatial data, and awareness of the spatial data

combined with human skills (Grimshaw, 1989). A main point of this dissertation is that if

online grocers handle their spatial data better, thereby resulting in the recognition of









zones that surpass profitability thresholds, even online grocers, with their notoriously thin

profit margins, could become successful.

In a later article, Grimshaw (1991) expounds upon his earlier literature and states

that GIS should be an integral part of overall corporate information systems and "strongly

linked to the business strategy. Indeed, this is the view that online grocers should take

when considering GIS as part of their logistical model. He also asserts that GIS should be

used as a not only an operational and managerial tool, but also as a strategic tool.

Although there could be debate as to whether he properly classifies the utilization of GIS

to facilitate the "delivery of information-based products or services" as a "strategic"

entity, as opposed to an operational entity, there is little doubt that GIS should be used to

perform the delivery function. Nevertheless, in the same article he asserts that

considerable value can be added to corporate information if the geographical aspect of

that information is included in the analysis. He expands his evaluation of GIS from

mostly a retail-based tool, to a more sweeping analysis of GIS in his more recent

publication (Grimshaw, 2002).

While considering GIS as a part of corporate strategy, Grimshaw (1991) says that the

use of the technology should be considered an opportunity rather than just a cost. Another

way to say this would be that there is an opportunity cost by not using the technology. In

the case of online grocers, if GIS technology is not used it may be infeasible to run a

successful delivery service, which means that by not using the technology the opportunity

to reach the large home grocery market will be lost.

Interestingly literature is still being published that calls GIS "new." For example

David Hakala's (2003) article entitles "Location: The New Killer App" tells how GIS









when used with databases can be used to make "powerful applications." This

phenomenon of calling two decade-old technology "new" means that there is still

significant ignorance within the business community about the potential of GIS as an

operational tool (Hakala, 2003). This was not the case with, say, enterprise resource

planning (ERP). Is the contrast between the adoption of GIS and ERP due to the

possibility that most people are geographically uninformed? Whatever the reason, the

stage is set for the adoption of GIS in industries such as online grocers because of the

need to view geographic information as a profitability enhancer

Despite the relative slow pace of adoption, GIS is gaining momentum in certain

sectors. Murayama (2001b) talks about the adoption of GIS in Japan. Interestingly, he

says that most of the editors of GIS-related academic journals are geographers. This in

itself shows the as yet lack of intellectual osmosis of GIS into other business-related

publications. One might think that integrally spatial fields of business such as logistics or

marketing would begin to publish spatially related journals. Evidently, this is not yet

occurring, which can also help to explain the lack of geographic solutions employed by

businesses in these sectors.

One reason that GIS is gaining popularity in Japan is that data is collected down to

very small regions. Also digital maps have been developed at a "rapid pace" since the

early 1990s (Murayama, 2001b). Also, according to the same article, point of sale (POS)

data is also being accumulated in large quantities in the country. Because of these facts, it

seems promising that this research can have strong implications to online grocery

delivery services in Japan. A countervailing situation exists, however. This is that Japan









has stricter privacy laws about the usage and dissemination of personal spatial data than

the United States does (Murayama, 2001b).

An interesting perspective about the use of GIS in cyberspace has been written by

Masanao Takeyama who says that geographers have just begun to study cyberspace, and

that the study of geographic "telepresence" holds important meaning for determining

"cyberplaces." Cyberplaces are places of "interactivity between cyberspace and physical

places (Takeyama, 2001). Getting a handle on the geography of daily or cyclical

cyberplace patterns may help online grocers decide between using home delivery or

communal drop boxes. Takeyama (2001) suggests that cyberplaces can cause spatial

metamorphism in cities. Online grocers that can detect these new spatial patterns might

be better able to cater to the growing market of mobile and stationary internet users.

O'Malley et al. (1997) tells that although store location is the "most critical decision

a retailer makes," the use of GIS to make such a decision by United Kingdom retailers is

not as pronounced as it could be. A reason for this is that the technology is not integrated

into strategic decision making to any great extent (O'Malley et al., 1997).

Other recent literature exists that supports the premise that GIS and spatially oriented

business practices are being adopted more. Lowe (2003) says, "The use of geospatial data

is moving from the mapping department in the back office to the existing business

practices of the enterprise." He also talks about supporting technologies that can enable

the non-GIS expert to better use spatial business solutions. One such technology is peer-

to-peer (P2P) computing which could allow any GIS application on a computer to

simultaneously be a client and a server. According to Lowe (2003), decoupledd peer-to-

peer spatial environments are removing old barriers that used to exclude non-technical









users." This means that it is becoming more feasible to initiate solutions such as the

delivery route system, which is suggested in this dissertation.

Despite the comparatively slow start GIS has had in taking a foothold in mainstream

business applications, the move toward adoption of GIS by science and industry is called

a "GIS revolution" (Yano, 2001). This literature says that there have been efforts to

increase the value of GIS by making it a geographic knowledge system (GKS). This

transformation has sometimes not been successful because some advanced quantitative

procedures that can be used with a GIS have not been adopted by mainstream businesses.

A factor that might be contributing to this relatively low acceptance of GIS is that United

States businesses have been influenced relatively little by academic geography. This is

because some of the general public do not fully perceive geography as being a bona fide

profession (Yano, 2001).

Enabling Geographic Information Systems Technologies

Because the purpose of this dissertation is to make suggestions that may increase

online grocers' profitability through strategic delivery zones, some review of literature of

the enabling GIS technologies will be of value.

For various reasons, a non-centralized GIS information systems platform can better

serve end users. One reason is that data that is not centrally located can be more readily

accessed by end users who utilize a distributed system. Also, it has been demonstrated

that nearest neighbor queries can be efficiently executed in multi-disk and multi

processor environments (Papadopoulous, 1997).

By the early 1990's GIS users where beginning to realize the redundancy behind

collecting and creating their own data sets, especially when they could relatively easily




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EW40WUKNK_F6O62P INGEST_TIME 2011-11-02T16:47:18Z PACKAGE AA00004678_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES



PAGE 1

GEOGRAPHIC SIGNIFICANCE OF DELIVERY ZONES INANECOMMERCE ENABLED GROCERY DELIVERY STRATEGY By KEITH CARL HERREL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2006

PAGE 2

ACKNOWLEDGEMENTS The completion of this dissertation would not have been possible if it were not for the loving devotion of my wife, Tsuneko Herrel (also known as Jaato ") and the enduring love and support of my mother Jean Herrel, who passed away before she could see me graduate with my Ph D. I know however that she will see me graduate as she looks down from heaven. My three darling children, Schawn (also known as "Chub"), Lina Lee (also know as "Screaming Meemie" or "Angel Princess "), and Aaren (also known as "BoupTo YouBoy "), have given me the will, perseverance, fortitude, and just plain heartwarming desire to achieve in life both academically and professional! y. ii

PAGE 3

TABLE OF CONTENTS ACKNOWLEDGMENTS ................. ........................................... .. ... .... .ii LIST OF FIGURES ......................................... ............. ....... ....... ......... ... V ABSTRACT ..................................................................................... viii CONCEPTS AND METHODOLOGY ........ ....... .......... . .... .............. ......... . 1 Purpo se ............................... ...... ...... .... .................................... . 1 Concept s ................... ......................................... . .. ................. 3 Methodol ogy .................................... .. ........ ......... ...................... 4 LITERATURE REVIEW .................... .......... .......... .. ........... ............... .... 8 Online Grocers' F a ilures, Difficulties, a nd P oss ible Solution s .................. . 9 Sp atia l Deci sio n Support Sy s tem s .................................................... 18 Online Retailer Geo gra phic I ss ue s . .............. .................................... 25 Grocer Web Site De s ign I ss ue s ..................... .. ........... . ................... 29 Artificial Intelligence ..... .............................................................. 30 Online Grocers' Store Locations ....... ............... . . ......................... .45 Geographic Information System s for Traffic Flow Simul a tion .. . ............ . .48 Geographic Information System s for Vehicle R o utin g and Lo g i stics ... .. . ..... 51 Network s .......................... . ....................... ............................ . 56 Data Vi s u a lization and Data Repre se ntation ....................................... 59 Geo gra phic D a ta and Surface Repre se ntation .................................... . 66 Geo gra phic Information Sy s tems in Bu s ine ss ...... .. .. ............... .......... ... 86 Enabling Geographic Information System s Technologies ......................... 92 Spatiotemporality .............. .. ...................................................... 101 Other Supporting Literature ............. ............................ ........ .... ..... 107 SIMULATION ELUCIDATION ................ .................. ........................ 122 THREE SIMULATION SEQUENCES THAT RESULTED IN EXCEPTIONAL A VERA GE AND MEDIUM PROFIT .. . .. ................. .............. ............... 178 Exceptional Profit .................. ...... .. .. .............. ........................... 178 Low Profit. ............................................ ............. ....... ............ 179 Medium Pr o fit ......................................... ............ .................... 179 CONCLUSION ......... ....... .. ................................ .............................. 186 SOURCES CITED .................................. ..... ........... ...... ......... ............ 199 iii

PAGE 4

BIOGRAPHICAL SKETCH ........................... ............................ ...... .... 215 iv

PAGE 5

LIST OF FIGURES Figure Hex ago n a l la yo ut. ..................... ............. .. .......... ........... ..... ........ 7 2 Durati o n of delivery in d ays ........... ........... . ....................... .. . .. 146 3 Profit opt io n s sett in gs .............................................................. 146 4 Minimum d ollar delivery ........ ........................................... .. . .. 147 5 Mandatory profit. .. .......................... ....... ................... .. .. ....... 147 6 Deli very c h arge per concentric region .......................................... 148 7 D e liver y charge p e r region interface ..................... ....... ................. 148 8 Deli very charge per re g ion pop up form ........................ .. ............ 149 9 Simulati o n o ption s form ............... ............................................ 149 10 Number of orders error catching .............................. .......... ........ 150 11 Number of deliveries drop-down lis t. ........................... .... .......... . 150 1 2 Durati o n of deliverie s error catching ..... .. . .. .............................. 151 13 Durati o n of deli ve rie s drop down lis t. ......................................... 15 l 14 Region a nd legend color coordination ....... ...................... ............ 152 15 Report d as hbo ard for the s imulation run on April 1 3 ....................... 152 16 Simulati o n main map re s ult with profit se ttin gs ........................ ..... 153 17 Pau se function . ........................... ................. ...................... 153 18 Profit options form ............................ ....... .................... ...... 154 19 Allow di s patch alerts radio button .............................................. 154 V

PAGE 6

Figure 20 Truck dispatch notification ...... ................................................ 155 21 Simulation main map result showing truck usage ......... ............. .. .. 155 22 Report dashboard for the simulation run on April 15 ....................... 156 23 Population report ................................................................. 156 24 Profitability chart. ................................................................ 157 25 Total mandatory profit per simulation report. .................... ........... 158 26 Re si dence report .................................................................. 159 27 Settings report. .................................................................... 160 28 Delivery report. ................................................................... 161 29 Simulation options form ........................................ ................ 161 30 Simulation main map result showing three si mulations ..................... 162 31 Simulation charts report ......................................................... 163 32 Simulation charts report showing second and third simulation ............ I 64 33 Settings report showing three simulations ...................................... 165 34 Total mandatory profit per simulation report s howing three simulations .. 165 35 Residence report showing orders from the second and third simulations .. 166 36 Mandatory profit per concentric region on profit options form ............ 166 37 Mandatory profit per concentric region on settings report .................. 167 38 Submit Delivery Charge button ................................................ 168 39 Delivery charge per region interface ........................................... 168 40 Delivery charge per region form ................................................ 169 41 Region A3 showing the delivery charge ...................... ................. 169 vi

PAGE 7

Figure 42 Profitability chart showing profit for region A3 .............................. 170 43 Mandatory profit per simulation report showing profit for region A3 .... 170 44 Settings used with Intra-Regional Drive Time function .................... 171 45 Intra-regional drive time interface ................... ................ ......... 171 46 Intra-regional drive time pop-up form ......................................... 172 47 Settings report showing 15 minute intra-regional drive time ............... 172 48 Delivery report showing increased intra-regional drive times .............. 173 49 Inter-regional drive time interface .............................................. 173 50 Settings report showing increased inter-regional drive time ................ 174 51 Delivery report s howing increased inter-regional drive times ...... ........ 174 52 Overall average population per region ......................................... 175 53 Popul ation per region interface ....... ............................ .............. 175 54 Main map s howing various amounts of orders placed ....................... 176 55 Region s showing higher and zero populations .................. .............. 176 56 Report dashboard showing various populations .............................. 177 57 Population report show ing various populations .............................. 177 58 Profit and simu lati on options for hig h-profit scenario .......... ............ 180 59 Profitability quotient and profit gauge for high-profit scenario ........... 181 60 Profit and sim ulation options for low-profit scenario ........................ 182 61 Profitability quotient and profit gauge for low-profit scenario ......... ..... 183 62 Profit and sim ulation options for medium-profit scenar io .................. 184 63 Profitability quotient and profit gauge for medium-profit sce nario ........ 185 vii

PAGE 8

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy GEOGRAPHIC SIGNIFICANCE OF DELIVERY ZONES INANECOMMERCE ENABLED GROCERY DELIVERY STRATEGY By Keith Carl Herrel May 2006 Chair: Grant Ian Thrall Major Department: Geography This work contains a geographic and logistic tool that will allow grocers who are contemplating to deliver groceries via e-commerce to better decide whether they should embark on that initiative. As web-based orders become more commonplace, this computer application will have the potential to assist both large and small grocers in the critical decision of starting a delivery service. The causes of failures of grocers are traced. An operational computer program is demonstrated, with embedded algorithms that make for a realistic simulation. The computer simulation together with the user-friendly interface "front end" provide a practical and also scholarly solution to a complex geographic problem viii

PAGE 9

CONCEPTS AND METHODOLOGY Purpose The potentially lucrative grocery home-delivery market largely remains untapped because of the financial risk and operational difficulty of running such a service. Also recent history is fraught with unsuccessful attempts at delivering groceries. The rapid bankruptcy of the high-profile Webvan company is possibly the worst debacle of all the grocery delivery attempts. This dissertation contains a logistical simulation application to solve one of the most plaguing problems that some grocers have, which is the problem of deciding to initiate a home delivery service or not. The primary purpose of the simulation is to reduce the financial risk involved with rolling out an e-grocery initiative. This means that grocers can take a quantum leap forward in satisfying the annual multi-billion dollar United States grocery market by using my application. This is also true for other industrialized countries. How is this possible? First, the great number of grocer home delivery failures is no secret. This could be resulting in many grocers hesitating to start a delivery initiative. Second, there could be a notable benefit to the people in much of the industrialized world if grocers can successfully deliver to time strapped customers. Traffic jams and fuel consumption could decrease if people do not use their cars to shop after work. Third, a grocer can decrease the number of employees, such as cash register attendants and stock persons needed in the brick-and-mortar store because customers who would otherwise go to the store will not do so if the groceries are

PAGE 10

home delivered Fourth, the infirm and elderly will benefit because the competition to satisfy people who accept home deliveries could very well drive down delivery prices that otherwise would be out of the economic means of these people. This list of benefits is not exclusive, but should give justification to the claim that grocery home delivery is a benefit to not only the grocer who could make a greater profit or acquire more customers, but also to society in general. No computer simulation application as presented here has ever been presented in a public forum, including academic journal articles or news media Grocers who are contemplating deliveries will find this profit estimating simulation application easy to use powerful in its functionality, and versatile to allow its use in the actual delivery operational operations Actual customer addresses can be input into the application to generate simulated deliveries to loyal and prospective customers. Actual times that are required to deliver to each region can be input before the simulation begins. The output of the delivery data is shown is easy-to-read automatically generated reports. Round trip truck times that include the average time required to deliver within each region can be input separately into the application to give an accurate report of how long it takes a truck to leave the store, deliver to the households, and return to the store. Also, the number of trucks required to deliver to the population of all the regions can either be specified or calculated. Because rarely is it possible to give a clear-cut yes or no answer about whether a business venture should be undertaken or not the conditions where a online e-grocer (or a grocer who is taking orders by telephone) will reach a predetermined amount of profit can be determined by running this application. In other words, questions like, "by what 2

PAGE 11

pathways can profitability be achieved?" and "what values do you need in combination to reach a predetermined profit?" must be answered before a grocer should begin delivering groceries. Because profitability is naturally the goal of any grocer, this application gives an indication to the grocer about what clusters or sets of attributes that will work to achieve the required level of profit. Each simulation that is run can portray a different scenario of what is needed to achieve profitability. Concepts In terms of this logistic application the term Mandatory Profit" is used. The Mandatory Profit is the profit that must be attained on each run to a region. A region, from a grocer's standpoint, is an area that has customers with a certain profile, or takes a certain amount of time of travel. A grocer must determine if deliveries to a region will be profitable or not. If the grocer decides that each dispatch of a delivery truck (or "run") to a region must result that revenues must exceed operating costs by $X, then all the grocer needs to do is input the $X profit figure into the Mandatory Profit dialogue box of this simulation The dialogue box is a data input box on the logistic computerized interface that I have developed Additionally, the Mandatory Profit can be specified according to different regions allowing myriad possibilities for different profitability attainment scenarios. The expected profit margin can be input into the application. This percentage is calculated in conjunction with the Mandatory Profit and the Delivery Charge to ascertain whether the initiative can be profitable. The Profit Margin can be varied from 0.1 % to 3 % of the value of the groceries delivered. 3

PAGE 12

The Delivery Charge is another variable that can be input. The grocer can specify any amount of delivery charge from zero to as high as he or she would like. This delivery charge is added to the calculation of each order generated and is used to calculate the Mandatory Profit. Another variable that can be added to each simulation scenario is the Minimum Dollar Amount. If a grocer feels that it is not worth his while to deliver anything less than $20 of groceries, the merchant could input that option into the simulation. Succinctly stated, an objective of the simulation is the determination of the conditions where an online e-grocer (or a grocer who is taking orders by a telephone) will reach a predetermined amount of profitability. When a grocer uses the logistical application, he or she will be in a significantly better position to decide whether to initiate a home delivery s ervice, or at least make a more calculated decision about whether more costly analyses are justified or not. Although the logistical simulation application contains sophisticated algorithms, and performs computational tasks, a main purpose of mine is to create an application that is intuitive and easy to use. This I have done. Methodology This presents a spatial decision support system for the delivery of groceries from fixed "brick and mortar locations to customers homes. The spatial decision support system (SDSS) will be demonstrated with a simulation. The simulation will perform the following functions. Determine the mix of variables that are necessary for the online grocer to generate a profit while delivering groceries 4

PAGE 13

Tell the total profit that can be generated through deliveries, given a certain profit margin. Determine the amount of trucks necessary to deliver to the customers. Show what regions normally reach a profitability margin, and at what times they do so. Provide very good, medium, and poor profitability scenarios using the simulation application I developed. The simulation includes identification of the algorithms required to implement the SDSS. When combined with reasonable data inputs, the information flow and algorithms will allow for the identification of which geographic settings the SDSS can be successfully deployed, and which are unlikely to be met with success Even if no geographic settings are indicated as being likely candidates for an actual SDSS of this type, the simulation will be a success as it will indicate to investors that profit margins in the contemporary grocery industry are insufficient to take advantage of the technology and online ordering and store distribution of groceries to customers. It is assumed that the grocery store has surveyed its current customer demand base to determine the amount of groceries the customers will order from home each week. In other words, the grocery store management is assumed to have reasonable expectations of market penetration by neighborhood under various pricing schemes. The geography of the delivery market will be show as follows. An urban region will be depicted on a map. A pattern that consists of honeycomb regions will be overlain on the map (Figure 1 ). The home store will be located at the center of the honeycomb area In Figure 1 diagram, the store is located in the center blue hexagon. The hexagonal regions will be identified by capital letters. The home store in the center region will be in region "H" meaning "home". 5

PAGE 14

Each customer will be associated with a hexagonal region. A random number generator will select customers for use in the simulation. The simulation will compress time. The contiguous hexagonal pattern will continue outward from the store, in a radial manner (Figure 1 ). 6

PAGE 15

7 Figure 1. Hexagonal layout

PAGE 16

LITERATURE REVIEW Although the main purpose of this dissertation is to emphasize the geographic significance of delivery zones in an e-commerce enabled grocery delivery strategy, the work necessarily encompasses more than geography and GIS (geographic information systems), and this multi-disciplinary approach is reflected in the literature review. An important part of this section is "Grocer web Site Design Issues. This is because, although the logistic application that I developed is run through Microsoft Excel, the application can be u sed as a web service, which ba s ically means it can be u se d by computers at remote sites for a fee. The section about artificial intelligence is included to better acquaint the reader with potentially important grocer decision-making tools such as intelligent agents that could traverse networks and read mobile computers in delivery trucks. The Literature Review would be incomplete without the incorporation of the two sec tions that include GIS for traffic flow simulation, vehicle routing, and logistics. These concepts are the important technologies for grocers who want to optimize their trucks routes, which can result in more satisfied customers because of timely deliveries and less fuel and manpower expenditures. The manner which polygons, points, and lines are represented on a GIS along with the way data is depicted near these geometric figures should be important to grocers who deliver through a GIS interface so as not to include superfluous data, or omit important data to delivery, operational, and managerial personnel. A review of the literature about these concepts is also included in this section. 8

PAGE 17

Online Grocer's Failures, Difficulties and Possible Solutions There is a lot of supporting and relevant literature that, when combined, can provide the building blocks for the utilization of delivery zones in geographic enterprise optimization systems (GEOS) for online grocers. After reviewing the below-described literature, the solution proposed in this dissertation for online grocer deliveries is not only necessary, but also feasible. In 2000, at the minimum seven pure online grocers were operating in the United States. By the end of that year the only two that remained were Webvan and Peapod. In an ominous statement, Ring and Tigert (2001) said no company has ever made a profit in this (pure online grocery) business." Perhaps the most visible of all grocery delivery companies demises is that of Webvan, the multi-billion dollar venture that started optimistically, but ended soon after it began. Much literature exists concerning the quick bankruptcy of this and other online grocery delivery companies (Partch, 1999; Saccomano, 1999b; Tweney, 1999; Briody, 2000; Bubny, 2000; Sanborn, 2000; Evans, 2001; Heun, 200 I; Partch, 200 l; Ring, 200 l ). The founder of Borders Books, Louis Borders, created Webvan in 1999 to deliver groceries and other household items to homes within 30 minutes of receiving the order (Tweney, 1999). George Shaheen, a managing partner of Anderson Consulting, whose father was in the grocery business, was the CEO in charge when Webvan went bankrupt (Partch, 200 l ). There was much hope and skepticism when $11 billion was raised in Webvan's 1999 initial public offering (Evans, 2001). The hope was that Webvan, optimistically called the spot where "the internet meets your doorstep," could capture a portion of the annual $650 billion United States online grocery, prepared food, and 9

PAGE 18

drugstore market (Sanborn, 2000). Of this amount, $450 billion of food is purchased annually in the United States (Ring, 2001). It was a grand scheme that depended on a network of large and expensive automated warehouse distribution centers as springboards for the delivery vans. Partch ( 1999) wrote in "Home Delivery Another Problem Solved," that a single distribution center would have the capacity to dispatch as many groceries as 20 supermarkets. The distribution centers were approximately 330,000 square feet each. The first distribution center was located in Oakland California; and the company planned to build another 25 similar centers in other cities ( Ring 200 I). A portion of Webvan's plan did have a customer allure to it. The fledgling company offered 24-hour free delivery (with orders more than $50) with prices up to I 0 % cheaper than brick-and-mortar grocers (Partch, 1999). Beside s the money raised from the IPO, deep-pocketed investors like Knight-Ridder CBS, and Softbank sa nk money into the company (Saccomano, 1999b ). But contrary to s tockholder and other investor optimism, literature shows that many people were also pessimistic about Webvan's viability. Dan Rabinowitz of Peapod Inc. another online grocer, called Web van's business model "a lot of smoke and mirrors" (Saccomano, 1999b ). In the same article, grocery consultant Cristopher Hoyt shows his doubt by sayi ng that Webvan would not be able to so lve the perishable food distribution problem, and that even if they can deliver the dry foods efficiently, the grocery market for dry foods is being supplanted by cheaper super centers. Webvan's investors and managers would have benefited from Michael Brown's commentary on automated grocery distribution centers. Brown said in 1993, the best 10

PAGE 19

strategy at present may be to postpone investment in a centralized warehouse with high levels of automation, correspondingly high capital costs, and a lengthy payback period (Browne, 1993 ) Miles Cook, vice president of consulting firm Bain and Company, said, "online grocery is a murderously difficult market, (Heun, 2001 ) Evidently Cook, Rabinowitz Hoyt, and other skeptics were right because only 20 months after its inception, Webvan fired 2 000 employees and filed for bankruptcy (Evans, 2001 ). It seems that the less-than-optimal business model employed by Webvan is not the only factor to blame for that company's demise. Paul Bubny ( 2000) explains that customer sentiment regarding online purchases is an inhibiting factor His PricewaterhouseCoopers study revealed that 21 % of internet users would never purchase groceries online. This says nothing of the percentage of current non-internet users that will not buy groceries online in the future when they begin to u s e the internet. Bubny (2000) calls grocery shopping a "care-giving function, and says consumers are not ready to allow a "detached process such as online shopping" to supply their food. Further, a survey conducted by NPD group revealed that a mere 2 % of consumers would buy groceries exclusively from an online site. The same article, however states that, with proper incentives, customers can be convinced to buy groceries online. Respondents to the PricewaterhouseCoopers' survey stated that price is their main concern when buying online groceries which includes free delivery. This implies that if other factors such as freshness and rapid delivery are fulfilled, consumers may be persuaded to buy groceries online if the price is right. 11

PAGE 20

Mainstream grocers are not turning a blind eye to online grocers. In fact, the threat posed to large grocers' markets by smaller online grocers is real. Alan Mitchell says that even a 5 to 10 % rise in home grocery shopping could "severely dent" a superstore's underlying profitability (Mitchell, 1999). This statement can be especially portentous to brick-and-mortar grocers if they consider that overall online purchases are estimated to keep on increasing (Killgren, 1999). Evidently with this in mind, large United States grocers like Albertsons and Safeway are initiating brick and click grocery business models (Briody, 2000). Safeway and Royal Ahold, a Dutch conglomerate, both formed joint alliances with pure online grocers. By joining with foundering Peabody, Royal Ahold saved Peabody from probable bankruptcy. Safeway invested $30 million in Groceryworks.com. Not long after Groceryworks.com built two fulfillment centers in the south-central United States, the company closed them down. Safeway subsequently decided that filling online orders from its store shelves was a better model and began doing so (Heun, 2001). Albertsons is moving forward with a purely internal brick-and-click delivery initiative. Possibly as a result of Webvan' s closure, Albertsons' on line sales in the Seattle area increased 300 % after Webvan stopped operating. Like Safeway, Albertsons fills orders from its stores' shelves. Safeway is also implementing a fleet management system to better route their trucks and more closely monitor driver and vehicle performance (Barnes, 2002) Ring and Tigert (Tigert, 200 l) enumerate six groups of dissuasive factors contributing to customer skepticism toward online grocery purchases. According to them, place, the first factor, is important because customers like to be in a place where they can 12

PAGE 21

touch and smell the merchandise. The next group, product, says that online grocers may not have enough stock keeping units (SKU) that the customer desires. An SKU is essentially an individual type of product. Third, value is "more bang for the buck," which is often sacrificed by relatively high delivery fees. Service comes next. When a customer visits a store there are hopefully more knowledgeable and courteous employees present than a single delivery truck driver. Communication, the fifth factor, comes in the form of advertisement or notification to the customer. Ring and Tigert (2001) do not expound on how this group is an inhibiting factor for online grocers. The sixth group, technical problem s, appears in the form of wrong orders or late deliveries, which can ruin online grocers. Another problem, as told by McKinnon and Tallam (2003), is the security of delivering to homes, especially when the home is unoccupied. Groceries can be delivered four ways to unoccupied residences. l. The driver can enter the house. 2. A drop box is available. 3 Another customer-designated location is available 4 The goods are left at a separate local agency which delivers the groceries when the customer is home. The article does say that overall, unattended delivery of groceries is relatively secure when compared to the unattended delivery of more valuable items. Of course, the worth of the goods matter little if the cause of the security breach is the delivery driver having access to the interior of the home. The article does say that although the unattended groceries may not be stolen, the fact that they are left visible implies that the homeowner is not present which can invite burglary into the premises. McKinnon and Tallams' (2003) security implications are relevant because a significant amount of the prospective home grocery delivery market may not by online 13

PAGE 22

because of the security implications. In the same article, the authors list different ways groceries could be left at an unattended home. Those ways are a fixed interior box with three compartments of various temperatures with a keypad-lockable door; a fixed external box connected to an external wall; a mobile box delivered to the customer and connected to a cable to the house; and communal collection boxes. The communal boxes can be of various types. Primarily they are in a central place like a popular parking lot or near a train station. Communal boxes can have numerous compartments, each with its own lock and changeable code to open the lock (McKinnon and Tallam, 2003). When deciding on a grocery delivery model, consideration should be given to each of these drop-off strategies to decide which would probably work best given the catchment neighborhood, type of customer lifestyle segment profiles (LSP) catered to, and logistical strategy used by the online grocer. Of the LSP databases used for this type of analysis, MOSAIC, ACORN and NDL' s Lifestyle database are three of the most highly regarded (O Malley, 1997). The type of drop-off method is not unrelated to the "skill the drivers have in delivering the groceries. Skill here relates to the ability of the drivers to learn routes on the fly. If an online grocer delivers to individual residences, especially when guaranteeing delivery within an hour, it becomes highly unlikely that any two routes taken within, say, one or two days, will be exactly the same. Michael Haughton expounds upon the relationship of uncertain daily deliveries with the drivers' ability to learn new routes on the fly (Haughton, 2002). Haughton (2002) states that fluctuating demand can result in inefficient delivery and dissatisfied customers He refers to a "learning burden," and proposes a correlation metric 14

PAGE 23

to assess the familiarity of a driver with certain rectilinear delivery routes. Because my research is focused on solving the problem of individual residence delivery which presupposes daily fluctuating demands, Haughton s (2002) literature is relevant to this dissertation. Cross and Neil (2 000) call online grocery shopping a "booming global phenomenon." Obviously this statement does not seem to include the plights of the failed United States online grocers. Evidently, these authors are referring to online grocers in Great Britain, Scandinavia and Iceland that are not doing as badly as those in the United States. Optimi stica lly however Cross and Neil tell how convenience is a driving factor toward the success of online grocers. They mention how technologically savvy customers are the most profitable ones because they are more inclined to shop online. The article also says how employers can use online grocers as a fringe benefit to their employees. The employer can pay for the delivery service to the office where the employees pick up their groceries when they leave work. Another factor that could provide inertia toward successful online grocer businesses is UCCNet. This i s a subsidiary of the Uniform Code Council which is the organization that developed the univer sa l product code. UCCNet allows e-commerce companies to register their supply-chain related data. UCCNet is an internet trading network that could benefit buyers and sellers of food, beverages, and other items ( Violono, 1999). Because the network aids collaborative planning, forecasting, and replenishment, smaller online grocers who use UCCNet should be better able to procure needed stoc k in times of unexpected spikes in online purchases. More literature pertaining to UCCNet is reviewed below 15

PAGE 24

Catering to the growing United States Hispanic market is an online grocer called LatinGrocer. Saccomano ( 1999a) states that by 2009, 40% of the United States population will be Hispanic, creating a very large grocery market. LatinGrocer's unique point is that its services are also offered in Puerto Rico. An effort to identify the success factors in e-grocery home delivery was made by Punakivi and Saranen (200 l) using simulation. The article supports the intent of this dissertation by explaining that the lack of a proper logistical home delivery infrastructure is the main deterrent to a successful online grocer. Using Finland as an example, Punakivi and Saranen (2001) concluded that home delivery can be as much as 43 % cheaper for customers when compared to the customer driving to the supermarket. Punakivi and Saranen (2001) write that the cost to deliver groceries will be less when the delivery time window is larger and customers live in densely populated areas These are intuitive factors. What might not be so intuitive, however, is that the number of needed delivery vehicles increases by 250% when a manned delivery model (when the customer must be home) is used instead of an unmanned model which uses drop boxes. Tanskanen and Punakivi (2002) write that the only factor more costly than the deliveries in a manned home delivery model is the actual picking and packing of the groceries One reason for the higher manned deli very costs is that up to 60% of the packages might be undeliverable because of people not being home. In Great Britain some main grocers are using unattended delivery methods. In Finland, grocers are using shared multi-temperature drop boxes for unattended delivery. A factor that affects the costs and profitability of the shared reception box model is the number of separate compartments in each box (Punakivi and Tanskanen, 2002). 16

PAGE 25

It is important to note that online grocers might have to operate somewhat contrarily to flexible production techniques. This is because although flexible production uses strategies such as just in time deliveries (Browne, 1993), online grocers who precisely forecast a significant proportion of their deliveries should consider pre-packing those groceries to save time during busier order packing times. Shorter delivery time windows notwithstanding, the ability for customers to choose what time the groceries are delivered is another factor that increases delivery costs. In other words, given a 40-minute delivery time window, the model that allows customers to choose what delivery time window is the more expensive delivery method (as opposed to the grocer deciding what delivery time window to dispatch the truck) (Punakivi and Tanskanen, 2002). Speaking of delivery costs, online grocers must consider the tradeoff of shipping from fewer sites, which increase delivery costs (Browne, 1993), or dispatching the groceries from more outlets, which could increase the packing costs. Browne writes that products with relatively low value densities should be distributed by local delivery. This would apply to most groceries. Gorr, Johnson, and Roehrig (200 I) talk about the difficulty of home delivery services and how GIS can help with delivery solutions. They describe the constraints imposed upon delivering hot meals to residents home. Their research is applicable to this dissertation because online grocers may offer hot delicatessen foods delivered to homes. Indeed, it might be this hot (and cold) food service that helps the online grocer to gain market share. But, as this particular article states, the time window that exists to deliver these foods is greatly decreased because of the necessity to maintain very hot or very cold 17

PAGE 26

temperatures. Gorr, Johnson, and R oehrig (2001) write that 45 minutes is the maximum allowable delivery time for these deliveries. Gorr, Johnson, and R oehrig (2001) also discuss the feasibility of using a spatial decision support system (SDSS) for delivering hot meals. The combination of GIS, algorithms, and managerial judgment are parts of SDSS. A particular problem the authors address is the situation created when customers change addresses. The SDSS algorithms should be robust enough to handle this type of change (Gorr, Johnson and Roehrig, 2001). One approach that online grocers might purs ue when considering their overall s trategic plan s to deliver groceries is to look at the composite customer equity that the delivery regions encompass. Customer equity, which is often used as an augmentation to brand equity, is the s um of the lifetime values of all of the customers a business has (N iraj, 2001 ) If, when contemplating a delivery plan the customer equity calculation doe s not look as if it could be increased or even s u s tainable, it might be advisable to rethink the grocery delivery plan Some aspects to look at when determining the sustainability of customers are complexity factors and efficiency factors of those customers (N iraj 2001 ). Succinctly s tated the se two factors are antithetical forces. Wherea s complexity factors, s uch as difficulty fulfilling a grocery order, decrease a customer's lifetime value, efficiency factors, which make a delivery to a particular address easier, can help increase the lifetime value of the customer. Spatial Decision Support Systems Research about u s ing GIS as an integral part of SDSS began years before the above noted article by Gorr, Johnson, and Roehrig (200 l ). Crossland et al. (2 001) s tudied the 18

PAGE 27

effectiveness of using GIS in SDSS to see if the time used to solve problems could be significantly decreased. The article reinforces the important point that most business data have "one or more spatial components." Three of these components, as stated in the article, are relevant to online grocer business models -customer addresses, delivery vehicle locations, and site selection. Calling GIS a "highly evolved technical toolbox," the article, says that GIS should not only be used to find solutions to problems, but also to guide the users to problems that might have been unforeseen before the adoption of a GIS-enabled SDSS (Crossland, l 995). The worth of decision support systems (DSS) in general is stated by (Ylachopoulou et al., 200 l ). They say that DSS "couples the intellectual resources of individuals with the capabilities of the computer to improve the quality of decisions and to support managerial judgment" (Ylachopoulou et al., 200 l). The value of spatial aids to managerial judgment should not be underestimated. Thrall (l 995) endorses the value of using GIS to optimize judgment in managerial decision making. Thrall writes that, ... using GIS to improve judgment is a crucial stage in reasoning with GIS because it links GIS to the market economy." Thrall expounds on his GIS-facilitated judgment maxim by describing the value of business geography. Thrall said, "Business geography integrates geographic analysis, reasoning, and technology for the improvement of the business judgmental decision. Without the demonstrated ability to improve the business decision, there is no business geography. This differentiates business geography from the traditional descriptive or explanatory objective of economic and urban geography" (Thrall, 2002). Gorr et al. ( 1995) tell how a GIS-enabled SDSS can allow decision makers, or their assistants, to use databases in real time. This is of course important for on line grocers that 19

PAGE 28

employ metamorphic delivery zones because, as this dissertation will s how, it is when a delivery zone reaches a ce11ain profitability threshold that the truck should be dispatched to that zone. A decision maker, or his or her assistant, must be able to make the real-time decision to send the truck. Gorr et al. ( 1995) show that spatially related decisions can be made faster and with less error when SDSS are used. Supporting the assertion that deliveries should be made only after a certain profitability level has been reached is Niraj et al.' s (200 I) research that says a company should "eliminate transactions that do not add value." If value to a company (co nsidered here as profitability) i s not created by a delivery to a zone, then perhaps delivery to that zone should be deferred to a later time when the value of that zone increases. Care should be taken that this strategy does not result in dropping customers who might be unprofitable in the short run, but might become valuable customers in the foreseeable future (Niraj et al., 200 l ). A SDSS incorporates GIS with a database management sys tem ( DBMS) and peripheral software tools that allow interaction with the analysts and other users ( Tarantili s et al., 2002). Tarantilis et al. call SDSS a new scientific area of information systems applications." Possibly, the relative newness of SDSS is in part the reason that little if any literature exists about the utilization of this technology to s uccessfully solve the problem of delivering groceries that are purchased online. I will demonstrate in this dissertation that spatial decision support systems can be the difference between the success and failure for any online grocer of significant size. The importance of routing information systems is underscored by Browne as he says that the end of the era of "cheap freight" transport is here where reliability used to be attained without modern information systems (Browne, 1993). 20

PAGE 29

Tarantilis and Kiranoudis (2002) use SDSS to solve a vehicle routing problem (VRP) using the backtracking adaptive threshold accepting (BATA) model. The constraints posed in these authors' research mirror those that would be inherent in a typical grocery delivery situation. They are minimize the distance traveled, do not exceed the capacity of a vehicle, and, one vehicle delivers each customer's order. Although the article does not specifically mention web services, it does say that reusable software components are important parts of a SDSS. Because, more recently, the importance and viability of web services (for example, services created by the .Net platform) have gained mainstream exposure, the plausibility of creating a component-based SDSS is greater than it is without using web services. Tarantilis and Kiranoudis (2002) give a detailed and graphical explanation of all the components of a SDSS. Arc View is the GIS used. BATA, in conjunction with the Dijkstra algorithm, determines which vehicle should deliver to which customers and the order that it should do so. BAT A is a "metaheuristic" method, meaning that it relies on more than one heuristic to arrive at a solution. The article states that the SDSS is used by Athens, Greece taxi companies, newspaper companies, and for meal delivery. Benefits attained by using the SDSS include the ability to make semi-structured or unstructured vehicle routing decisions, transport cost reduction, and better control of distribution (Tarantilis and Kiranoudis, 2002). Also, when routing decisions are made using a SDSS, the decisions made might be more intuitive if some sort of multimedia capabilities exist within the system. Cartwright and Hunter explain how decision support systems can be enhanced by the incorporation of multimedia These authors say, "users of geographic information need far more than the printed map, and they are likely to need rapidly-21

PAGE 30

produced and well designed spatial displays combining a number of different types of data" (Cartwright, 200 I). As pertains to this dissertation, the key words expressed by Cartwright and Hunter are "rapidly-produced ... spatial displays." Another notion proposed by these authors is that although the incorporation of multimedia into GIS is very valuable, care must be taken to not "overw helm the users. Ogao and Kraak (2 002) agree with this by saying that animation can distract the users if it is overused. These cautions should be adhered to by online grocers who incorporate delivery zones into their geographic enterprise optimization systems Ralston, Tharakan, and Liu ( l 994) espoused the virtues of SDSS before GIS was considered a main s tream technology. The authors sa id that GIS must be merged with spat ial analysis tool s for GIS to be used to its fullest potential. Agreeably, GIS alone, without being augmented with tools such as simulation or artificial intelligence is a powerful technology but remains a mere shell of what it could be when used with these tools. In Ralston et al.'s article it is said that transportation network s are comprised of physical and logical links. Physical links are the actual roads, railways, or inland water routes. The Logical link s are point s along the phy sica l links for deliveries, pick ups, and inter-modal transfers. Phy sica l links of the same type can be divided into classes depending on factors s uch as cost, capacity, and speed. Paths between a so urce and destination always have both physical and logical links (Ralston, 1994). Nasirin and Birks expound upon the types of SDSS used by retailers in Great Britain. Knowing customer proximity to the store and those customers' purchasing behaviors are two common functions performed when SDSS is used by retailers. The literature states 22

PAGE 31

that one reason retailers are adopting GIS-enabled decision support systems is because of the increasing availability and portability of spatial data which often exists on compact disks. Despite this growing adoption of GIS by retailers there has been little research about the methodologies that retailers employ when using GIS to make decisions, even though successful usage of GIS by retailers involves "enormous numbers" of factors (Nasirin, 2003). An important point made in this article is that GIS should be used by retailers to determine who the competition is and where they are located. Keenan (1998) says that GIS is a technology that can "take advantage of other technologies," when used to make routing decisions. Keenan also states that straight-line distance is insufficient for routing calculations, and that time traveled is a better indicator of the route. Using a decision support system (DSS) is important when ascertaining the best route because sometimes the best route may actually look circuitous (Keenan 1998). According to Keenan ( 1998), this is where the "soft" constraints such as decision makers judgment, apply The importance of the decision-makers' input is why solutions to routing problems require the use of DSS more and expert systems less (Keenan, 1998). Many times data and information applicable to the routing problem may already be in the corporate management information system (MIS). Often information such as customer demand and vehicle size is business-specific, thereby requiring a customized link to the data from the GIS (Keenan, 1998). Because spatial data is getting less inexpensive to purchase, the availability of comprehensive highway databases will become more readily available Also, according to Keenan, highway data within a GIS when used in with sophisticated management science routing algorithms can result in useful SDSS. 23

PAGE 32

Genetic algorithms, although not strict ly a spatial decision support system component, are intriguing problem sol vers that can help with difficult and time consuming geographic problems. Clarke (1998) believes that Genetic algorithms (GA) utilization will soon become more prominent in GIS. Genetic algorithms use a "Darwinian" method to find feasible solutions for problems by forcing solutions to "evolve" by allowing only the "strongest" genre of variables to "survive" repeated iterations through the problem. Chromosomes, consisting of genes, which themselves contain alleles, "evolves the population of solutions by repeated selection and mating" (Van Dijk et al., 2002). Van Dijk et al. demonstrate the use of GA to solve a map labeling problem. Some of the dynamics of the GA problem solving method are Selection, which means better parents produce fit children. By using tournament selection iteratively choosing parents until n sets of two parents are chosen -a set of parents is only chosen if their children are fit. Reproduction, which means each set of parents produce two children. Crossover refers to a certain amount of genes from each parent which are imbued into the children. Replacement means some less fit children are exterminated to make room for other stronger offspring. Finally Termination occurs when, at some point, the population "converges, or become as close to optimal as possible. Obviously, this is a very abstract explanation of GA. As a concrete example, Van Dijk et al. show how map labels are optimally placed using GA while specifying the direction of the label in respect to the city, degree of congestion of labels, and how large the labels are in proportion to city population. Online grocers should investigate the utilization of GA in their routing and customer selection optimization strategies. This is because if delivery time is considered a scarce resource, the profitability of customers 24

PAGE 33

becomes a paramount issue when deciding routes Customers can be chosen using GA by "natural selection" using profitability as one of the gene traits as described above. Online Retailer Geographic Issues "Microscale" competitor analysis is possible with GIS (Sadahiro, 2001 ). Sadahiro (2001) uses this approach as he looks at retail location in Japan and discovers that stores are usually clustered around railroad stations. Clustering, although often used to denote a sort of condensation of entities within a region, is also a group of techniques use to analyze complex geographic data to ascertain if spatial relations exist or not. Pattern spotting and data mining are two ways to determine the presence or absence of clustering (Murray, 1999) While looking at two types of clustering problems the median clustering problem and the central point clustering problem Murray discovered that the median clustering problem requires less computational power to solve than the central point clustering problem. An aspect of retail clustering that pure online grocers (those without a brick-and mortar store) should consider is that they might not be able to capitalize on the important aspect of being within customers' comparison shopping region. Sadahiro (200 I) recognizes this important factor and develops a probability density function, which quantifiably tells the degree of agglomeration of retail stores of the same classification. One reason stores in the same classification locate near each other is to provide customers with the opportunity to shop comparatively (Sadahiro, 2001 ). Although this is not the only force that results in retail agglomeration, it should be recognized as a force that might take customers away from online grocery sites. Guy describes comparison shopping as one factor to consider when investigating the grouping of retail stores He 25

PAGE 34

says that comparison shopping has an "enjoyable" aspect to it (Guy, 1998). Can comparing prices of competing online grocers online be "enjoyable?" This question is worth considering by online grocers. Baker addresses another problem that is encountered by online retailers. He says that online retailers do not adequately consider central places when selling online. This is because the fast speed of most internet communications lends itself to the underestimation of the importance of spatiality in cyperspace. He calls the failure of retailers to comprehend distance minimization strategies such as the gravity model "a major barrier to successful marketing and profitability for internet retailing (Baker, 2001). Recently, consumer activity patterns have been given greater scrutiny by retailers. Consumer activity patterns could be relevant to online grocers for various reasons. For instance, activity patterns could help online grocers decide whether to deliver to the home or use communal drop boxes. Decision tree induction which is used by artificial intelligence and statistics, has recently been said to be useful to help model activity-based models (Arentze, 2003). Albatross models model travel demand using a rule-based decision tree. Arentze (2003) says that decision tree induction can be useful for modeling spatiotemporal behavior such as the behavior exhibited in activity patterns. Activity based travel behavior (ABTB) is the term coined for this type of activity monitoring (Frihida et al., 2002). The concept of "space-time paths" is covered by Frihida et al. (2002) in their article about ABTB. According to these authors, space-time paths can help explain, predict, and plan for past and future activity patterns of commuters. 26

PAGE 35

The reason for the attention given to activity patterns is that previously, retailers looked at discrete trips when calculating the amount purchased by consumers. The current consumer trend, however, is to engage in multi-stop "tours." The itinerary of these tours is influenced by retailer policies, travel costs, and other services and entertainment consumed on the tour. These "hybrid" multi-stop tours are blurring the distinction between retail shopping trips and other types of trips (Roy et al., 200 l ). Online grocers might be interested in the efforts to classify hybrid trips. Some classifications enumerated by Roy et al. are tours that a) start and end at home, b) start and end at home with a work stop c) start and end at work, and d) tele-orders which are orders placed over the telephone. Of course online grocers would be interested in the fourth hybrid trip category tele orders. But there should be ample reason for online grocers to concern themselves with the other three categories as well. This is because, according to Roy et al. (200 l ), consumers consider aggregate properties of tours. This means that if groceries ordered online can become part of the tour (by picking up the groceries at communal drop boxes), or might not become part of the tour (thereby saving the consumer time by having the groceries delivered at home) The duration and sequence of activities undertaken during tours is important; so much so that various models of activity duration during trips have been developed. Two of the models are the unconditional and conditional risk models. Some of the many determinations made by Popkowski et al. (2002) when using these models are shopping after work decreases with age; single people shop often after work; poorer people are more likely to indulge in leisure after work; after leisure people are likely to conduct 27

PAGE 36

personal business, then shop. Online grocers could look at these determinations when deciding web page content, delivery routes, and more. Task allocation is similar to activity patterns. Task allocation is the allocation of subsets of tasks by household members while considering context variables such as constraints and imperatives relevant to those household members such as car availability, and urgency (Borgers et al., 2001). Other issues to consider when attempting to predict if a household member will perform a certain task are whether the task is mandatory or discretionary; or is shared or performed alone. Borgers et al. (2001) assume that task allocation can be correctly simulated. Although this assumption is debatable, it may be worthwhile for online grocers to consider situations when household members must perform a mandatory task, such as shop for groceries, while at the same time be constrained by only having one vehicle because a spouse or child is using the sole car when the shopping trip must be performed The logistical requirements of online grocers are different than that of brick-and mortar supermarkets. Although there is an overabundance of literature about retail supermarket logistics, there is little that applies solely to the logistic requirements of online grocers. Third party logistic firms (3PL) can supply all or a portion of a companies warehousing and distribution needs (Balakrishnan, 2000). This could be a novel idea for online grocers who do not want to pack and ship their own groceries. It might also be an innovative service provided by a new kind of 3PL that caters exclusively to grocery delivery. A difficulty in using 3PLs for grocery delivery could be in the compensation scheme. There are many variables, such at type of goods (frozen or hot), 28

PAGE 37

weight, changing demand, and expedited foods to name just a few. According to Balakrishnan (2000) achieving fairness in compensating multiple distributors can be challenging. Fee tables usually are made to calculate delivery compensation. Two primary ways to determine fee values are tariff and cost-based approaches. The former gauges the delivery compensation by geographic regions while the latter estimates the delivery costs using a function of distance and weight of the goods (Balakrishnan, 2000) Microsoft s Solver contains linear programming capabilities that can be useful to even smaller online grocers that are developing fee schedules or in-house delivery driver compensation. Grocer Web Site Design Issues Wang and Gerchak (200 l) support the importance of stock keeping units (S KU) in the decision of a customer to shop at the store. Of course, customers prefer stores with more SKUs. Therefore, a tradeoff has to be reached when allowing greater "virtual" shelf space for more profitable items or showing more SKUs per web page Wang and Gerchak's ( 2001) research discusses the relationship between demand and the allocation of retail shelf space. Because, according to the article, demand increases when more shelf space is allocated to a product, an online grocer that more prominently displays the higher profit items on its web site might have lower delivery co sts relative to the cost of the groceries delivered. This issue deserves more research Interestingly, Wang and Gerchak (2001) say that for any two identical retailers, the total profit made between them depends only on their total inventory level not on how the total inventory is allotted between them. This could be significant when two online grocers, or an online grocer and a brick and mortar grocer compete in the same market area. The word 29

PAGE 38

"identical" can be ambiguous. But, if identical is defined as the same amount of SKU's, then, if this proposition is correct, an on lin e grocer that has the same amount of SKUs as a brick and mortar grocer should have a reasonably good chance of selling as much as the brick and mortar store. In "Designing an Effective Cyber Store Interface," Kim and Eom (2002) talk about the importance of clearly showing on a web site that risk-free and on time delivery is offered. They propose four elements of customer satisfaction. The elements are product and/or service, support, bad experience recovery, and extraordinary service. The factors of risk-free and on time delivery could be classified under any or all of Kim and Eom's (2002) elements of customer satisfaction. Throughout their article it is said that an e-commerce site should allow customers to comparison shop. Taken in conjunction with showing the greatest possible amount of SKUs on a web site, if a site allows, or even encourages customers to compare the prices of the host grocer's SKUs with the prices of other grocers products ( online or not), customer satisfaction and sales might increase. Artificial Intelligence The study of neural networks is a branch of artificial intelligence that pertains to roughly mimicking the brain to make calculated decisions. Literature exists that shows the attempts to include the decision-making power of artificial neural networks (ANN) in GIS. Although I have not located any literature that specifically uses ANN to determine delivery routes for online grocers, ANN has been used with GIS for other types of applications, which shows the feasibility of using these technologies together. The predictive power of the combination of GIS and ANN to forecast land use changes in Michigan is documented by Pijanowski et al. (2002). The article says the use 30

PAGE 39

of GIS and ANN can "aid in the complex process of land use change." Also, "ANNs are powerful tools that use a machine learning approach to quantify and model complex behavior and patterns This is precisely the type of functionality online grocers need in order to predict demand and create tentative delivery zones to better prepare for that demand. Pijanowski et al. state that because of the spatial nature of the variables, the incorporation of ANN with GIS is "essential." The same should hold true for an online grocer's logistical dynamic routing application. Black ( 1995) writes that using ANN with the gravity model. He uses a gravity artificial neural network (GANN) to show commodity flows between nine United States census regions. Using as input values regional flow of commodities, regional flow attraction, and interregional distance, he discovered that the accuracy of the gravity model improved when he moved from the unconstrained (conventional) gravity model to the GANN. Black concludes that flow modeling can be revolutionized by using the GANN model. What Black's research contributes to the purpose of this dissertation is the clarification that ANNs increase the flexibility and usefulness of not only GIS, but also spatial modeling in general. Black ( 1993) states that an industrial firm must make three decisions after it has decided to operate. I. Where to locate, 2. What technology to use, and 3. What the scale of production will be. Although Black was mainly referring to manufacturing firms, his three points can also apply to an online grocery initiative. The technology inferred by Black can include GIS, computer simulation, SDSS, intelligent traffic systems and more. Scale could refer to the scope of the business model as referring to delivery methods and delivery zones, while firm location is of course where the grocery store and its 31

PAGE 40

distribution center(s) are located in comparison to the customer base. These three factors should be considered within the entire online grocery delivery model. It should be noted that a concern of both Pijanow s ki and Black in the above mentioned research about ANN and spatial technology is the problem of spatial autocorrelation. Re searc h exists addressing thi s problem. One example is from Duckham (2000) who discusses "error se nsitive GIS. Relevantly this literature also touches on the realm of artificial intelligence because it uses induction an established artificial intelligence technique. According to this article, "an inductive learning algorithm s hould be able to automatically deduce rule s that embody the patterns in that data ... According to Gahegan (2000), inductive machine learning is gaining popularity in the geographic community. Intere s tingly, even though, according to Duckham et al. (2000), the inductive algorithm deduces; it does not induce, this literature is noteworthy because it endorses an artificial intelligence (A I ) so lution to s patial problems s uch as autocorrelation that could be otherwise intensified by the incorporation of another AI so lution namely artificial neural network s (ANN). AdditionaJJy, because of the low profitability and labor intensivene ss of ascertaining the amount of error in geographic data, the article says that there probably wiJJ not be many companies s trictly concerned with quality issues in geographic data. Duckham et al. 's (2000) research therefore recommends the inductive learning algorithm because it i s a relatively low cost way of ascertaining spatial data quality. Neural spatial interaction model s are somewhat related to gravity models. Three gravity model input variables relate to measures of "origin propulsiveness, destination 32

PAGE 41

attractiveness and spatial separation" (Fischer, 2002). Neural s patial interaction models relate to gravity models not only because they expedite mathematical modeling but because they are adaptive enough to "deal with incomplete, inaccurate, distorted, missing, noisy, and confusing data ( Fischer, 2002). Although the field of artificial intelligence includes technology s uch as expert syste m s, neural networks and fuzzy logic, intelligent agents possess a particularly promising niche in the field of online grocery s upply chain optimization and visualization. Because of their ability to "automatically" scan intranet s, extranets, and the internet, intelligent software agents (or just intelligent agents"), or bots can be programmed to periodically monitor databases, data marts, and directories The gleaned data could then be made available in a more informative format, such as in GEOS-enabled online grocery logi s tic solutio ns Although using bots to collect data can be difficult ( Fontana 2002), companies, s uch as SAP, are aggressively pur s uing the utilization of bots in ERP (e nterprise resource management ) and SCM (s upply chain management ) solutions. Steve Ranger tell s how SAP is using bots to gather supply chain partner data and make the resulting information available through portals ( Ranger, 200 l ). One of the forerunners in intelligent agent re sea rch is the Carnegie Mellon University. A look at their intelligent software agent website gives an inkling of the branches of this field of artificial intelligence (Carnegie Mellon, 200 l ) The Carnegie Mellon si te s how s a few bot technologies that could be of particular value to online grocery logistic plans. Local area discovery and multi-agent learning are 33

PAGE 42

two such technologies. In multi-agent learning the bots learn in "dynamic environments" such as that which exist between different networks of companies interior supply chains. Business rules markup language (BRML) is an intelligent-agent facilitated technology that could help heterogeneous e-commerce-related applications to exchange "rules." It is referred to as an "interlingua" technology that was developed by IBM while the company was working on the Business Rules for E-Commerce project. What this has to do with intelligent agents is that bots have been relying on the knowledge interchange format (KIF) to transfer knowledge. BRML is supposed to add "prioritized conflict handling" capability to KIF, which is important for maintaining business rules and when transferring information between systems such as those between different internal supply chain partners. West and Hess (2002) say that software agents (intelligent agents) should be employed to help users with interactivity and difficult spatially relevant jobs. This also pertains to the user friendly, or procedural aspect of GIS knowledge management, even when the procedures to use the specific GIS package are contained within the instruction manual that is geared for GIS programmers and analysts (West and Hess, 2002). Intelligent agents can facilitate comparison shopping. Kim and Eom (2002) suggest using intelligent agents to not only allow customers to compare product prices, but also to compare products on multi attributes. For online grocers the various attributes could be product freshness, nutritional information, average delivery time per product, and more. The article states that just as important as allowing your customers to do comparison shopping through agents, is the need to not block competitors' intelligent agents from entering your site to perform comparisons (Kim and Eom, 2002). This is because if a 34

PAGE 43

store does not appear on an online shopper's online comparison list because that store blocked the incoming agents, that shopper might think that particular store does not carry the item that the shopper is looking for. Intelligent agents are discussed by Tsou and Buttenfield (2002) in their literature about GIServices (GISServices are explained below). Agents can be used on a distributed GIS-enabled computer architecture to find and bind spatial data objects across networks. According to these authors, three intelligent agents perform the distributed GIS system find and bind functionality. They are filer agents, interpreter agents, and decision agents. Collectively these agents find the requested information, filter out unnecessary data, bridge heterogeneous data environments, and autonomously make decisions such as what server to use to send the requested information to the GIS user. Sometimes spatially enabled artificial agents communicate between themselves using the knowledge query and manipulation language (KQML) (Tsou and Buttenfield, 2002). Online grocers with more than one distribution center should consider using this type of technology when adopting GIS-enabled delivery solutions. Although not strictly necessary for the development of delivery zones, swarm intelligence and complexity are scientific fields that have recently come to the fore in some business strategies. A primary reason that this literature is reviewed is because of the intuitive value that these two concepts can have for the construction of delivery regions in a geographic enterprise optimization system (GEOS). Literature exists that explains the combination of GIS with "swarm intelligence." A project undertaken by the Department of Geography at Southwest Texas State University 35

PAGE 44

incorporates swarm intelligence with GIS to "study and simulate" various entities in supply chain situations (Zhang, undated). Swarm intelligence can loosely be defined as a neural network of interacting intelligent agents. Whereas a neural network is commonly a closed network, meaning that inputs and outputs are controlled, swarm intelligence can denote the utilization of intelligent agents moving over disparate systems to detect unforeseen patterns (Payman, undated) Eric Bonebau is one of the pioneers of applying the concept of swarm intelligence to managing business He is a coauthor of "Swarm Intelligence: From Natural to Artificial Systems 1999. According to Fredman, ants set up supply chains to accomplish sophisticated ta s ks, and the emulation of these processes can help companies deal with complex environments. Her literature explains how logistics is a natural application of these ant-foraging algorithms." Conventional supply chains use more or less centralized methods to forecast demand and fulfillment. According to Roy ( 1998) centralized planning methods can be disrupted by fluctuations in customer demand. More suitable than centralized planning is conceptualizing the supply chain as a supply web, and incorporating swarm intelligence to traverse the web to better calculate and compensate for demand fluctuations (Roy, 1998). One particular statement by Roy ( 1998) summarizes a main point of my emphasis on profitability thresholds as driving forces behind the shape of delivery zones. This is "any changes in customer demand ... can easily affect how feasible or profitable a given plan may be. These unplanned events will always happen, so it is desirable to create a management system that adjusts to them more gracefully" (Roy, 1998). He uses a vehicle 36

PAGE 45

analogy is used to demonstrate this point. Simply stated, instead of having a "centralized" vehicle and adding parts to it it can be better to conceptualize having thousands of part s that "floc k together" to become a vehicle (Roy, 1998) This concept may be very u s eful when parceling groceries together for individual orders or routes. According to Roy ( 1998) if swarming intelligent agents learn about regions pecific pricing dynamics they could help "develop new s trategie s to take advantage of changing market conditions. Roy ( 1998) says that each swa rming agent can individually generate "internal make s uborders. It i s not difficult to visualize how this could facilitate packaging and s hipping of groceries purchased online. After some time the agents can attain the ability to "c reate new forecasting techniques and learn which of the techniques are most accurate" ( Roy, 1998). When it comes down to it, accurate forecasting i s the linchpin that online grocers depend on. At an extreme, if forecasting could be I 00 % accurate, no wasted efforts in packing or shipping would be incurred, which would of course result in significant competitive advantage. Enterprise Reso urce Planning (ERP) is an established technological methodology that is used to facilitate and assimilate information flows between and within departments in an enterprise. ERP uses centralized databases, common user interfaces and sophisticated middleware to optimize the information flow within a company, or between supply chain members thereby increasing the accuracy and timeliness of inter-corporate and intra-corporate decision making. ERP vendors are increasingly incorporating forecasting functionality into integrated applications. J .D. Edwards sells the Demand Planning 4.0 collaborative application module to enable partners to better make forecasts throughout the supply chain (Ferguson, 2002a). According to Ferguson the software uses 37

PAGE 46

"market intelligence" and demand information to predict product demand. Although the article does not state from where the market intelligence originates, it would be desirable if the information comes from current industry-specific marketing research data. Demand Planning 4.0 also accounts for holidays and promotions when forecasting demand. A key word in the description of this supply-chain centered product is "collaboration." This effort by J.D. Edwards shows that there is importance in having forecasting information readily available when collaborating throughout the supply chain. More collaboration related literature is reviewed below. Southwest Airlines is saving $10,000,000 annually by using swarm intelligence to help route cargo (Fredman, 2003). Problems were arising from ground crews using their judgment about what flights they should forward luggage and other cargo to. Naturally, the cargo handlers would wait for the next flight that had space in the cargo hold that was going in the same direction as the luggage. This procedure, however, was not optimal, as discovered after using swarm intelligence. This technology revealed that many times it is better to leave the luggage on a plane than to use ground crew manpower to change the luggage in mid route. Southwest's freight transfer rates have reduced by up to 80% the necessity for storage facilities has been decreased, and payroll has been reduced because of the decrease in necessary manpower (Fredman, 2003). Other noteworthy companies that are using swarm intelligence-related solutions effectively are Unilever and Capital One. As stated by Fredman (2003), the use of swarm intelligence in complex organizations helps these organizations harness their tacit knowledge. Tacit knowledge is knowledge that is difficult to codify in databases, but has high value because of the exclusivity it has to the company that is using it. In other words, tacit knowledge can give 38

PAGE 47

a company its competitive edge. This is precisely what is needed for businesses to succeed in the low profitability margin online grocery industry. As expected, "swarming" intelligent agents perform tasks that mimic life forms such as ants, bees, or birds which exhibit swarming behavior (Resnick, 1998). Foraging for data (like ants foraging for food) is the primary task of swarming agents. It is not hard to imagine how swarming intelligence could help create delivery zones. Swarming agents could collect data from tens of thousands of entities in delivery regions such as PD As, trucks, databases, and directories, and present this information in real time to strategic decision support systems. Deborah Gordon (2002) published an interesting article about task allocation within an ant colony. Amazingly, although essential tasks such as caring for the young ants, nest construction and maintenance and foraging are accomplished in a very efficient manner, nothing is in charge of managing" these operations (Gordon, 2002). According to the literature, it seems like some of the task allocation within a nest depends upon the number of ants that are already involved in performing that particular task. For example, if an ant leaves the nest to forage for food and does not return, no other ants leave the nest that day. But if it does return, successively increasing amounts of ants will leave to find food and perform other tasks like carrying any dead ants back to the nest. Communication between ants is performed when the insects touch their antenna. Also, ants emit an odor that is specific to the task that they are performing. The ants do not tell the other ants what to do; instead it is the interaction pattern that determines the probability that an ant will perform a job (Gordon, 2002). Another relevant fact is that because ants live only a year, there is no handing down of knowledge from one 39

PAGE 48

generation to another. The work that gets done is due to indigenous knowledge that the nest collectively possesses. Gordon's (2002) research alludes to ongoing work that uses intelligent agents to simulate the collective jobs that are performed by social insects such as ants and wasps. It should be relatively easy to imagine the benefit that the incorporation of these types of agents into an information system that creates delivery zones. The potential decision making accuracy, resilience, forecasting ability, and longevity of such a system are some of the benefits that make the utilization of swarm intelligence a technology that is worthwhile of study by online grocers. A relevant project undertaken at the department of Zoology at Michigan State University is the Multi-Agent Based Economic Landscape (MABEL). The project simulates the group dynamics of individuals in institutional and organizational environments by using swarm intelligence with GIS (Mabel, 2002). Although the incorporation of intelligent agents into GIS-augmented online grocery delivery routing applications has yet to develop, many of the technological tools already exist to make this combination possible. Ants and bees have been studied to determine how these insects efficiently work together to accomplish sophisticated tasks such as procuring nutrition or building tectonically durable nests and hives. These types of studies are not only the underpinnings of swarm intelligence theories and practices, but also integral aspects of a field called complexity science. Van Uden et al. (2001) call complexity science a body of knowledge that is "not trivial" for business, and the "contender for the top spot in the next era of management 40

PAGE 49

science." Carol Kennedy (2000) says that complexity science has "exciting possibilities" for various fields including logistics. The Southwest Airlines example above was an instance of a company using artificial intelligence to solve a "chaotic situation. Complexity science recognizes that systems, whether those systems are biological, technical, or organizational, exist in a chaotic world where forces are constantly impinging upon the system. Regions, such as delivery regions serviced by online grocers, can clearly be considered examples of systems existing in a chaotic and complex environment. Whenever online orders are coming into the grocer s system, the regions that need delivery should change shape or composition. Various complexity issues are inherent when considering the delivery, not least of which is the profitability threshold that encompasses myriad variables such as packing time, crime rate along the route, fuel costs, driver familiarity with the area customer(s) lifetime value, competitive pricing, traffic en route, position of other delivery drivers, customer payment history and payment method, customer profile, and much more. The boundaries of the region are a constantly changing factor because the shape and geographic area of the region's boundaries can allow, or disallow the region to meet a profitability threshold Many properties make a system complex, such as "incompressibility" which is the inability to explain a system in a level that is less complex than the complexity level of the system itself, without losing some of the explanatory aspects of the system. Another notion that is perhaps more relevant to online grocery delivery methods is that a complex system is both deterministically chaotic and anti-chaotic. Deterministically chaotic means being "incredibly sensitive to small disturbances," and anti-chaotic means being "incredibly insensitive to large disturbances" (Van Uden et al., 2001). It is probably best 41

PAGE 50

for a delivery zone delineation syste m to be resilient enough to withstand large, unprecedented spikes in orders, while being selectively reactionary to small patterns or states that could make a region profitable or not. Yan Uden et al. (200 l) sum up complexity science by saying that it is the body of knowledge that claims that "everything is connected to everything else." This author's literature also tells how unwise it i s to concentrate on select part s of a system while ignoring the interactions between these parts If we consider delivery zones as parts of a logistics sys tem, we might find that interaction s between these zones are important in determining if the zones should be combined into one deliver zone to meet the profitability threshold or remain separated. Complexity scie nce tells u s that permanent boundaries never exist ( Yan Uden et al., 2001). Here again is an argument in favor of changing the boundaries of delivery zones in real time, depending on the overall logistical, operational, and profit habitat at hand. "In complexity science, all boundaries are emergent and temporary" ( Yan Uden et al., 2001). Another complexity science maxim that should be followed by online grocer delivery zones is the property of self organization, which is sy nonymous with anti-chaos. As was explained in the section about swarming insects, complex systems have the trait that they can self organize. Concisely stated, when there are a large amount of elements in a large state space, when the conditions are random, these elements tend to converge into small areas of this space" ( Yan Uden et al., 2001). The delivery zones should do exactly this. Instead of considering each and every customer as an individual profit creating entity, the system should converge the customers into profitability zones, 42

PAGE 51

regardless of the geographic proximity of the customers with each other and with the store (up to a certain range). Interestingly enough, this literature says that a trait of self-organization is that if the starting conditions are similar, "a quantitatively similar pattern will always emerge" (Van Uden et al. 2001 ). This can be somewhat reassuring for online grocers because there will of course be times when a certain proportion of the deliveries will be similar to previous deliveries This phenomenon should significantly aid in forecasting and pre-packing the groceries. Stephan Toffler's book "Adaptive Corporation is based on many complexity science maxims. He advises the usage of "complex adaptive systems to cope with the "unpredictability of the modern marketplace (Lloyd 2000). Similar to the self organizing nature of complex systems, Toffler s "Sense-and-Respond" system allows the realization of benefits such as mass customization and the ability to capitalize upon wild marketplace swings. This book shows the growing awareness of the viability of running organizations as complex systems, instead of operating the organizations hierarchically The ability to mass customize can be very important for online grocers, especially if we consider packaging as an operational entity that also can be mass customized which indeed it is because the contents of each package is different than each other package. Therefore, the concepts offered in Toffler's book can apply to online grocers. Both the meaning of "complexity science" and complex" can vary depending on who is defining the term(s) (Van Uden et al., 2001). Christoph Adamnil (2002) said "Nobody knows precisely what is meant by the word 'complexity."' John Casti (200 l) attempts to clarify what the meaning of complex is. By using the examples of infinitely continuing repeating 43

PAGE 52

decimals as opposed to finite decimals, he says that of course the infinite non-repeating decimals are more complex He goes on to say that randomness is also measured by degrees, while Adami! (2002) asserts that randomness is the opposite of periodical occurrences. The relevance of this literature is that an organization might want to determine the degree of complexity or randomness of occurrences in an environment before deciding on what type of system to implement. If one agrees with Adamil's (2002) assertion that "randomness does not give rise to organisms," then before a company establishes a system to cope with complexity, the randomness of the environment should be determined in order to predict if the system will be viable that is, if any type of analogy between an information system and an organism exists. Casti gives a general outline about how to determine complexity and randomness in various environments. Adami! (2002) views complexity as the amount of information that an organism stores about its environment. Global companies are beginning to become aware of the value of complexity science in their business operations. General Motors, Capital One, and Ford are three such companies (Wujciak, 2003). Perhaps the example most germane to this dissertation is that of Ford. Ford allowed a large array of automobile configurations to be ordered by distributors, which created difficult to manage manufacturing schedules and inaccurate demand forecasts. Ford solved the configuration problem by using methods to reduce complexity. One possibly useful way of thinking about "populating" a routing information system that creates delivery zones on the fly is a concept offered by Adami! (2002). He says that entropy is "potential knowledge" and that "sequence entropy" is a length of a 44

PAGE 53

tape, and the marks on the tape is the information Measurement which in the case of online grocers would be the geographic measurement s of the delivery zones, populates the tape. In other words, the geographic measurement s of the delivery zones tum the online grocer's entropy into information. Most relevantly, Adami! (2002) states, "the information-filled tape allows you to make predictions about the state of the syste m that the seq uence is information about." In this case the s tate would be the sta te of online orders at any particular time, and the predictions would be forecasts of those orders. Intuitively Adami! (2002) says that a "complexity catastrophe" is a rapidly changing environment. We could view all of the above-stated online grocer failures as resulting from complexity catastrophes. That is if we consider "rapidly changing" to be the state caused by the reception and re s pon se to online grocery orders. "If the changes are fast and extreme, not only will the organism be maladapted to this new environment, but also its measurable phy s ical complexity will have decrea se d commensurately" (A dami! 2002). This is exactly the s ituation Webvan found itself in, which of course led to its demise. Therefore any online grocer that begins a delivery initiative must keep it s physical complexity at the utmost minimum that is nece ssary to fulfill the orders. In other words, multimillion dollar di s tribution centers, such as tho se created by Webvan, may not be able to cope with the dynamic environment created by online grocery sales. Online Grocers' Store Locations If an online grocer has not yet decided on a site to open the brick-and-click s tore, the site se lection proce ss becomes highly relevant to the businesses' overall strategy. Thrall (2000) asserts that investment in retail location is one of the most important investments a bu s iness can make It can also be the most costly. 45

PAGE 54

Moreover, when one considers that the central s tore is the epicenter of grocery dispatches, which in turn can contribute to the success or failure of the company because of the high delivery expense and the low profit margin the site selection process takes on a high magnitude of importance. There i s much literature that extols the value of GIS for the retail industry (Grims haw 1989; Brown 2000; Murayama, 2001 a; Thrall 2002a; Hakala 2003). Although I have found no literature that deal s with the u se of GIS to create metamorphic delivery regions for online grocers, an abundance of literature exists telling how important GIS is for retail location delivery, a nd retail marketing. Online grocers s hould u se heuri s tic s and techniques that are different from brick and mortar grocers when deciding on the location of their store or distribution center. Although the term "online grocer" is used throughout this dis se rtation I suggest that any store selling online s hould be of the brick-and-click kind. Thi s paper will refer to exclusively online grocers as pure online grocers." Ylachopoulou et al. (200 1 ) enumerate transportation type(s ), customer type(s ), competitor location (s), and sa les levels as some of the factors to consider when choosing a site. The authors recommend using GIS to facilitate the site selection process because of the data and s patial modeling complexity involved. Site selection tools fall in various categories. These categories include analog model s, rules of thumb, linear programming, simulation, checklists, gravity models, and linear programming. The "paras itic" approach is mostly used by smaller s tores that copy the location deci s ion s of larger retailers (Clarke, 1998 ) A broad rule of thumb for online grocers to consider is that goods that are consumed frequently should be distributed from a dense network of locations ( Borchert, 1998 ) 46

PAGE 55

Groceries certainly fit this category, so an online grocer will have to reach a tradeoff in the s ite selection decision as to have more distribution centers, or s tore s, and less delivery truck s, or vice versa. Hernandez et al. ( 1998 ) say that because retailers have notoriously been "cavalier in their approach to store location. Techniques for deciding on the location of a store range from using checklists to sop histicated artificial neural network (ANN) systems. Moreover, some location deci s ions are in response to location decisions made by the competition. To better s trategize retail location some companies have made semi-a utonomou s property divisions within the company to manage s tore location ( Hernandez et al., 1998). This means that smaller scale online grocers might have to have some type of "equalizer" because the se s tores will probably not have the resources to create an entire department devoted to store location or even marketing. Therefore, online grocers could use Hernandez et al.' s ( 1998) tactical, or local marketing, strata to help compete. Hernandez et al. divides retail location techniques into three strata s trategic monadic (location mix, such as relocation re-fascia and re-merchandi s ing ), and tactical (local marketing ) One tactical technique that online grocers s hould be able to efficiently capitalize upon is changing online food price s in respon se to competitors' prices. Although brick-and mortar s tore s can do this by using the UPC ( universal product code) and point-of-sale technology these sto res must rely on periodic advertisement to inform customers of prices. Online grocers who have developed a significant customer ba se can inform these customers of price changes and coupons by e-mail or as soo n as the customers log on to the website. 47

PAGE 56

The maxim that those who do not know history are doomed to repeat it can be applicable to online grocers' site selection strategies. Thrall gives a detailed account of the family "club," showing how the amount of people per dwelling has decreased though time in the United States (Thrall, 2002a). Borchert ( 1998) supports this fact, while also talking about decreasing birth rates of families. The ramification of these and other demographic trends should be looked at closely by online grocers when deciding where to locate. Borchert ( 1998) writes about how lower level stores that were located in city centers are losing their place in established retail hierarchies. It might be possible, however, for online grocers to stem the tide of location displacement by virtue of their delivery strategy. This conjecture remains to be seen, however. Nevertheless, the reinvigoration of many downtown areas, plus the governmental dissuasion of commuters to use cars in downtown areas (primarily in Europe) (Borchert, 1998), can also have an influence on online grocers location strategy. Clarke (1998) takes the utilization of GIS for retail location planning and divides it into three eras, before approximately 1985, when the technology was not used at all for retail location; from that time to the late l 990's, when GIS gained a foothold in many retail organizations; and the present time when data mining and optimization techniques are used with GIS for retail site selection. He supports his threefold division by showing that retail location has become much more complicated recently, necessitating the usage of artificial intelligence and optimization techniques. 48

PAGE 57

Geographic Information Systems for Traffic-Flow Simulation Recently efforts have been made to use GIS for s imulating traffic flows for various reasons. Considering the great amount of money to be made -or lost with an online grocery, simulating the time to delivery should be an integral aspect of any strategic plan. Above, it is described how Punakivi and Saranen (20 0 l ) used simulation to determine the delivery time s to houses using various criteria such as attended and unattended deliveries. Simulation is also used to determine traffic flows within cities' main commerce areas and between different city centers (Medda, 2003). Traffic simulation can be divided into macroscopic, mesoscopic and microscopic traffic flow simulation. Mesoscopic simulation s hows traffic flows at a scale between macroscopic and microscopic. Nokel et al. (20 02 ) have published research about mesoscopic traffic s imulation, while Wahle, Chrobok, Pottmeier, and Schreckenberg re sea rched microscopic traffic flow si mulation. The research about me soscop ic traffic s imulation s tate s that there are many times when the area s imulated must be wide enough to be considered macroscopic, but the finer level of roadway detail nece ssi tate s a granularity that the authors classify as mesoscopic (Nokel et al., 2002). On the other hand Wahle et al. (2002) contend that because of the increasing processing power of computers, microscopic traffic sim ulation is becoming more feasible. The article also says that microscopic traffic s imulation is valuab l e as a transportation planning tool which helps vehicles navigate the roadway easier. If a comprehensive grocery delivery plan is to be implemented, sim ulation u si ng all three scopes of simulation (macro-, me so-, and microscopic ) s hould be considered. Wahle et al. (2002) say that the output, s uch as travel times from traffic simulation 49

PAGE 58

systems can be used as input for other intelligent systems. This is precisely what is necessary for a dynamic (real time) grocery delivery sc hedule information system. The system tested by Wahle et al. (2002) s upposedly functions in real time Simulation can be used to overcome the current analytical and dynamic modeling inadequacies of GIS applications (Wu, 1999). He recognizes the importance of modeling space-time (s patiotemporal ) processes through simulation with GIS, and is s upported by literature by Bernard and Kruger (2000). Integration of GIS with s patiotemporal models will allow the "forecasting of space-time patterns in a reasonable, repeatable and consistent way," precisely what would be required of an application that shows delivery trucks' performance based on profitability thresholds. While noting that GIS-based s imulation is quite complex, Wu ( 1999) says that three contributions can be made to deci s ion maker s by this type of s imulation description prediction and prescription. Relating to the macro, me so, and micro sco pic s imul a tion scopes state d by the authors mentioned above, Wu says that GIS-enabled s imulation may reveal relationships between micro and macro sco pic behaviors. As applied to grocery delivery, the profitability ramifications related to s mall deviation s in order time s might not be discernable in a microscopic s imulation, but might be noticeable if the s imulation was run on a macroscopic scale As stated by Wu (1999), the purpose of s imulation is to see how a global s tructure is evolving from uncoordinated individual behavior s." Another benefit of GIS-enabled simulation noted by Wu i s that the simulation result s can be s tored in GIS "scenario libraries for use in later si mulation s (Wu, 1999). Bernard et al. discuss the integration of GIS with s imulation in so me detail. The difference between coupling and integration of GIS applications is explained. Coupling is 50

PAGE 59

transferring data between two GIS, while integration is a "monolithic" model where GIS and simulation tools are implemented on top of a common data and method base (Bernard, 2000). The literature goes on to explain common ground between initiatives undertaken by both the OpenGIS Consortium (OGC) and the Simulation Interoperability Standards Organization (SISO). The OGC's mission is to facilitate the worldwide proliferation of "geoinformation services." This is to be accomplished in part by the utilization of two models and a specific architecture called the Open Geodata Model, the Information Community Model, and the Service Architecture. Much of this initiative is fashioned after the web services paradigm. Of similar importance is the High Level Architecture which was developed by the US Department of Defense, and is supposed to be a standard for interoperable simulation components (Bernard, 2000) According to this literature, these initiatives are necessary because there is a "gap between the offered interfaces and services in the GIS area and the requirements of the simulation area The integration of simulation and GIS can be highly relevant for online grocers who deliver. This integration should allow the use of "what if' analyses to ascertain if delivery regions will meet their profitability threshold or not which, in turn, will allow managers to better decide when to pre-pack the groceries that have the most probability of being purchased by customers in those regions. Geographic Information Systems for Vehicle Routing and Logistics Pain Stubing of Ernst and Young said, "Execution is the most important thing for any retailer" (Reardon, 2000). This can not be overstated when considering online retailer delivery dispatches GIS, and the role the technology plays in routing, can help online grocers dispatch delivery trucks in the most cost-effective manner. 51

PAGE 60

Closely related to traffic simulations are routing scenarios. There are significant amounts of literature that pertains to the role GIS plays in vehicle routing that can apply to online grocery delivery. Sometimes the GIS can be a component in a spatial decision support system that helps determine the best routes. Campbell et al. (2001) tell how GIS has helped create new logistical solutions. Their paper talks about a hybrid distance approximation solution to assign routes to snow removal trucks in Montreal. Their hybrid model strives to "combine accuracy of shortest paths ... with the simplicity of approximations." One aspect tying this research to my dissertation is that the shortest path algorithm alone is perhaps insufficient to determine the best route for delivering groceries. There are other factors that might make a longer distance path preferable. Essentially, these authors' hybrid method speeds up the routing process by creating a reduced road network that eliminates superfluous roads. According to the authors, the calculation of routes in very large networks can be cost and time prohibitive (Campbell et al., 2001). A problem with many routing tools is that they view the network and traffic costs as "static" entities, or entities that have attained an "equilibrium" state (Wu, 200 l ). This is clearly unacceptable for fast changing online grocery delivery schedules. The article reinforces this idea by saying" ... a static model of congestion is an oxymoron." The need for forecasting traffic flows and congestion is addressed in the article. The acronym GIS dynamic traffic assignment (GIS-DT A) is coined by the authors. Aided by a GIS, analysts can change delivery routes according to changes in predicted traffic flows. This simulation was performed using Arc/Info GIS. The importance of coordinating delivery schedules with traffic congestion is underscored 52

PAGE 61

by Browne as he says that delivery reliability decrease s as congestion increases (Browne, 1993) A notable amount of recent literature exists about using simulation in conjunction with spatial decision s upport syste ms (SDSS) to so lve the vehicle routing problem. While explaining their Hierarchical Path View Routing (HPVR) model, Huang breaks down vehicle routing sys tem s into two broad types. One is where individual vehicles perform their own routing calculations using CD ROM based maps and on-board computers. The other method u ses a centralized path (ro ute ) discovery model. The centralized path discovery model was demon s trated by Huang ( 1997 ) to be le ss expensive to implement The HPYR model is relevant to thi s dissertation. The model uses only a s ub se t of all the roads possible to s ugge st a route for a vehicle They call this method "fragme ntation of arteries. The fragments are classified by road type ( highway, s ide road, etc). This method makes routing computational costs cheaper because not a ll the roads are used to calculate any particular path from origin to de st ination. Becau se of constantly changing road conditions, the s tatus of routes must be recalculated frequently. If HPVR is used this frequent recalculation of routes can be performed more quickly because fewer roads are u sed within each hierarchy. It is worth mentioning here that the hierarchical se t that includes highway s would be used to calculate paths between regions, which may not be necessary for local grocery deliveries However this model ha s merit because if it is employed by local grocers, unnecessary or superfluous roads may be omitted from local route calculations, thereby speeding up the route-optimization process. Alternately, if the online grocer wishes to 53

PAGE 62

branch out (for example, on weekends) to other customer regions, it could incorporate the main artery hierarchy into the analysis. Also, although I don't propose using a hierarchical method per say to form delivery routes the hierarchical segregation of roads for logistics and transportation planning has intriguing potential to be used for further grocery delivery route optimization efforts. GIS can also assist with determining accessibility measures for roads depicted with ITS (intelligent transportation systems). Miller et al. have developed a method of ranking the accessibility of arteries using Arclnfo (Miller, 2000b ). Their space-time accessibility model (STAM) calculates the accessibility of nodes in a network based upon drivers activity schedules. Activity schedules are essentially the times that a specific set of users will use the roads. Accessibility to roads will be greater or less depending upon the time of day. Based upon travel diaries ," drivers' activity schedules are divided into mandatory and discretionary travel times. A similar concept might be used to group online grocery customers into mandatory or discretionary delivery time windows Another consideration online grocers might have when developing a logistics strategy is the type of traffic jam that occurs during a certain time of day. Mahnke and Kaupuzs (2001) classify traffic flow into three "regimes." The first is a free flow of small densities of vehicles, which is obviously the preferable time to embark on a delivery route. Next is the "coexisting phase," where traffic jam clusters coexist within free flowing traffic. And the "viscous overcrowded situation," where a high density of cars move at low velocities." These broad classifications could be assigned as variables to the delivery zones to better ascertain if the groceries can be delivered on time. Quite possibly, for example, easternmost delivery zones could be experiencing the first category, while 54

PAGE 63

northwest sections of town may be undergoing the viscous overcrowding condition. With close inspection it could be discovered that this pattern repeats itself daily. This information should help to enable the scheduling of grocery deliveries with close regard to factors such as congestion patterns. An inseparable aspect of online grocers' delivery solutions should be the incorporation of GIS into their logistic plans. The entire concept of delivery zones is merely a hollow conception if not executed with a GIS. A company called IVU Traffic Technologies uses GIS as an integral part of its e logistics solution (IVU 2000). This company shows foresight by creating a solution that allows visualization of logistical spatial information to allow informed management decisions. The company asserts that GIS functionality is important fore-logistics solutions because of the importance of location visualization for management decisions. IVU claims, as this paper also asserts, that data visualization as facilitated by G IS, is especially important fore-businesses that utilize e-logistics solutions. An example of advances made in the GIS visualization field is 3D Analyst. This software application is an Environmental Science Research Institute (ESRI) product that allows realistic animations such being able to "fly through" a region (Thrall, 2002b). A Korean governmental organization that incorporates GIS into an e-logistics solution is the Electronics and Telecommunications Research Institute (ETRI) which develops GIS / e-logistics postal solutions in their Postal Technology Research Center (Electronics and Telecommunications Research Institute, 2003). Kipling Holding AB a German mobile internet technology company also develops e-logistics solutions using GIS. Kipling's niche is the utilization of global systems for 55

PAGE 64

mobile communication (GSM) in their e-logistic solutions. GSM is a prominent technology used for second and 2.5 generation wireless telecommunication services (GEO Community, 2001). Global systems for mobile communications is showing itself to be a valuable technology in the advancement of e-logistics solutions as evinced by an Indian e-logistics solution provider called, most appropriately, e-Logistics Ltd. According to the owner of e-Logistics Ltd., Mr. V. Sanjeevi, the usage of GPS to monitor trucks is more expensive than using GSM in thee-Logistics solutions (GIS Development, 200 l). Sanjeevi also asserts that en-route delay of delivery trucks can be reduced significantly if e-logistics solutions are utilized According to IVU Traffic Technologies, "about 80% of all data in a company is location driven" (IVU Traffic Technologies, 2001). The company' s FilialWeb technology enables the selective presentation of spatial information. The word selective is important here because even if companies are able to garner and store real-time spatial information, it is another matter to be able to present it selectively to the operational employees that must make a decision about whether a grocery delivery route is profitable or not. Networks LeHeron et al. ( 200 l) explain how supply chains should be networks of tacit and codified knowledge; and that within the last decade there has been a large-scale change in the networks of food consumption. If localized food consumption, especially the consumption that is fulfilled through online grocers, is thought about as being fulfilled through a network of supply chain knowledge, then more innovative models of delivery may be contrived. Using this mode of thought, online grocers might be able to devise 56

PAGE 65

ways of having their suppliers coordinate with the delivery drivers to possibly deliver more bulky items to the drivers as they are en route to or from the customer premises. According to Balakrishnan et al. (2000), there has been "increasing emphasis in recent years on customer orientation providing the right product at the right price, time, and place And this emphasis has propagated up the supply chain. Industrial networks have been defined as the "spatial pattern of sales linkages" (Benenson, 1998) According to this literature, dense and complicated networks can exist even when small sets of linkages (such as those established by small to medium sized online grocers with their suppliers and customers) exist. These types of networks are not unlike matrices, where the dimensions are equal to the number of interplaying objects. Benenson et al. s ( 1998) comparison of networks to multidimensional matrices lends itself to the thought that some problems encountered by online grocers can possibly be solved by viewing these problems with the assistance of multidimensional databases and multidimensional visualization methods such as virtual reality These thoughts should be discussed in further research about online grocery delivery solutions. Indeed, albeit in an abstract sense, Benenson et al. ( 1998) talk about discovering "visually meaningful patterns" in multidimensional information about business networks. They give an example of sales patterns that, they say, could be too complicated to interpret by visual means if depicted on a one or two dimensional map. Moreover, this research cautions to be aware of sub -n etworks that can exist within networks. More in line with the intelligent transportation definition of networks, Zhou et al. (2000) talks about modeling networks with objects Under this model, objects that embody both spatial and thematic information are used to show spatial entities on the GIS. 57

PAGE 66

This literature says that most GIS transportation (GIS-T) applications use a feature-based model to show roads and other pertinent entities. Features are like classes in object oriented programming and modeling. Therefore GIS-T applications should go beyond strictly representing features, and use the objects that result from the features. This is similar to the widely accepted object oriented modeling and programming paradigm. By using this object oriented approach, transportation networks can be more holistically" depicted because each entity such as roads, points, intersections, etc can be objects that contain the data and functions that work upon the data (Zhou et al., 2000). Whereas Zhou et al. (2000) recommend the use of "virtual networks" to better depict multi-modal traffic network analyses, virtual multi-mode networks exist by sharing nodes and sharing points in a network that does not necessarily reflect the exact shape of the real multi-mode transportation network. The virtual network abstractly shows the topological relations of the multi-mode network. Evidently this is a more efficient way of representing multi-mode networks, as the representation of multi-mode networks, but according to Southworth et al. (2000), this is very difficult. Actually Southworth et al. (2000) while attempting to model intermodal (similar to multi-modal) freight networks discovered that it was easier to model these networks without using GIS Online grocers may find that a multi-modal delivery system consisting of s uppliers trucks, company trucks (or cars), and even possibly bicycles or employees that walk the groceries to the nearest customers. A delivery model where these "walkers and/or bicycle delivery employees might even meet delivery trucks at predetermined locations to more rapidly deliver to the "last mile" of customers may be depicted using multi-modal classifications. Although I do not necessarily suggest this type of break-of-bulk delivery strategy in this 58

PAGE 67

dissertation, the possibility of modeling multi-mode logistics might make it more feasible to use simulation to better ascertain the cost effectiveness of such last mile break-of-bulk points. Southworth et al. (2000) claim that because most GIS packages depict networks geometrically, the use of these packages to show multi-modal logistics models is practically useless. This literature says logical intermodal route representations are more appropriate. Data Visualization and Data Representation A scientific discipline exists called "visualization in scientific computing" that contains collections of methods that are used to perform high-definition simulation (Bernard, 1998). The existence of this discipline demonstrates the current emphasis on the importance of data visualization. Gahegan (2000) restates the definition of visualization as "the process of creating and viewing graphical images of data ... with the aim of increasing human understanding." He also lists visualization as an important means to achieve collaboration by experts (Gahegan, 2000). Spatial data visualization, especially when adopted by companies that use integrally spatial strategies, such as routing and delivering groceries, should put significant emphasis on adopting visualization technologies to make operational, managerial, and strategic decisions. The value of clear interand intra-corporate visibility is articulated well by John Fontana (2002), He said, "Companies must learn to read their supply chain's performance ... as a gauge of the health of their business." He goes on to tell how a reflexive phenomenon occurs when something goes wrong in the supply chain (Fontana, 2002). This occurrence, often called the bullwhip effect (Turban, 2002) should be 59

PAGE 68

detected as soon as possible to preempt undesirable ripple effects throughout an online grocer's interior s upply chain. Some undesirable outcomes from the bullwhip effect could be overstocking or stock-outs in peak periods. The bullwhip effect can cause operations to stock unnecessary inventory to hedge against unforeseen online spikes in demand. Robert Malone (2002), executive editor of Inbound Logistics magazine said, "companies need greater visibility along as much of the supply chain as possible." Clyde Witt (2002) supports the notion of the value of supply chain visibility His article about Ford's six-sigma quality control program tells how visibility along the supply chain is important in achieving six sigma logistical operations. If a supplier fails to deliver on time, or throughput on the factory floor drops below a certain amount, the proper people can be notified creating a clear picture of the events that occur throughout the supply chain. Jan Bowland managing director of KPMG Consulting, agrees with the notion of overall supply chain visibility as an important aspect of logistics. She asserts that shipment visibility is rapidly gaining recognition as an important priority for shippers, particularly inter-modal carriers (Kuhel, 2002). Bowland adds that the value of supply chain visibility goes further than just the awareness of what product is at what location at what time. Full spatiotemporal knowledge should help to alleviate the necessity for safety stocks which are kept in case of emergencies. This is because emergencies are less likely to occur if full visibility into the supply chain is attained. Despite the relative awareness of the importance of visibility into supply chain operations, the amount of executives that are actually using sophisticated spatiotemporal 60

PAGE 69

monitoring models or spatiotemporal monitoring technologies is surprisingly low. According to John Zipperer (2002), less than 10% of senior executives surveyed properly track supply chain performance, less than 7 % collect correct or relevant information, and less than one third of the executives track supply chain performance beyond the home corporation. According to Zipperer (2002), telephones and fax machines are still the preferred method for many buyers to send orders to manufacturers, which creates problems such as unnecessary errors The article goes on to mention some outstanding problems with supply chain monitoring, while noting that the problem of lack of visibility throughout the chain is the most important one. Because the visualization of the internal supply chain of online grocers is of primary importance to pare unnecessary efforts, the lack of awareness of visualization solutions must be addressed when incorporating a grocery delivery plan. Cartographic animation is valuable when it results in more intuitive judgments. Using cartographic animation users can see "geospatial transitions as they happen in time, as opposed to simply viewing the end states" (Ogao, 2002). According to these authors, cartographic animation can allow spatial phenomenon to be viewed holistically instead of disparate "instances of time." The increase of bandwidth because of optical fiber networks can facilitate the rendition of three-dimensional rendering technologies such as virtual reality markup language across networks (Shi ode, 2001 ). Goodchild (2002) states that transportation research requires many models but comparatively few types of data. The data that is used, however, is chiefly geographic. Defining the combination of GIS and transportation science as "GIS-T," Goodchild 61

PAGE 70

(2002) elucidates the three viewpoints of GIS-T, the map view, navigation view, and behavioral view. Bergougnoux (2000) supports this standpoint and adds that these three viewpoints can help researchers gain a perspective on the development of GIS Transportation (GIS-T) solutions. He also refers to GIS-T as "a significant area for research and development." Goodchild (2002) uses nodes and links to describe the map view. This is similar to Keenan's physical links and logical links. But Goodchild (2002) notes some flaws with the link/node view such as features that overshoot or undershoot the endpoints. Also, because streets are one-dimensional lines, information pertaining to the respective sides of the streets is lost. Interoperability between spatial databases is another problem. These undesirables are characteristics of the map view. The navigation view "requires a massive extension of attributes provided in the map view (Goodchild 2000). These attributes include factors that show hindrances to the flow of traffic such as turning restrictions (for example, illegal U turns) and one-way streets. An inhibitor to successful representation of streets in the navigation view is that collecting data about the flow of traffic and representing that data in lane views is more expensive than just showing single line, one dimensional streets The behavioral view adds the time dimension and takes into consideration how discrete transport objects such as vehicles, boats, trains or people move on the network. Goodchild (2002) says that all the different methods of representation necessary in the behavioral view are not yet included in any one GIS package Whether the logistics model used by an online grocer needs the navigation view or the behavioral view can be a topic for further research. 62

PAGE 71

Research performed by Winter (2002) underscores the point made by Goodchild that the representation of turning restrictions are an important aspect of the navigation view. According to Winter (2002), every turn has a cost to it such as time used to decelerate and then accelerate, along with waiting time; and these costs should be incorporated in any representation of a transportation network used for logistical purposes. However, incorporating turn costs in a network is expensive because they greatly increase the data necessary to represent the graph, and makes it necessary to create more sophisticated routing algorithms (Winter, 2002). One particular point made by Winter (2002) could especially apply to online grocers delivery plans. He states that a circular tour is when an entity leaves a point, makes a trip, and then returns to that point. This is of course what the grocery delivery truck will do, so estimating the cost, amount, and location of turns that will facilitate, and not hinder, the delivery truck making a circular tour back to the main distribution center should be important factors in the overall logistical plan for grocery delivery. Data modeling features of GIS should be improved to allow better user interfaces such as "exploratory" data analyses that allow users to progress through non predetermined paths (Spaccapietra, 2001 ). This functionality would obviously be useful for grocery delivery. West and Hess (2002) explain that metadata pertains to both the technical and business aspects of a business, and that the business metadata is especially useful to the end users. Much of what West and Hess (2002) say about the importance of intelligent agents and metadata for end users is reflected by Tsou and Buttenfield (2002) as they discuss both of these aspects in the context of distributed geographic information services. 63

PAGE 72

According to Tsou and Buttenfield (2002) two type of metadata are required to implement geographic information services (literature pertaining to GIServices is below) operation metadata and connectivity metadata. Operation metadata facilitates the representation of cartographic information, such as coordinate projection and spatial footprints, across heterogeneous networks. Data connectivity metadata specifies data connection protocols such as java database connectivity (JDBC) and open database connectivity (ODBC) that are used to access spatial data across the network(s). According to this literature, the use of both operation and data connectivity metadata enables geographic objects to be "more accessible, self-describing, and self-managing. Dashboards are tools used to quickly and concisely visualize trends in corporate performance by employees that have responsibility to monitor those trends. Often referred to as digital dashboards (Ricadela, 1999a), these tools s end important operational indicators to the computers of corporate decision makers. Some of the uses of dashboards, as stated by Whiting (2002a), are mitigating threats, highlighting business opportunities, and depicting performance figures. The underlying technology used in dashboards are data warehouses that amalgamate information from disparate data stores. According to Whiting, GE is a prominent user of dashboard technology. The company incorporates uses a corporate-wide dashboard called the "cockpit" (Whiting, 2002a). Dashboards are intra-network or internet-enabled business intelligence tools A difference between dashboards and portals is that dashboards seem to be more task exclusive whereas corporate portals can include links to a diverse amount of information sources. Nevertheless, dashboards are becoming popular and are worthwhile investigating as an integral part of GEOS-enabled online grocer logistic plans. Also, the 64

PAGE 73

term "dashboard" is appealing because it connotes control mechanisms in a functional layout. The incorporation of spatiotemporal "instruments" into dashboards such as the depiction of delivery zones could be useful for online grocer logistics strategies and profit maximization. PeopleSoft and Microsoft are two companies that have begun developing dashboard solutions for businesses (Ricadela, 1999a ; Ewalt, 2002). These applications, often classified as business intelligence solutions, show potential to be incorporated into online grocer logistic plans. According to Ewalt (2002), dashboards can give front-line employees better decision making power. An integral part of a GEOS should be some type of dashboard that shows the color-coded regions to be delivered to changing dynamically. At times, operations managers might have to incorporate their judgment about what route a dispatch should go on, especially if the dashboard-enabled GEOS shows that two routes exist, but only one driver is available at the moment. An instance of managers using their GIS-assisted judgment to dispatch delivery trucks is a reflection of Thrall s assertion that, "G IS is one part of a larger information technology that may be drawn upon to improve our judgment" (Thrall, 1995b) Some similarities can exist between dashboards that contain GIS and geographic information portals. Indeed, some literature refers to portals and dashboards synonymously (Ricadela, 1999b ). A differentiating factor between the two technologies could be that geographic information portals include smart maps. Many aspects of smart maps are similar to the functions that I said should be contained in GEOS in my geography master's thesis entitled "The Conceptual Integration of Geographic lnformation Systems into Enterprise Resource Planning." Those functions essentially 65

PAGE 74

facilitate the retrieval of information pertinent to an area selected on a map. Called "center piece s of graphical information portal s," smart maps can show information not usually included in a GIS such as real time information and interactive regions. As an example, when a user places a cursor over a region of interest on a smart map an associated "subject tree or pop up menu appears, providing the user with choices germane to the region ( Peine! and Ro se, 200 l ). If an information source contained in the tree is selected, a file will be downloaded from a web browser. According to Peine! and Rose (2001), geographic information portals can help non-GIS profes s ionals use relevant spatial information because of the intuitive, visual nature of the information retrieval aspects of the information trees. Their literature explains that logistics is an application area where geographic information portals can be utilized. The real-time information depiction qualities of s mart maps might be suitable for the representation of online grocer delivery zones. Geographic Data and Surface Representation Pienarr and Van Brake! ( 1999) summarize the value of data contained within GIS by saying, "the value of the GIS is dependent on the quality of the data contained within the system, undoubtedly making it the most important component of a GIS ." It could be financially prohibitive for online grocers to generate their own data about roads and customer profiles at least during the first few years of operations. This is why online grocers should be aware of the various types of geographic data available to them. Pienarr ( 1999) divide online geographic data so urce s into five categories. They are, educational, commercial, information providers interactive mapping, and online data searching. Educational resources include universities and other schools; businesses and 66

PAGE 75

organizations sometimes provide GIS data, often on a subscription basis; information providers include online GIS journals. Generator services, map browser services, and real-time maps comprise the interactive mapping geographic data source of data. Online data searching providers can have search capabilities using metadata, and permit transfer of data by Ff P (Pienarr and Yan Brake!, 1999). Declarative knowledge exists within metadata. When end users have to choose between hundreds of map themes, declarative knowledge can help with their choices (West, 2002). This could be especially important if the end users are not GIS professionals, which is often the case when the average manager or end user is not trained in cartographic principals, which are necessary to create professional GIS coverages. Literature by Miller underscores this point as he discusses the potential for improved geographic representation in spatial analysis through the use of GIScience (Miller, 2000a). Spaccapietra (200 L) says the type of end GIS user is changing. Most end users are not GIS specialists, and GIS has not become interactive enough for inexperienced GIS users to use quickly (Spaccapietra, 2001 ). Swink and Speier ( 1999) also highlight the fact that the effectiveness of SDSS is dependent upon the cartographical representation of the data Swink and Speier ( 1999) go on to tell that because of the increasing complexity of business data that must be incorporated into maps (such as the data that pertains to logistics and grocery delivery) the level of cartographic detail increases greatly. Many times it is impossible to show the complete level of detail that is available, so GIS must be used judiciously to show data patterns that help users make better decisions Essentially, as the number of data points to be considered increases, the complexity of the decision at hand increases Swink and 67

PAGE 76

Speier, 1999). Despite this, Swink and Speier ( 1999) have discovered that better transportation-related decisions are made when more customer zones are included in the analysis instead of less. Results vary, however, if the decision makers have varying degrees of "spatial orientation skills." Also worthy of consideration is literature by Bittner and Frank (l 999) who argue that although most of the spatial representations rendered by GIS are currently based upon analytical geometry other ways of representing geographic space might be considered. Most relevant is their proposition that "constraint-based" GIS representation be considered. They also say that language can be a way to represent formal models on a GIS. This too is noteworthy because the results of the calculations for delivery zones can be expressed in words superimposed upon color-coded zones. Somewhat supporting Bittner and Frank's (1999) research is research by Marble (2000) who states that GIS inadequately uses the spatial tools available to it. Calling current GIS-related computational approaches "myopic, he says that the dimension of time has often been omitted in spatial analyses (Marble, 2000). Granted since the year 2000, more literature has surfaced that pertains to spatiotemporality; but Bittner predicts that GIS and spatial analysis are on the "brink of a major revolution In a sense I agree with him because much of the utilization of GIS to determine optimal (profit maximizing) routes for dynamic online businesses like grocers has not yet begun to be realized. This is an important point that is emphasized in my dissertation. Miller (Miller, 2000a) further expresses the inadequacies of current geographic representation through GIS by saying that the inception and rise of GIScience might herald in a "comprehensive re-examination of geographic representation in spatial analysis" (Miller, 2000b). Much of 68

PAGE 77

his assertions are based upon the premise that because computational power is increasing, so should the power of decision makers who use this spatial information. An example of the utilization of computerization to help make spatial decisions is the SpaceStat and DynESDA extensions for ArcView, which perform sophisticated statistical analyses, and depict those analyses graphically (Anselin, 2000). Anselin's (2000) research also says that much of the spatial analytical capacity of applications, especially if those applications can be employed by mainstream GIS packages such as Arc View, Maplnfo, or Arclnfo, will allow non GIS experts to better perform these spatial analyses If this is true decision makers in industries such as online grocery will benefit from the increased ability to peiiorm comp l ex statistical spatial calculations to ascertain profitable deliveries and more. Also, possibly as a way to make end users of grocery delivery spatial applications acclimated to these types of applications, the users and decision makers could be trained first on more simplistic location-allocation programs such as NEWLAP that have intuitive menu-driven interfaces (Lindquist, 2002). Because data (such as the data in relational databases) is tabular, an important aspect of GIS is that it can combine tabular data with cartographic data (on maps). When one considers that much customer transaction data is initially stored in tables, this functionality apparently becomes valuable. The value of GIS is increased further when combined with lifestyle data about customers in retailers' catchment areas. In a United Kingdom survey most respondents preferred the MOSAIC lifestyle database, with ACORN the second-most preferred (O' Malley, 1997). Further when extensions to GIS networking software, such as those that have location-allocation functions, are implemented, catchment areas can be more accurately determined. Geographic location-allocation software assigns demand areas to 69

PAGE 78

supply centers while simultaneously maximizing supplier coverage and minimizing the travel time for customers ( Figueroa, 2000). O' Malley et al. ( 1997) s tate that the most important data for retailers are the data that are generated internally. Online grocers can have one important a dvantage pertaining to internally generated data, which is the generation of internet -ge nerated data. All data relating to purcha ses made o nline can and should be added to corporate databases to allow customer analyses. Moreover, many of the se analyses can be performed in real time u s ing procedure s like online analytical proce ssi ng. Where as brick-and-mortar grocers u se analytics (more o n analytics-related literature is below) to ge nerate coupons on the back of sa le s receipts to be u se d on the next purchase; online grocers can generate incentive s s uch as co upon s to be u sed on the current purcha se depending on what is in the customers virtual o n line s hopping cart at the time. O Malley e t al. ( 1997 ) s upport this notion, say ing that the future is likely to see g reater u se of re a l-time d ata for analyzing ... Geographic data is inherently multidimen s ional a nd recent developments in constraint database s are allowing better representation of multidimensional dat a Succinctly s tated, constraint databases attempt to s how infinite collections of points in a finite number of dimen s ional spaces (Grumbach, 2001 ). An example would be to s how three-dimen s ional polygons as conjunctions of linear inequalitie s Depending o n the s tipulated constraints, query language s can perform operations on the point s that constitute the poly go n s Assessing the exact extent of catchment areas ha s always been problematic The reason for thi s is that the outer boundaries of geographic "sales cones" have fuzzy 70

PAGE 79

boundaries that can only be estimated (Loffler, 1998). Sales cones are areas that radiate away from the store, with the vertex of the cone representing the store (Christaller, 1933). The outer range represents the farthest point at which a customer will travel to a store or a store can deliver to a customer (King, 1984). This outer range is the boundary of a region, where the geographic extent of the region is dictated by the distance decay influence on the customers of a particular retail store that is located within the region. Thrall (2002b) said, "The rate at which demand declines with distance to the retail center is known as the distance decay of store patronage." I claim that the shape of customer sales cones for online grocers will be different than for strict brick-and-mortar grocers at that same location. Further, the sales cones for an exclusively brick-and-mo11ar grocer will change if that grocer begins an online sales and delivery initiative. Additionally, because of the data gleaned from georeferenced on line sales the boundaries of the sales cones will become less fuzzy. Also, speaking of borders of sales cones, it is relevant to note that a "zone of indifference" exists between market areas where the attraction of neither market area is greater for a particular customer (Akwawua, 2001). Also, it is important for online grocers to calculate the decrease of the customers' friction of distance that occurs because of the zero drive times and how those non existent customer drive times affect the gravity model as pertaining to the "law of retail trade gravitation," which is, the ratio of shares of turnover of two shopping centers or central places is proportional to the ratio of their attractiveness and vice versa (Loffler, 1998; Thrall, 2000b ). 71

PAGE 80

Customer drive times should not be overlooked when developing an online grocery delivery strategy. Thi s is because brick-and mortar grocers' first criteria u se d when ascertaining a stores prospective customers' trade areas is the customers' drive times to the sto re ( O Malley 1997). Therefore if, through GIS, an online grocer can discover areas where the re s idents drive times to the nearest brick-and-mortar grocer are comparatively high the online grocer could u se this information to cater to those residents because they would benefit most by home-delivered groceries bought online in comparison to people living closer to the store. Loffler ( 1998 ) expects the relative cost of s hopping trips to increase in the future because of increased drive times to stores combined with greater fuel consumption. Thi s means that the value of purchasing groceries online s hould increase proportionately with the increases in fuel prices and fuel consumption. Luoma et al. ( 1993 ) s upply a somew hat antithetical theory by sayi ng that because of the increasing wealth of the customers, the customers will drive even farther, regardless of the s ize of a shopping center that he or s he i s going to ( Luoma et al. 1993). Still another per spec tive is given by Borchert (l 998), who says that despite any improved customer s pending ability, the amount spent for retail goods increases only marginally. Although all of these postulates may be debatable online grocers s hould investigate and test these assumptions in their regions to better decide upon marketing and delivery s trategie s Buor (2002) demonstrate s how distance decay is relevant for medical services in Ghana. It s hould also be mentioned that literature ha s been published that attempts to refine some aspects of the gravity model. For instance Hu and Pooler (2 002) propose a competing destinations model that tries to remove some bias from distance decay 72

PAGE 81

parameters in conventional gravity models The underlying proposition in the competing destination s model i s that an alternative location can "compensate for the bad features of another alternative". Said another way, "destinations compete for the attention of decision makers, (A kwawua, 200 I). Luoma et al. ( 1993) support thi s by saying that some models do not consider competing destinations that lie within a customers "personal threshold." The theory of intervening opportunities takes the competing destinations model a step further. According to thi s theory, the number of persons going a given di s tance is directly proportional to the number of opportunities at that di s tance and inversely proportional to the number of intervening opportunities" (Akwawua et al., 2001). Akwawua et al. (200 l ) go on to say that some people use the internet to prioritize their search before embarking on a route. Although there are competing destinations and intervening opportunities within the region(s) that a person is willing to travel, a per so n will limit the destinations to which he or she will travel to tho se that are perceived clearly. Online grocers should capitalize on this assumption by having their web sites outstanding in clarity and functionality Chen and Jiang (2 000) while emphasizing the importance of integrating disparate spatial data across networked sys tem s, tell how an "event driven" model of spatiotemporal database changes can be an effective way to utilize spatial decision support syste ms ( SDSS). The idea of using events as the basis for computer applications is not new. Visual basic for applications (VBA) is an event driven programming language. This means that an event, such as a mouse click, pressing of a command button etc, triggers the underlying code to run. Chen and Jiangs' (2000) literature, however, is an 73

PAGE 82

attempt to use the event driven model to represent spatiotemporal data. The authors say that representation of events in databases is a "hot research topic," and "this event-driven approach .. provides a new way for simulating system workflow." Although their research pertains to events within a land parceling SDSS, it is not difficult to extrapolate from this literature how valuable an event based database schema could be for online grocery delivery businesses. If a sudden or gradual increase or decrease of online purchases occurs within two groups of geographically distinct customers, it would be important to determine the cause(s) of this change in ordering patterns Possibly the increase of online orders is occurring because certain sections of two adjacent regions are serviced by two or more delivery trucks on certain days, decreasing delivery times and increasing customer satisfaction. The event here would be the superimposition of parts of two or more delivery drivers routes (a union of area). The end "state of the region served by two routes simultaneously would be increased orders (possibly including new customers). Therefore, the connection between the event and the end state, as shown by the event-driven database schema, could show to management the importance of periodically doubling up drivers in certain areas (possibly in areas of high-value customers). Alternately, the correlation of events and end states, in this scenario, might show management after a cost/ benefit analysis is performed, that the amount of increase in customer orders is not sufficient enough to pay for the added expense of doubling up of delivery trucks for that region or time period. For online grocers interested in possibly initiating an event-driven database schema, Chen and Jiang (2000) list useful operators (in augmentation to conventional SQL operators such as AND, OR, etc). Three of these operators are the sequence operator(;) -showing the sequence of events, the 74

PAGE 83

periodic operator (P) designating an event that occurs periodically, the FIRST operator which designates that an event occurred before another event ( Chen and Jiang 2000). Roddick et al. (2001 ) address other schema issues while talking about the spatiotemporal receptivity (o r non-receptivity) of databases. By using "schema versioning," temporality can be added to a spatial database. This literature makes a distinction between sc hema versioning and schema evolution by say ing that schema versioning is a special type of schema evolution. Schema evolution is the modification of the database schema without any loss of data, while schema versioning i s the ability to query all data within a database "both retrospectively and prospectively ." For dynamic delivery zone related businesses, such as online grocers, knowledge of schema versioning and evolution might be of so me value. However, as Roddick et al. (2001 ) s tate, sc hema versioning has not been incorporated well into s patiotemporal databases, but also say, "spatial schema versioning would be a useful adjunct to many sys tems ." Spatially extended structured query languages (SQL) that u se operators such as overlap, direction contains and distance are better at representing data, rather than analyzing the data ( Huang 1999). Two more recent SQL efforts attempt to allow more spatial analytic capabilities. They are SQL3 MultiMedia Specification ( SQL/MM), and Open GIS Simple Features Specification for SQL (OpenGIS SQL) SQLffemporal is an international standard for spatiotemporal data modeling (Peuquet, 200 I). Many of the functions contained in SQL/Spatial are similar to those in Arc View. Some operations are VORONI CONVEXHULL, DISJOlNT, WITHIN, and DIFFERENCE. According to Huang et al., the functions that create new types of spatial features are the most difficult to define in SQL. SQL/Spatial can be run on a client/ server architecture, with an 75

PAGE 84

interactive front end for making queries and a SQL/Spatial server on the back end (Huang, 1999). While considering employing any database schema technique, online grocers should not forget that not all street databases are directly compatible with each other. Interpretation of features (ontology) and level of detail are two aspects where cartographic data can differ (Noronha, 2000). Objects such as turn lanes and traffic circles might not be represented on all maps. Also, Noronha (2000) says that most databases store addresses as ranges rather than individual address points. This clearly could be a problem for online grocers. Another consideration is that if delivery drivers will be navigating while using mobile cartographic devices (which they should be doing in order to respond in real time to online grocery order changes), is that maps take much longer to download than other types of representations. Berto lotto and Egenhofer (2001) propose a more expeditious way to download maps, which allows initially only a partial representation of the map that is being downloaded. This literature explains the difficulties of segmenting portions of a vector map to be downloaded. These difficulties do not necessarily apply to raster maps. One difficulty is that when maps are overlaid in a GIS, the process of sending those maps over a network is computationally expensive. Map generalization, which is decreasing a vector map's detail, is a "complex and timeconsuming process (Berto lotto and Egenhofer, 200 l ). Nevertheless this type of problem must be solved by online grocers who want to send maps in real time over the network to delivery drivers. Surface representation of elevation also may have relevance to routing solutions The point at which a delivery zone meets a profitability threshold can be calculated by 76

PAGE 85

incorporating some principles of fluid dynamics that use interpolated depressions and relief of a map using a GIS. Atkinson says that measures of elevation are relatively simple to apply to maps, and that this measurement "is of fundamental importance for a range of applications." Moreover, supporting the assertion that profitability can more intuitively, and perhaps more accurately, be depicted and calculated over a region is literature by Chang and Harrington (2000) say that profit can form a "landscape." Their literature goes on to discuss how different degrees of "fitness" can affect the profitability landscape. If Chang and Harrington's (2000) concept of "consumer preferences" can be considered analogous to changing consumer online grocery purchases, we can see how the spatial representation of profitability might be worthwhile endeavor. Etzelmuller (2000) discusses different ways of showing changes in surfaces by using grid-based digital elevation models (DEM). By using DEM, it is possible to quantify surfaces to compare the surfaces for aspects such as roughness, changes to the surface(s), and "noisier" or less noisy surfaces (Etzelmuller, 2000). Some of this calculation is performed by assigning wavelengths to surfaces, and having the higher amplitudes representing relief, and wavelength valleys representing depressions. The DEM calculations result in a coefficient that when equal to 1, represents small topographical changes, and 0 or a negative number means larger topographical changes. This type of allocation of coefficients to topography could have relevance to the "terrain" transcended by grocery delivery vehicles. Therefore, this research deserves further consideration when developing logistical applications for businesses. Additional-ly, besides just quantifying surfaces in DEM using Etzelmuller's (2000) wavelength technique, data pertaining to various features can be overlaid on digital 77

PAGE 86

elevation models. This might be useful in a trough and relief method for calculating profitability thresholds. I propose that "conventional" representation of polygons on GIS are not suitable for many business decisions, such as those decisions that must be made by online grocers about mos t profitable routes. Cressie et al. (2 000) reinforce this idea by stating, "polygons are typically counties, health districts, or s tates, which are politically chosen entities that often have nothing to do with the etiology of the phenomenon. In many ways, using predetermined polygons for routing profitability analysis is too constraining, and cannot lead to accurate determinations of the value of any particular route. Also, attributes related to polygons change all the time. Stagnant cartographic polygons may not be able to adequately show these changes. Indeed Cressie et al. (2000) say that causation of events can be better determined when data relating to polygons are modeled according to time. The literature pertaining to fuzzy points, fuzzy polygons, and spatial uncertainty is provided because of the pos s ibility of applying these type s of uncertainty measurement to online grocery delivery systems. Besides the actual representation of polygons the relationship between the polygons and the point s (c ustomer nodes, delivery truck s, etc) must be considered. When some sort of non-definitivene ss or imprecision exists about the nodes within a polygon we can say that the point s exist with some sort of uncertainty (Leung, 1997 ; Morris, 2003; Robinson, 2003). In essence, fuzzy se ts add a dimension to Boolean algebra that proposes every proposition is either true or false. Fuzzy sets add a "middle," subjective possibility called a membership function (Robinson, 2003). These membership functions can be shown on 78

PAGE 87

an x and y graph as triangular, trapezoidal, or Gaussian. Fuzzy sets can provide an infinite set of values that apply to variables (Morris, 2003). In other words, the degree of membership of a variable to a class is expressed rather than the probability of membership of a variable to a class (Peuquet, 2001 ). Robinson (2003) predicts that as the sophistication of GIS increases the need for human input for the definition of the fuzzy set function should decrease We could consider grocery orders from customers within a region that come in to the online grocer with no determinable pattern (completely s tocha s tic) as random orders. This could lead to decisions to be made under uncertainty. Some of the se decisions would be to have a buffer amount of inventory or have extra drivers on s tandby in case forecasted orders actually come in. Decisions made under uncertainty relate to fuzzy set concepts, which have been attempted with GIS ( Leung, 1997). A point or polygon represented on a GIS can be considered fuzzy if either object's s hape or location is imprecisely recognized. Basica lly stated, a point or polygon can have degrees of belongingness to sets of properties The body of literature pertaining to using GIS to make deci s ions under uncertainty is growing, however, the literature pertaining to allotting degrees of fuzziness to points and polygons is limited Robinson ( Robinson, 2003) s tated that although the u se of fuzzy sets in GIS has grown in the la s t decade commercially available GIS that support fuzzy information processing i s rare The above cited authors' research can possibly be applied to grocery deliver zones, especially if the shape of those zones are considered random or fuzzy. Moreover because fuzzy sets allow the meanings of natural sentences (that contain indefinite words like near, where, how, large, and small) to better be projected on a map, the advancement of fuzzy sets in GIS 79

PAGE 88

seems inevitable, and the online grocers that become early adopters of the technology may gain valuable first mover advantage in this highly competitive market. Cheng et al. ( 1999) say that most natural phenomena are bounded by fuzzy transition zones. Can the changes in customer catchment region boundaries be considered natural phenomena? If not, Cheng et al. s ( 1999) assertion could also apply to "unnatural" phenomena. Nevertheless since online grocers' delivery areas may change, sometimes hourly, the conception of fuzzy transition zones could apply to these regions. However it is difficult to measure the changes in transition zones, especially since fuzzy transition zones may overlap in contiguous regions (Cheng et al., 1999). According to these authors, regions change shape during "epochs; and the changes in the regions are measured by comparing their size at different epochs, during which time they might have shrunk, shifted, or expanded. This is called state transition. Peuquet (200 I) supports this view by stating that regional boundaries, along with boundary categories can be fuzzy. This mode of thought can be of use to online grocers who want to calculate to different degrees about if a region is profitable or not before dispatching a truck. Allan and Lowell (2002) propose another way of dealing with spatial uncertainty by using "abjects. They say that all points, lines, and polygons on a map can belong to several map classes, which denotes that the prominent map class to which these entities belong is not always certain. Polygons whose classes are indeterminate are abjects. Abjects that are surrounded by objects (which definitively belong to a particular class) can, under certain circumstances be joined or merged with the class thereby becoming part of the object itself (Allan and Lowell, 2002). 80

PAGE 89

Funamoto (2000) endorses the notion of multi-level "clumps" of points that exist within geographic polygons. According to him, geographic clumps develop because of various phenomenon such as concentrations of cancer victims within a region. Defined, a clump is a cluster of point s, with each point having a circle of a certain radius emanating from it. The "clump radius" is the collective radii of the circles. What this research seems to be is another way of depicting geographic concentration, which could be of value to logistic models. Funamoto's (2000) research is noted here because of the potential of incorporating his theorie s into online grocery delivery models. Clumps should not be confused with clusters, however. Clu s ter s are the much debated macroeconomic theory by Michael Porter about how related businesses "cluster" around each other ( Martin, 2003). When geographic data is represented in layers, as is commonly done by a GIS, the resulting spatial entities (po ints lines and polygons) are quantified to arrive at certain conclusions. Once the data is quantified, it can be s tudied by order, topology or algebraic methods ( Yongli, 2000). An online grocer might tran sfo rm visual cartographic information into quantitative information by layering a coverage containing the distance from the neare s t distribution center with a coverage containing the amount of purchases per week to derive a new coverage that can tell if closer customers order more often. One way to s how the quantification of map layers is through map algebra, which is a way to sample continuous space, and show those samples on a template that contains rows and columns. Interestingly algebra, which is most often associated exclusively with numbers, is actually a way to "facilitate integration and cross-fertilization of abstract models," (Pullar, 2001 ). In this case, the abstract model is a map. 81

PAGE 90

MapScript implements map algebra using C++ class libraries which include functionality for arithmetic, trigonometric, exponential, and relational operations on templates where template s refer to arbitrary neighborhoods. Groups of cells within a template are a l so neighborhoods. According to Pullar (2 001 ), by u s in g the cell / template representation of neighborhood values, a "large class of high level s tati s tical a nd m a thematical operations" can be performed including time-dependent analyses. This form of modeling can be u se d to examine changes across space and time. Pullar (2 001) says, travel time is related to the cost, or work effort, to move a certain di s tance. My addition to thi s po s tulate i s that s ince time essentially equals money, if the work effort re s ult s in enough profit then the cost of the travel will decrease o r even become non frictional. The r e l a tionship of map algebra to the method I a m s ugge s ting i s that in s tead of conceptualizing neighborhood s using two-dimen s ional template s, s uch as tho se u sed by map algebra we can conceptualize the neighborhood s (delivery zones) as three dimensional entities, incorporating the profitability variable (or o ther var iable s) as the third dimens ion. Thi s could be called map calculus, but s ince the online grocer delivery zones could change, often in re a l time fluid dynamics may be the more appropriate way t o conceptualize this re gio n a l quantificati o n As menti o ned by Bernard a nd Kruger adaptive geographic grids can be used to simulate the movement of fluid. Al so, if cartographic repre se ntation s a re three dimens ional, hydrodyn a mic movement m ay be simulated ( Bernard 2000). Therefore, if innovative thinking is u se d modeling methods s uch as map algebra a nd fluid dynamics may be very important s tep s for quantifying delivery region s for online grocers. It s hould be noted, however, that becau se three dimensional data i s not readily available on the commercial marketpl ace ( Koninger and 82

PAGE 91

Bartel, 1998) online grocers might have to pay for expensive data collection if this option is employed. Also, according to Koninger and Bartel ( 1998), there exists a way to represent solids called constructive solid geometry, which uses Boolean operations to make more complex solid objects from simpler three dimensional objects. Pullar (2001) explains that "mathematical morphology" within map algebra is the theory for analyzing spatial structures. Two morphological operators, dilation and erosion, change the region under study by either increasing or decreasing the size of the region. Also, cells within a neighborhood template can either die, be born, or continue to live, depending on the properties of the nearest cells. [t seems feasible that the dilation and erosion operators, or some semblance of them, can be used to create delivery regions. When incorporating map algebra, or virtually any spatial quantification method, some thought should be given to linear referencing, which can render the necessity of locating points on the earth's surface unnecessary when performing spatial quantifications (Scarponcini, 2002). This is done by using anchor points, such as benchmarks, which are georeferenced points used as initial reference points. Then, from the anchor points arbitrarily spaced points are placed. The points are one-dimensional entities used to specify locations. Once points are arranged in a route or network topology from an anchor point or between anchor points, a linear data topology exists. One trait inherent in a linear topology is that if an anchor point changes because of, for example, the change in the location of a traffic intersection, the linear topology does not necessarily have to change The origin of the topology merely begins at a different anchor point. A section bastioned by an anchor point is called an anchor section. Scarponcini (2002) says that anchor sections are more stable than routes because routes become 83

PAGE 92

unusable as the entire linear datum topology changes. This is not the case in linear referenced topologie s becau se in linear referencing measurements relative to the anchor point (s) are made when depicting the datum topology as opposed to absolute mea s urement s s uch as me as urement s made to place mile marker s at permanent interval s from a state or county line for example. When a route comprised of absolute mea s urement s changes, each measured point on the route mu s t c han ge Thi s is n o t so with linear referenced topologies (Scarponcini, 2002). Ju s t as abso lutely measured di s tance s can cause problems when route changes occur, s p a tial sca le poses its own problem s when u s ing GIS to decide o n best routes, s uch as delivery route s u ed by online grocers. Even the definition of "scale can be a mbiguou s, me a ning both p at i a l extent" a nd "amount of detail (Atk in so n 2000). An aspec t of sca le that i s important t o online groce r s i s that entities s hown a t o ne sca le can acq uire con iderably different relev a nce when viewed at another sca le. As an example, areas not se rviced might not be v isible when viewed on a GIS in s mall sca le whereas the se areas mig ht become highly noticeable when v iewing a s ub sec tion of the o riginal area a t a larger scale. According to Atkinson heterogeneity and regularity of cartographic entities can change when the sca le i s adjusted. The same grocery delivery routes that look elliptical at a s mall sca le might look poly go nal at a larger sca le. An exception to thi s phenomenon i s fractals. Fractal s are pattern s th a t do not s i g nific a ntly deviate when viewed a t different sca les. Fractal pattern s exist in both phy s ical and human geography (A tkinson, 2000). One way to decide o n a sca le appropriate for th e job a t hand s uch as viewing the mo s t grocery delivery zones, while see ing the clo ses t points any two delivery truck s are at the sa me time, i s to scale the map down, then scale it up until robu s t parameters are 84

PAGE 93

discovered. Robust parameters are those parameters that are visible and relevant at all scales within the study area. The area studied is also known as the "support" (Atkinson, 2000) Adjustments in scale can also affect the way spatial dependence between regions is viewed. The fact that households of certain LSPs (lifestyle segmentation profiles) live closely together within a region can be lost if the regions are viewed at too small a scale. Some of this information loss can be mitigated by kriging, or "smoothing" of information from neighboring regions to adjacent regions. Moreover, because scale can be divided into spatial and temporal realms (Maruca, 2002) the problems associated with scale become that much more complex. Somewhat related to the above-stated micro-, macroand meso-scopic traffic flow simulation scales is micro-, macro-, and mesoscale scale representations (Bernard, 1998). Mesoscale representations extend from a few to several hundred kilometers. Microscale and macroscale representations are of course, respectively less than, or greater than this range. According to Haining et al. (2000), quantitative measures that are incorporated into GIS packages are not sufficiently inclusive or sophisticated. Nevertheless, it is important that online grocers become aware of some of the existing quantitative functionality. There is no shortage of literature pertaining to this topic although no writings could be found that are exclusively focused toward online grocer decision-making using GIS enabled statistical processes. Although considerable breadth could be allotted to this subject, an account of just a few of the more pertinent spatial statistical packages is included in this literature review. 85

PAGE 94

Spatial analysis in a GIS environment (SAGE) is a statistical package that can be integrated with Arc/lnfo. SAGE works with polygon coverages and includes data management, graphical drawing, querying, and classification tools (Haining et al., 2000). Polygons input into SAGE can be "dissolved" to create coverages for sets of regions. Linear regre ss ion, a common s tatistical method u sed by geographers and other scientists, is enabled by SAGE. Ordinary linear regre ss ion, linear regression with s patially corre lated errors, and linear regression with a spatially l agged response variab l e are the three types of linear regression that can be performed by u s ing SAGE. Dynamic brushing, the operation of quantifying the movement of a s hape over a region, is not handled very well by SAGE, however ( Haining 2000). a s patial data analysis package that can be integrated with ArcView i s called S-PLUS. Besides being a quantitative sys tem with more than 2000 functions, S-PLUS i s also an object-oriented lan g uage. By creating customized ArcView Avenue scri pt s, u sers can increa se the a n a lytical capabilities of S-PLUS. Data tran sfe r between Arc View and S PLUS is supposed to be relatively seamle ss (Bao, 2000). S-PLUS graphs include boxplots, histograms variograms, and multiple three-dimensional layouts. Geographic Information Systems in Business The literature cited lea ves little room for doubt that GIS in combination with deci s ion-making tools and artificial intelligence will continue to be useful as a facilitative system to allow optimal deci s ions in many business se ttings including the online grocery bu s ines s Busines s geography and managerial geography" are gradually being recognized as a mean s to a competitive edge ( Ri s to, 1998 ) Dr. Grant Thrall (2000) has recently published an important book about bu s iness geography, which give greater 86

PAGE 95

impetus to the recognition of the field of business geography as a bona fide aspect of geography and an important aspect of business location strategy. Following is a review of other literature that asserts the usefulness of G 1S as a managerial tool. Grimshaw (200 I) says that the primary business benefits attained through scheduling and routing systems is gained by linking geographic knowledge about customers to these systems. Although there have been great efforts by small and medium enterprises to integrate corporate-wide data, such as through ERP systems, most businesses have ignored this "key dimension" of data. Businesses concentrate on the "what, how, and why" of strategy but to a much less extent consider the "where" dimension (Heinritz, 1998; Grimshaw, 200 l ). One reason for this non-integration of geographic data is that smaller businesses, such as many online grocers, do not have the technical or intellectual resources to do this. Despite this, O'Malley et al. ( 1997) have found that 48 % of survey respondents (United Kingdom retailers) develop in-house spatial databases. Yet, these authors have stated that retailers are "data rich .. but .. information poor" (Hernandez, 1998). Nevertheless, as Grimshaw (200 I) says, because "benefits from improved routing and scheduling go hand in hand with benefits from improved customer relations," the geographic aspect of data should not be overlooked, especially by the online grocery industry, which has low profit margins. Perhaps one of the most important point Grimshaw (2001) makes is that geographic knowledge about who the most profitable customers are. ln order to better calculate the profitability of each delivery route, georeferenced customer profitability variables should be incorporated into the equation. Ni raj et al. (200 I) bolster the notion of the different value of customers. Their literature also explains that the lifetime value of a customer 87

PAGE 96

should be calculated. Although their research concentrates on the lifetime value of supply chain partner s, some of their findings can be applicable to the lifetime value of online grocery customers. One thing they say is that the cost of serving each customer can be different, thereby affecting the lifetime value of the customer (Niraj et al., 200 l ). This becomes clear in a n online grocery delivery context when variables like time to walk to the back door, pre se nce of mean dogs, and high crime neighborhoods are added into the lifetime value calculation for individual customers. Of the three categories of GIS s tated by Brown (2000), the category of GIS as an analytical sys tem best fits the need s of online grocery delivery initiative s. As an analytical system, the use of GIS ha s various benefits for busine ss decision makers. Hickman goes as far as to say that by converting large amounts of geographic and tabular information into a more useful form, G IS "enhance s every aspect of the business" (Hickman, 1999). Mos t pertinent to this dis se rtation is the article's mention of the usefulness of GIS to help track routes in real time. As "early as" 1989 the value of G IS for business was predicted by so me literature. While reinforcing the fact that much industrial data have geographic dimensions, Grims haw ( 1989) said that retailers can revolutionize their marketing plans by georeferencing customer data. Two of the four factors that he sai d would influence the adoption of GIS in business are particularly true for the online grocery industry. Those factors are better techniques for handling spatial data, and awareness of the spatial data combined with human skill s ( Grimshaw, 1989) A main point of this di sse rtation i s that if online grocers handle their spatial data better, thereby resulting in the recognition of 88

PAGE 97

zones that surpass profitability thresholds, even online grocers, with their notoriously thin profit margins, could become successful. In a later article, Grimshaw ( 199 l) expounds upon his earlier literature and states that GIS should be an integral part of overall corporate information systems and "strongly linked to the business strategy. Indeed, this is the view that online grocers should take when considering GIS as part of their logistical model. He also asserts that GIS should be used as a not only an operational and managerial tool, but also as a strategic tool. Although there could be debate as to whether he properly classifies the utilization of GIS to facilitate the "delivery of information-based products or services as a "strategic" entity, as opposed to an operational entity, there is little doubt that GIS should be used to perform the delivery function. Nevertheless in the same article he asserts that considerable value can be added to corporate information if the geographical a spect of that information is included in the analysis. He expands his evaluation of GIS from mostly a retail-based tool, to a more sweeping analysis of GIS in his more recent publication (Grimshaw, 2002). While considering GIS as a part of corporate strategy, Grimshaw ( 1991) says that the use of the technology should be considered an opportunity rather than just a cost. Another way to say this would be that there is an opportunity cost by not using the technology. In the case of online grocers if GIS technology is not used it may be infeasible to run a successful delivery service, which means that by not using the technology the opportunity to reach the large home grocery market will be lost. Interestingly literature is still being published that calls GIS "new. For example David Hakala s (2003) article entitles "Location: The New Killer App" tells how GIS 89

PAGE 98

when used with databases can be used to make "powerful applications." This phenomenon of calling two decade-old technology "new" means that there is still significant ignorance within the business community about the potential of GIS as an operational tool (Hakala, 2003). This was not the case with, say, enterprise resource planning (ERP). Is the contrast between the adoption of GIS and ERP due to the possibility that most people are geographically uninformed? Whatever the reason, the stage is set for the adoption of GIS in industries such as online grocers because of the need to view geographic information as a profitability enhancer Despite the relative slow pace of adoption, GIS is gaining momentum in certain sectors Murayama (2001 b) talks about the adoption of GIS in Japan. Interestingly, he says that most of the editors of GIS-related academic journals are geographers. This in itself shows the as yet lack of intellectual osmosis of GIS into other business-related publications. One might think that integrally spatial fields of business such as logistics or marketing would begin to publish spatially related journals. Evidently, this is not yet occurring, which can also help to explain the lack of geographic solutions employed by businesses in these sectors. One reason that GIS is gaining popularity in Japan is that data is collected down to very small regions. Also digital maps have been developed at a "rapid pace" since the early 1990s (Murayama, 2001 b). Also, according to the same article, point of sale (POS) data is also being accumulated in large quantities in the country. Because of these facts, it seems promising that this research can have strong implications to online grocery delivery services in Japan. A countervailing situation exists, however. This is that Japan 90

PAGE 99

has stricter privacy laws about the usage and dissemination of personal spatial data than the United States does (Murayama, 2001b). An interesting perspective about the use of GIS in cyberspace has been written by Masanao Takeyama who says that geographers have just begun to study cyberspace, and that the study of geographic "telepresence" holds important meaning for determining "cyberplaces." Cyberplaces are places of "interactivity between cyberspace and physical places (Takeyama, 200 l). Getting a handle on the geography of daily or cyclical cyberplace pattern s may help online grocers decide between using home delivery or communal drop boxe s Takeyama (200 l) s ugge s t s that cyberplaces can cause s patial metamorphism in cities. Online grocers that can detect these new s patial pattern s might be better able to cater to the growing market of mobile and s tationary internet users. O Malley et al. (1997) tells that although store location i s the mo s t critical deci s ion a retailer makes ," the u se of GIS to make s uch a deci sio n by United Kingdom retailer s i s not as pronounced as it could be. A reason for thi s i s that the technology i s not integrated into s trategic deci sio n making to any great extent ( O Malley et al., 1997). Other recent literature exists that s upports the premise that GIS and spatially oriented business practice s a re being adopted more. Lowe ( 2003) says, The u s e of geospatial data is moving from the mapping department in the back office to the existing busines s practices of the enterprise. He also talks about supporting technologies th a t can enable the non-GIS expert to better use spatial bu s iness solutions. One such technology i s peer to-peer ( P2P ) computing which could allow any GIS application on a computer to simultaneously be a client and a server. According to Lowe (2 003 ), decoupled peer-to peer spatial environments are removing old barriers that used to exclude non-technical 91

PAGE 100

users. This means that it is becoming more feasible to initiate solutions such as the delivery route system, which is suggested in this di sse rt at ion. Despite the comparatively slow s tart GIS ha s had in taking a foothold in mainstream business applications, the move toward adoption of GIS by science and industry i s called a GIS revolution (Ya no 200 l ). This literature says that there have been efforts to increase the value of GIS by making it a geographic knowledge system (GKS). This transformation ha s sometimes not been s ucce ss ful because so me advanced quantitative procedures that can be u se d with a GIS have not been adopted by main stream bu si ne sses. A factor that might be contributing to thi s relatively low acceptance of GIS i s that United State s bu s ine ss es have been influenced relatively little by academic geography. This i s because some of the ge neral public do not fully perceive geography as being a bona fide profe ssio n ( Yano, 200 I ) Enabling Geographic Information Systems Technologies Because the purpose of this dis se rtation i s to make s ugge st ions that may increase online grocers profitability through s trategic delivery zones, some review of literature of the enabling GIS technologie s will be of value. For various rea so n s, a non-centralized GIS information syste m s platform can better serve end u sers. One reason i s that data that i s not centrally located can be more readily accessed by end u se r s who utilize a di s tributed sys tem. Also, it has been demons trated that neare s t neighbor queries can be efficient l y executed in multi-di s k and multi proce ssor environments ( Pap a dopoulou s, 1997). By the early l 990's GIS u sers where beginning to realize the redundancy behind collecting and creating their own data se ts, especially when they could relatively easily 92

PAGE 101

tap into the same type of data that existed across distributed networked environments (Hunter, 2002). Within geography, this type of awareness provided inertia to adopt distributed object and data models such as component object modeling (COM), common object request broker architecture (CORBA), and web services to be used as decentralization facilitators. The Open GIS Consortium attempts to use these technologies by addressing geographic data interoperability problems (Hunter, 2002). The exploitation of the internet al.ong with the advent of spatial data infrastructures are two developments that have great significance to the utilization of distributed GIS (Abel, 1999). According to these authors, the high capital investments that are required for some sophisticated GIS operations are putting a drag on the proliferation of geographic analysis in business, industry and academia. As a possible response to this, spatial internet marketplaces should allow the geographic distribution of data sets and geographic services by a subscription / request paradigm. In this paradigm, value added resellers would sell the data sets to the customers as soon as it is requested. The literature compares this scheme to pay-per-view television (Abel 1999). To accomplish this task, federated databases must feed sophisticated distributed processing infrastructures that link together heterogeneous networks. Geographic communities of interest will be formed upon the realization of this type of technological and strategic plan (Abel, 1999). Two important services that should be offered by int ernet marketplace infrastructures are query services and function services. Query services supply a subset of data to the customer depending on the customer's needs. Function services perform computations on the requested data on software owned and operated by the service provider. 93

PAGE 102

Before continuing with the review of distributed GIS systems literature, a short synopsis of literature related to interoperability is provided next to better become aware of some current technological advances that can facilitate the implementation and deployment of distributed GIS systems. The term "interoperability" can encompass the use of many different technologies and methods that focus on integration of disparate computer applications. web services, as mentioned above, show great promise to integrate applications. An integral part of web services is extensible markup language (XML). This is a growing standard that is used for creating markup languages and describing data (Deitel, 2003). A main purpose of XML is to keep the data independent of the applications that use it. Taking the utilization of XML one step further is extensible stylesheet language transformation (XSTL). What XSTL does is use XML to basically allow data to be reformatted into different forms depending on the situation. For instance, XSTL can transform a purchase order into a web page or a printable invoice (Dettelback, 2002). This can be a significant step in the ability to visualize data along interior online grocers supply chains, especially if documents can efficiently be converted to web pages that are accessible to departments over corporate intranets. According to Paul Sholz (2002), XML will make data queries possible on a "global scale." This adds emphasis to the premise that XML can be an important part of interoperability between mobile delivery units and the back end databases. The term virtual enterprise network (VEN) has recently appeared, giving but another term to the lexicon of terms that relate to interand intraenterprise interoperability. An indicative aspect of VENs is that they concentrate on using inter-enterprise directory 94

PAGE 103

services to facilitate access to intranets and extranets, while making those networks more secure an important part of any supply chain interoperability so lution (Lewis, 2002) Wang and Nig (2000) say that distributed GIS can fulfill users' needs better than centralized sys tem s Some reasons are greater system reliability, faster response times, better data s haring capability, and facilitated system growth (Wang and Nig, 2000). Their article shows how CORBA i s a technology that allows geogra phic objects to be stored at remote locations ; and how the se objects can be called from networked computer nodes. CORBA also help s to so lve another problem of di s tributed geographic data which is tha t geographic data s tructure s on two different nodes (co mputer s) can be different. CORBA can mitigate thi s problem by the ability of the technology to u se a common databa se sc hema on the distributed sys tem. Moreover, the use of CORBA help s to maintain concurrency control, which i s the protection against two s tation s (peop le or computers) manipulating the sa me d a t a a t the sa me time ( Wang a nd Nig, 2000). What doe s thi s mean for online grocery delivery se rvice s? For one, delivery people can have access to a richer amount of computer se rvice s that orig inate from various points The se se rvice s can include customer relation s hip management (CRM) capabilities s uch as cross se lling up se lling or automatic generation of coupons based upon the customer's profile. Because Wang states that greater number of GIS s will be networked thi s technology can be important for even s maller online grocers. Thi s can also help to ascertain a customer's lifetime value to a company by better s howing the customer's tran sac tion pattern s and preference s ( Ni raj, 200 l ). A review of literature pertaining to di s tributed GIS sys tem s would not be complete without some account of literature that pertains to rniddleware. Middleware is the set of 95

PAGE 104

computational devices both computers and software components within the applications on those computers that logically and or physically sit in the "middle" of end user applications and back-end systems such as databases. Middleware technologies are used to integrate computer systems in disparate departments or enterprises into a working functional system. ln an article where the investment firm Morgan Stanley states that one of the firm's chief priorities is the integration of its applications two mainstream middleware technologies are compared (DolGicer, 2002). Java servlets and enterprise java beans (EJB) are two middleware technologies that are described in the article. Although the article is quite inclusive in its comparison of these two powerful middleware technologies, one comment is particularly noteworthy. That is, when referring to e-commerce, "there are really only two middleware platforms to choose from -J2EE Uava 2 enterprise edition) and .NET ( dot net). Enterprise java beans and java servlets are incorporated in J2EE. In an effort to give order to the different business processes that must be marshaled into the middleware platforms a technological model called business process management systems (BPMS) takes the end-to-end business processes that exist within an organization or across organizations and establishes a set of rules that the software must abide by to route" the information through the middleware platform(s). A language called business process modeling language (BPML) is an XML-based language that helps coordinate data within the BPMS (Baker, 2002). Baker, Smith, and Fingar's (2002) research show how complicated underlying technologies make implementing end-to-end business processes difficult. The technology became a barrier because the technology is a 96

PAGE 105

"manifestation of the complexity and diversity of the sys tem s it ( the enterprise) had acquired and developed over the years." According to the article, BPML can help to model business processes at the bu s iness and not the technical level. This essentially isolates busine ss de s ign deci s ions from the complexity of the technical middleware infrastructure ... Taking the potential of BPML one step further, Renee Fergu so n (2002 ) writes about how the modeling language can play a part in offering busines s proce ss integration services over the web. Version 1.0 of BPML, which was released in August, 2002, use s a protocol called web se rvice s choreography interface (WSCI). The two technologie s can help provide softw are engineers and busine ss analysts with views of how busine ss proce sses perform in different sce narios. In the article, howe ver, there i s so me s keptici s m as to the amount of companies that will actually u se BPML a nd WSCI. Enterpri se information inte gra tion (Ell) i s the u sage of middleware to g i ve access to multiple dis parate databa ses (A pril 2002). A common term for this type of "tra n s parent data access is "fe derated d a tabase s Enterpri se information integration perform s data coalescing by u sing metadata repositories from various backe nd so urces. What thi s means for visualization within online grocers' interior s upply chains is that any department who have access to another department's data, through an intranet or portal for example, can access cross corporate data as easily as if that data was on just one database. This i s relevant in companies where terabytes of data might exist across the company on different platform s According to the article, Microsoft's new Yukon SQL Server d a taba se and Oracle' s 9i have Ell capabilities built in. GIS can be distributed among disper se d u sers by using 97

PAGE 106

what is called geographic information services ( GIService s) (Tso u and Buttenfield, 2002). Clearly this is modeled after the web services paradigm. Recent research s how s, however that the provision of GIS as a web se rvice require s different technical obstacles to be overcome. The benefits of GIService s -enabled distributed sys tems parallel the benefits set forth by Wang for CORBA-enabled distributed syste ms. A basic tenet of GIS se rvices is that every computer in the distributed system i s both a client and server. Thi s architecture help s make the sys tem application independent ( meaning the variety of applications that can be used by the sys tem is not constrained by those applications contained by the core system itself). Instead of adopting only CORBA, T so u and Buttenfield (20 02 ) s ugge s t u si ng CORBA in conjunction with the distributed component object model (DCOM) a nd remote method invocation ( RMI). It i s easy to see how GIService s can benefit a n online grocer, even if that grocer ha s only two distribution locations. If there was no interoperability between the respective s ites' GIS-enabled routing applications, each di str ibution center (store or warehouse) might have to deliver to its own geographic region By enabling the distributed geographic components (see below) to communicate, routes can overlap much like unions in Venn diagram s. This could allow better customer service s uch a backup deliveries initiated by another distribution center (store) if the other s tore' s delivery truck becomes delayed. Of course, this scenario assumes that different grocery orders are routed to respective stores ba sed upon georeferenced delivery addresses. Components, as mentioned in the previous paragraph, are "ready-to-ru n modularized programs that are dynamically loaded into a network-based sys tem to enable GIS functionality" (Tsou and Buttenfield 2002). Essentially components are simi lar to 98

PAGE 107

objects in object-oriented programming, but components are more sophisticated, occasionally containing objects. Sometimes components themselves are encased within "containers." Containers are abstract grouping entities used to better accomplish the abandonment of traditional client / server architectures. This is because each GIS node (computer) is both a client and server in the GIServices paradigm Componentization of GIS can alleviate the need for batch conversion of data and system import and export difficulties. This is particularly relevant because spatiotemporal data representation necessitates the exchange of great amounts of data between systems (Bernard et al., 1998). Bernard et al. ( 1998) expound upon the use of component models such as DCOM and CORBA along with application programming interfaces (API) to achieve better interoperability between GIS Abel also talks about the importance of these component models to enable interoperation within spatial internet marketplaces (Abel, 1999). A problem with distributed GIS is that different applications may use different data dictionaries and taxonomies for the data (Pundt, 2002). For advanced decision making capabilities and better data accuracy, some online grocers may wish to collect their own data. For example, the online grocer might want to know how long, on average, it takes a particular residence to open the door in order to better determine the route (people who answer the door slower might be delivered to at specific times in the route). Generally speaking, older people might take longer to receive their packages than families with younger children. It is possible to input this data with mobile GIS devices (Pundt, 2002). However, Pundt (2002) says that the issue of semantics "the meaning of the geographical object rather than the geometrical representation" should be considered too. Zhou (Zhou et al., 2000) agrees with this but say that semantic 99

PAGE 108

information is difficult to represent on a GIS. If an online grocer operates from a single store, the problem is less troublesome than if various stores must access the same semantically classified data. Pertaining to the time it takes to receive grocery packages, if one store uses the semantic data tags STR, ATR, and FfR, meaning slow to receive, average to receive, and fast to receive, all other stores that use or input data into this distributed system must be aware of and use the same semantic style. Pundt (2002) says that it is more efficient for GIS workers in the field if they know the semantic correctness of the data as they put it in. The input of semantic data by grocery delivery persons can be done by using text box masks, which force the person inputting the data to input it in a certain style such as only numbers within a specific range, or only a specified amount of letters per text box. Semantic problems can manifest themselves in vague terms such as long, near, old, far and close; or in statements such as "two roads 'cross' each other" (Hunter, 2002). Online grocers should take care to not incorporate these types of ambiguous terms in analyses of delivery routes. The value of mobile GIS users (grocery deliver persons) inputting data into the system should not be underplayed, especially in the very low profit margin grocery industry. Certain seemingly insignificant information, such as that about streets having inordinate amounts of deep potholes that slow drive times or result in cracked eggs could be easily left out of consideration if drivers are not instructed to record this information while on route. Further, if the drivers are versed on inputting the data into mobile GIS systems, the possibility of incorrect data entering the database could be lessened because the necessity for repeat data entry is lessened. Possibly, an online grocer could make its 10(

PAGE 109

own ontological classifications for delivery drivers in order to expedite the data entry process in the field. According to Pundt (2002), ontological (naming and grouping) considerations are very important when using mobile GIS. Grimshaw (2001) also wrote about the value of having delivery drivers add to the corporate spatial knowledge base. He says that information about customers that have narrow drives or poor turning areas can greatly help to improve customer service. This type of information can be used to create impedance tables or tables that geographically relate to the road network and show hindrances to rapid traversal of the network (Figueroa, 2000). Dunkin Doughnuts is using wireless devices to profile their customers by using gathered spatial information. Handheld devices used by Dunkin Doughnut employees transmit customer survey information to home-office servers (Barnes, 2002). Another problem with semantics is that data used to visualize geographic entities go through various representations from the time it is captured to the time it is visualized on a GIS. During this process data semantics can be lost and the final geographic representation rendered unclear to the user (Bernard, 1998). Supporting this viewpoint is literature by Frihida et al. (2002) that says geographic "semantic richness" can be lost by database normalization rules. Spatiotemporality The amount of literature pertaining to spatiotemporal issues has increased significantly in the last five years. This is probably because of the new emphasis put on spatiotemporal phenomenon (Spaccapietra, 2001 ). It is only natural that any firm (not just online grocers) that moves goods or people be interested in both the spatial and temporal aspects of the data they use. 'Time geography," which of course encompasses 101

PAGE 110

spatiotemporal concepts, is defined in the Dictionary of Human Geography and includes transport geography, which is a subcategory of human geography that includes household space-time budget s (as mentioned above). According to Lenntorp ( 1999), during the 1980s more people incorporated time geography into their thought s, and ha s thu s become more widely accepted by geographers. Intricate and in se parable processes" can be observed through the u sage of time geography ( Lanntorp 1999 ) One of the problems with representing discontinuou s ly moving s patiotemporal entities on a GIS i s that unle ss every si ngle latitude and longitude coordinate of the object represented i s captured, so me "jerkiness" will re s ult because of the l ac k of s mooth arcs that re s ult because of the paucity of point s. Saglio and Moreira (200 I ) attempt to rectify thi s while demons trating software benchmarking by u s ing the Oporto si mulator. A benchmark i s attained by u s ing software to test a sys tem to see if it i s functioning adequately As opposed to di sco ntinuou s ly movin g objects objects that change (location direction s ize o r s hape ) continuously are more easily represented because there could be a function of time that exists that when used can more accurately s how the change with no di sco ntinuity points. It i s important to make the di s tinction between moving point s and moving polygon s because moving points are more easily represented (Saglio and Moreira, 2001 ). Online grocers might be intere s ted in simulating online s hape s, especially if they are s imulating the increase or decre ase in popul a tion flow to their marketing catchment areas for example. Zhang and Hunter (Zhang and Hunter, 2000) s upport Saglio a nd Moreira's (2 001 ) discontinuity claim by saying that a problem with incorporating temporality into GIS i s 102

PAGE 111

that unless the "continuity assumption" is used, it is difficult to accurately show the trajectories of spatially dynamic objects (Zhang and Hunter, 2000). Most relevantly in Zhang and Hunter's (2000) literature is the assertion that instead of using the z axis for spatial height, as in DTM, the z axis can be conceptualized as a time entity (in which case it would assume the letter t, not z). Even though geographic data and data representation literature are discussed above, the distinctive nature of spatiotemporal data necessitates allotting some more space in this section for a review of the subject. Despite the literature that is devoted to spatiotemporal issues such as that about spatiotemporal databases, advancements of practical applications in this field have been slow (Tryfona, 1999). Nevertheless, the need for integration of spatial and temporal dimensions is recognized in various current literature such as that written by (Claramunt et al., 200 l ). Spatiotemporal database functionality was first attempted in the 1980s, but has not gained much practical momentum, and industrial spatiotemporal database usage is still in its infancy (Peuquet, 200 l ). Also, many of the attempts at creating spatiotemporal databases consist of merely extending the relational database model by tacking temporal data onto spatial data, which results in sub optimal representations of regions and moving objects on a GIS (Peuquet, 200 l ). Spatiotemporal data representation exploits the capabilities of modern computers to represent more than traditional geographic entities such as points, lines, and polygons. High-powered computers can also show the temporal nature of these objects (Renolen, 2000), if the data is modeled and queried correctly. Renolen (2000) describes a spatiotemporal information system (STIS) as an "information system where the spatial location and temporal history ... are of interest to organizations This definition can be 103

PAGE 112

critiqued as not explicitly stating that the information syste ms actually contain s patiotemporally referenced data, instead the literature only says that the spatiotemporal dat a i s of intere st to organizations." Nevertheless this literature provides a good analogy of map s to conceptual data models (such as data flow diagrams), and s how s how some conceptual modeling methods can be used to construct STIS. The models selected as being useful are entity-relationship, activity, object-oriented, and activity model s ( Renolen 2000) The author discusses entity relationship (ER) and time modeling languages which are language s that time stamp conventional entities. Tryfona et al. ( 1999 ) takes ER modeling a step further by adding temporal demarcations to entities in ER diagrams. STER is an acronym meaning s patio-temporal entity relationship diagrams ( Peuquet 200 I). Adding to the growing li st of s patiotemporal acronyms are, ST AM, which means spatiotemporal access method s, and spatiotemporal database (STDB) ( Tzouramanis, 2002). One goal of STDB is to use efficient indexing method s to reduce the time it takes to query s patiotemporal objects. ST AMs include the ability to generate "synthetic" data, which is statistica lly produced data relating to real-world phenomena; sets of real data; query processors; and visualization tool s (Tzouramanis, 2002) According to Frihida et al. (2002), very large data sets are accumulated when monitoring traffic flows, and this data should be monitored both geographically and temporally. But monitoring spa tiotemporal data ha s its difficulties. The SQL queries neces sary to model spat iotemporal data are long, and databa se seek and retrieval times can be prohibitively high (Frihida et al., 2002).

PAGE 113

Perhaps one of the most interesting types of GIS is multimedia GIS. Multimedia GIS can improve the end user's perception of reality while using the system, and might be able to "open a whole new dimension" to the way GIS is used (Cartwright, 200 l ) Pissinou et al. (2001) discuss geographic digital libraries and the use of these in multimedia GIS. However inadequacies in spatiotemporal data paradigms are resulting in the need for new spatiotemporal data models (Pissinou et al., 2001). In digital video, the intra-object spatial composition and intra-object temporal composition are two methods of spatiotemporal description. These authors propose a geographic topological directional model (GTDM) to incorporate these and other concepts into spatiotemporal geographic video sequences. In this model they specify data operators, such as meet, overlap, and during. This literature states that the GTDM can be used for vehicle routing, which might make it a useful tool for online grocers. Spery et al. (200 l) say that geographic objects change in thematic and temporal ways Thematic changes can be more readily handled than temporal changes, however. Objects can have "filial links to each other by being parents and/ or children. This is called the lineage model. An important aspect of the lineage model is that it allows remote databases to be updated with spatiotemporal changes that have occurred since the last update, without transferring the whole database in the process (Spery et al., 200 l ) This model is referred to as the update model by Zhang and Hunter (2000), who list three other models. They are the snapshot model, which simply superimposes a new layer onto the coverage at successive time intervals; the event based model, which links cells by ordered sets of events; and the 4-D model that incorporates time in a similar manner as other spatial dimensions. 105

PAGE 114

Tryfona et al. ( 1999) highlight the need for spatiotemporal data manipulation solutions for applications such as routing and forecasting. Their literature discusses the peculiarities of spatiotemporal data and applications by classifying objects into categories of objects that change continuously, change discretely, or have continuous motion accompanied by changes in shape. Investigation of this type of research may show a correlation with this type of object analysis and the delivery zones that I am suggesting in this dissertation. Another perhaps relevant point of Tryfona et al. s ( 1999) literature is the description of chrons, which are individual time values that are taken from "time domains." "Evolving regions are discussed by Claramunt and Jiang (2001). This concept is close to the grocery delivery regions that I am suggesting, except Claramunt and Jiang s (2001) evolving regions do not address profitability thresholds in the context of online grocery stores. Despite this, these authors' research does address the temporal validity of evolving regions as being only valid during the intersection of the "life spans of the regions. The important point here is that that "life spans of regions does not infer that the regions disintegrate after they "die." [t means that certain properties of the regions in question cease to exist. Perhaps profitability can "die" for an online grocer's delivery region if it decreases under a certain threshold. Abstract data types are used by Erwig et al. ( 1999) to show moving and evolving spatiotemporal objects. Abstract data types are many-sorted algebras, which contain object types and the operations that influence the objects. For example, two abstract data types are mpoint and rnregion. Both of these types are influenced by time as they move through space (Erwig et al., 1999). According to these authors, two steps are required to 106

PAGE 115

create a spatiotemporal data model using many-sorted algebra. First, data types and the operations between them are provided names. Names of data types are called sorts, and names of the operations are called operators. Assigning sorts and operators is referred to as giving the abstract data type a signature. Erwig et al. ( 1999) say that by using spatial algebra, which essentially consists of mathematically depicting moving points as arcs and moving polygons as polyhedra, modelers can describe spatiotemporal movement with "unlimited expressive power." Other Supporting Literature [tis hard to imagine any logistical solution that does not incorporate G[S. However there is an opulent amount of literature that pertains toe-logistics which does not necessarily address GIS. Because the subject of e-logistics is germane to delivering groceries via online orders, some e-logistic related literature is reviewed in this literature review. There are various definitions of what "e-logistics" actually means. Generally, E logistics focuses on using internet technologies to make supply chains easier to manage. [t could also be safe to define it as using e-commerce technologies, such as web-enabled logistically related flows of information and images, to facilitate and optimize the logistic efforts of an organization. A valuable contributor to the body of e-logistic literature is the automotive industry. Recently, e-logistics has taken on a very large role in allowing car manufactures to cut their lead time, or the time it takes to get a new car to market, to a fraction of the time it was a few years ago. Add to this fact the recently acquired capability of consumers to specify online what kind of automobile they want, and the picture should become slightly 107

PAGE 116

clearer about the magnitude of how intricate the logistics must be to get the parts to the manufacturer, and the car to the customer within a specified period of time. Although the processes involved in packing and delivering groceries may not be as intricate as those involved in manufacturing cars in accordance with web orders, many of the principles involved in satisfying customized automobile orders can apply to customized grocery orders. Automotive Design and Production magazine has various articles that recognize the essential role e-logistics plays in car manufacturing (Gould, 200 l ). "Today's logistics involves sophisticated supply chain management applications infused with collaborative capabilities, all running on top of an internet-based e-commerce infrastructure and fully integrated with back-office information systems to create a seamless system from product development to delivery." Gould's (200 l) article also provides a comprehensive, yet succinct, account of various aspects of logistics, such as third party logistics providers, fulfillment service providers, lead logistics providers, logistics exchanges, and logistics visibility providers. Naturally, the terminology in the field of logistics is growing as the discipline becomes increasingly more sophisticated. This is most likely because of the growing amount of information technology that is required to use logistical solutions. Of course, package delivery firms should be implementing modern logistical solutions. UPS is one delivery firm that is aggressively utilizing e-logistic technology. A fully-owned subsidiary of United Parcel Service is called UPS e-Logistics. It offers a "supply chain solution" for B2B (business to business) and B2C (business to consumer) companies. The solution includes automated inventory availability, delivery requirement, and fulfillment capacity information. The subsidiary handles web, telephone, and mail10~

PAGE 117

based customer orders and delivers orders to a majority" of United States addresses within two days (UPS Business Solutions, 2002). The description of these UPS services can be considered a compact sub-definition of e-logistics. The UPS e-Logistics services are like a "plug and play" solution for businesses. Of course the solution as explained by UPS is a significant oversimplification of the average corporate logistical business strategy; nevertheless, the idea of using an e-logistic provider is definitely intriguing, and is catching on in popularity. Documented instances of large corporations saving millions of dollars through e-logistic solutions are rife. For example, Winn Dixie used Manugistics to recently implement an e-logistics solution that ties together 1 ,000 stores with 13 distribution centers. In addition, Wal-Mart and Dell are incorporating e-logistic s trategies to keep inventories minimal and allow for better build-to-order capabilities. This early adoption of e-logistic technologies and processes should indicate that not only the concept of e-logistics is viable, but should also suggest that the omission of e-logistics from any comprehensive retailing e-commerce strategy, such as that of an online grocer, could render that strategy incomplete and possibly insufficient. Jennifer S. Kuhel (2002a), writing for Supply Chain Technology News, explains how the execution of logistical plans cannot take place in a silo (Kuhel 2002a). According to Kuhel (2002a), logistics is a participatory discipline that, in order for it to succeed as an inter-organizational tool must be implemented by all participating companies. Encapsulating this thought is a term called "collaborative logistics," which also implies using internet technology to coordinate inter-organizational logistical entities. The recognition of this term can help underscore the value of collaboration among people 109

PAGE 118

who pick and pack the groceries, the employees who monitor the GEOS, and the delivery truck drivers. An important point in Kuhel 's article (2002a) is the quote by Bob Obee, CIO of Roadway Express logistics, who said, "I'm shocked by the single-digit percentages of bills of lading that are electronically tendered to us. Thi means that most companies are not even close to being ready to collaborate logistically, at least not on any type of technological level. This fact bolsters the necessity for this type of research that aims at informing online grocery startups about the many aspects e-logistics that can make grocery delivery a profitable business. Kuhel's article (2002a) continues with the "Seven Immutable Laws of Collaborative Logistics" which are stated by John Langley, logistics professor at the University of Tennessee. The seven laws essentially state how important collaborative logistics is for the secure movement of "information, products, assets documents, and capital." One particularly noteworthy item in Langley's list is the secure movement of "information. Langley, whether knowingly or not, touches on the new concept of the "information supply chain. This is the awareness and exploitation of information as it progresses through the supply chain. Because supply chains do exi t within a company, such as between warehouse operations and outbound logistics in a supermarket, the information supply chain is important to online grocery initiatives. One notable company that has especially been under the gun to establish an e logistics system is Unilever. Nash (2002) tells of how the giant corporation, which has more than 900 brands in 150 countries, is earning less profit than its closest rival, Proctor and Gamble (Nash, 2002). Unilever's warehouses are not optimally located, and trucks 11 C

PAGE 119

are covering too much ground to deliver products Although the company has 100 se parate SAP ERP sys tems a lot of money i s being lo s t due to the logi s tical inefficiencies. This fact emphasizes the point that although ERP has been implemented to streamline information flows throughout the enterpri se, the ERP sys tems are in s ufficient when it comes to optimizing the movement of goods. The article s tate s that the location of Unilever's customers i s a large determinant of what e-logistics sys tem the company will employ. To reemphas ize the fact that even s uch a large company s uch as Unilever, which had the providence to implement sophisticated ERP sys tems ha s discovered like many other organizations, that essential cost sav ing functionality is not contained in the ERP systems. Therefore, in the case of already established online grocers that do have ERP initiative more emphasis must be placed on the optimal movement of groceries and trucks by sophisticated spatiotemporal e-logistic systems. In the same article, Fred Berkheimer, director of Unilever's North American logistics department sa id that tran s portation costs will be decrea se d between 10 and 12% after distribution centers are relocated based upon customer location. Needle ss to say, for a company of this s ize, this cost sav ings will be great. Much of the sav ings will be because trucks will be able to run at 20% greater capacity th a n they have been before the distribution centers are optimally located. The value of efficient intra-corporate logi s tics, which of course i s the main concern of online grocers that deliver, is accented by Hackler (2 002) who says, the level of efficiency in routing components and packets (e nvelop s containing important documents ) within a company becomes more important when ( inter-corporate) s upply chain partners are depending on the efficiency of any individual company ( Hacker, 2002). This is where 111

PAGE 120

internal parcel tracking systems come in. Many parcel tracking systems use bar code scanners, wireless devices, online lookup, and e-mail as their internal logistical tools. But Hacker (2002) says nothing about the incorporation of a GIS system in the parcel tracking system to visually show where any component or packet is within the enterprise The article does say, however, that efficient parcel tracking systems can reduce delivery times by 80 % A brief mention of how grocers can generate coupons for customers depending on their purchase s using a technology called "analytics" is important. Because analytics should play a n integral part of an online grocer's operations, a review of so me analytics related literature follows. As with most cases of information technology-related terminology analytics can have a different definition depending on who is doing the defining. Wayne Eckerson (2002), director of education and research at The Data Warehou s ing Institute says that analytics, or "enterprise analytics" are terms that really encompass most business intelligence technologies s uch as content management, GIS enterprise application integration (EAI), text mining, OLAP, collaboration, and more Jennifer Ma se lli (2 002) writes about how Siebel Sys tem s, a leading customer relationship management (CRM) solution vendor, is witnessing growth in demand for its analytics functionality more than any aspect of its CRM solutions. In the sa me article, a Gartner survey shows that more than half of businesses surveyed say that they are planning on increasing their CRM analytical capabilities. Reduction in the time to analyze and redistribute data is a competitive advantage (Foley, 2002). According to Foley (2002), it is important to provide reports to senior 112

PAGE 121

executives that include real-time information about thing s s uch as tracking the process of orders as they originate from the retail customer, to the movement of products from the warehouse, and on to store s helves. The article calls bu s ine sses that u se this type of model as "rea l-time busine sses." This i s of course s hould be one of the objectives of online grocers as they disp a tch delivery truck s in accordance with the generation of real time delivery zones. The need for better analytics among the supply chain ha s not gone unnoticed by major ERP/ SCM (supply chain management) vendors. For example, Manugistic s ha s unrolled their Proce ss Analyzer so lution that u ses OLAP capabilities in Microsoft's SZL Server 2000 database. Rick Whiting (2002b) tells how Manugistic s i s concentrating on incorporating new analytical capabilities into s upplier management ap plications. The se ap plication s are s uppo sed to facilitate the rating of s upplier s by usin g metrics s uch as percentage of on-time deliveries and de gree of product quality. M anug istics' bu s ine ss intelligence software includes predefined benchmarks for cycle time, work in progre s productivity workload, and resource usage David T o mpkin s (2002 ) s upports the a rgument that effective s upply chain m a nagement requires up-to-theseco nd monitoring a nd total visibility across di s parate s upply chain sys tems. In his article, Tompkins (2 002 ) explains that nece ssary real-time a n a lytics s hould be based on a decentralized or multie nterpri se sys tem s configuration, which basically refer s to implementing real-time monitoring so lution s on an "extended enterprise" model 113

PAGE 122

Because of the importance of web services for not only interoperability, but also because of the potential of the technology to facilitate the implementation of GEOS, a further review of web services-related literature follows. Johanna Ambrosio (2002) provides some perspective to the course of events that are taking place that can lead to the adoption of web services by enterprises. She says that ERP has approximately 70% worldwide penetration, with supply chain and customer relationship management (CRM) at 40 and 30 % respectively. This give a good indication of the fundamental trend of companies to optimize decision making by marshalling data, and making that data available in real time to supply chain partners. Enterprise application integration (EAi) is one way to connect software from these general types of information systems. But EAi is not always efficient enough to perform impeccably well. Ambrosio's (2002 ) article tells how web services could become the "next EAi generation," by filling in many of the EAi' s. David Truog, a Forrester Research principal analyst, states in Ambrosio s article that the cost of application integration should drop because of web services. Reinforcing the assertion that web services will indeed be a next generation of distributed applications is Peter Fischer's (2002) article, which gives a succinct and clear picture of how valuable web services can be for information visualization. According to Fischer, web services will permit collaboration in a peer-to-peer manner. This primarily means that the middle layer (middleware, EAi, etc) will become transparent to computers communicating over the internet, even if the communicating computers are using disparate and theretofore incompatible applications. According to Fischer (2002), web services will become the "glue" that makes components (applications) interoperable. 114

PAGE 123

Albeit minimal, Fischer (2002) gives some mention to the construction of a dashboard made from web services. Although the term "dashboard" can have different meanings, the article de cribes dashboards as corporate portals comprised of "portlets." A portlet could be considered a central collaboration zone for supply chain partners, providing the partners with access to data groupware, spatiotemporal images and more Electronic data interchange (E DI ) ha s been one of the ways supply chain partner s transferred data between themselves. EDI has been notoriously expensive to implement, however. The high cost of EDI has effectively limited the utilization of the technology to larger s upply chain partners who could afford the expense. Web se rvices might be used as an alternate or augmentation to EDI solutions. For example, s uppliers could be queried using web serv ice s to discove r the amount of ins tock items before se nding EDI purchase orders ( Harreld Krill and Schwartz 2002). Possibly because EDI is so expensive, approximately 80 % of product s purchased over the s upply chain are from suppliers not using EDI. According to Harreld Krill, and Schwartz (2 002), the time lag of hooking up a new supply chain partner to the EDI network could take months. web se rvice s could be used either as an interim technology before the EDI sys tem is fully implemented or totally used in place of the EDI so lution. That i s, if web se rvice s prove to be as useful in inter-corporate data exchange as EDI ha s shown itself to be to tho se corporations that implemented the technology David Litwack and Peter Fingar (2002) ponder whether web se rvices will really eventuate into being the "next big thing. In attempting to answer thi s que s tion Litwack and Fingar (2002) postulate that companies will first link internal computer systems with web services, and then concentrate on tying in business systems into the web services 115

PAGE 124

paradigm. As an example, the article tells how much easier it is for a retailer and a logistic provider to coordinate and track the location of parts if both companies are using web services solutions. Web services can allow maps to be composed of layers from disparate internet cartographic providers (Lowe, 2002). Moreover, the layers can be changed in real time without requiring the user to reconnect to the map-providing service. This functionality holds great promise for online grocer interior supply chain applications where the geographic flow of groceries, people, information and money are relevant to all departments associated with the packing, loading, and delivery of groceries The location recognition capabilities of wireless have the potential to result in diverse and meaningful applications and strategies for online grocery deliveries. Michael Cohn describes the potential benefits and some of the existing hurdles that must be overcome before location based services ( LBS) become a widely-adopted technology. As for the benefits, Cohn (2002) explains how LBS can be used to optimize supply chain practices. He tells how, by using LBS, defective merchandise can be prevented from completing the assembly line process in order to prevent the defective merchandise from being sent to customers. Three big GIS companies ESRI, Maplnfo, and Autodesk are developing LBS services. In one article Maplnfo's business unit general manager, Larry Delaney predicts that many businesses, especially supply-chain related firms, will adopt LBS solutions after the technology becomes prevalent in consumer markets. Microsoft has also become an important player in the market for LBS. Microsoft's MapPoint LBS platform is used by companies such as Hewlett Packard, Clarity, Action 116

PAGE 125

Engine, and lntrado. MapPoint@ uses .NET technology, and i s a web service that uses XML and" ... lets developers integrate maps driving directions, distance calculations, proximity searches and other location intelligence into applications, business processes and web sites." Especially interesting is the "friend finder that allows anyone connected to a wireless service to know the location of other people connected to the service. This type of service can possibly be applied to all entities in online grocery operations. Some of the hurdles that must be cleared before LBS can truly be widely adopted are the compatibility i ss ues of the different wireless standards such as global system for mobile communications (GSM) and code division multiple access (CDMA). An intere s ting wireless technology exists that is s upposed to allow centimeter preci se location detecting capabilities (Werb and Sereiko 2002). Ultra-wideband (U WB ) radio technology differs from conventional radio signal technology by not continuously broadcasting. Instead a UWB transmitter sends pulses, which represent ones and zeros, in less than one billionth of a seco nd. Basic UWB radios will be able to transmit at 20 megabits per seco nd with gigabit transmissions possible as the technology matures Werb and Sereiko (20 02) explain how many UWB radios can communicate without cross-interference a feat not possible with conventional radio technology. Additionally, the technology allows for the manufacture of low power utilization low cost, and high penetration ( the ability to go through obstacles) radios. So what does this mean for location discovery applications? Although an infrastructure capable of supporting UWB radios is still a few years off, companies have begun to develop applications that use the technology. Indoor wireless tracking l l

PAGE 126

application s for hospital s and manu fa cturer s are currentl y bein g de v i sed. The Marine s a r e w o rkin g on a per so nn e l lo c a ti o n s y s t e m u s in g UWB. UWB w as a pp roved for u s e in the United Sta t es b y the F e deral C o mmunic atio n s Commiss i o n in Febru a r y 2003. Thi s a pproval combine d w ith the cent imet e r-preci s e locati o n cap a bilities o f the t e chn o l ogy, can put th e tec hn o logy o n e s t e p closer t o b e in g u sed b y o nline g rocer y deli ve r y s oluti o n s. C ar ol y n April a nd He a th e r Harreld (2002) expl a in h o w c o llab ora ti ve "applic a ti o n s" a r e beginnin g t o exi s t th a t "allow multiple a pplications a nd d a t a sources to t a lk. Thi s d e finiti o n o f colla b ora ti o n include s only the a pplic a ti o n s i nvol ve d a t t h e intra o r g anizati o n o r int e r -organizati o n level. C ompre hen s i ve colla b o rati o n so lu t i o n s are, h o wever n o t limi te d t o t h e und e rl y in g t ec hn o logy. Ind eed, th e need for huma n m a n agers a nd exec uti ves t o colla b ora t e dri ves th e de ve l opme nt of t he techn o l ogy. T h e refor e care mu s t b e t a k e n wh e n di ffe r e nti a tin g b e t wee n compre h e n s i ve colla b orat i o n mo d e l s a n d t h e techn o l og i es th a t fac ilit a t e th a t coll a boratio n Neverth e l ess, in A pril an d H ar r e ld's (2 0 02) c o lumn ins i g ht i s p rov id e d a b o ut so m e o f th e techn o logical compo n e nt s t h a t ena bl e a ppli catio n s to fac i l it a t e colla b o rati on. S o m e s u c h com pone nt s are XML a nd web se r v i ces. Julie H a hnk e a n d D a n Sulli va n (2002) t a lk a b o ut how colla boration i s categori zed into three p r im a r y gro u ps l. P ro du c t m a n age m e nt a n d sup pl y c h a in m a n age m e nt ( S C M ) 2 Int e rn e t t ra din g exch a n ges a nd m a r ket pl aces 3. Informa t io n exch ange ap pli catio n s. Onlin e groce r s s h o ul d con ce rn th e msel ves w ith a ll th ree face t s. These a u t h o r s d efi n e s uppl y c h a in m a n age m e nt as und e r sta ndin g d e m a nd tr e n ds a nd con s u mpt i o n p a t te rn s, as well as s uppl y lead ti mes t o o ptimi ze p ro du ctio n sc h ed ul i n g, m a nu fact ur i n g, a n d 118

PAGE 127

fulfillment." This definition also relates to managing partial fills, backorders, substituted products, warehousing, transportation and delivery scheduling. The article states that, of the three groups above, the supply-chain collaboration group is the oldest, and the information exchange group is the newest. The specifics of this third group, which is still emerging and maturing, is hard to define, but the group is related to business intelligence, portals, content management, CRM, and knowledge management. It seems difficult, however, to clearly separate the groups, especially the first and third group. This is because comprehensive SCM solutions should include virtually all of the functions listed in the information exchange application group. Collaboration application vendors are obviously becoming receptive to the needs of businesses, as evinced by the functionality incorporated in new collaborative portals. Mears explains that needed capabilities such as document sharing, colleague and expertise locating, and threaded discussions are incorporated in portals (Mears, 2002). Mears (2002) says, portals are becoming the centerpiece of how companies do business." This statement concisely implies a trend of allowing employees to interact increasingly through portals and decreasingly thorough face-to-face contact. This trend could give rise to a lot of technological and strategic ramifications. One such ramification can be the difficulty of adoption of collaborative solutions. To address this issue, Corechange is supposed to allow users to both simply use the application while allowing business users to "set up community portal pages, decide who can be a member of that community and what content, applications, and collaboration features will be available." Yet, despite the product s varied functions, evidently no geographic capabilities exist such as the ability to locate stakeholders or other physical entities. 115

PAGE 128

The amount of literature pertaining to portals i s growing quickly. Indeed this is probably a corollary to the amount of expenditure going toward this technology. Giga Information Service s predict the current $8 0 ,000,000 total a nnual portal expenditures will grow to $2,000,000,000 annually in 2005 ( Frye, 2002). Frye (2 002 ) explains that because a portal is a personalized" browser interface users of portals will be able to communicate with, and be a distinct player within multiple communities. A great potential lies here in the simplicity of employees not having to personalize or adjust their computer's desktop or network interface before participating in the delivery zone delivery scenario underway at any particular time. Frye's article goes on to say that web services will add thrust to the usage of portals becau se of the ability of web services to overcome interoperability glitches that thwart the unimpeded implementation of portal s between disparate systems within the supply chain. While saying that portals should include presentation, per so nalization, collaboration proce ss(es), publishing and distribution, Frye (2002) also said that a current trend could be the movement toward "federated portals. This i s essentially the ability to access a s ingle enterprise portal by multiple portal "communities ." Although the article does not clarify if the communities are inter-or intraorganizational, it is assumed that the federated portals could exist both within and between corporations ( in the s upply chain). Steve Konicki (2002) discu sses how retailer supply chains can save $40,000,000,000 annually by using better collaboration. In his article the importance of UCCnet is described UCCnet is a standards body that is building an all-inclusive central registry, called the UCCnet Global Registry that will contain information about all retail products. 12(

PAGE 129

This registry could facilitate the dissemination of valuable retail product information to smaller retailers, s uch as small online grocers, and make them more competitive. The UCCnet Global Registry will contain 62 piece s of information about every product. Package s ize, product number, and manufacturer are so me of the information that will be available about each product. Konicki s (2 002) article declares the registry to be a historic breakthrough" that could decrease the time it take s to update changes about products throughout the s upply chain from weeks to po ss ibly minutes. Konicki (2002) de scri bes how collaborative planning, forecasting, and replenishment part s of an inte gra l s upply chain optimization model -can be greatly facilitated by a central regi s try Collaborative planning forecasting, and repleni s hment can allow retailer s and manufacturers to tran sfe r data a bout product s to coordinate forecasts, replenishment sc hedule s, a nd overall collaboration. A central product regi s try s uch as the UCCnet Global Re g i s try could s ignificantly increa se the viability of the collaborative planning forecasting, a nd replenishment model. Obviou s ly there i s no paucity of new technological methodologie s that, if u se d by astute online grocers, could make their logi s tical and overall fulfillment operations competitive and profit a ble Thi s Literature Review necessarily encompasses a wide variety of up-and-coming and established technologies and delivery practices. A problem doe s exist, however for online grocers to decide which technologies and practices to choose from. A main purpo se of my incorporation of the myriad facets of related technologies and logi s tical trend s is to at least provide online grocers (a nd potential online grocers) the background knowledge nece ssary to intelligently begin to decide what to implement in their delivery initiative. 121

PAGE 130

SIMULATION ELUCIDATION This section demonstrates, by way of images taken from my logistical computer application, how a grocer can simulate the delivery of trucks to a hypothetical delivery area. The application uses Microsoft ExcelTM as the underlying program. The user of the program, however will have little indication that he or she is actually using Excel. This is because I customized the application with Visual Basic for Applications (VBA) code. The application, when u sed by a grocer, will be run on a high-powered back-end computer. What follows i s just one of virtually thousands of pos sib le profit-maximizing (or profit-ascertaining ) sce narios that a grocer who has plan s to deliver can s tep through to determine such things as the amount of truck s required time to deliver to a region and more. The following sc reen shots document that dozen s of variables can be input into the application such as de si red profit interand intradelivery times (w ithin and between regions), regional population and more. Also the simulation user can specify how many time s the s imulation can be run. The application also creates various reports "on the fly ," which can be viewed by the user after the simulation(s) run through. The s imul a tion can be run from one time to thousands of time s. The Simulation Options form i s opened with the pre ss of a button The logistical simulation i s a sophisticated computer application that I programmed, which include s 645 pages of Visual Bas ic for Applications code ( measured in 12-point single-spaced type), and is comprised of 337 macros, 88 spreadsheets, 66,000 data elements in the form of addresses, scores of functions, and procedures, and random number generators which are all 122

PAGE 131

compiled into a fully functional logistical decision-making application that is driven by algorithms that I devised. The application can be utilized by both beginning and established grocers. All that is required of these grocers, besides the necessary computer platform, is to perform a survey of at least its current customers to collect some idea of the amount of deliveries that might occur in each delivery region. A grocer who is considering a delivery initiative should first take a survey of at least his existing customers to see if there is intere s t in having groceries delivered to the home or region that the customers live in Even though the purpose and scope of my dissertation i s not at dependant on survey results, some of the questions on a survey taken by a grocer who is considering using my computerized logistical decision facilitating tool might be s imilar to the following. How long have you been a customer at our store? What i s your average weekly grocery bill at this store? What is your average weekly grocery including purchases at this sto re and other stores? What is your zip-plus-four code? If you do not know your zip-plus-four code, what is your address? Would you be interested in having our store delivery groceries to your house? If you would like groceries delivered, how much per week (in dollar s) would you guess that you might order from us to deliver to you? May we call you about this ? If we can call you, what is your phone number? Would you be using the internet to place your order? 123

PAGE 132

The grocer could also ask other questions that he or she believes is relevant or appropriate. For example, asking the customer to supply his or her name may, or may not, be conducive to getting more accurate responses. The survey forms may be generated with the sales receipts, and possibly something, such as two or three free apples, may be offered as an incentive to fill out the survey. Once the surveys have been accumulated, and the addresses (or zip-plus-four codes) are gleaned from the surveys, the addresses are simply plugged into the application s under! ying database. From the information on the surveys the grocer can determine, within bounds, how many orders will be generated within a certain time period. For example, if, according to the surveys, it is determined that there will be 1500 home deliveries per week, the grocer simply inputs 1500" into the Amount of Orders dropdown list ( Figure 1 ) and inputs "7" into the Duration of Delivery in Days dropdown list (Figure 2). A "Pause Every" function is also provided. Because the simulation can run quite fast, it might be hard for the user to see the exact status of the trucks or the amount of deliveries sent to each region. The Pause Every function allows the user to specify when to temporarily halt the simulation so he or she could better see the status of the simulation The user can make the simulation pause when every few orders are generated, or every thousand, or more orders are generated. To do this the Pause Every function is used on the user interface. The Main Map shows a prototypical delivery area divided into 66 hexagonal regions (Figure 1). Naturally, each grocer s customer market area will have a different geographical shape. But, for purposes of this simulation, a somewhat key-shaped area is 124

PAGE 133

depicted on the screen (Figure 2). This is not to suggest that the pre-determined shape of the customer market area is a trivial matter, however, my assertion is that this shape is a reasonable manifestation of what could be a collection of geographic regions that has, say, a river to the west, mountains to the north, and possibly a sea coast surrounding the east and south perimeters. Although this is an arbitrary shape, the underlying principle of visualizing and recording the deliveries can be applicable to any area if the hexagons are rearranged accordingly But the rearrangement of the hexagons is not part of the functionality of this application. As mentioned a grocer should want to discover the conditions where an online e grocer (or a grocer who is taking orders by a telephone) will reach a predetermined amount of profitability. By predetermined, I mean that before each simulation sequence is executed, the user can input exactly what profit is required per delivery run to a region, from 0.1 % to 3 % The Profit Options button i s pressed to show the Profit Options form. Then the Profit Margin (in % ) drop down list is pressed, showing the range of profit as described below (Figure 3) One-tenth of 1 % profit is represented by 0.0 I, and is the lowest setting. 0.1 means I % profit. While 0.3 means 3 % profit, the highest allowable in this application. The Minimum Dollar Delivery drop down box allows the merchant to specify from $5 to $100 (Figure 4 ). This of course means that the grocer has created a business plan, and has notified customers that any orders of less than this amount will not be able to be delivered Any order amounts generated by the random number generator that is less than 125

PAGE 134

this amount will be rejected, and the random number generator will ge nerate another number. The Mandatory Profit calculation for a region i s ( Profit Margin x amount of groceries ordered at an individual residence)+ ( the delivery charge for an order se nt to that region). Therefore an order of $100 with a profit margin of l % sent to a region with a $5 delivery charge per o rder will generate ( 100 x I ) + (5) = $6 in profit. If the Mandatory Profit for that region has been s pecified as $18, it will require three of these identical orders to reach that region's Mandatory Profit and h ave a truck dispatched to that reg ion. The application allows for an overall Mandatory Profit amount to be input, or different Mandatory Profit amounts to be input for "co ncentric regions." The hexa go n s are geometric s h a p es which can naturally be concatenated into rings o r "b uffer s" around a central point. My application a llows the merchant to specify different Mand a tory Profit s for each concentric region. Perhaps, a lesser Mand atory Profit would be spec ified for clo se r concentric regions, allowing truck s to be dispa tched to those regions with greater frequency. Figure 5 s how s a $45 Mandatory Profit se ttin g for all of the region s Figure 6 s how s gra du a lly increasing Mand a tory profit set tings for all concentric regions radiating out from the Home Region The incre ases are gra du a ted in $5 increments. As mentioned above, the delivery charge can be from zero to any a mount the merchant may require In Figure 6, the text box next to the Delivery Ch a rge (in dollar s)" allows for the input of a global charge that i s added to every o rder delivered to every region. A 5 put into that box will assign a $5 delivery charge to every region When the 126

PAGE 135

"Submit Delivery Charge" button is depressed, the value in the text box is relayed to the Delivery Charge Per Region map. As seen in Figure 6, each region contains 5 since thi s amount was s ubmitted from the Main Map sc reen. The Profit Option s form s tays visible in case the user wants to reenter another global value. The Set Regions to Null button on the Delivery Charge Per Region map in Figure 6 clears out all of the delivery values in every region. To set the delivery charge of any individual region, the user simply has to click on the hexagonal region and a form will pop up Figure 8 s how s the results of a u ser clicking on Region E4 to enter a Delivery Charge of $10. The Delivery Charge per Region form pops up with the phra se Region E4 Delivery Charge" to clearly show what region's delivery charge i s being changed. Pres s ing the "Submit" button will put a" 1 O" in Region E4. Notice areas se lected turn red when clicked on. After se tting the Delivery Charges, pressing the M ain Map button will bring the u se r back to the main control panel. The Simulation Option s button ( Figure 8) on the Main Map opens the Simulation Options form and also displays the words "Simulation O near the top left of the screen. Two tabs are on the Simulation Option s form the Number of Runs tab and the Population Option s tab. Using the Number of Run s tab the u ser can choose to run the simulation from two to 1 ,000 time s The Number of Run s tab ha s two mutually exclusive radio buttons. Choosing the Same Time Window button will make each successive simulation run through the sa me times. For example if two day s are chosen and the beginning time is I am, May 1, all the si mulations for that seq uence will run from 1 am, May 1 to lam, May 3. 127

PAGE 136

If, however, the Successive (concatenated) Times radio button is chosen with a two day, two s imulation sequence, then the first simulation will run from l am, May I to I am May 3; and the second s imulation will run from l am May 3 to I am May 5. The Run the Project button of course initiates the simulation. If the Run the Project button is pressed without selecting any options, as i s the case in Figure 9, various error catching functions will initiate. As shown in Figure l 0, the first error message that occurs tells the user that no simulations can be run without choosing the Number of Orders After pressing "OK" on the message box, the u se r chooses "5 00 from the "Num ber of Order s" drop down list (F igure 11 ). The Run the Project button i s again pressed But as Figure 12 s hows, the simulation will not run without s pecifying how many days the s imulation s hould run for. The OK" button is pre ssed on the error message box and a two day Duration of Deliverie s is selected ( Figure 13). Figure 13 also s hows the Gross Order s per Simulation legend. This legend s how s how the hexagon s will change color in res pect to the amount of total (gross) orders that were placed in a s imulation. Thi s mean s that the color of the hexagon has nothing to do with the number that is in the hexagon after the first delivery truck returns from making deliveries to that hexagonal region. Here are two important definitions. 1. RUN. A run is the dispatching of a truck to a region, the delivery of groceries to all addresses that placed orders in that region, and the return of that truck to the grocery store. The grocery s tore i s located in Region H. 2. SIMULATION. A s imulation can be comprised of many runs to each and every hexagonal region During each simulation, after a truck returns from a delivery to a specific region, the number inside that region's hexagon will become "O." This means 128

PAGE 137

that there are zero orders called in so far for the next run. As soon as one order comes in for that region the number inside that region will become "l ," and will increase as more orders come in, and will continue to increase until that region reaches the predetermined Mandatory Profit, at which time a truck will be dispatched; and when that truck returns the number in the hexagon will again become "0. This sequence will occur for the duration of the simulation for every hexagonal region in the entire area. A hexagon s color, on the other hand, will continue to change in proportion to and in concurrence with the colors shown on the legend The colors in the hexagons tell the user of the application the total number of orders called in by households in that region. The color continues to change for the duration of the simulation. The colors of all hexagons become purple when the Reset Regions to Zero button is pressed. In other words as a simulation is running, the Main Map shows, in color-coded regions, the total amount of orders that has come in from that region. When regions meet their Mandatory Profit trucks will be dispatched to those regions. The number of the truck that is dispatched appears clearly to the left of the simulation screen. Truck dispatch notifications tell when a truck is dispatched, and when it returns. Above the truck dispatch notifications is the time the simulation began, and the time it ends. The color-coded legend is provided to conveniently see how many orders have been generated from a region at any time. The numerical value that appears in a region is the amount of undelivered orders that have been generated from that region. The number turns to zero when a delivery is made to that region Figure 14 shows the Main Map after a two-day, 500 order simulation was run. 129

PAGE 138

At this point it is important to emphasize that it is impossible to capture the dynamic nature of this simulation application by looking at the static scree n s hots in this dissertation write up. Because the s imulation i s also a fully functional working prototype of a logistical interface the dispatching and returning of trucks occur in real time which is completely unapparent to anyone merely looking at s tatic scree n s hots. Figure 14 s how s the number 500 in the top left (to the left of "ORDERS PLACED THIS SIMULATION"). This i s the total (gross) number of all the orders called in or placed through the internet to all the 66 hexagonal re g ions. Thi s number includes orders that were delivered to, along with orders that have not as yet been delivered to residences in the regions The numbers that remain in each of the hexagon s are the number of orders within each hexa go n that have not been delivered yet. Thi s information is a lso s tated in white letters to the right of the junction of regions E8 and F7. Notice the Fir st Order Was Placed at:" and La s t Order W as Placed At:" times. These time s will always reflect the duration s pecified in the Duration of Deliveries of Days drop down list. The Truck ( number l, 2, 3 ... 15) dispatched to Region ( H Al, A2 ... 04)" notice s on the upper left of the sc reen appear every time a truck is se nt to that region The red Insufficient Amount of Trucks mean s that the 15 truck s available during thi s s imulation were not enough to deliver one order per run Figure 15 s how s the results as displayed on the Report Dashboard. Notice the NO SIMULATION OPTIONS WERE CHOSEN BEFORE RUNNING THIS APPLICATION. Thi s i s because no options were chosen from the Simulation Option s form. 130

PAGE 139

The vertical bars show the total of orders called in from each region, with each bar representing a different region (hexagon in Figure 14). Notice the time the simulation ran is clearly displayed near the top of the Report Dashboard. The Main Map button returns the user to the Main Map. The other buttons will be explained below. So far, we have created a non-optimal situation, meaning that although we looked at the Profit Options form, we have not mandated any Profit Options yet for any simulation sequence. The colors of the hexagons in Figure 14 are associated with the colors legend in the same figure. The values of the hexagonal regions match the values of the associated regional bars in the Figure 15. This is to make it easy for the users of the application to see how many orders were generated from each region. For the next simulation, the profit margin, minimum dollar delivery, average dollar delivery, mandatory profit, and delivery charges, will be set in order to stipulate and fine tune the exact criteria that we want, or in other words, dictating the "pathways by which profitability can be achieved." This is an important point, and one of the underpinnings of the purpose of this logistical application. We are creating a progressively more sophisticated "combinatorial situation as we progress. Figure 16 shows results of the simulation. The preset profitability settings are clearly visible in the Profit Options form. Notice in Figure 16 that fewer trucks were needed when Profit Options were specified than when no Profit Options were set. This is because in the prior simulation, a truck was dispatched as soon as a single order came in. 1n this simulation a truck is not dispatched until a region reaches its Mandatory Profit. 131

PAGE 140

Figure 16 s how s that at the time the 500th order was placed, truck s number one and three were at the home s tore. The other trucks were out delivering to their re s pective region s as clearly described near the top left of the Main Map Figure 17 s hows the Pau se function being u se d. Figure 18 s how s the Profit Option s used to run the simulation shown in Figure 16. The rea so n for the Pause Every" option i s to a llow the si mulation observer to h ave a kind of a breathing space" to let him or her see the s tate of the s imulation in whatever increment that i s desired. Thi s i s a new thing that I have devised to allow managers to make better judgments about how they might want to change so me or all of the variables on s ub se quent s imulations. Thi s is s ignificant in that it is jus t one of the various innovation s I am providing to better ascertain the geographic sco pe of what is h a ppening at any desired step of a ny simulation. Managers will be able to see at a ny predetermined interval exactly where the orders are coming from. This will g ive an intuitive feel about what regions, or clusters of regions are generating orders, meeting the mandatory profitability ; and also of what truck s are out, and of what trucks are waiting. The Figure 17 s imulation's se ttings are lO days, 1,000 orders, and the Profit Option s are s hown in Figure 18. It was instructed to Pause at 500 orders. Take a close look at the me ssage box in Figure 17. It tells how many orders have been generated, the amount of s imulated hours that have pas sed since the beginning of the simulation ( in this case 120 which is one half of the lO day duration because the application was instructed to pause at 500 deliveries, or one half the total 1,000 delivery specification for the entire 10-day period. It also shows the current simulated time. 132

PAGE 141

The application allows the user to specify pauses in increments from every 10 deliveries to every 5,000 deliveries. Also notice that the amount of trucks used is far less than the above simulations because the Mandatory Profit is set at $70 (Figure 18). A zero delivery charge was also used (Figure 18). So far the Suppress Dispatch Alerts radio button has been activated. This radio button, along with the Allow Dispatch Alerts radio button just below it allows the user to toggle between having a message box appear every time a region reaches its Mandatory Profit, and not having the message box appear. The Allow Dispatch Alerts is a useful tool that tells the person monitoring the logistical dashboard (the actual application on the screen) that it is time to immediately send a truck to that region. This is just some of the functionality that I incorporated into this application in order to create a prototype of an actual logistical interface for managers to use. My application is not just a simulation package. Look at Figure 19 to see the Profit Options, and to see that the Allow Dispatch Alerts radio button is activated for the next simulation. Figure 20 shows the first Dispatch Alert that was generated for this simulation sequence. The "OK" button has to be pressed for the application to continue running. Figure 21 shows the same simulation upon its completion. Just by looking at the Main Map managers can determine how many trucks were used, the time span of incoming orders, the amount of orders generated by each region, the amount of undelivered orders, and, of course, the Profit Options. But this information is only the beginning. The preponderance of useful data is in the underlying reports that are generated on the fly for each simulation sequence. 133

PAGE 142

Figure 22 shows the Report Da s hboard for the simulation s hown in Figure 21. Pres s ing the Population Report will s how the following data (Figure 23). The Popul at ion Report s how s that each region ha s 4 ,000 hou se holds as shown by the bar gra ph and the data in the chart. Also, the population chart can be updated in the software to reflect actual population numbers. Notice that you can return to either the M a in Map or the Report Da s hboard from a ny report. The rel a tive mandatory profit earned for all regions, for all s imulations is s hown in the Profit ab ility Chart (F igure 24). Not many of the s lice s of this particular chart are remarkable, but thi s will change as we further fine tune so me varia ble s s uch as popul a tion and individual Regions' Mandatory Profit s pecific atio ns. The total Manda tor y Profit per region is s hown in Figure 25. Thi s information is reached by hitting the Profit Chart button. The Profitability Quotient ( Figure 25) i s a numb e r between zero a nd one that uses the s um total of the profit for a ll regions for th a t particular s imulation and divide s that number by a divisor that results in the number between zero a nd o ne. It i s used to compare the relative profit between s imulation s. The maximum, minimum, average, s tandard deviation, mode, and median values pertain t o the s imulation which column they fall under. Figure 26 s how s a s ub se t of the data that i s attained by clicking on the Re s idence Report. Notice that the amount ordered for every hou se hold the time of delivery, the run number, a nd what regi o n the hou se hold exists in i s clearly shown. Notice that no orders of le ss than $10 were called in as per our s pecification The Setting s Report for thi s si mulation i s shown in Figure 27. The top of the Settings Report reflects the information the user input on the Main Map interface. 134

PAGE 143

The chart in the Settings Report shows the names of the regions on both the left and right hand sides, population that the user has input, the delivery charge (in this case, as specified is $5) the mandatory profit (in this case, as specified is $60). The One-Way Drive Times and Intra-Regional Drive Times are also specified. A region's Intra-Regional drive time is the average time it takes to stop at a residence, drop off the groceries and drive to the next residence. The application s default ( pre programmed) time for all the regions intra-regional drive times is five minutes, as shown in Figure 27. Specifically, if there are 10 deliveries to be made to a region on a run, the total intra-regional drive time for that run is (IO deliverie s x 5 minutes per deli very) 50 minutes. If the round trip drive time to a region (the time it takes to drive from the store to the region, plus the time it takes to drive back to the s tore ( not including making the deliveries) ) is, say 70 minutes, then the total time from the dispatch of the truck from the store, until the time the truck returns to the store is (70-minute round-trip drive time + 50 minutes intra-regional drive time) 120 minutes. We will look later at the ability of this logistical application that I developed to allow the user to "fine tune" the Intra-regional and One-waydrive times. But first, the Delivery Report i s discussed. I ran a simulation using the same Profit Options and time window (two days) with 500 deliveries to generate the Delivery Report shown in Figure 28 Figure 28 shows a subset of the information about the 500 deliveries made during the above-mentioned simulation. Notice that, again, this is an on the fly generated_report, as all of the reports are. This is an important and valuable feature of the simulation I developed. Of course, by simply using the Print button on any of the reports, that report 135

PAGE 144

can be printed through the print preview window. The Delivery Report shows the information as de sc ribed in the column heading s, which can be u se d by a grocer to ascertain further information about cyclical order time s per residence. Figure 29 shows the Simulation Option s form open with three si mulations chosen u s ing "How Many Time Would You Like To Run The Simulation" drop-down menu Up to 1000 s imulation s can be chosen using this menu. Notice the Simulation# O" near the top left of the screen. Thi s will change in accordance with the actual number of the s imulation in the seq uence. The "in Succe ssive (Co ncatenated) Times" radio button is checked. This means that each s uccessive simulation will run in the following period (in thi s case, three days). If the Same Time Window" radio button was se lected the application would continuously s imulate the deliveries s tarting and ending at the sa me time. Successive (Co ncatenated) time s mean s that the next si mulation in the sequence begins at the immediate seco nd after the prior s imulation ended. Figure 30 s how s the s imulation at the end of the third sequence. Notice in Figure 30, the "Simulation#" is "3", which in itself s how s that the sim ulation se quence ha s either entered or completed its third simulation of this three-simulation concatenated sequence. In thi s case it indicates that the si mulation ha s comp l eted its third simulation in this three s imulation seq uence. Figure 31 and Figure 32 are two Figure s taken from the same report that is automatically generated when the Simulation Chart s button is pressed from the Report Dashboard. The first bar chart i s the same as is see n on the Report Dashboard. It is replicated on the Number of Orders per Region report for convenience. The following 136

PAGE 145

three charts show the amount of orders generated from each region for each simulation in this three-simulation se quence Each bar in each graph of Figure 31 and 32 repre se nt s a region. Region H i s the leftmost bar and gradually progres ses to Region G4, the rightmost bar. The top graph called Composite Total of All Orders for All Simulations" graphically depicts how many orders for a region was generated by adding the number of orders for the same region from the graphs shown under "Orders Per Individual Simulation. By looking at Region H we can see that in the first simulation 5 orders were called in from that region. In sim ulation 2 Region H generated 8 orders, and in the third s imulation it generated 4 orders. Therefore the Composite Total ( the topmost graph) for Region H s hows (5 + 8 + 4 = 17) that the tot a l amount of orders for all si mulation s i s 17 for Region H. Thi s functionality will work even if 5,000 or more simulations are run meaning that 5,000 Orders Per Individu a l Simulation graphs will be automatically generated and totaled in the Composite Total of All Order s for All Simulations graph on the top of thi s part of the logistical dashboard. First, notice in Figure s 3 l and 32 that the times are concatenated for s imulation s one and two. The chart under the "Composite Total of All Orders For All Region s For All Simulations i s the total of all the charts that follow it. Each bar in each chart from s imulation one through s imulation three is the amount of orders generated from that region for that s imulation. Notice that the times above each bar chart represent the time window that that simulation was run. These times, like everything else on the se bar charts are automatically generated. 137

PAGE 146

Notice in Figure 33 that the Settings Report which i s automatically generated, shows the settings as input at the beginning of this s imulation sequence. Figure 34 s hows that the total mandatory profit for the three s imulations was automatically recorded, proving that the logistic application that I developed can generate data for more than one simulation on the fly. Figure 35 s hows a s ub se t of data from the Residence Report. The multiple page s of the Residence Report could be s hown here but the purpo se of thi s Figure is to show that data from all three s imulation s ha s been recorded. For each s ucce ss ive s imulation in thi s simulation seq uence Figure 35 s hows the amount of groceries ordered, time of order, address of the ordering hou se holds run number and regions from all three s imulations. Figure 36 s how s the Mandatory Profit being set by regions. Setting the Mandatory Profit by Concentric Region i s another way the grocer can fine-tune his business model. For example, if the grocer would rather deliver to farther region s only when a greater profit i s realized from those di s tant regions he could se t the concentric regions accordingly; preci se ly as was done in the simulation related to Figure 36. Fig ure 37 s how s the Settings Report for this s imulation Notice that the Mandatory Profit h as been recorded as being set by $5 increment s, as we s pecified at the start of this simulation ( Figure 36). Figure 38 shows the sa me Profit Option s as the prior s imulation, but for this s imulation the Submit Delivery Charge" button will be hit bringing the Delivery Charge per Region interface into focus. This interface is s hown in Figure 39 The Delivery Charge per Region interface allows the grocer to further fine tune the business model. When a user clicks on any region, a form will appear allowing that 138

PAGE 147

user to change a delivery charge for what ever region (or regions) he wants to. Figure 40 shows the result of clicking on Region A3. Notice that the region clicked on turns red, reminding the user what region ( s) he has changed Figure 41 shows that the new Delivery Charge of $15 has been input into region A3 after the Submit button was depressed. When the simulation is run, a delivery charge of $15 will be calculated into Region A3' s profit and dispatch functions. The Profitability Chart for this simulation, as displayed in Figure 42 provides visual evidence that Region A3's profit is proportionately greater than the other regions because of the five-times greater delivery charge that is calculated into every delivery to Region A3. The Profitability Chart can be thought of as a profitability pie ," which at a glance gives managers an intuitive, yet accurate, visualization of what delivery regions are generating the greatest least or mid-range profit. It is a part of the overall decision facilitating logistical dashboard that is integral to this application. The numerical representation of the proportionate regional profit is shown in the Profit Chart for this simulation, as displayed in Figure 43 As with all of the data and images displayed in my dissertation, the data displayed in Figures 42 and 43 are generated on the fly with every simulation, and is very simply viewed by clicking the buttons of the intuitive application that I developed. Figure 44 shows the Main Map and associated settings that were used for a simulation where the Intra-Regional Drive Time functionality was used. A region s Intra Regional drive time is the average time it takes to stop at a residence, drop off the groceries, and then drive to the next residence. 139

PAGE 148

Recall that all the above Settings Reports (such as the one in Figure 25) showed 5 .00" in the Intra-Regional Drive Time column. The application's default ( pre programmed ) time for all the regions' intra-regional drive time s i s five minutes This means that if there are l O deliveries to be made to a region on a run, the total intra regional drive time for that run is ( 10 deliveries x 5 minutes per delivery)= 50 minutes. Using this same scenario, if the round trip drive time to a region ( the time it takes to drive from the store to the region, plus the time it takes to drive back to the store (not including making the deliveries)) is say, 70 minutes then the total time from the dispatch of the truck from the s tore, until the time the truck return s to the store is (70 minutes round trip drive time+ 50 minutes intra-regional drive time ) = 120 minutes. Pressing the Intra-Regional Drive Times" button bring s up the sc reen in Figure 45. Clicking in any region will allow the user to change the Intra-Regional Drive time for that region. Figure 46 shows the result of clicking in Region D l and inputting an Intra Regional Drive Time of 15 minutes. Figure 47 s hows a s ubset of the data in the Settings Report that was automatically generated when this s imulation was run. Notice that Region D l's Intra-Regional Drive Time is 15 minutes as was s pecified before the simulation was run. The Delivery Report that was generated for this s imulation ( Figure 48) shows the Round Trip for two delivery runs to D regions. Even though both deliveries have comparable Gross Regional Profits the delivery run to Region DI took substantially longer than the run to Region D6. Of course, this is becau se it took on an average of three times longer to drive between residences in Region D 1 than it did in Region D6. 140

PAGE 149

The same start-up settings that were used in the prior simulation sequence were used for this next simulation sequence. The One-Way Drive Times for two regions were set manually for this simulation. The application sets one-way default drive times to each region in minutes, as follows. Region H = 10 A Regions = 15 B Regions = 20 C Regions = 25 D Regions = 30 E Regions = 35 F Regions = 40 G Regions = 45 A truck's round trip drive time is the one-way drive time multiplied by two. The application calculates a trucks' total time to return to the store every time a truck is dispatched to a region. The total time to return to the store is the round trip drive time plus the amount of deliveries made within the region multiplied times the intra-zone delivery time. Figure 49 shows the interface for the One-Way Drive Times with two regions' times already changed to 90 minutes, with preparations for the same amount to be input into Region C4. Figure 50 shows the settings report with the One-Way Drive Times generated on the fly, and in accordance with the settings that were specified at the beginning of this simulation. 141

PAGE 150

Figure 51 s how s the Round Trip Times for the manually changed regions in proportion to C 1 and region Fl 0. Note that although Region Cl i s in the same "concentric zone" as other C region s, and th a t Region Fl O i s geographically farther away from all "C" regions, both of the se region s have s ubstantially le ss Round Trip Drive Times. Figure 52 s how s the Population Option s tab on the Simulation Option s form with "250" entered into the dialogue box. Figure 53 s hows the Population Per Region interface that appeared after the "Submit" button was pre ssed. Popul at ion per individual region can be entered through this interface by clicking on any hexagon. Thi s si mulation gives an example of regions Bl and C2 having no hou se hold s, and regions A4, A3, and B4 having 9,000 households. Figure 54 s how s all but Region B4's population changed. Figure 54 s how s the s tatus of the population input form after region B4 was clicked on. Figure 55 s hows the Main Map upon completion of this simulation. From this interface i s easily seen by comparing the region s to the legend that Region Bl a nd Region C 1 generated no orders. In contrast, regions A4, A3, and B4 generated far greater orders than any other region. The contrast in the amount of orders generated by all of the regions in this simulation can be see n even more clearly in the automatically-generated bar chart on the Report Dashboard ( Figure 56). The Population Report for this si mulation s how s, both graphically a nd through a table the proportionate population per region ( Figure 57). Each bar s how s the a mount of hou se holds in the region They axis values are multiplied by 10. Therefore, if the y axis shows "500" for a region, the amount of households in that region is 500 x 10 = 5,000 household s 142

PAGE 151

The above discussion is intended to describe the various functions that are contained in the simulation software. The above also demonstrates the uniqueness of the application in that although companies such as SAP and Manugistics have developed sophisticated logistic decision making software, neither of these companies have developed anything specifically for the grocery delivery industry. In addition, the software application actually runs very well. This is not a trivial statement when considering that 650 pages of computer code, along with myriad algorithms underlie the application. The application can be run thousands of times without any faults, operational or logical, occurring. Additionally, when generating the screen shots in this section, the application ran without a single hitch, or system failure occurring. The above sequence of simulations is intended to provide the reader with the overall scope of the logistical decision-making application that I developed. First the application s interface (main map drop down menus, control buttons, and radio buttons ) are intended to allow decision makers to use the application with relative ease. This logistical application, despite containing sophisticated functionality, should be able to be adopted by online grocers with minimal amounts of information technology background. Indeed, many software packages contain tutorial manuals that are hundreds of pages long. If this is contrasted with the mere dozens of pages of screen shots (above) and associated written explanations it should be evident that this application is relatively easy to use, thereby solving one of the integral problems of software application adoption by businesses a high learning curve to use the applications Another very significant reason the Simulation Elucidation section is incorporated in this dissertation is to clearly demonstrate that the various reports, which essentially 143

PAGE 152

cons titute the gro cers logi s tical d as hb o ard are generat e d "on the fly," w h ic h means th at no m a nu a l data retrieval process e s a r e n eeded, s u c h as writ in g sophi s t ica ted SQL (s tructured query l a n g u ag e ) queries. This i s a n impo rt a nt p o int beca u s e it all e v i a te s the need t o u s e d a tab as e pers onnel to retrie ve informati o n s u c h as deli ve r y t i mes and profitability per regi on. This d a t a i s clearly s h own in th e r e p o rt s w hi c h can b e prin t ed after simula tion s are run Al s o, the Simulat ion Elucid a ti o n s hould s h o w t h a t t he applic a ti o n i s ae s tetic a ll y pleasin g, whi c h i s a nice a m e nit y t o m a n agers a nd o peratio n a l p e rsonn e l th a t h ave t o u se it on a d a il y b as i s An a ctual g rocer y s t o r e c a n input th e addresses g leaned fro m th e surveys it h a d it s c u s t o m e rs' compl e t e int o th e a pplicati on The a pplicati o n as i s ex i s t s n o w h as I 0 ,000 h y p o th e tic a l addresses p e r regi o n in th e und e rl y in g spread s heet s t h a t are n o t v i s ibl e t o the u se r w h e n th e a pplic a ti o n i s run ln o th e r word s, th is ap pl icatio n h as 66 ,000 addresses th a t i t hit s upo n w h e n it run s. A g rocer w h o i s con s id e rin g a n o nlin e d elivery initi a ti ve would fir s t input i ts cu s t omers' addresses int o th e s pre a d s h eets. Beca u s e of th e a bilit y of t h e a p p licatio n to a llow the u se r t o s p ec i fy th e p o pulati o n p e r indi v idu a l region whe n th e s i m ulati o n run s, th e g rocer w ill get a realis ti c estima t e of ho w m a n y o rd e r s a r e generat e d per regi on. Whe n thi s i s combine d w ith th e M a nd a t o ry Profit fun c ti o n th e r es ult s of each s imul a ti o n o r combin atio n o f s imul atio n s, w ill s h o w if those regions meet th e profi t req uir e m e nt s th a t t h e groce r need s t o deliver t o th e reg i o n s One o th e r imp o r ta nt fun c ti o n th a t t h e grocer mu s t use i s th e Minimum D olla r D e li very a m o unt. This a m o unt could b e fro m $ 0 t o as hi g h as th e grocer deem s feas i b l e. 144

PAGE 153

B y altering the Minimum Dollar Delivery variable, along with the Mandatory P rofit, and P rofit Margin, the grocer can see whether the initiative may be profitable. 145

PAGE 154

146 Figure 2. Duration of delivery in da ys Figure 3. Profit options settings

PAGE 155

Figure 4 Minimum dollar delivery :::.=T F :wca::::sdflg (),Spe,cifyMandat:oryProfitby CancenvicR.egion 1 1 1 < 1 D I E I F I G I Figure 5. Mandatory profit 147

PAGE 156

==t r-:-a:::::'Nlli'tG Or. Speofy Mllnd,Qlry Proftt by COnca,tric Region H 11 A f,s S r,. C jB or,. r,.~ F Figure 6 Delivery charge per concentric region =~ r-::v-~=IMtbng Or, SpedfyflWlda1DryP'rofttb-Conce,11C:Ra,gieln 1 181 c l 01 1 1 1 Figure 7. Delivery charge per region interface 148

PAGE 157

= i :--...:.::-' --...,-1,yCoann:" 1 1 1 < 1 01 1 1 1 Figure 8. Delivery charge per region pop up form Figure 9 Simulation options form 149

PAGE 158

150 Figure 10. Number of orders error catching Figure 11. Number of deliveries drop down list

PAGE 159

151 Figure 12. Duration of deliveries error catching Figure 13. Duration of deliveries drop-down list

PAGE 160

152 Figure 14. Region and legend color coordination Figure 15. Report dashboard for the simulation run on April 13

PAGE 161

153 = ::w~=Ntlng Or Spedfy MllndatDry Profit by c:onm,n; Regio,, H I A l8I C I 01 1 1 1 ---.awveJ Figure 16. Simulation main map result with profit settings Figure 17. Pause function

PAGE 162

Proftt~On,) fo.iii'""3=:;0G1ar-=r;o-===.-,,g Or. Specify Mandatory Proftt by Corun1ric Regk,n 1 r r c r 01 1 1 1 Profit Opt ions Pl'ofitMorginf,,%) fo.oe3=:;0a1ar fio3 Mrdaa,ryProfit F ...... boo<_, ...... regions) by f9)0 0-, Speofy Mrdaa,ryPl'ofitbyCana,ntnc~ I A l81 c l D I E I F I G I =-:;,a-r,,~ ,St.lmtDIM,y*9'1 Figure 18. Profit options form Figure 19. Allow dispatch alerts radio button 154

PAGE 163

155 Figure 20. Truck dispatch notification = ~:----~=INl!a'IQ Or. ~tort Profit by Ccrantric Region I l l c l 01 1 1 1 Figure 21. Simulation main map result showing truck usage

PAGE 164

156 Figure 22 Report dashboard for the simulation run on April 15 POPULATION REPORT l[I K j Q:QSI! ...... At A2 .., M ... ,.. at 82 83 ... .. POPUlAl10H -... ... ... ... ... ... .. ... PfRaNT Of TOTAL POPOI.ATION 0 _0151.515 00151615 0 _0151515 0 _0151515 0 0151515 0 .0151515 o .o,s,s,s 01151516 0 0151515 0015151S 00151515 0 0151515 IEGION .. 87 88 .. ., .,, Cl C2 0 C4 C$ ca OOf'UlAl10H -... ... --... ... -... -... ... PlRCfHT Of TOTAL POPULATION 0 0151515 00151515 0 ,0151515 0..0151515 o .01s1s1s 0 .0151515 0 .0151515 0 .0tStSIS 0 _01s1s1s 0 .0151515 0 0 151515 0..015'515 AEOIOff C7 CB cs C10 Ctl 01 02 03 04 DS 08 07 POPUlATDH ... ... ... ... ... ... 400 -... ... ... -PERCENT 0# TOTAL POIIUlATION 0 .0151515 0 .0t51SI S to1s1&1s 0.0151515 00151515 0 .0151515 0.Dt51515 0 .0161515 0.0151515 0 .0151515 00161515 0 .0151511 ...... DO DI 010 Oil 012 .. f2 D .. .. .. f7 POPUlAllOH ... ... ... ... ... ... 400 ... ... ... ... PlRCEHT Of TOTAL POAJLATIOH 0 .0151515 00151515 0 0151515 0 .0151515 00151515 0.0151515 0 .0151515 1 .e1s1s1s OOtStStS 00151515 1 .0151515 0 0151515 REOION .. .. ... rn n ,, .. ,. .. n .. ... ... ... 400 ... ... ... .. ... ... ... ... 0 .1151515 0015151' 0 0151S15 0.01S1515 0 .015151 5 0 .015151S 0 0151S15 00151515 00151 51S 0 .0151515 0 ot-5151S 00151515 IE""' .. FIO G I G2 G3 G4 OOf'UlATC>lt ... ... .. ... ... 400 POK:VfT Of" TOTAL POfVLA110N 0 .0151515 0015151S 0 0151515 0 0151.~15 0 .0151515 0 .0151515 1 1 l IJ i-'"""' i-Figure 23 Population report

PAGE 165

157 -If X Profitability Chart Figure 24. Profitability chart

PAGE 166

158 TOTAL MANDATORY PROFIT PER SIMULATION ..., Zone -1 --3 -4 -__,, --10 T_.... 21.N A 1 21.00 .. 52.31 A2 52.37 _,., 31.83 Al '9.13 ... ...,. ,.. ..... ... ..... Al ... .. .. 2700 Al ~:f; .. 51 .17 '" .. .. ,,, .. .. ... _., 2<.00 ., 24. 00 .. .... 1 .. " --.... .. 24.00 -21.00 21 00 __ 27 .DG 97 27.00 -...., SL32 --60.17 -40.,1 --31 51 IHI 31 .51 _.,, Jt. 12 ... 31. 12 -C1 40.24 c, 40.24 -C> 42.11 C2 42.11 -Cl 2700 C3 27. 00 -C4 50 .03 C4 50. 03 -Cl .... a ..... -Cl ..... co ..... -CT 45.21 CT 45.21 -Cl s:2.1, ca 52.10 -Cl ..... a ..... -C1 24. 1'1 c .. 2.111 -C11 St.52 C11 Sl. 52 -01 Sl.42 01 Sl.'2 .. .... D2 ..... -03 43 N .. 43 N -04 12.0I "' UOI -00 20.20 .. ..... -00 .... .. .... __ 57.24 01 .,.,, --,s.,. .. 15 14 -00 27.00 .. 27. 00 ... '5. 12 ... 3S 12 _01, ..... 011 31.02 ... 51.80 012 .. ... .. 42.10 .. '2.10 -0 31. 73 u 31.13 _., 21.00 .. 27.00 .. ..... .. ...,, .. 27.1/G .. 21.00 .. 24,00 .. 24.00 _., .... E7 SH3 .. 45. 15 .. "5.15 .. 31.10 .. 31.90 ... 37. N ... 31.81 1-~~Et L ... 01c_ ... _'8, 01 Region fl 32.83 fl-32.113 Region f2 35.32 f2 35.32 ,. Region f3 42.10 f3 '2.10 Region f4 ,s .62 f4 45.62 Region Fl 311.42 f5 llUl Region H 45.61 H 45.61 Region n 30.22 n '.!0.22 Region H 4Hl fl 43.43 Region 11 19.19 11 19.19 Region f10 42.71 flt '2.11 Region 61 28.0II GI 28.0II Region G2 45.113 G2 45.113 Region G3 31.17 G3 31.17 Region G4 S32S G4 53.25 .... -~ 0 00 0 00 0 .00 0 00 0 .00 0 .00 0 .00 0 .00 T.llif -511.63 0 .OII 0 .OII 0 .OII 0 .OII 0 .OII 0 .OII 0 .OII 0 .OII 0 .00 W.MUI 15.14 0 .00 0 .OII 0 .00 0 .00 0 .00 0 .00 0 .00 0 .00 000 A-39.SO --11.:M 2700 -41.53 Pnl--521 Figure 25. Mandatory profit per simulation report

PAGE 167

159 Residence Report Simulation # Delivery Amount Delivery Time Address Run# Region simulation1 201. 76 4/16/05 11 : 05 302 Orange Lane RegionA1 simulation1 213.62 4/17/05 4 : 11 705 Orange Lane RegionA1 simulation1 60 86 4/17/05 2 : 27 897 Orange Lane Region A1 simulation1 73 62 4/16/05 22 : 37 174 Grapefruit Lane RegionA1 simulation1 273 .81 4/15/05 17 :31 696 Kiwi Lane 1 Region A1 simulation1 270.01 4/17/05 1 : 35 359 Grape Boulevard 1 Region A1 simulation1 296.92 4/15/05 16 : 22 900 Grape Boulevard 1 Region A1 simulation1 63 .51 4/17/05 5:14 547 Joe Boulevard 1 Region A3 simulation1 209 13 4/16/05 14 : 27 839 Joe Boulevard 1 Region A3 simulation1 210 .61 4/16/05 8 : 18 860 Joe Boulevard 1 Region A3 simulation1 161. 05 4/16/05 8 : 53 204 Pete Boulevard 1 Region A3 simulation1 183 34 4/15/05 23:22 372 Pete Boulevard 1 Region A3 simulation1 289.03 4/17/05 8 : 02 640 Francis Way 1 Region A4 simulation1 129 67 4/15/05 19 :43 874 Francis Way 1 Region A4 simulation1 275 84 4/16/05 15 : 48 618 Sally Way 1 Region A4 simulation1 103.41 4/16/05 2 : 32 848SallyWay 1 Region A4 simulation1 12 68 4/17/05 1 : 58 240 Jennifer Street RegionM simulation1 155.38 4/15/05 19 : 26 423 Jennifer Street Region A4 simulation1 46.42 4/16/05 18 : 12 570 Jennifer Street Region A4 simulation1 162 53 4/16/05 18 : 06 692 Jennifer Street Region A4 simulation1 196 15 4/15/05 10:42 193 Nutmeg Road Region A6 simulation1 152 .81 4/16/05 11 :57 623 Nutmeg Road RegionA6 simulation1 64.84 4/17/05 0:49 642 Nutmeg Road Region A6 simulation1 161. 65 4/15/05 20 : 12 878 Nutmeg Road Region A6 simulation1 82 .44 4/17/05 10 : 14 22 Pepper Road Region A6 simulation1 246 95 4/16/05 22 : 25 847 Pepper Road RegionA6 simulation1 73 .01 4/16/05 4 : 45 18 Salt Road Region A6 simulation1 218 19 4/15/05 20 : 35 399 Sugar Parkway Region A6 simulation1 25 78 4/17/05 6:01 862 Sugar Parkway 1 RegionA6 simulation1 198 16 4/15/05 11 : 16 246 Seattle Street 1 Region B1 simulation1 173 79 4/15/05 14 : 03 106 Miami Street 1 Region B1 simulation1 32 80 4/16/05 12 : 03 321 Miami Street 1 Region B1 simulation1 294 74 4/17/05 6 : 24 572 Miami Street 1 Region B1 simulation1 115 38 4/15/05 20 : 24 12 Austin Street 1 Region B1 simulation1 290 70 4/15/05 18 : 34 302 Austin Street 1 Region B1 simulation1 273 74 4/17/05 5:49 197 White Boulevard 2 Region B2 simulation1 221. 09 4/16/05 4 :51 568 White Boulevard 1 Region B2 simulation1 187 85 4/16/05 2 : 15 624 White Boulevard 1 Region B2 simulation1 223 25 4/15/05 10 : 13 393 Red Boulevard 1 Region B2 simulatlon1 202 23 4/16/05 23 : 52 290 Blue Boulevard 2 Region B2 simulation1 241. 49 4/15/05 17 : 48 727 Blue Boulevard 1 Region B2 simulation1 277.32 4/15/05 15 : 59 947 Blue Boulevard 1 Region B2 simulation1 228.22 4/16/05 13 : 12 57 Yellow Parkway 1 Region B2 simulation1 11. 47 4/17/05 7 : 15 394 Yellow Parkway 2 Region B2 simulation1 153 79 4/15/0521 : 27 812 Yellow Parkway 1 Region B2 simulation1 22 35 4/16/05 20 : 18 368 Rose Avenue 2 Region 84 simulation1 278 93 4/16/05 9 : 10 476 Rose Avenue 1 Region 84 simulation1 197 78 4/17/05 9 : 40 520 Rose Avenue 2 Region 84 simulation1 119.57 4/16/05 5 : 19 154 Tulip Avenue 1 Region 84 simulation1 96 90 4/15/05 13:35 424 Tulip Avenue Region 84 simulation1 286 20 4/15/05 19 : 20 77 Pansy Avenue Region 84 simulation1 134 52 4/15/05 16 : 56 592 Pansy Avenue Region 84 simulation1 147 66 4/15/05 10 : 53 602 Pansy Avenue Region 84 simulation1 259 .21 4/15/05 18 : 46 211 Sunflower Avenue 1 Region 84 simulation1 163 94 4/15/05 10 : 19 268 Sunflower Avenue 1 Region 84 simulation1 162 26 4/16/05 22 : 54 827 Sunflower Avenue 2 Region 84 simulation1 76 06 4/16/05 11 : 17 549 Ow1 Lane 1 Region B5 simulation1 122.10 4/16/05 9 : 22 701 Ow1 Lane 1 Region B5 simulation1 224 64 4/17/05 0 : 09 983 Ow1 Lane 1 Region B5 simulation1 291.89 4/16/05 8 : 30 53 Sparrow Lane Region B5 simulation1 150 64 4/15/05 12 : 08 370 Sparrow Lane Region B5 Figure 26. Residence report

PAGE 168

160 400 5.00 10 500 80 400 5.00 15 5.00 80 400 5 .00 15 5 .00 80 400 5 .00 15 5.00 80 400 5.00 15 5.00 80 400 5.00 15 5.00 80 400 5.00 15 5.00 80 400 5 .00 20 5.00 80 400 5.00 20 5 .00 80 400 5 .00 20 5.00 eo 400 5.00 20 5.00 80 400 5.00 20 5 .00 80 400 5.00 20 5.00 eo 400 5.00 20 5.00 80 400 5.00 20 5.00 80 400 5 .00 20 500 eo 400 5.00 20 5.00 eo 400 5.00 20 5.00 80 400 5 .00 25 5 .00 80 400 5 .00 25 5.00 eo 400 5.00 25 5.00 eo 400 5 .00 25 5.00 80 400 5 .00 25 5 .00 eo 400 5.00 25 500 eo 400 5 .00 25 5.00 80 400 5 .00 25 5.00 80 400 5.00 25 5.00 eo 400 5 .00 25 5.00 eo 400 5.00 25 5.00 80 400 5 .00 30 5.00 eo 400 5.00 30 5 .00 80 400 5.00 30 5.00 eo 400 5.00 30 5 .00 80 400 5.00 30 5.00 eo 400 500 30 500 80 400 5.00 30 5.00 80 400 5 .00 30 5 .00 eo 400 5 .00 30 5.00 80 400 5.00 30 5.00 80 400 5.00 30 5.00 eo 400 5.00 30 5.00 eo 400 5 .00 35 5.00 eo 400 5 .00 35 5 .00 eo 400 5 .00 35 5.00 eo 400 5.00 35 5.00 eo 400 5 .00 35 5 .00 80 400 5 .00 35 5 .00 eo 400 5.00 35 5 .00 80 400 5 .00 35 5.00 eo 400 5.00 35 5.00 eo 400 500 35 5.00 eo 400 5.00 35 5 .00 eo 400 5.00 40 5.00 eo 400 5.00 40 5.00 eo 400 5.00 40 5.00 80 400 5 .00 40 5.00 eo 400 500 40 5.00 80 400 5.00 40 5.00 80 Figure 27 Settings report

PAGE 169

161 Delive Gr-Regional Profit Slml,ladon I Trudt and R Sen1Time Retum Tune RoundT Time Gr-Re lonal Profit II Profll -1 Trvck1-lo~03 411710510:00 411710511: 38 1:37 '8207 S2.07 -1 Trvck1_1o_06 411710514.07 4117/0515 :51 1 :'3 $6909 19 09 -1 Trvcl1-to-H 4117/05 15:51 4117/05 18:55 1 :03 584.41 SUI Sffulolion1 Trvck 2 -ID 87 4117/0518:14 4117/05 17:35 1:20 S82.2S S2.2S -1 Trvck t -to -"'6 411710520c16 4117/0521:20 1:G3 $6148 S1.41 -1 Trvck2__D2 411710520:28 4117/0522:11 1 :43 584.90 SUD -1 Trvck3_1D_U 411710520.45 4117/05 22:35 1 :49 $60.81 S0.81 -1 Trvck1 _to_E6 411710522:29 4111!1050,18 1 :41 $60 .70 S0.70 -1 Trvct2_.,_e10 4117105 22.40 411MJ50:0t 1:20 $67.21 S721 -1 Trvck 3 -to 01 411710522:58 411111050: 36 1:37 $63 .71 S3.71 _, Trvck4-lo-F8 411710523:21 411MJ51:27 2:06 Mt ao s1. ao -1 Trvck1__C2 411MJ5'24 4111111)51:56 1:32 $61.12 sa.12 -1 Trvck2 ___ t1 4/l&'DS 1 : 16 4111111)53.11 1:55 $61.SO SUD -1 Trvd:3 ___ ,_. 4111111)51"45 4111111)5 3:00 1 :14 $68G3 $6.Gl -1 Trvck4_1o_E9 4118/051:51 411811153:46 1:55 S6541 SS.41 _, Trvck 1 -G4 411MJ52:t4 41181115 4:28 2:12 S6t 52 St.52 -1 Trvckt_lo_H 41181115 5:53 41181115 7 : 19 1:26 $63.61 SJ.81 -1 Trvckt-lo-C1 4118/05 7:25 411811158:51 1:26 S82.82 52.12 -1 Trvd:2-toReg,onAS 411811157:31 411811158:45 1 :14 584.42 $4.42 -1 Trvd:2-to-GJ 411811158:45 4118111511:10 2:24 S63.19 S3.19 -1 Trvck1_to_C3 411811159:55 4111111)511:38 1 :43 563 .63 $3 .63 Smllllion1 Trvckldll-lD-119 411MJ5 t0:06 4118,'f)S 11:27 1:20 S71.38 STl.38 _, Trvcl4dilj>oldled-A1 4118111510:12 41181115 11:33 1:20 $68.&4 sa.&1 -.1 Trvck 5-to -C4 4118111510:23 4118111511:14 1:32 584.38 $4 39 _, Trvck1_1o_06 411MJ512".01 411M>514; 08 2:06 $63 04 $3. 04 Figure 28. Delivery report = :"ca:::: Nlbng Or Spedfy Ma,datDry Pl"offt by Concllnlf'tcRCQiiDn I l l c l 01 1 1 Figure 29. Simulation options form

PAGE 170

162 =~ rso=-~=Ntbng Or,SoedtYMrdatarvProfttbyCOnotnCncReglcln 1 1 1 < 1 01 1 1 1 Figure 30. Simulation main map result showing three simulations

PAGE 171

163 Figure 31. Simulation charts report

PAGE 172

164 Figure 32. Simulation charts report showing second and third simulation

PAGE 173

165 -II X Figure 33. Settings report showing three sim ulations TOTAL MANDATORY PROFIT PER SIMULATION Z one Siftlllbon1 -2 -3 -5 --.a --.1 _, SmationlO Tolol _.... ~ -,,,_ ,, "' 1f 1:: ~ llegion Al 2510 <0..22 ,o.os A1 llegion A2 35 97 *" 2..00 A2 IOUI lloglon Al 21 00 .. 78 IUD Al es.ee llegionM 2,00 '8.11 37.23 M ::~ llegionAS 47.80 2,00 3021 Al lloglonM 27 00 <012 2,. 00 M 91.ll lloglon 81 22.05 11.45 18.97 81 52.'7 llegionll2 33,9 '500 23.13 112 ':: : llogionBl 25,1 30 00 30.01 Bl llegionM 29,81 ... 12 38.<6 114 11027 llegionll6 2,00 25 37 ,s.s1 116 au, lloglon 88 1758 '935 40 .30 Ill 10'24 Region 87 '921 28.93 '9.33 111 119.<1 lloglon 88 1187 ZllB 6.81 88 '6.5E Aogionll 2735 29.51 38.91 Ill .::~ llegion IHI 3307 ,e,. 31. n ... llegion911 45.'3 48.1S 18.7' IHI 112.31 llegionC1 '3.75 42.00 2,.00 Cl 109.75 Aogion C2 3741 4UI 32.29 C2 ::.~ lleglonCl JIS2 31107 31.99 Cl AoglonC4 35 09 35. 19 31. 95 C4 103.22 llegion cs 32 .94 311. 08 37 .11 cs 109 ,12 llogion Cl 27.90 311.21 '9. 89 Cl Ill.Ill! Region C7 4187 31 43 311. 00 C7 112.09 lloglonCI 42.00 39.83 42.52 Cl 124 .15 llegionct 2919 21 00 30 67 ct IOM lloglon Ctl 40. 04 ~ .01 40. 84 C10 1211.811 lloglonC11 1542 31100 27 89 C11 112.31 lloglon 01 4710 12.29 2400 01 83 .311 Aogion D2 43 .78 '-C,01 40, 70 02 12S .49 llegion03 4062 2,00 ,e.,o 00 111 .02 lloglon 3175 32.115 211.S< 902 Region 06 24.00 21. 08 27 .00 06 72.08 llegion 06 2700 3112 28.48 06 6' n lloglon07 3709 3000 3&48 07 10551 lloglon 01 27 00 '911 n 01 112.95 lloglon 09 31115 4111 5'.00 09 142.03 !logion D1I 24,41 ,s.48 28.'3 011 !111.32 llegion 011 35.51 '7.89 ... n 011 lloglon 012 31 07 Zl.19 18.51 012 Aogion f1 2,00 37 .88 23.61 E1 18.57 lleglon E2 42.03 40 .,1 48 59 E2 131.02 lloglon D 4807 ,511 40. 03 D Ill.IHI llegion f4 49, 51 27 92 21.00 f4 !111.41 lloglon f5 3491 JUI ,1.as f5 114.8 !legion ff 49.51 25.07 ,s.1& ff 120.3! lleglonf7 40 .07 22.05 42.11 f7 1~! !legion fl 32.50 3281 28.59 fl lleglon f9 16.99 4510 JU f9 97.' lloglonf10 42.22 29.25 12.43 ftl 13.9( lleglon fl1 4800 40.112 47 .12 f11 134.5< .., __ .. ~-,,_ 30311 ffi "\: -;: _., 32.90 47.85 32.02 n ~~--_ .. ,, ... '4.Z, .... ,, fl 35.7S 45.06 ,a.12 ,. :-:: .. ., .. 13.7 3510 ,. .. .. .. ... ,. = ,. 11~: _., '5.311 2700 2100 :,. .. 3338 ,. .. .... .. .. 20-2, >Ul 4303 .. 161.IO ~,10 35.N 3131) ., .. ... 115.24 -G 27 00 42.18 47.>S G 1 117.03 -G> :,000 -.... G> '~r -Gl .._ .. .. _,., ,,_ .. Gl -04 ..... 2700 3522 G4 ,o, --'"" ..... --'-" """ 00 -"" o .oo -50. 44 41. N ..... 0 .00 0 .00 00 000 0 .00 0 .00 0 .00 MlMnum .. ., 11 .45 . 0 .00 0 .00 0 .00 0 .00 0 .00 0 .00 0 .00 --,. .. 3500 33.80 --. ... 8 .75 ,o ... -24.00 ..... 2A-OO ....... ,SJ)J ,. .. >2.81 ... ---4 .50 n ... ee54.1STOIIIIAI~ I 5400....,,._M~ O .OOllilUmr.lfflM~ 3'.t/.2 A"'999AI~ 0 .0457 fln:ifll'> .. .-Figure 34. Total mandatory profit per s imulation report showing three simula tion s

PAGE 174

166 simwtion3 293.36 4/19/05 23.07 259 Diorite Street 2 Region F1 simulation3 99.31 4/18/05 17:00 727 Oiorita Street 1 RagoonF1 simulation3 161.58 4/17/05 11:20 739 Oiorite Street 1 Region F1 oimulation1 5182 4/19/05 18 35 723 Genesis Road 1 Region F2 simulalJon 1 235.06 4/18/05 10:37 86 Exodus Road 1 Regoon F2 llmulation1 218.20 4/17/05 14: 53 105 Exodus Road 1 Region F2 simulation1 250.51 4/18/05 14.47 249 Exodus Road 1 Region F2 simtllation1 281.54 4/18/05 2:07 614 Exodus Road 1 Region F2 Stmulalion1 151.42 4118/05 7 :18 809 Exodus Road 1 Regoon F2 simulation1 294 80 4/20/05 6 15 951 Exodus Road 2 Region F2 .. simulation1 131.26 4119/05 8 21 328 Joshua Road 1 Region F2 srnulaion1 35.51 4/18/05 16:48 837 Joshua Road 1 RegoonF2 simulalion2 178.46 4/18/05 16:33 166 Genesis Road 2 Region f2 simulalion2 262.36 4/19/05 18 46 85 Exodus Road 3 Region F2 srnulation2 295.07 4/18/05 3 :18 134 Exodus Road 2 RagionF2 simulation2 72.21 4/18/05 12:40 916 Exodus Road 2 Region F2 simulation2 222.04 4/19/05 21 47 257 Josnua Road 3 Region F2 simul.-ior\2 297.79 4118/05 13:14 850 Joshua Road 2 Reg,onF2 simul.-ior\2 247.23 4119/05 4 :48 875 Joshua Road 2 Regoon F2 simulation2 12668 4119/05 046 12RuthStreet 2 Region F2 svnulation2 247 95 4120/0S 5 34 671 Ruth Street 3 Region F2 simulation2 90.68 4/19/05 8 32 699 Ruth Street 2 Reg,onf2 simulation2 9669 4117/05 1455 757 Ruth Slreet 2 Region F2 simulation3 135.91 4119/05 21.40 551 Genesis Road 3 Region F2 s,mulatJon3 225.05 4118/05 5 :46 695 Genesis Road 3 ReviOfl F2 9imulation3 228.09 4/20/05 3 43 810 Genesis Road 3 Region F2 limulaoon3 103.34 4/18/05 23 29 188 Exodus Road 3 Region F2 s,mulallonJ 203.47 4117/05 14:115 856 Ruth Street 3 Region F2 simulwon1 136.92 4/19/05 8 :56 566 Solomon Road 1 Region F3 s i mui.ion1 202.44 4/20/05 7 16 620 lsaioh Road 1 Region F3 Slmulalion1 90 00 4/18/05 1321 493 Hosea Road 1 RegionFJ s,mulalion2 119.74 4/18/05 2:52 276 Solomon Road 1 Region F3 simulation2 66 73 4/18/05 23 19 402 Solomon Road 1 Region F3 aimulation2 68 90 4/19/05 0 02 467 Solomon Road 1 Region F3 s,rnuation2 276 86 4/17/05 23.08 132 Isaiah Road 1 Region F3 aimulation2 155.98 4/17/05 23 42 264 Isaiah Road 1 Region F3 aimulation2 187.03 4/17/05 13 20 832 Isaiah Road 1 Region F3 simulation2 288.16 4/18105 16:24 466 Hosea Road 1 Region F3 simulation1 294.47 4/18/05 2:33 877 Shang Road 1 Region F4 simulalion1 207.91 4/19105 21 11 943 Shang Road 1 Region F4 simulation1 287.38 4/20/05 6 :24 137 Qin Street 2 Region F4 smulation1 50.90 4/19/05 5 ;29 170 Qin Street 1 Region F4 simulation1 118.85 4/17105 16 10 337 Qin Street 1 Region F4 F igure 3 5 Re s id e nce r e port showing orders from the second and third s imulation s = r-::v-co::::..-. Or, Specify ~ti:ryProflt byC.onc:entric Reoon H 11' A I' S l'5 C r,;0 135 E F F F G r,o ""'-.0woe I F i gu r e 36 Mandato ry profit per conc e ntric r egi on on profit options form

PAGE 175

167 .. .. uo uo uo MIO ... uo uo ... ... ... .. IUD uo uo .... uo I.GO ... 25 -UD 21 uo -uo -uo UI uo -uo I.OD uo -uo UD 211 ll.00 -UD ll.00 -MD 3D ll.00 -uo 3D f.GO 40D uo 3D a.oo 400 ., I.GO a.a uo -'MO 3D uo to 4IIO a.a -uo UD 3D uo IO = uo 311 uo -uo 30 uo .. 3D 15.00 -UD 35 ... .. UD uo a. .. .. ut uo ue -... -a.GI II.GO -UD II.OD .... uo -a.a uo -UD uo 400 uo uo -uo 40 uo 4D ,... IUID .... 40 uo -40 !ill .. 40 uo Ml 40 .. -4D uo -4D UD uo 40 5A .. uo U8 UII ue UI 45 5.aD uo Figure 3 7 Mandatory profit per concentric re g ion on settings report

PAGE 176

168 =:.:! r---:w~=setq -....... ..., ..... byC'.onantrc-H I" f,o 12' C F D 13' E F F F G r,o --0wvel Figure 38 Submit Delivery Charge button = r---:w==::fHttftg Or, Speofy MridlitDryProftt by C'.oranb'tc Reocn "I" r,or,,cF Dl"E FFf,o Figure 39. Delivery charge per region interface

PAGE 177

169 ==~t r-==aetlng Or, Spedfy Mrdamry Profttby Conanrc~ H J" A r,;, B l'5 C r,o 0 f,s E F F 145 0 [,o Figure 40. Delivery charge per region form =~ r-:w~=..-. Or, Soeofy ~tort Profit t,y C.onancnc Revar, "fts fi,r,.cr,o or,,e r,o- 145[,o ::::J"-1--.0wvo I Figure 41. Region A3 showing the delivery charge

PAGE 178

170 Profitab ility C hart n.. chart shows the ........, total mandatory profll -!of Ill n,goons !of al ..........,, -Fig ur e 42. Profitabili ty chart sho w in g profit for re gi on A3 TOTAL MANDATORY PROFI T PER SIMULAT ION -:;;.... -t _, ..._, SiffMlltio,n4 Sirn.almone _, s....tion 10 r..., -" m : n A 1 .... --A1 52373 ~= 558.28 .u 558. 28 .. ... . 601.ea ... -M ..... .. 165. 00 .. ..... ..... _., l07.18 112 )07. 16 .. 22121 ., 221.21 .. ... .. .. .,._ .. --...... .. .., .. --l3UI .. 139.41 -SJ 545.70 87 5'5.70 --....... .. 1135.,e --49CU8 .. 490. 48 ... 120.00 ... 1211. 00 .. 124..22 .,. 12'22 ReglOr'IC t tt. n C1 411. 73 _c, 1'7.32 Cl 127.32 -c, _ 71 CJ ... 71 -C4 1SU5 C4 1SJ.85 _c, IZS.00 a 1ZS.OO _.,. 161.31 Cl "'I.JO _c, 33825 C1 3'll25 -Cl 111 .00 Cl 111 .00 -ct 500.32 Cl 509. 32 ~c,, S>S. 12 c,o S>Sll2 Jliegk:w'IC11 ... .,.. C 11 ...... -01 1&2.01 01 142.01 -02 ...... I ...... -0, 321. 76 D3 32178 ... 21'.22 ... 214.22 -06 .,..., 06 US. 92 .. 3'<31 06 344 .31 3711'1 D7 377.87 .. ..,_ .. .. 503.SI .. ..,.,,. DI ...... _01, ...... 010 ...... _01, S50I 011 .... ,. _01, 25324 01, 253.24 .. ..,.,,. ., 343.&t -D 405..33 .. ..... ,, -D 13'.'1 D 134,42 .. 633. 30 .. 633.JO .. .... .. 5S4 .89 .. ....,,, .. 400.57 _., ... ., E1 156.43 I ... .,..,., f.l 470.50 -.. 1S4..37 ... 1"4.37 F igure 4 3. T ot a l mandato ry profit p e r s imulation r e port s ho w in g profit for re g ion A 3

PAGE 179

=:a=T ::'9ltwl:X:,'settnsJ o.,--,__,.., ..... byc.ran"1cH I I I C I D I E l I G I Figure 44. Settings used with Intra-Regional Drive Time function ==r :w~=eeuna Or. 5peofy M.ndacriry Profit by Ccna!ntrlc 1 1 1 < 1 D I E I F I G I """"""""'0wve I Figure 45. Intra-regional drive time interface 171

PAGE 180

=~~:w~=fsetq C.-, ~Mandatcry ProfitbyConcentricR~ t l81 c l 01 1 1 1 Figure 46. Intra-regional drive time pop-up form 1110 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 400 3.00 3.00 3 00 3.00 3.00 3.00 3.00 300 300 3.00 3.00 3.00 3.00 3 00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3.00 3 .00 3 00 3.00 300 300 3.00 3.00 3 .00 3 .00 10 15 15 15 15 15 15 20 20 20 20 20 20 20 20 20 20 20 25 25 25 25 25 25 25 25 25 25 25 30 30 30 30 Time 500 5.00 5.00 5.00 500 500 500 500 500 500 500 500 500 5.00 5.00 5 .00 500 5 .00 5.00 500 500 500 5.00 500 5 .00 500 5.00 500 5 .00 15.00 5.00 500 5 .00 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Figure 4 7. Settings report showing 15 minute intra-regional drive time 172

PAGE 181

173 Delive Groll Regional Profit Simulation I TruckandR Sent Time Return f1111e Protk Sffllllliln1 Truct1dispaldledloRegi:,nE 8 4122J052:59 4122105S:OO Si!ulliJn 1 Truct 2 dispaldled lo Regi:,n 811 4l22J05 4 :08 4122105 5:29 $12.17 Smation1 Nd: 1 dilpadled lo Regi:,n E7 4l22J05 5:29 4122105 7 :18 1:49 51927 SirailliDn1 Truct2~toAegill!G3 4122J05S:46 4/22/05 7 :35 1:49 $110.95 SID.95 Si11111io111 TNd:3dilplb::tledlDRegi:,nE6 4l22J05 6 : 49 4121/0S 8:45 1:55 S102.54 S2.54 Si!ulliJn 1 TNd:4diljloldledlDRetiJnE4 4l22J05 7 :12 4122J059:08 1:SS S101.59 $1.59 Si11Uiiion1 Tllld: 1 diljloldled ID lleglln F4 4122A)St25 4l22A!S 1 1 :32 2:06 S107.06 $7.06 Si!ulliJn 1 Truct 2 dilplll:lled lo lleglln C9 4l22J05 9:48 4l22J05 11:14 1:26 $104.76 $4.76 Si1Ulliln1 TNd:3dilpalc:hedlDRlti>nC6 4122/DS 11:09 4122AJ512:35 1:26 S111.29 S11.29 Si1111111Dn1 TNd: 1 lilpadled lo Regi:,n C11 4122/DS 13:15 4122AJ514:42 1:26 S10521 $521 Saulliol11 Tllld:1 dilplll:lledloRegi:,nE8 41221051516 412210517:46 2:29 S107.58 ST.58 Sinllltial11 T Nd: 2 dil9llc:fled ID Ragixl 81 D 4/22!0517:17 4l22A!S 18:38 1:20 $100.98 S0.98 Si11Uiiion1 Tn1ct I dllplldled ID Region F1 4122J0520:S1 4122/DS 22:57 2.16 $102.74 $2.74 1 1 ...... II 1:55 111.15 SillJllion1 Truct 3 dipae:fled ID RegiJn 4l22J05 2125 4l22ID5 23:32 2 :06 S107.99 S7.99 Siniolion 1 Truct4dilplldledlo~D3 4122/DS 22.'34 412l'D50:18 1:43 S112.04 S12.04 Si1Uation1 Truct1dispaldledloRegi:,nF2 41231050'.01 4/ZWS2:13 2:12 S11Q.91 S10.91 Siniolion1 Truct 2 dilpadled ID RegiJn F 10 4/23/050:24 4/ZWS225 2.111 S113.78 S13.78 Si!ulliln1 T Mi 3 dilpadled to Rl!gpl F7 4/23/051:56 4Q3/053:S1 1:55 S101.97 S1.97 Si11Uiiion1 T Nd: 1 dilplldled ID Regi:,n 06 4/23/05122 4123/0SS:41 2:18 S109.90 $9.90 Si11Uiiion1 TNd:2dilpeldledloRegi:,nC4 4l23ID5 4:31 4l23ID5 i-58 1:26 S106.64 SS.64 Smllliixl1 T ruct 1 dilpaldled lo lleglln A 1 41231D5i'52 4Q3IDS 7 :07 1 1 4 16.20 SilUlliln 1 Truct2dilpc:t,edloRegi:,nF4 4/23/056:32 .OJ/059:02 2:29 $4.72 t ... DI 4illl57:k 4illl5tt 2:47 Slllillli)ri1 Truct 3 dilplldled lo Regi:,n F8 4/23/05 7 :53 412:W510:00 2:06 S104.08 $4.08 .. .. -Figure 48. Delivery report showing increased intra-regional dri v e times = = r :v-~=setbr4 Or, SoeofyMrdatDryProftt by Cona!ntricRei;,on 1 l91 c l D I E 1 1 1 Figure 49. Inter-regional dri v e time interface

PAGE 182

174 -Sett i ngs Report Nlfflber rA Sinua~ons: 1 suntvess Oi"""rch Alel1S Yes I Duration rA Det\ca, at the Same Time : No Profit Marcin fin oercenll : 0 .06 Mirinun Doler : -v: 50 Nt.mber rA Orders ,_ Sinuation: 500 Reaion Pnnu.,bon x 1 0 .,_Char"" One-Wav Dnve Time lntra.uAnional Oriw Time Mandatnn1 Profit Reaion Region H 400 3 .00 10 5 .00 100 Region H R.g;o,,AI 400 3 .00 1 5 5 00 100 RegionA1 Rag,o,!A:J. 400 3 .00 15 5 .00 100 RegionA:J. RegionAJ 400 3 .00 15 5 .00 100 RogionAJ RegionM 400 3 .00 15 5 .00 100 ReglonM RegionA5 400 3 .00 15 5 .00 100 RegionA5 RogionM 400 3 .00 15 5 .00 100 RogionM Region Bl 400 300 20 500 100 Region Bl RegionB2 400 3 .00 20 5 .00 100 Region B2 Region 83 400 3 .00 20 5 .00 100 Rogion83 Region 84 400 3 .00 20 5 .00 100 Region 84 Region 86 400 3 .00 20 5 .00 100 Region BS Region 86 400 3 .00 20 5 .00 100 Region86 Region 87 4 00 3 .00 20 5 .00 100 Region 87 Region 88 400 3 .00 20 5 .00 100 Region 88 Region 89 400 3 .00 20 5 .00 100 Region 89 Region BIO 400 3 .00 20 5 .00 100 Region 810 Region BIi 400 3 .00 20 5 .00 100 Rogion 811 Region C1 400 3 .00 25 5 .00 100 ~onCI Region C2 400 3 .00 90 5 .00 100 Region C2 Region CJ 400 3 .00 90 5 .00 100 Rogion CJ RegionC4 400 3 .00 90 5 .00 100 Region C4 Region cs 400 3 .00 25 5 .00 100 Region cs Figure 50. Settings report showing increased inter-regional drive time Delive rt --Simulllionl T111c:hndR on Sent Time Return f1111t SiUliln1 T rucl: 1 dilplldMd ID ileglon D7 4/22Al5 2:44 4/22Al54:22 It -41m51:31 SiUliJn1 T 111cl: 2 dilpllclled ID Regbn FB 4/22Al5 9:10 Wlll511:06 1:55 S183.05 $3.05 SiUliln 1 Tfllcl: 1 dilpllclledllileglonD5 41221'514'.50 Wlll51&-.28 1:37 $183.74 SiUliJn 1 T111cl: 1 dilpm:hed to RegiDfl 84 Wlll518:00 l/111/)519'11 120 s110.n 1 llll2 It 22:21 SiUliJn I T rucl: 1 dilpllclled ID ilefon E6 4122/11521:57 41221'523:52 1:55 S108.0I 18.0I SilUllian1 Trucl:2 dilpaldled II Region U 41221'5 23:12 4/231051:18 2:06 $105.43 SS.43 Sialiln1 Trucl:3dilpaldledlDRegbn810 4llll05 23:40 4/231051:13 1:32 S104.91 S4.91 Silllillianl T111cl:4dilpm:hedlDllegiollA2 41221'5 23:46 412:WSl:OI 1:14 $115.39 S15.39 Silllillian1 Trucl:1dilpe:llldlDllegiollM 4llll05 23:52 4/231051:07 m $111.02 $11.02 .... TIIIIH It .... -144 11111 SinulliDn 1 rucl: 4 11 llefon F9 "23.VS 1:0I 4/231053:08 2:06 ST.37 .... I a 231 ll74 SillMliDnl Trucl: 2 dilpaldled ID llegioll 82 4/23/052:04 412:WSJ:31 1:26 $101.95 St.95 Sium11 Trucl: 1 dilpaldled ID Regbn C11 4/23/053:06 4/23/05 4 :57 1 :49 Sl1420 Sl420 Si!Uliionl Trucl: 2 dilpaldled ID ileglon 89 4/23/054:23 4123105 5:43 1:20 $101.71 S1.71 Si!dlliDll1 Trucl: 3 dilpllci1ed to Regbn E l 4/23/054:40 412:WSS:35 1:55 $110.41 S10.41 Siluial 1 Trucl: I dilplld1ed lo RegiDfl C5 4/23/056:01 4123105 7:21 1:20 $101.88 SUl8 Silulliarl1 Trucl:2 dilpaldled ID Regbn ll6 4l23IOS 6 : 12 41231058:02 1:49 S 103.78 $3.78 Sildiil,1 Trucl: 4 dilpm:hed ID Regbn 88 4/231056.10 4/23105 7 :56 1:26 S114.00 S14.00 Siouliioll1 Tlllcl:JdilplldledllllogmD2 4/23/05 7:04 4/23105 8:42 1:37 S110.63 $10.63 SinulliDnl Trucl:1dispaldledlDRegbnC9 4l23IOS 7 :44 4123105 9:11 1:26 S116.81) S16.110 Sildiil,1 Trucl: 2 dilpe:llld ID Rega, /IB 4123/0S 8 :07 4123105 9:22 1:14 S101.07 Sl.07 Slulilll1 Trucl:4dilpm:hedloRegi)ll/>J 4123/0SS:19 41231059:40 1:20 S103.0I $3.0I -. ... -. . Figure 51. Delivery report showing increased inter-regional drive times

PAGE 183

== ::v-~=:..-; Cr. Specify Mandalz:lry Proftt by Cona:ntrie Rega, r r r c r o r r r G r Figure 52. Overall average population per region ==~:w~=setmg O. ,Specify_Pn> .. byConolnOic-" r r r c r o r r r G r Figure 53. Population per region interface 175

PAGE 184

176 = ~:w~=iMilalg Or. Spedfy Mendtltory Proftt by Concfflbic Region H I A I B I C I D I E I F I G I Figure 54. Main map showing various amounts of orders placed =~~:w~=Ml&Wlg Or. Spmfy Mr,de10ry Profit by Conaentric Region H I A I B I C I D I < l I G I Figure 55. Regions showing higher and zero populations

PAGE 185

177 Figure 56 Report dashboard showing various populations POPULATION REPORT ~ikW:IDKllHm .. ..,.. H ... A2 A3 M A5 M ., B2 03 ... 95 PORILA1101t 250 250 250 900 ... 250 250 250 250 900 250 PERCENT Of TOTAL POPUI.AlllN O .Ot31278 0 .01382:71 0 0138278 0 0501383 0 0501383 0 0139211 O .OtJ1271 0 0131278 0.1131278 0 0501393 0.0131278 AfGl)N .. 07 ea 119 .,. "" C1 C2 C3 CA C5 CB POFIJLA1101t 250 250 250 250 250 250 250 250 250 250 250 P(RC[.NT OF TOTAL POPULAlllN 0 01J9278 0 .01311271 0 0139276 0 .0131'278 O ot3921& 0 0131271 0 01J8Z78 0 ttJ9278 0 Ot31278 0 013B278 0 0139218 .. ..,.. r:, CB Cl C10 C11 D1 D2 03 "' D6 D7 PORILA1101t 250 250 250 250 250 250 250 250 250 250 250 250 PEACfNT Of TOTAL POPUl.AlDI 0 0139278 0 .01311279 o o,mre 0 01)1'278 001l8271S 0 0131278 O .Otll278 0 .11ll278 00138278 0 .01312,. 0 0139271 O .ttl1271 Af(ll)H ll8 DO D10 D11 012 El E2 E3 .. E5 E7 PORILA1101t 250 250 250 250 250 250 250 250 250 250 250 250 PERCENT Of TOTAL POPULAllON 0 .013127& Oltll278 0 01J8278 0 .0138'271 OOl'8278 0 otlll27& 0 .01312711 0 .01l8276 0 0131271 001Jll278 D 0131271 0 .0138278 REOOH .. Et EIO f11 F1 F2 F3 ,. .. ,. F7 ,. PORILA1101t 250 250 250 250 250 250 250 250 250 250 250 250 PRaNT OF TOTAL POPU'l.ATDN 00138271 O..ot31'278 0 .01'827t 0.0131218 0 013827a 0 0139271 0.0131278 0 0131271 0 0138219 0..0131278 1 0t3'127e 0 0131271 "'"'"'" "' F10 01 02 03 "' PORILA1101t 250 250 250 250 250 250 PERaNT Of TOTAL POfU.ATl)N 00139271 0 .0lllV& 0 .0131271 0 0138278 0.0138276 0 0138276 ... IOO 700 ... 500 ... 300 200 100 0 Figure 57. Population report showing various populations

PAGE 186

THREE SIMULATION SEQUENCES THAT RESULTED IN EXCEPTIONAL A VERA GE AND LOW PROFIT ABILITY This chapter contains three scenarios that compare the overall Mandatory Profit of three runs. The first scenario will show a high overall Mandatory Profit. The second scenario will show a low overall Mandatory Profit. The third scenario will show a medium overall Mandatory Profit. What is demonstrated here is that by altering the various variables inherent in the application, the grocer can create scenarios where profit will be inferior, average, or superior. This valuable capability lets the user who may be uncertain about cyclical customer demand see the outcome of attempting deliveries under various circumstances. This is but another trait that is unique to this logistical application Exceptional Profit Figure 58 shows the Main Map with the criteria set to run a simulation that will generate a high profit. Notice that the population is set to 9,999 households per region. Figure 59 shows the Profit Report for this simulation. At the very bottom of the Profit Report is the Profit Gauge The Profit Gauge is a relative measure that can be used to compare the Profit earned between SERIES of simulations. It does not, for example compare the profit between simulation 1 and simulation 2 of the same simulation series. It does however compare the profit between series of simulations This is because each series of simulations (the simulations that occur in sequence after you hit the Run the Project button and until the last simulation finishes) has different Profit Option setting. 178

PAGE 187

The Profit Gauge is a simple, but effective number between 0 and 1 which is the total profit earned during a series of simulations divided by the number of grocery orders for that simulation sequence divided by 100. The higher the number the greater the relative Mandatory Profit earned. The Profit Gauge for this simulation sequence is 0.2018 (very bottom of Figure 59). Low Profit Figure 60 shows the Main Map with the criteria set to run a simulation that will generate a low profit. What is not shown is that the population is set to 500 households per region a much lower density than the prior simulation that had 9 999 households per region. Also compare the lower profit margin, and delivery cost with those of the prior simulation. The very bottom of Figure 61 shows the Profit Gauge of 0.0138 for this simulation. Compare this with the Profit Gauge of 0.2018 for the prior simulation. Medium Profit Figure 62 shows the Main Map with the criteria set to run a simulation that will generate a profit intermediate between the high and low profit levels that were generated in the two prior simulation sequences. The population is set to 4 000 households per region -which is of course between 9 999 and 500 households that was set in the prior two simulations Also compare the intermediate profit margins and delivery costs with those of the two prior simulations. The very bottom of Figure 63 shows the Profit Gauge of 0.0808 for this simulation. Compare this with the Profit Gauge of 0 2018 for the high profit simulation sequence and the Profit Gauge of0. 0138 for the low profit simulation sequence. 179

PAGE 188

180 ProfftMer;i-,(ln%) ~=Dollr fioo3 =~ rm:v-~=fsriing Or, SpeofyMa'idaloryPl"ofitbyCOl"ICWltricAegion H I I I C I 01 1 '1 Figure 58. Profit and simulation options for high-profit scenario

PAGE 189

181 T OTAL MANDA T ORY PROFIT P E R SIMULATIO N ,. z--1 -2 -3 -4 -..s _, __ _, -10 T* .... w uu. w .,.._ -"' 140.00 200 00 "'' 3400 -A2 240 00 144.14 A2 314.1 4 -,u 160 00 140. 00 A3 300 .0C .... 220 00 100 00 M 320 00 -A6 160 00 1IO .OO A6 320 0 -Al 200 00 12509 Al 325 0 200 00 ,oo.oo ., 3000 -112 100 00 180 00 112 : : : -83 120 00 180 00 83 -84 10000 18000 84 2IOO 120.00 ZI0.00 400 0 180. 00 111.'7 32U _., 16000 11324 81 3432 140 00 60 00 20000 20 .00 120 00 1400C -811 200. 00 140.00 B1I 3'1UO 811 117.52 224.19 811 312.41 -C1 200 00 !20. 00 C1 ~ : : -C2 Ill.DO 180. 00 C2 -C3 220 00 121.00 C3 340 0 -C4 14000 185, 10 C4 JOS.1 -C5 I0.00 10000 a; 1IO O -cs 40 00 121.13 Cl 1'1.1 : Aogior>C7 120 00 240.00 C 7 380 00 -Cl 180. 00 2111. 00 ca .c20.ao -ca 140 00 I0. 00 Cl 200 .0C lloglanC11 140 00 1'8.85 C11 ::: -C11 100 00 240 00 C11 -01 190.00 147 M 01 327. 14 -02 100 00 II0. 00 02 200 00 -03 I0. 00 Sl.00 03 140 .0C -1M 121 00 180.DO 1M =: lloglanDS 40 00 260 00 lloglonDS 200 00 II0. 00 Ill ::: -D7 141 N 100 00 D7 lloglanlll 220 00 100.00 Ill 320 .IO -09 180 00 ,-.,s DI 341.4! lloglanD11 22000 203.31 D1I 423.31 _01, 268.80 126.87 011 :le.4l lloglon012 167.53 180.00 012 347 .S3 Region El 160-:00--300.00 1Et 460.0C -Region E2 268.84 80.00 E2 34&& Region El 180.00 80.IIO El 260.0I Region E4 200.00 285.30 E4 465.JC Region E5 180.00 140.00 E5 320.0C AeQlon Ef 110.00 40.00 Ef 160.0C Region E7 110.00 147 .53 E7 267.53 AeQlon El 80.00 180.00 El 260.00 Region H 180.00 180.00 H 360.011 Region E10 200.00 1110.00 E10 JOO.DO Region E11 110.110 60.111 E11 180.0I AegionF1 267.67 l01l.S8 ft 374.Si Region f2 140.00 160.00 f2 JJ0.110 Aegionfl 140.00 110,00 fl 260.110 Region f4 1n.11 60.110 f4 232.1 Region rs 80.00 180.00 FS 260,01 Region K 180.00 210.00 K 400.011 Region n 20000 146.65 n 346.65 Aegion f1 110.110 140.00 f1 260.011 Region f9 160.00 148.43 308.4: Region FIi 128.96 180.00 F10 308.91 Region G1 180.00 110.00 G 1 JOO.DO Region G2 160.00 110.110 G2 280.011 Region Gl 140.00 140.00 G3 280.01 Region G4 140.00 203.82 G4 34311: lolll 1wo,.>< 111113.4:> U .1111 0 .00 0 ,00 0 .00 0 .00 0 .110 0 ,011 0.00 IIIU1IIIII BM JOO.DO 0 .110 0 ,00 0 .00 0 .110 0 .00 0 .110 0 .00 0.110 llmn 10.00 40.IIO 0 .110 0 .00 0 .110 0 .00 0 .00 0 .00 0 .00 0.00 AYlflllll 152.54 153.23 sin.dllevlllioll 52.85 57.36 Mode 140.00 180.110 llediln 160.00 148.13 l'lo-Ouolieli 10.14 10.23 10180.118 TolllAISiMtin 3110.00 lllxilum Al SiMlions 0 .00 llfflllrnAISinlllllons 152.89 AYlll!III Al SmlllliJns 0.20181'1oftr ..... F i gure 59 P rofita b ility qu o t ient an d p rofit ga u ge for hi gh profit scenario

PAGE 190

182 =:.=r J154l ~this:-:~ Or. Specify MirldatDry Proftt by Corantnc Reoion 1 l81 c l 01 E 1 1 1 Figure 60. Profit and simulation options for low-profit scenario

PAGE 191

183 T O TAL MANDA TORY PROFIT PER SIMULAT ION 1 i Zone _,, _, _, SifftAldDn,4 SfflAlltion 7 Siln..flltioftl .__,o ., __ T_ .. u ..... ,.... U ,IO ..... ~: _.., .... 10.17 A2 _.., 11.51 .... A3 22.11 -M 13.52 ... M : : ~ ... S .12 1321 ... ... 13.34 12.00 Al 20.03 .... 1.70 .. ,,us .. 10.02 8.20 02 V.22 _., u, 13.32 .. .... .. ,..., ... .. '!; .. 1 12 2.13 .. .. uo 7,04 IS.Z _., &21 11 .75 07 -i 17 18SS ---..... 11. 11 21.11 .... . , ..... IHO 2, ... Aeglorl81 1 .... 13.IM .. 11.SS _c, 11 ta ILl& c ~!: -C2 1.52 C2 -C3 22.75 ... C3 29. 70 -"' ..,. ISAl3 C4 28.31 -c 5.311 ... Ci 12.U _,. 11.52 11.23 Cl 21 .1, -C7 14.71 ,. .. C1 ,..., -Cl 12.23 2.11 Cl 14.l: -Cl 11. N .... Cl fll,gio,IC,. .... 13 10 C11 ...__.c 1 13.21 .... C 11 ,.: .. -01 ... .. ,. 01 10 .71 _., 15.87 5.20 02 20.1 -03 .. ., UI 03 ~!: ... ... .... 1M .. 4.65 13,42 .. 20. 0 .. 1a..sc II.A .. 30. 1 _o, 1100 10.71 D7 ~! --10 11.00 .. --1112 IS.TO DI 278 --1.Dl 7 01 010 .... _01, 14.20 11.45 011 ::~ _01, .... 11.818 012 .. 1113 .... ., 21:.i .. 11.12 11.tO [2 21.22 _., .... 11.3111 .., :1 -fA ... 3.12 ,. ... ,. n s, .. .. .... 12.51 .. 31.10 -"' -u S.73 12.17 ., 17 .IO Region fJ 13.30 10.59 EJ 23,: Region r, 8 .93 1.54 Et 10.41 -Aogion Elf 12.48 6.911 Elf 19_4j -Aogion E11 s.1a 11.2S E11 20. 0 Aogion f1 6 .04 5.33 F1 11. 3 Aogion Fl 3.116 14.71 Fl 18.5 Region FJ 6.88 10.56 FJ 1U Region F4 15.116 11.64 F4 21.50 AogionF5 11.74 12.78 f5 24.Sl !logion K 11.13 6.18 K 18.11 llegionfl 5.911 9.11 n 15.81 Aeglonfl 14.79 13.13 Fl 27.'1, Aegion Ft 15.07 13.70 fl 28.n Aegion FIi 10.58 6 .44 F 1 1 17.01 Region G1 111!5 6.11 G 1 20.5! llegionGl 10.60 9 .64 G2 202 AeglonGl 14.50 1024 G3 24.7 AogionG4 7 .79 16.28 G4 24.CII ,_ ,,._,. .. .,. D .IIO o .uu u .uu u.uu 0 .00 0 .00 0 .00 u .uu u22.75 18.36 0 .00 llDO 0 .00 0 .00 0 .00 0 .00 llOO 0 .00 u1.52 1.54 0 .00 0 .00 D .80 0 .00 0 .00 D .00 0 .00 0 .00 A-10.92 10.02 Sllnda'dOMllirl 4.n 190 Wodt' IIIIA fff/A llediln 11.08 10.88 Ptollll>llyO-1.44 1 .32 1381.74 TallllAI S1n1a1iDn1 22.75 UIXiulAI0 .00 llia.nAJ-. 10.47 Avnge Al Smllolions 0 .0138 Ptoll ""'F i g ur e 61. Profitabilit y quoti e nt and profi t g au g e for lo w -profit sc e nario

PAGE 192

184 ==T 175 ::w~=iHtlir,g Or, Sc>ectfy MM1d&tDry Proftt by Concentric:Reijon 1 1 1 < 1 01 1 1 1 Figure 62. Profit and simulation options for medium-profit scenario

PAGE 193

185 T O T AL MAN DA TORY PROF IT PER SIMULA TI ON Ai ,_ -1 ......... ~3 -4 SiffUll!iDft5 ~Sin'Ullllioff7 ~10 T- =-... ""' -A1 41.00 70. 51 A 1 111.51 -A2 49 .00 41.00 A2 ... ... ... ..... Al 122. -M 63.00 .... M 129 _,.. 14.00 41.80 ,.. 1:13. -Al 51.00 51.00 Al 112. .. 21.00 70.00 '" ... .. 51.57 11.00 .. 128.S _.., 11.47 ..... .. 130.1 .. 51. 00 21.00 .. ... --,,.oo 49 .00 .. ... -21.00 81.23 _., ,. .. ..... ., 127.&t --70 .00 63.00 133.00 -4900 42.00 11 .00 --42.00 ... .... 111 .34 .. 1 35-00 ..... 911 125 -c1 14.00 58.33 C1 142. -C2 .D 11.31 C2 147. -Cl 58,00 84.29 C3 120 -CA 13.00 55 .32 CA 111. _.,. 11 .00 75.19 Ci 1N. 1 -a ., .. 51.11 ca 124 _c, 51. 00 14.73 C1 1,._ -ca "' 13.00 a 1:,0 -c, 63.00 17.15 Cl 130c -c,g 03.00 42.00 c,g 105. -C11 35. 00 ..... C11 as -'" n .oo n .os 01 154 -02 73.11 70 .00 02 141 .. ..... ., .oo .. 1115s -04 "1. 27 5723 1H 124 50 .. 14.00 55.81 .. 139.11 .. ..... 42.00 .. 17' _o, 11112 41.00 07 132JI .. 70 .00 70 18 .. 140 7 .. St.75 42.00 .. ~:.: --54-13 14.00 011 -011 70.00 41.(10 01 1 ltt. C _01, 41.10 17.73 012 ,,,.. .. 63.00 131 Ct E1 HM.et -u 42.00 ,u, u 1U1 -(l ..... 75.15 u -~ 17.45 N.00 ~= .. 13. 00 71.IS .. .. 35.29 41.15 .. -f7 ... 00 ts.SI f7 181. 51 ~=:-42.00 ~~7.14_ .. 88. 14 1-12.,-17]10 .. ,~.17 -:;:-_, .. 73.20 41.00 .,. 122.l Aegb'IE1t ..... 51.00 El1 112.4 .. 29.00 noo F1 1~ _., 41.00 N.00 f2 147. .. .. ... 14.00 fl Ill. Allgiaf'lf4 51.00 73.03 f4 129 .. ..... 13.00 ,. 118. .. 35.00 70.00 .. 105. -" 47.30 21.00 n 75. .. 71.12 1100 .. 142.1 .. 29.00 ... 128 C ... 14.00 41.00 F1I 133.00 ~Gt ..... 47 70 G1 10176 _.., 71,75 72.0I GZ 143.11 -Gl 35.00 51.00 Gl : : : -G4 47.51 42.00 G4 1~oo 0 .00 -0 00 ... 0 .00 0 .00 0 .00 -131.81 000 0 .00 0 00 .... 0 .00 0 00 .... 0 .00 -21.00 29.00 0 .00 D .00 0 .00 0 .00 0 .00 D 00 000 O.OD A-5Ul 13.13 -, ... 18.29 -5100 41.00 Pn,-------0= ..... 03.00 7 .83 1.33 IOSZ.21STotalAI~ 131.81........,..AI~ o .oo u-.....AJSffl.llalilJna 91.23Avwap1AI~ OJ)IOI Proft .--Figure 63. Profit a bil ity quotient and profit ga u ge for m e dium-profit scenario

PAGE 194

CONCLUSION The chief distinction that my dissertation has is that it shows that total revenue and total profit can be maximized by selected inclusion or exclusion of submarkets of a larger trade area. A computer application that contains various algorithms is presented that solves for which submarkets are to be included and which excluded Characteristics of the consuming population as well as the delivery business are exogenous inputs to the simulation. Characteristics of the consuming population include population density, rate of consumption and total population per region. Characteristics of the delivery business includes required profit margin, delivery charge per region, intra-regional drive times, one-way drive times from the store, and mandatory profit (for all regions, or set by distance from the home store). The results are sensitive to the specification of the parameters of the simulation. The simulation can be calibrated to closely correspond to real world parameters and trade areas, thereby allowing delivery businesses to identify when to open and serve various submarkets of its larger trade area, and when to exclude those submarkets. This dissertation is the first published work that includes the concept of mandatory profit by submarket. Mandatory Profit is the profit that must be projected before a submarket will be served by a home delivery service. This is an important addition to business management procedures because a main cause of online grocer failures is lack of profitability. The simulation and algorithms presented in this dissertation introduces a tool to allow home delivery services in general, and online grocers in particular, to 186

PAGE 195

correctly assess as well as decrease the risk of offering their business services. The application pre se nted in this di sse rtation can be used to assist business ventures in acquiring venture capital becau se total revenue and total profit can be calculated before the actual expense of s tarting business operations. After the model is calibrated with reasonable values for the parameters of the simulation, the delivery service can calculate the customer base. Actual customer addresses and their geographic coordinates can be input into the simulation. The time in which submarkets turn profitable can be calculated. The simulation can also calculate the number of delivery trucks needed, and the cycles ( time s per day, week, or month) that regions became profitable. Conversely, a delivery service can calculate which submarkets provide insufficient profit to include within active delivery areas. This means that the grocer can choose to defer delivery to those region s until a later time. Po s tponing deliver to some re gio n s from the inception of the delivery initiative is s ignificantly preferable than denying some customers their groceries "occas ionally because the regions in which those customers live are not profitable enough to deliver ( have not met the required mandatory profit). Denying groceries to customers living in a region that has previously received groceries from the grocer can very well result in losing tho se customers for good. Therefore my dissertation, by virtue of its unique mandatory profit functionality, can allow a potential online grocer (o r other similar delivery service) to better forecast the amount of delivery-related employees it must have, more accurately determine the amount of trucks it will need, po ss ibly attain start-up capital more expeditiously, and preempt losing customers through denied orders once the delivery initiative has begun 187

PAGE 196

Therefore, my dissertation should be considered to be a trade area expansion bu s ine ss model and not a routing optimization application such as ESRI's Network Analyst. Sears has made remarkable improvements in tracking assets, including appliances to be delivered; the optimization of delivery patterns ; and the prediction of road volumes ( Environmental Science Research In s titute, 2005a). ArcView Network Analyst i s used as the backbone technology for this logistical route optimization program. Network Analyst's functionality includes route optimization drive time analytics, s horte st path algorithms, and directions for delivery drivers (Enviro nmental Science Research In s titute 2005b). According to ESRI, dispatchers can create routes for the appliance delivery drivers in 10 % of the time that it took them before the implementation of the new GIS initiative. Customer satisfac tion has increased because of more timely and reliable deliveries. Also merchandise can be returned to the store more expeditiously (Environmental Science Re sea rch In s titute, 2005c ). The Network Analyst-facilitated home delivery initiative is called the Enhanced Home Delivery Service ( EHDS). The sys tem automatically assigns drivers and trucks to customers and records drive times, driver stops, and delivery times. EHDS includes algorithms for so lving "vehicle routing problem s within time windows" ( Weigel 1999). EHDS generates report s that include the driver's name, orders delivered late direction reports ( to the customers premi ses), stop reports that include the homes visited and in what order, travel time s including waiting time s, and average stops per route ( Weigel, 1999). 188

PAGE 197

Although the Enhanced Home Delivery Service has sophisticated algorithms for minimizing di sta nce traveled and maximizing customer sat i sfact ion, the application doe s not allow Sears to predetermine how profitable a delivery mu s t be in order to defray the costs of delivery such as wages, fuel charges, and depreciation on the vehicles and operational equipment. The ESRI product doe s not allow the user to solve for tho se submarkets of a larger trade a rea that can be profitably se rved This is a differentiating factor of my application in that the m a ndatory profit ( that which much be attained by a run to a region ) i s not possible to mandate with the EHDS sys tem. There are significant factors that differ from delivering groceries as opposed to delivering big-ticket appliances. One s uch factor is that many large appliances easily run into hundreds, if not thousands of dollars. Therefore the nece ss ity of calculating the predetermined profit for delivery of big ticket item s s hould be le ss than th a t of figuring the profitability of bags of groceries worth tens of dollars. In other words, a refrigerator sold at $800 may already reflect a 40 to $ 100 profit. Also, the EHDS sys tem does not allow for the input of delivery charges in the calculation of fees earned by the delivery of an appliance. But thi s i s of course attributable to Sears offering free delivery of it s appliances. All costs are included in the base charge of the groceries to the customers. These charges include, but are not limited to wages and salaries; vehicle fuel costs; utilities for the home s tore (e lectric, gas, water); taxes incurred on income; taxes incurred upon the purcha se of land, fixed machinery, structures, computers and computer software, and more ; depreciation (wear and tear) of vehicles and fixed machinery; wholesale and retail cost of groceries to the grocer; insurance; repairs to machinery and vehicles; services incurred such as for outsourcing and services (legal, accounting, etc .); employee health 189

PAGE 198

care s ubsidizing ; registration costs of vehicles; s tart up costs including s tore shelves and associated structures, refrigerators, ovens, employee uniforms ; advertising; and more. After acco unting for all of these initial and ongoing costs, a percent profit i s derived. This should be considered similar to the way a brick and mortar grocer calculates the percent profit (excep t a brick and click grocer would have the additional cost of delivery truck s and delivery driver wages, whereas a pure online grocer would not have initial costs of cash re g ister s and ins tore advertising or sales personnel). For example, after the store calculates all of the above costs and finds that for every dollar of groceries so ld 98.5 cents are costs, the profit margin is 1.5% If the grocer decides that a delivery charge should be added to thi s, the grocer can then choose to employ the delivery charge function in my a pplication to see when a nd how often the regions s imulated will reach the mandatory profit. The grocer can then choose to run s imilar si mulation s with a zero delivery charge. After both of the se s imulation sequences are run, the grocer could then juxtapose the result s of the self-generated report s to see which business model (t hat which uses delivery charges, or that which uses a zero delivery charge) i s more feasible considering both customer satisfaction and overall profitability. It i s also pos s ible to enter smaller and larger amounts of delivery charges to ascertain a customer sat i sfac tion ( in terms of timely grocery delivery)/overall profit dichotomy. This is not to say that thi s profit-related business model is the only way to incorporate cost/ profit into the overall model. Another possible way would to be to include all operational expenses into the grocery charge, then compensate for the costs incurred by delivering groceries by adding on a delivery charge to some or all orders placed. Because my application allows for delivery charges to be added on to orders by a 190

PAGE 199

regional basis, and s ince orders sent to more distant regions are more costly, the grocer can compensate for the graduated delivery costs (less for closer region s, more for further regions ) by applying a higher delivery charge to customers who live farther away from the store, thereby creating a correlation between delivery costs and delivery charges. To make my logi s tic application more attractive to grocers I would add the following functionality to the application Metamorphic zones Handoff point algorithms Better database co nnectivity (to run queries) A user interface that allows point and click se t up of the hex ago ns to conform to the trade area of the grocery store Overlay the hex ago ns in item 4 above onto a air photo or sa tellite image Delivery regions s hould not be s tagnant or immutable polygons. Ins tead delivery zones s hould be metamorphic entities. Metamorphic delivery zones s hould encompass the most profitable addresses that require delivery to at any particular ins tant. Bes ides profit maximization functionality metamorphic real time delivery zones should minimize delivery cost and delivery times. To elucidate, if a si ngle region ha s met its mandatory profit (say region X), and regions geographically adjacent to that region (regions Y and Z ) have not yet met their mandatory profit ( provided those regions have at lea s t one outstanding undelivered order), the sys tem would automatically connect region s X, Y, and Z together and inform the dispatcher to send a single truck to the three regions. Factors to include in this decision are whether, historically regions Y and Z have not met their mandatory profit in a cyclical manner that is chronologically near to the 191

PAGE 200

dispatch of the truck to region X If past customer behavior shows, beyond a reasonable statistical doubt that regions Y and Z will meet mandatory profit shortly, it might be preferable to have the truck make a run to only region Z, return to the store quickly and load up again with the full amount of orders for regions Y and Z, which by that time will have met their mandatory profit, and would result in higher earnings for the grocer. Metamorphic regions do not have to be adjacent. Relatively distant, but as yet non profitable regions could be combined by joining regions between the two regions to create a "corridor of profitability. Even though my application has the potential to greatly decrease the risk of not making a profit on a delivery dispatch, the ma x imization of the profit potential of a particular run along with the problem of reaching each and every customer has not been adequately addressed yet. One reason that profit earning potential is not maximized in my application is that regions are rigidly defined geographic areas. What will happen using only rigid delivery regions is that a specific delivery truck might deliver to all the customers within a zone, then either go to the next zone that has outstanding orders or return to the home store. Although the route traversed on the way to the zones and the route traveled within the zones can be minimized by common GIS routing functions, no consideration is given to the optimization possibilities that could exist by combining or breaking up the zones. It could be the case that zone #23 and zone# 24 are adjacent. Both zones have surpassed their profitability thresholds. But on this particular delivery dispatch 70% of the customers in zone 24 are located in the eastern part of the zone In this case, by not adhering to strict hexagonal areas, optimization possibilities exist. If the total amount of 192

PAGE 201

deliveries to zones 23 and 24 exceed the capacity that one truck can carry, the simplest strategy might be to first deliver to one of the zones, come back to the store restock the truck, and deliver to the other zone. After all, both zones reached mandatory profit. However if the truck has enough capacity to deliver to all of the eastern part of zone 24 (which contains 70 % of the orders for that zone) plus enough capacity to deliver to all of zone 23, the profit for that specific deli very could be 70 % greater than if it delivered to region 24, returned home, then went to region 23. What about the remaining 20% of the customers in the western part of zone 24 that haven t received their groceries yet? There has to be some probability that those customers can be grouped with another zone that lies somewhere to the west of zone 24. If so, those remaining customers could be serviced with a truck that goes to that western area. Of course this type of dynamic routing is not as easy as it seems. What if the customers in the western part of zone 24 were promised their groceries at the same time as those in the eastern part? They would be very unhappy is they had to wait for the next truck. This means that rigidly defined zones are not the most efficient geographic areas to deliver to. The key is to have a logistical application with the capability to morph delivery zones to whatever shape is necessary at a particular time while taking into consideration numerous variables. Possible shapes could be elongated, finger shaped areas, or even archipelago-shaped areas. Delivery archipelagos may very well be the most common type of delivery delineation because of their profitability potential. Each "island" in the archipelago might just be a small fraction of a region (maybe one or two blocks). The application should dynamically generate and delineate optimal "island 193

PAGE 202

hopping" delivery dispatches instead of just showing blocks of zones that have met mandatory profit. Succinctly stated, it may be much more efficient and profitable to deliver to elongated or archipelago shaped zones instead of rigid hexagonal zones. One important reason for this postulation is that instead of waiting for zones of a predetermined shape to reach the profitability threshold, the online grocer can be delivering to irregular shaped zones continually because the probability of an indeterminately shaped region reaching mandatory profit can be significantly greater than a predetermined region reaching that same threshold, even if both the indeterminately shaped region and predetermined region are of the same area (as measured in square miles, for example). The only caveat to this rule is that any indeterminate region be allowed to assume the exact shape and location as any predetermined region in the universal region. Also, because more customers would be delivered to using metamorphic zones, customer satisfaction will increase. Moreover, customers who are denied orders may be lost for good. Metamorphic regions can have the potential to "grab" these otherwise un served customers and include them in the overall profitability plan. My application allows for the grocery store to evaluate which regions to "open" and serve, and which not. Once a decision is made to serve, they must serve regardless of the profit. So a merchant can use my application to assist in making a decision of when to open regions. Various conclusions can be derived from the concept of attempting to keep the profit level, as high as possible on a delivery. The first conclusion is that the metamorphic 194

PAGE 203

delivery zones will change often, if not constantly, as new orders come in and old orders are delivered. A second conclusion is that by using archipelago-shaped zones, the profit level can become much higher than the usage of predetermined zones. This is accomplished by including orders from regions that might not have met the mandatory profit if the regions were stagnant shaped hexagons. The point being made here is that by allowing the logistic application to dynamically create polymorphous regions, the profit per delivered route can increase along with the amount of customers delivered to. The inclusion of handoff points, or the locations of where trucks in the field receive groceries from other trucks in the field, is important. The attainment of this functionality would necessitate the inclusion of shortest-distance networking algorithms and forecasting functionality. The forecasting functionality would determine historically what addresses ordered what groceries in the past, and when they did so. Larger trucks in the field could be pre-loaded with these goods and meet with other trucks so they do not have to return to the store. This is a very important function that should be incorporated. The application that I developed is based on using ExcelTM spreadsheets as the back-end database The application would be much more powerful if the data was in an Access, 0B2, or Oracle database. This is because sophisticated queries could be run to answer questions such as what regions become profitable when, what customers have waited the longest to receive their groceries and what trucks are most efficient in traversing the network. Therefore, this is the third aspect I will improve on when I acquire more programmers. 195

PAGE 204

My application consists of a predetermined arrangement of hexagonal regions. Thi s s hape will have to be a ltered t o create an appropriate overlay on a real town, or sect ion of a town Therefore the ability to drop and drag the hex ago n s to whatever shape is nece ssary i s an important improvement that mus t be made The region s s hould also be able to be manipul ated over air photos and sa tellite image s so the underlying topography streets, and buildings could be better s hown under the regions. The reason I programmed it using VBA in ExcelTM i s that Excel i s practically a ubiquitou s app licati o n A preponderance of companies are familiar with the a pplication which could result in a s horter learnin g curve for companies that want to programmatically change the application ExcelTM i s a l so equipped with s preadsheets that can easily be u sed for a back-end database. Visual Basic for Applications, despite havin g its limitati o ns, i s still a quite powerful programming language. For reasons of demonstrating the potenti a l utiliz a tion of my logistic application, the Excel / VBA combination was s ufficient. It is recognized however that other languages, such as Visual Basic, C++ or Java would be preferable, because the sim ulation s would run faster. However a main purpose of mine was not to have the application run fast, but instead have it run s lower so those viewing it could better see the trucks being dispatched, the region s changing color in concordance with the number of orders generated, and the regions meeting their respective mandatory profit. For these purpo ses a s l ower running application has its benefits. The main weaknesses of my model i s that it does not contain the functionality for metamorphic regions, handoff points, and better databa se connectivity. It also is not 196

PAGE 205

optimized for speed, which should be changed by re-programming it in a higher-level programming language. The main contribution that my dissertation has contributed to academic knowledge is my application automates trade areas. Thrall (2002b) wrote about trade area expansion and drew upon the actual geographic expansion of Red Lobster out of central Florida. Thrall discussed the costs and benefits of a multi-branch retailer choosing various expansion strategies, including hierarchical dispersed versus spatially contiguous regional saturation. However Thrall's discussion while seminal did not provide an operational algorithm or procedure that allowed the decision maker to calculate the differences in profit between alternative geographic expansion scenarios. My dissertation fills this gap in the literature by taking trade area analysis the next logical step. My dissertation contributes to the academic body of knowledge by answering questions like, "what markets and submarkets should be included to insure profitability by delivering groceries?" and "what values do you need in combination to reach a predetermined profit?" These questions have not been adequately answered either in academic literature or in industry. Because profitability is naturally the goal of any grocer, this application gives an indication to the grocer about what clusters or sets of attributes will work to achieve the required level of profit. Each simulation that is run can portray a different scenario of what is needed to achieve profitability. These facets are lacking in the realm of current academic knowledge. The best way to optimize the profit ascertaining potential of my application would be to incorporate the metamorphic zone functionality stated above. Aside from that, goal-197

PAGE 206

seeking functionality would have to be added. To do thi s, region s would have to be specified and the mandatory profit would be calculated by the amount of profit a truck earns on a run to more than one region. For example, a truck is dispatched with an extra amount of groceries. While the truck i s in the field, its route changes according to the location of the customers who are constantly plac ing orders. In this sce nario the trucks are the dependant variables, and the customers are the independent variables. A grocer who delivers would certainly hope for inela s tic dema nd as it pertains to delivery charges Thi s is because if the amount of groceries ordered will decrease a t a decreasing rate in proportion to the increase of delivery charge. [n thi s sce n a rio the grocer could be relatively ass ured that a s mall increase in delivery charge would result in a s maller monetary decrease in grocerie s ordered. Becau se food is a n o n-di sc r e tionar y item meaning that people h ave to eat, food ha s a propen s ity to have an inelastic demand assoc iated with it. However, the choice to have the food delivered i s discretionary, giving some impetu s to elastic demand with regards to home delivery of groceries. Therefore grocers s hould strive to keep the delivery charge as minimal as po ss ible, or attempt to offset the deli very charge with excellent se rvice. Another point th a t could se rve to offset the elasticity of demand when it comes to delivery service that it does cost customers money, in fuel, wear and tear on their vehicles and time to s hop. [f the cost of delivery does not s uper sede these three cost point s then the elasticity of demand in regard s to deli ve rie s could be effectively nullified. Although [ sa id much about my logistical application being a s imulation it i s really more than that. After a grocer uses my application to intelligently decide whether to 198

PAGE 207

begin delivering groceries or not, he or she can use the application as an operational front end to actually dispatch the trucks to the regions. Of significant importance is the value this application can have to the failure prone grocery delivery bu iness. With judicious usage of this application grocers who might otherwise fail to adequately make a profit can predetermine the plausibility of success before capital is invested in delivery operations. Therefore, the application that I developed is not merely a concept, but a unique working application that has yet to be introduced into the decision-making realm of prospective, and already operational egrocers. In other word the Main Map interface can be used as a logistical dashboard where orders that come in are automatically input into the database, causing the logistical dashboard to notify the dispatcher to send a truck to the regions that have met their predetermined profitability levels. This will of course require connecting the application to a back end web interface. 199

PAGE 208

SOURCES CITED Abel, D.; Gaede V.; Taylor, K.; and Zhou, X., "SMART: Towards Spatial Internet Marketplaces, Geoinformatica, Vol. 3, No. 2, pp 141-164, 1999 Adami!, C., What is Complexity? Bioessays, Vol. 24, No. 12, pp 1085-1094 2002. Adaptive System ," Management Science Vol. 46, No. 11, pp. 1427-1440, November, 2000. Akwawua, S., and Pooler J ., The Development of an Intervening Opportunitie s Model With Spatial Dominance Effects," Journal of Geographical Syste m s," Vol. 3, pp 69-96, 2001. Allan, R., and Lowell K., Managing Spatial Uncertainty Using Attribute, Geometric and Neighborhood Measures in an Empirical Rule-Ba se d Model Approach," Journal of Geographical Systems Vol. 4, pp 141-156, 2002. Ambrosio, Johanna. June, 2002. Web Services: Report from the Field. Application Development Trends. Vol. 9, No, 6. pp 25-28. Anselin, L., "Computing Environments for Spatial Data Analysis," Journal of Geographical Systems Vol. 2, pp 201-220, 2000. April, Carolyn; Harreld Heather. September 2, 2002. "Information Aggregation." InfoWorld, pp 33-34, volume 24, issue 35. Arentze, T., and Timmermans, H., Measuring Impacts of Condition Variables in Rule Ba se d Models of Space-Time Choice Behavior: Method and Empirical lllustration, Geographical Analysis, Vol. 35, No. l 2003 Atkinson, P., Surface Modeling: What's the Point ?" Tran sac tion s in GIS Vol. 6, No. 1 pp l-4 2002. Atkinson P., and Tate, N., Spatial Scale Problems and Geostatistical Solutions : A Review ," Profe ss ional Geographer, Vol. 52 No 4, pp 607-623, 2000. Baker, Jeanne; Smith, Howard ; Fingar Peter. July 2002a. BPML at Work: A Case Study." Internet World p 29. Baker, Jeanne; Smith Howard; Fingar Peter. July, 2002b. "Integrated Value Chains. Internet World, pp 24-29. 200

PAGE 209

Baker, V., "On Modeling Internet Transactions as a Time-Dependent Random Walk: An Application of the Retail Aggregate Space-Trip (RASTT) Model," GeoJournal, 53: pp 407-408, 200 I. Balakrishnan, A.; Natarajan, H. ; and Pangburn, M. "Optimizing Delivery Fees for a Network of Distributors ," Manufacturing & Service Operations Management, Vol. 2, No. 3, pp 297-316, 2000. Bao, S.; Anselin, L.; Martin D.; and Stralberg, D., "Seamless Integration of Spatial Stati s tics and GIS: The S-PLUS for ArcView and the S+Grassland Links ," Journal of Geographical Systems Vol. 2, pp 287-306, 2000. Bardinet C, and Bournay, E., The Use of 30 Mapping in Geological Research and Risks Analysis: Evaluation of a Water Supply Project in the Kathmandu-Melamchi Area," GeoJournal 49: pp 159-163 1999. Barnes S., "Business GIS Flows to the Mainstream ," GeoSpatial Solutions, Vol. 12, No. 7, p 12, 2002. Benenson I.; Sofer M.; and Schnell, I., Geographical Information System s as a Visualization Tool, Applied Geography, Vol. 18, No. 2, pp 117-135 1998. Bergougnoux, P., "Editorial: A Perspective on Dynamic and Multi-Dimensional GIS in the 21Century," Geolnformatica, Vol. 4, No. 4 pp 343-348, 2000a. Bergougnoux, P., "Editorial: Geographic Information Systems and Intelligent Transportation Systems, Geoinformatica, Vol. 4, No. 2, pp 123-125 2000b. Bernard L., and Kruger, T., "Integration of GIS and Spatio-Temporal Simulation Models: Interoperable Components for Different Simulation Strategies," Transactions in GIS, Vol. 4, No. 3, pp 197-215, 2000. Bernard L.; Schmidt, B.; Streit U.; and Uhlenkuken C., Managing Modeling, and Visualizing High-dimensional Spatio-temporal Data in an Integrated System, Geoinformatica, Vol. 2, No. I pp 59-77 1998. Bertolotto, M., a nd Egenhoffer M., "Progressive Transmission of Vector Map Data Over the World Wide Web, Geolnformatica, Vol. 5, No. 4, pp 343-373, 200 I. Bittner, T., Frank A., On the Design of Formal Theories of Geographic Space ," Journal of Geographic System s," Vol. I pp 237-275, 1999. Black W., "Spatial Interaction Models Using Artificial Neural Networks," Journal of Transport Geography ," Vol. 3, No. 3, 1995. 201

PAGE 210

Black, W. "Transport Route Location A Theoretical Framework ," Journal of Transport Geography, Vol. I, No. 2, 1993. Blacka, W., "Spatial interaction modeling using artificial neural networks," Journal of Transport Geography, Vol. 2, No. I, pp 3-18, 1994 Borchert J., Spatial Dynamic s of Retail Structure and the Venerable Retail Hierch y," GeoJournal, Vol. 45, pp 327-336, 1998. Borger s, A; Hofman, F.; Ponje, M.; and Timmermans, H., 'Towards a Conjoint-Based, Context-Dependent Model of Task Allocation in Activity Settings: Some Numerical Experiments," Journal of Geographical Sys tems Vol. 3, pp 347-367, 2001. Brainarda, J.; Lovetta A.; and Batemana, I., "Using i soc hrone s urfaces in travel-cost models," Journal of Transport Geography, Vol. 5, No. 2, pp 117-126 1997. Briody, D., Grocer s Stock Online Shelves, lnfoWorld, Vol. 22, Iss ue 17, 2000. Brown, L., The GIS/SA Interface for Subs tantive Re sea rch (ers): A Critical Need," Journal of Geographical Sys tem s, Vol. 2, No. 1 pp 0043-0047, 2000. Browne, M., "Logistics Strategies in the Single European Market and Their Spatial Consequences," Journal of Tran s port Geography, Vol. l No. 2, pp 75-144, 1993. Bubny P., "Not Yet Clicking ," Supermarket Busine ss, Vol. 55, No. 7, pp 65-66, 2000. Buor, D., Distance as a Predominant Factor in the Utilization of Health Services in the Kumasi Metropoli s, Ghana," GeoJournal, Vol. 56, pp 147-157 2002. Campbell, J.; Labelle A.; and Langevin, A., "A Hybrid Travel Distance Approximation for a GIS-Based Decision Support System," Journal of Business Logi s tics, Vol. 22, No. 2, pp 165-182 200 l. Cartwright W ; and Hunter G., "Towards a Methodology for the Evaluation of Multimedia Geographical Information Products ," Geo Informatica, Vol. 5, No. 3, 200 l. Casti, J. Rooting Out Randomness, Complexity, Vol. 6, No. 4, 2001. Cheng, T., and Molenaar, M. Diachronic Analysis of Fuzzy Object s," Geolnformatica, Vol. 3, No. 4, pp 337-355, 1999 Christaller, W (1933) Die zentralen Orte in Siiddeut sc hland. Jena: Gustav Fi s her. Claramunt, C., and Jiang, B., "An Integrated Representation of Spatial and Temporal Relationships Between Evolving Regions," Journal of Geographical Systems, Vol. 3, pp 411-428, 200 l. 202

PAGE 211

Clarke, G. "Changing Methods of Location Planning for Retail Companies, GeoJournal, Vol. 45, pp 289-298, 1998 Cohn, Michael. August, 2002. TaylorMade-Adidas Takes a Swing at Wireless." Internet World, pp 32-34. Cressie, N.; Stern H.; and Wright, Deanne, Mapping Rates Associated with Polygons, Journal of Geographic Systems, Vol. 2, pp 61-69, 2000. Cross, R., Neal, M., Grocery Shopping Goes Portable," Direct Marketing Vol. 62, No. 11, pp 42-45, 2000. Crossland, M.; Wynne, B. ; and Perkins W., "Spatial Decision Support Systems: An Overview of Technology and a Test of Efficacy, Decision Support Systems Vol. 14, pp 219-235, 1995 Deitel, H; Deitel, P; DuWaldt, B; Trees, L. Web Services A Technical Introduction. New Jersey: Prentice Hall 2002. Dettelback, Bill. July 2002. "Why You Need ODS. Internet World. Pp 50-52 DolGicer, Max, Bayer, Gerhard Bardash, Michael. July 2002. Servelets and EJBs: Friends or Foes ?" Application Development Trends. Vol. 9, No. 7 pp 40-45. Duckham M.; Drummond, J.; and Forrest, D., Spatial Data Quality Capture Through Inductive Learning ," Spatial Cognition and Computation, Vol. 2, pp 261-282, 2000. Eckerson, Wayne. July, 2002. "Analytics: Hang the Imposter! Application Development Trends. Vol. 9, No. 7 pp 26-27. Environmental Science Research Institute. (2005a, June 29). GIS for Routing and Logistics. [Online. Accessed December, 2005], Available: http://www.esri.com/industries/business/busi ne ss/del i very routing.html Environmental Science Research Institute. (2005b, October 26). ArcGIS Network Analyst. [Online. Accessed December, 2005], Available: http://www.esri.com/software/arc gis/ex tensions/networkanal yst/index. html Environmental Science Research Institute. (2005c, December 6). ESRI International User Conference 2005. [Online Accessed December, 2005], Available: http://events.esri.com/uc/OandA/index.cfm?fuseaction=printall&ConferencelD=DA4945 55-C04F-A070-2407CB34C9CB9287 Erwing, M.; Hartmut, R.; Schneider M.; and Vazirgiannis, M., Spatio-Temporal Data Types: An Approach to Modeling and Querying Moving Objects in Databases," 203

PAGE 212

Electronics and Telecommunications Research Institute. 2003. R and D Areas. [Online. Accessed June 2004], Available: http://www.etri.re.kr/english/tec eco/e rnd/rndl.htm Etzelmuller, B., On the Quantification of Surface Changes Using Grid-based Digital Elevation Models ( DEM s)," Transactions in GIS Vol. 4, No. 2, pp 129-143 2000. Evaluating central point and median approaches," Journal of Geographic Systems, l: pp 367-383, 1999 Evans, B., Webvan: Who's to Blame ?" InformationWeek Issue 846, p 80, Jul 16, 2001. Ewalt, D., Built-in Analytics: Good, but not Alway s Better ," lnformationWeek, Issue 918, p 20, Dec 9, 2002. Ferguson Renee. July 8, 2002b. JOE Opening ERP Apps for Integration. Eweek, volume 19, number 27, pp l and 12. Ferguson, Renee. July 22, 2002a. "Specs Unite Partners ." Eweek Vol. 19, No 29, p 31. Fischer, M. Learning in Neural Spatial Interaction Models: A Statistical Perspective, Journal of Geographical Sy s tems Vol. 4, pp 287-299. Fischer, M., and Reismann M., Evaluating Neural Spatial Interaction Modeling by Bootstrapping ," Networks and Spatial Economics, Vol. 2, pp 255-268, 2002. Fischer Peter. July 2002. "Answering the Critical Web Service s Que s tions ." Application Development Trends, Vol. 9, No. 7, pp 51-57 Fleming D., Spatial Characteristics of Transportation Hubs: Centrality and lntermediacy, Journal of Transport Geography Vol. 2, No. l, pp 3-18, 1994. Foley John. September 16, 2002. "Real Time ." Informationweek, No. 906, pp 35-42. Fontana John. July 22, 2002. "Use r s, Vendors Try to Manage Web Services ." NetworkWorld, Vol. 19, No. 29, pp 28-29. Frihida A.; Marceau D.; and Theriault, M., "Spatio-Temporal Object-Oriented Data Model for Disaggregate Travel Behavior ," Transactions in GIS Vol. 6, No. 3, pp 277294, 2002. Frye, Colleen. October 2002. "IT Still Spending on Portals. Application Development Trends, Vol. 9, Vol. LO, pp 45-49. Gahegan, M., The Case for Inductive and Visual Techniques in the Analysis of Spatial Data ," Journal of Geographical Systems Vol. 2, pp 77-83, 2000. 204

PAGE 213

Goodchild, M., "GIS and Transportation: Status and Challenges," Geolnformatica, Vol. 4, No. 2, pp 127-139 2000. Goodchild, M., "Bringing Geographical Information Systems into Business," Economic Geography Vol. 78, No 4 pp 519-521 2002. Gordon, D., "The Organization of Work in Social Insect Colonies, Complexity Vol, 8, No. 1, pp 43-46 2002. Gorr, W., Johnson M. Roehrig S. "Spatial Decision Suppo1t System for Home Delivered Services," Journal of Geographical Systems, Vol. 3, pp 181-197, 2001. Gould, Larry. (200 l, October 26). Looking at Logistics. Automotive Design and Production. [Online Accessed March, 2004], Available: http://ww w.a ut ofie ldguid e.co m/ a rticl es / 110 I OS.html Grimshaw, D., Geographical Information Systems: A Tool for Business and Industry?" International Journal of Information Management, Vol. 9, pp 119-126 1989. Grimshaw, D., Geographical Information Systems as Part of the Corporate Information Strategy," International Journal of Information Management, Vol. 11, pp 292-297, 1991. Grimshaw, D., Harnessing the Power of Geographical Knowledge: the Potential for Data Integration in an SME," International Journal of Information Management, Vol. 21, pp 183-191, 2001. Grnmbach S.; Rigaux P.; Segoufin, L.; "Spatio-Temporal Data Handling with Constraings, Geo Informatica, Vol. 5, No. I, pp 95-115, 200 l. Guy, C., Classifications of Retail Stores and Shopping Centres: Some Methodological Issues, GeoJournal Vol. 45, pp 255-264 1998. Hacker, Joan. March, 2002. Are Those Parts In? We Think So .. Frontline Solutions, Vol. 3 No. 3 pp 17-22. Hahnke, Julie ; Sullivan Dan. May, 2002. "Lets Share, Application Development Trends Vol. 9 No. 5 pp 33-37. Haining, R.; Wise S ; and Ma, J., Designing and Implementing Software for Spatial Statistical Analysis in a GIS Environment," Journal of Geographical Systems, Vol. 2, pp 257-286, 2000 Hakala, D., "Location: The New Killer App," V ARBusiness, Issue 1902, p 18, 2003. Harreld, Heather; Krill, Paul; Schwartz, Ephraim. July 22, 2002. "V ANs Shift EDI Layer." lnfoWorld. Vol. 24 No. 29. pp 31-32. 205

PAGE 214

Haughton, M., "Measuring and Managing the Learning Requirements of Route Heinritz, G., "Methods in Retail Geography," GeoJournal, Vol. 45, pp iii-iv, 1998. Hernandez, T.; Bennison, D.; and Sarah, C., "The Organisational Context of Retail Locational Planning ," GeoJournal, Vol. 45, pp 299-308, 1998 Heron, R Penny, G., Paine, M., Sheath, G., Pedersen, J. Botha, N., "Global Supply Chains and Networking: A Critical Perspective on Learning Challenges in the New Zealand Dairy and Sheepmeat Commodity Chains, Journal of Economic Geography, Vol. l, pp 439-456, 200 l. Heun C ., "Delivery, anyone?" Information Week, Issue 846, pp 22-24, 200 l. Hickman, C. Revolutionizing the business through GIS," Transmission & Distribution World, Vol. 51, No 11, pp 82-85, September 1999. Hu, P., and Pooler, J ., An empirical test of the competing destinations model," Journal of Geographic Systems," 4: pp 301-323, 2002. Huang, B., Lin, H., Design of a Query Language for Accessing Spatial Analysis in the Web Environment, Geoinformatica, Vol. 3, No. 2, pp 165-183 1999. Huang, Y ; Jing N. ; and Rundensteiner, E., "A Hierarchical Path View Model for Path Finding in [ntelligent Transportation Systems," Geolnformatica, Vol. l, No 2 pp 125159 1997. Hunter, G. "Unders tanding Semantics and Ontologies: They're Quite Simple Really If You Know What I Mean ," Transactions in GIS, Vol. 6, No. 2, pp 83-87, 2002. [VU, October 6, 2000. Filiallnfo: Geolnformation Flexible and Timely. IVU Presents Efficient Location Management." [Online. Accessed April, 2004], Available : http ://www. i v u d e/cms/cms seite s u c h e n .pl? seiten id= I 6 2 l e n Keenan P. "Spatial Decision Support Systems for Vehicle Routing, Decision Support Systems, Vol. 22 pp 65-71 1998 Killgren, L., "Online Grocery Sales to Ripen," Marketing Week, Vol. 22 No. 26 pp 2829, 1999 Kim, E. and Eom, S. "Designing Effective Cyber Store User Interface, Industrial Management & Data Systems, Vol. 102 No. 5 pp 241-251 2002. King, L. Central Place Theory, Beverly Hills : Sage Publications 1984 Konicki, Steve. July l, 2002. "Shopping for Savings," Informationweek, pp 37-45. 206

PAGE 215

Koninger, A, and Bartel S. "3D-GIS for Urban Purposes ," Geolnformatica, 2: 1, pp 79-103 l 998. Kuhel, Jennifer. July/August, 2002a. Sharing the Load." Supply Chain Technology News. Vol. 4, No. 6. pp 20-24 Kuhel Jennifer. September 2002b. A Formula for Logistics Succe ss." Supply Chain Technology News, Vol. 4, No. 7, p I and 11. Lenntorp, B. "Time-Geography at the End of its Beginning," GeoJournal Vol. 48, pp 155-158, 1999. Leung, Y., and Yan J., Point-in-Polygon Analysis Under Certainty and Uncertainty," Geolnformatica, Vol. I pp 93-114, 1997. Lewis Jamie; Blum, Dan. Virtual Enterprise Networks." Technical Suport Vol. 10, No. 9, pp 18-22 Lindquist, P., Visualizing Alternative Model s and their Objective Function s in the Solution of Optimal Location Problems ," Journal of Geography, Vol. l O l No. 2, 2002. Litwack, David, Fingar, Peter. June, 2002. "In the Fast Lane. Internet World. Pp 22-24. Loffler, G., "Market Areas -a Methodological Reflection on Their Boundarie s," GeoJournal, Vol. 45, pp 265-272, 1998. Lowe, J ., "Technical Trends in the Geospatial Industry ," Geospatial Solutions Vol. 13, No. l pp 33-37, 2003. Luoma, M.; Mikkonen K.; and Palomgki M., 'The Threshold Gravity Model and Transport Geography ," Journal of Transport Geography Vol. 1 No. 4, pp 240-247, 1993. Mahnke, R., and Kaupuz s, J., Probabilistic Description of Traffic Flow ," Networks and Spatial Economic s, Vol. l pp 103-136 200 l. Malone, Robert. June, 2002. "Elite System Boosts Supply Chain Visibility." Inbound Logistics. Vol. 22, No. 6. pp 12-13. Marble, D., "Some Thoughts on the Integration of Spatial Analysis and Geographic Information System s," Journal of Geographical Systems, Vol. 2, pp 31-35, 2000. Martin, R., and Sunley P., "Deconstructing Clusters: Chaotic Concept or Policy Panacea?" Journal of Economic Geography, 3: pp 5-35, 2003. 207

PAGE 216

Maruca, S., and Jacquez, G., Area-Based Tests for Association Between Spatial Patterns ," Journal of Geographical Sys tem s, Vol. 4 pp 69-83 2002. Maselli Jennifer. June 24 2002. "Better Data Analysis Meets Customers' Demands ," lnformationweek, No 894. p 40. Mazen, S., and Jon s on G., "Package Handling Evaluation Methods in the Grocery Retail [ndustry," Packaging Technology and Science, Vol. 14, No. 1 pp 21-29, 200 I. McKinnon A. and Tallam, D.; Unattended Delivery to the Home: an Asse ss ment of the Security Implic a tions ," International Journal of Retail and Distribution Management, Vol. 31, No. I, pp 30-41 2003. Medda F., and Nijkamp, P., Urban Land Use for Transport Sy s tems and City Shapes ," Geographical Analy s i s, Vol. 35 No. 1 200 I. Miller H., Geographical Representation in Spatial Analy s is, Journal of Geographical Sy s tem s Vol. 2, pp 55-60 2000a. Miller H., and Wu Y., GIS Software for Measuring Space-Time Accessibility in Transportation Planning and Analysis ," Geoinformatica, Vol. 4 No 2, pp 141-159 2000b Mitchell A., Home Shopping Boom a Threat to Market Leaders," Marketing Week, Vol. 22 No. 9 pp 26-27 1999. Morri s, A., A Framework for Modeling Uncertainty in Spatial Databases, Tran s actions in GIS Vol. 7 No 1 pp 83-101 2003. Murayama Y., The Contribution of GIS to Geographical Re s earch ," GeoJourna1 Vol. 52 pp 163-164 200 I a. Murayama Y., Geography With GIS ," GeoJournal Vol. 52, pp 165-171 200 lb. Murray, A., Spatial analy s i s using clu s tering methods: Mushinski, D., Weiler S., "A Note on the Geographic Interdependencies of Retail Market Area s," Journal of Regional Science Vol. 42 No 1 pp 75-86 2002. Nasarin S., and Birks D., DSS Implementation in the UK Retail Organizations: a GIS Perspective, Information and Management, 40: pp 325-336 2003. Nash Kim. June 2002. Slimming Fast, Baseline. No. 7. pp 63-64. Niraj R.; Gupta, M.; and Narasimhan C., "Customer Profitability in a Supply Chain," Journal of Marketing, Vol. 65, No. 3, pp 1-16, 2001. 208

PAGE 217

Noke}, K., and Schmidt, M., "Parallel DYNEMO: Meso-Scopic Traffic Flow Simulation on Large Networks," Networks and Spatial Economics, Vol. 2, pp 387-403, 2002. Noronha, V., "Towards ITS Map Database Interoperability-Database Error and Rectification," Geolnformatica, Vol. 4, No. 2, pp 201-213, 2000. O'Kellya, M. "A Geographer's Analysis of Hub-and-Spoke Networks," Journal of Transport Geography, Vol. 6, No. 3, pp 171-186, 1998. O'Malley, L., Patter so n, M., Evans, M. Retailer Use of Geodemographic and Other Data Sources: an Empirical Investigation," International Journal of Retail & Distribution Management", Vol. 25, No. 6, pp. 188-196, 1997. Okabe, A. and Funamoto, S. "An exploratory method for detecting multi-level clumps in the distribution of point s a computational tool, VCM (variable clumping method)," Journal of Geographic Systems," 2: pp 111-120 2000. Papadopoulos, A., and Manolopoulos, Y., "Nearest Neighbor Queries in Shared Nothing Environments," Geolnformatica, Vol. l No. 4, pp 369-392, 1997. Partch, K., "Home Delivery, Another Problem Solved," Supermarket Business, Vol. 54, No. 8, 1999. Partch K. "The Trouble With Online Grocery," Supermarket Business, Vol, 56, No. 8, p 74, 2001. Peine} G. and Rose, T., Graphical Information Portals The Application of Smart Maps in GeoNet 4D," Geolnformatica,Vol. 5, No. 4, pp 327-344, 200 I. Peter, L.; Popkowski, L.; and Timmermans, H. "Unconditional and Conditional Competing Risk Models of Activity Duration and Activity Sequencing Decisions: An Empirical Compariso n, Journal of Geographical Systems, Vol. 4, pp 157-170, 2002. Peuquet, D., "Making Space for Time: Issues in Space-Time Data Representation, Geolnformatica, Vol. 5, No. 1 pp 11-32, 2001. Pienarr, M., and Van Brake!, P., "The Changing Face of Geographic Information on the Web: A Breakthrough in Spatial Data Sharing," The Electronic Library Vol. 17, No. 6, 1999. Pijanowski, C.; Brown, D.; Shellito B.; and Manik, G, "Using Neural Networks and GIS to Forecast Land Use Changes: a Land Transformation Model," Computers, Environment and Urban Systems, Vol. 26, pp 553-575, 2002. 209

PAGE 218

Pissinou, N. ; Radev I.; and Makki, K.; "Spatio-Temporal Modeling in Video and Multimedia Geographic Information Systems," Geo Informatica pp 379-409, 2001. Pullar, D., MapScript: A Map Algebra Programming Language Incorporating Neighborhood Analy s is," Geoinformatica, Vol. 5 No. 2, pp 145-163 200 l. Punakivi, M., and Saranen J. Identifying the Success Factors in E-Grocery Home Delivery ," lnternational Journal of Retail and Di s tribution Management, Vol. 29, No. 4 pp 156-163, 2001. Punakivi M. and Tans kanen K., Increasing the Cos t Efficiency of E-Fulfillment Using Shared Reception Boxes," International Journal of Retail and Di s tribution Management, Vol. 30, No. 10 pp 498-507, 2002. Pundt, H., Field Data With Mobile GIS : Dependencie s Between Semantic s and Data Quality, Geolnforma tica Vol. 6, No. 4, pp 363-380, 2002. Ral s ton B.; Tharaka n G.; and Liu, C., A Spatial Decision Support S ystem for Transportation Policy Analysis in Banglade s h ," Journal of Trans port a tion Geography, Vol. 2, No 2, pp 1011 l 0, I 99 l. Ranger, Steve. December 6, 200 l. "SAP Add s Intelligent Agent s to SCM Project. Computing. [Online Accessed May, 2004], Available: http://www.computing.vnunet.com/ ews/ I I 234 1 6 Reardon M. ; Wilder C.; a nd Nel so n M Online Retailer s Le a rn From Their Mi s take s," InformationWeek I ss ue 769 pl l l Jan 17, 2000. Renolen A., Modellin g the Real World: Conceptual Modelling in Spatiotemporal Information Sys tem Des i gn," Transactions in GIS, Vol. 4, No. l, pp 23-42, 2000. Resnick, M. "Unblockin g the Traffic Jams in Corporate Thinkin g," Complexity, Vol. 3, No 4, pp 1076-2787 1998. Ricadela A., Microsoft Relea ses Dashboard Tool s," InformationWeek, I ss ue 754, p 36, 1999a. Ricadela A., Fre e Access to Bu s ine ss Data, InformationWeek, I ss ue 763 pp 85-86, Nov 29, 1999b. Ring, L. and Tigert, D., Viewpoint: The Decline a nd Fall of Internet Grocery Retailers ," International Journ a l o f Retail & Distribution Management, Vol. 29, No. 6, pp 264-271, 2002. Risto, L. "What About Managerial Geography?" GeoJournal Vol. 44, No. l pp l-7, 1998 210

PAGE 219

Robinson, V., "A Perspective on the Fundamentals of Fuzzy Sets and Their Use in Geographic Information Systems," Transactions in GIS, Vol. 7 No. I, pp 3-30, 2003. Roddick, J.; Grandi, F.; Mandreoli, F.; and Scalas, M., "Beyond Schema Versioning: A Flexible Model for Spatio-Temporal Schema Selection," Geolnformatica, Vol. 5, No. 1, pp 33-50, 2001. Roy, B., "Using Agents to Make and Manage Markets Across a Supply Web," Complexity, Vol. 3, No. 4, 1998. Roy, J.; Smith, N.; and Xu, B. "Simultaneous Modeling of Multi-purpose/Multi-stop Activity Patterns and Quantities Consumed," Journal of Geographic Systems, Vol. 3, pp 303-324, 2001. Saccomano, A., Online Goes Latin, Traffic World, Vol. 259, No. 9, p 16, 1999a Saccomano, A., Virtual Supermarket," Traffic World, Vol. 259, No. 3, pp 16-17, 1999b Sadahiro, Y., "A PDF-Based Analysis of the Spatial Structure of Retailing, GeoJournal Vol. 52, pp 237-252, 200 I. Saglio, J., and Moreira, J. "Oporto: A Realistic Scenario Generator for Moving Objects ," Geoinformatica, Vol. 5, No. 1, pp 71-93 200 I Sanborn, S. "Webvan Senior VP Takes on Retail Challenge," InfoWorld Vol. 22, Issue 29, p 42, 2000. Scarponcini, P., Generalized Model for Linear Referencing in Transportation, Geoinformatica, Vol. 6, No. l pp 35-55, 2002. Shiode, N., "30 Urban Models: Recent Developments in the Digital Modelling of Urban Environments in Three-Dimensions," GeoJournal, 52: pp 263-269, 2001. Southworth, F., and Peterson B., "Intermodal and International Freight Network Modeling ," Transportation Research Part C, Vol. 8, pp 14 7-156, 2000. Spaccapietra, S., "Editorial: Spatio-Temporal Data Models and Languages, Geolnformatica, Vol. 5, No. I pp 5-9, 2001. Spery, L.; Claramunt, C.; and Libourel T., "A Spatio-Temporal Model for the Manipulation of Lineage Metadata, Geoinformatica, Vol. 5 Vol. 1, pp 51-70, 2001. Swink, M., Speier C., Presenting Geographic Information: Effects of Data Aggregation, Dispersion and Users' Spatial Orientation," Decision Sciences, Vol. 30, No. l pp I 69195, 1999. 211

PAGE 220

Takeyama, M. "Geographical Conceptualization of Cyperplaces, GeoJournal, 53: pp 419-426, 2001. Tarantilis, C., and Kiranoudis, C., Using a Spatial Decision Support System for Solving the Vehicle Routing Problem ," Information and Management, Vol. 39, pp 359-373, 2002. Thrall, Grant. "Retail Location Analysis, Step Seven : Judgement," Geo Info Systems, Vol. 9. No. 2, pp 36-37, 1999. Thrall, Grant. Business Geography and New Real Estate Market Analysis. Oxford: Oxford University Press, 2002a. Thrall, Grant. "Fast Processors Furious Hard Drives Fire Up ArcGIS 3D Analyst," Geospatial Solution s, Vol. 12, No. 10, pp 44-47, 2002b. Toffler, A., "The Adaptive Corporation," London : Gower, 1985. Tompkins, David. September, 2002. Building a High-Velocity Supply Chain Network." Inbound Logistics, volume 22, number 9, p 36 Tryfona, N., and Jen sen, C., Conceptual Data Modeling for Spatiotemporal Applications ," Geolnformatica, Vol. 3, No. 3, pp 245-268, 1999. Tsou, M., and Buttenfield B., "A Dynamic Architecture for Distributing Geographic Information Services ," Transactions in GIS, Vol. 6, No. 4, pp 355-381, 2002. Turban, Efraim ; McLean Ephraim; Wetherbe James; Bolloju Narashima; Davison, Robert Information Technology for Management. New York: John Wiley and Sons, 2002. Tweney, D., "Webvan Delivers Logistics Lesson to Online Vendors, lnfoWorld, Vol. 21, No 32, p 46, Aug 9, 1999. Yan Duk, S.; Thieren s, D., and De Berg, M., "Using Genetic Algorithms for Solving Hard Problems in GIS ," Geolnformatica, Vol. 6, No. 4, pp 381-413, 2002. Violino, B., "The Knowledge Merchants ," InformationWeek, Issue 722, p 20, 1999 Vlachopoulou, M.; Silleos, G.; and Manthou, Y., "Geographic Information Systems in Warehouse Site Selection Decisions," International Journal of Production Economics, Vol. 71, pp 205-212, 2001. Wahle, J.; Chrobok R.; Pottmeier, A.; and Schreckenberg, M. "A Microscopic Simulator for Freeway Traffic, Networks and Spatial Economics, Vol. 2, pp 371-386, 2002. 212

PAGE 221

Wang, F., A Distributed Geographic Information System on the Common Object Request Broker Architecture (COREA)," Geolnformatica, 4: l pp 89-115, 2000. Wang, Y., and Gerchak, Y., "Supply Chain Coordination When Demand is Shelf-Space Dependent ," Manufacturing and Service Operations Management, Vol. 3, No. l, pp 8287, 2001. Weigel, Don. ( 1999 January February). Applying GIS and OR Techniques to Solve Sears Technician-Dispatching and Home-Delivery Problems. [Online], Available: http://pubsonline.informs.org/main/pdfstore/ Appl yingG ISand_article. pdf West, L. Jr., and Hess, T. Metadata as a Knowledge Management Tool: Supporting Intelligent Agent and End User Access to Spatial Data," Decision Support Systems, Vol. 32, pp 247-264, 2002. Whiting, R., April 22, 2002a. "GE Capital's Dashboard Drives Metrics to De s ktops ," lnformationWeek I ss ue 885, p 54. Whiting, R., September 2, 2002b. Operational Apps Get Smarter and Easier to Use." lnformationWeek. No. 904, p 24. Winter, S. Modeling Costs of Turns in Route Planning," Geo Informatica, Vol. 6, No. 4, pp 345-361, 2002. Witt, Clyde. June, 2002. In Search of the Perfect Order. Supply Chain Technology News, Vol. 4, No. 5, p 8. Wood, R., "Managing Complexity," London: The Economist Books, 2000. Wraige, H., "Go With the Flow," Professional Engineering, Vol. 14, No. 14, pp 35-36 2001. Wu, F., "GIS-Based Simulation as an Exploratory Analysis for Space-Time Processes," Journal of Geographic Systems, Vol. l pp 198-218 1999. Wu, Y.; Miller, H. ; and Hung M. "A GIS-Based Decision Support System for Analysis of Road Choice in Congested Urban Road Networks," Journal of Geographical Systems Vol. 3, No. 3, 2001. Wujciak, M., and Patton, S., "Outsmarting Complexity," Optimize, pp 50-55, January, 2003. Yano, K., GIS and Quantitative Geography," GeoJournal Vol. 52, pp 173-180, 2001. Yick, M. and Pang C., Development of a Process-Based Model for Dynamic Interaction in Spatio-Temporal GIS," Geoinformatica, Vol. 6, No. 4, pp 323-344, 2002. 213

PAGE 222

Zhang W., and Hunter G. Temporal Interpolation of Spatially Dynamic Object s," Geolnformatica Vol. 4, No. 4, pp 403-418 2000. Zhlin Li; Yong, Li; and Yong Qi Chen "Basic Topological Model s for Spa tial Entities in 3-Dimensional Space ," Geolnformatica 4:4, pp 419-433, 2000. Zhou, C ; Lu F.; and W a n Q. A Conceptual Model for a Feature Ba se d Virtual Network ," Geolnform a tica, Vol. 4, No 3, pp 271-286, 2000. Zipperer John. September 2 002. Supply Chain Knowledge is the Answer." Internet World, volume 8, i ss ue 9, pp 28-29. 214

PAGE 223

BIOGRAPHICAL SKETCH After attaining his Bachelor of Science degree in journalism at the University of Florida in 1990, and working as a journalist for a few years, Keith Herrel decided that his writing skills would be best utilized in the field of business geography. Therefore he attained a Master of Arts degree in decision and information sciences before attaining his Master of Science Degree in Geography. Both of these degrees are from the University of Florida. Then while working as a geographic information systems programmer analyst, he completed the Master of Business Administration course at Florida State University. After completing his doctoral degree in Geography he plans to use his knowledge to further research the exciting field of geography's relevance to logistically delivering goods via the internet. 215

PAGE 224

I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality as a dissertation for the degree of Doctor of Philosophy. /(,~// Professor of Geography I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality as a dissertation for the degree of Doctor of Philosophy. J!U--Professor of Geography I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate, in scope and quality as a dissertation for the degree of Doctor of Philosophy. Profes so r of Geography I certify that I have read this study and that in my opinion it conforms to acceptable standards of scholarly presentation and is fully adequate in scope and qualit y as a dissertation for the degree of Doctor of Philosophy. ck t\. Thompson Lecturer, Deci s ion and Information Sciences

PAGE 225

This dissertation was submitted to the Graduate Faculty of the Department of Geography in the College of Liberal Arts and Sciences and to the Graduate School and was accepted as partial fulfillment of the requirements for the degree of Doctor of Philosophy May 2006 Dean Graduate School

PAGE 226

(. -f I i i ijf i li!IITirri lllI I I i 1 111 3 1262 07332 062 3