<%BANNER%>

Sourcing Strategies in a Supply Chain


PAGE 1

SOURCINGSTRATEGIESINASUPPLYCHAINByGERARDJOSEPHBURKEJR.ADISSERTATIONPRESENTEDTOTHEGRADUATESCHOOLOFTHEUNIVERSITYOFFLORIDAINPARTIALFULFILLMENTOFTHEREQUIREMENTSFORTHEDEGREEOFDOCTOROFPHILOSOPHYUNIVERSITYOFFLORIDA2005

PAGE 2

Copyright2005byGerardJosephBurkeJr.

PAGE 3

Thisworkisdedicatedtomyfamily,andespeciallymywife,Amy,foryourloveandsupport.

PAGE 4

ACKNOWLEDGMENTSThehardworkanddedicationrequiredbymetocreatethisdissertationweremadepossiblebythepersonalandpracticalsupportofmyfamily,friends,committeemembersandothermembersoftheDecisionandInformationSciencesDepartment.Iwishtoexpressmygratitudebyspecicallyacknowledgingeachgroupofsupporters.IwishtothankGod;myparents,JerryandCarlyn;mywife,Amy;mychil-dren,Maddie,Marley,EllaandJames;mygrandfather,FrankBurkewhosesupportispriceless;andmymother-in-law,DebCrenshawfortheirlove,inspira-tion,andsupport.IalsothankmyHoosierfamily,HankandBettyTallman,JimandJoniPing,JeannieMeenach,andDrewandPamKisselfortheirunwaveringencouragement,loveandmuchneededbreaksfrommystudiousendeavors.Althoughthisnextgroupforgratitudeislegallynotconsideredfamily,theyareclosertomethanfriendshipdescribes.IwishtothankLouandSandyPaganini,andEdandAtwoodBrewtonfortheirsupport,faith,andperspective.Also,manythanksgototheKempersandNelsonsfortheirfriendship.Mydoctoralstudiesweregreatlyenrichedbythecomraderyofmyfellowdoc-toralstudentsMarkCecchini,SelcukColak,EnesEryarsoy,LingHe,JasonDean,YuwenChen,ChristyZhang,FidanBoylu,andMichelleHanna.Additionally,IwouldliketothankPatBrawner,ShawnLee,andCindyNantzforassistingmethroughoutthistribulation.Mycapabilitytocompletethisdissertationwasdevelopedinlargepartbytheseminarsandcourseworktaughtbymembersofmydissertationcommittee.IwishtothankJaniceCarrillo,SelcukErenguc,AnandPaul,andJoeGeunesfortheir iv

PAGE 5

instructionandserviceonmydissertationcommitteeduringmydoctoralstudies.IalsowishtothankJaniceCarrilloforcandidlysharingherexperiencesintheacademicprofession.Finally,IwishtothankProfessorAsooVakharia,myadvisorandcommitteechair,forhisexpertguidance,timelyresponsestothedraftsofeachchapterofmydissertation,andgenuineinterestinmypersonalandprofessionalwell-being. v

PAGE 6

TABLEOFCONTENTS page ACKNOWLEDGMENTS ............................. iv LISTOFTABLES ................................. ix ABSTRACT .................................... xi CHAPTER 1INTRODUCTION .............................. 1 1.1SupplyChainManagement-AnOverview .............. 1 1.2StrategicIssuesinSCM ........................ 3 1.2.1ProductStrategy ....................... 4 1.2.2NetworkDesign ........................ 5 1.3OperationalIssuesinSCM ...................... 7 1.3.1Transportation ......................... 7 1.3.2Transformation ........................ 8 1.3.3InformationSharing ...................... 9 1.4FocusofthisResearch:StrategicSourcing ............. 10 1.5OrganizationofthisDissertation ................... 14 1.6StatementofContribution ...................... 14 2LITERATUREREVIEW .......................... 16 2.1Overview ................................ 16 2.2StrategicEvolutionofSourcing ................... 16 2.3Buyer-SupplierRelationships ..................... 17 2.4StrategicSourcing ........................... 18 2.4.1QualicationCriteria ..................... 18 2.4.2SelectionandAllocation ................... 19 3STRATEGICSOURCINGDECISIONSWITHSTOCHASTICSUP-PLIERRELIABILITY ........................... 27 3.1Introduction .............................. 27 3.2SourcingModel ............................ 28 3.3ModelDevelopment .......................... 30 3.4Analysis ................................ 30 3.4.1HeterogeneousSuppliers ................... 31 3.4.2HeterogeneousCostSuppliers ................ 35 vi

PAGE 7

3.4.3HeterogeneousReliabilitySuppliers ............. 36 3.4.4HomogeneousSuppliers .................... 37 3.5NumericalAnalysis .......................... 39 3.5.1ExperimentalDesign ..................... 39 3.5.2Results ............................. 40 3.6Conclusions .............................. 44 4IMPACTOFSUPPLIERPRICINGSCHEMESANDCAPACITYONSOURCINGSTRATEGIES ........................ 48 4.1Introduction .............................. 48 4.2SourcingModel ............................ 50 4.2.1Preliminaries .......................... 50 4.2.2SupplierPricingSchemesandCapacity ........... 51 4.3AnalysisandInsights ......................... 52 4.3.1ConstantPrice ......................... 52 4.3.2LinearDiscountPrice ..................... 53 4.3.3IncrementalUnitsDiscountPrice .............. 56 4.3.4AllUnitsDiscountPrice ................... 59 4.3.5SummaryofInsightsfromAnalysis ............. 61 4.4Application .............................. 62 4.5OptimalAlgorithmforIncrementalQuantityDiscountedSourcing 64 4.5.1AlgorithmDescription ..................... 65 4.5.2ComputationandValidityofLBpandUBf ......... 66 4.5.3TheBranchingProcess .................... 67 4.5.4FormalStatementoftheAlgorithm ............. 67 4.6Conclusions .............................. 68 5STRATEGICSOURCINGWITHDIVERSIFICATIONCONSIDERA-TION .................................... 70 5.1Introduction .............................. 70 5.2IntegratedSelection/AllocationModel ............... 71 5.2.1Preliminaries .......................... 71 5.2.2ModelDevelopment ...................... 75 5.3Analysis ................................ 77 5.3.1NoDiversicationBenet ................... 77 5.3.2DiversicationBenet ..................... 81 5.4ModelExtensions ........................... 83 5.5NumericalAnalysis .......................... 86 5.5.1ExperimentalDesign ..................... 86 5.5.2Results ............................. 87 5.6ConclusionsandImplications ..................... 91 vii

PAGE 8

6SUMMARY .................................. 94 6.1KeyResultsandDirectionsforFutureResearchfromChapter3 94 6.2KeyResultsandDirectionsforFutureResearchfromChapter4 96 6.3KeyResultsandDirectionsforFutureResearchfromChapter5 98 APPENDIX ..................................... 100 APROOFSFORCHAPTER3 ........................ 100 A.1ProofofCorollary3.1 ......................... 100 A.2ProofofTheorem3.1 ......................... 101 A.3ProofofCorollary3.2 ......................... 102 A.4ProofforCorollary3.3 ........................ 102 A.5ProofofCorollary3.4 ......................... 102 A.6ProofofCorollary3.5 ......................... 103 A.7ProofofCorollary3.6 ......................... 103 A.8ProofofTheorem3.2 ......................... 104 A.9ProofofTheorem3.3 ......................... 104 A.10ProofofTheorem3.4 ......................... 104 A.11ProofofCorollary3.7 ......................... 104 BPROOFSFORCHAPTER4 ........................ 106 B.1ProofofTheorem4.1 ......................... 106 B.2ProofofResult4.1 .......................... 107 B.3LinearDiscountPricingTestProblemsData ............ 109 B.4IncrementalandAll-UnitDiscountPricingTestProblemsData .. 112 CPROOFSFORCHAPTER5 ........................ 142 C.1ProofofTheorem5.1 ......................... 142 C.2ProofofTheorem5.2 ......................... 144 C.3ProofofTheorem5.5 ......................... 146 REFERENCES ................................... 148 BIOGRAPHICALSKETCH ............................ 154 viii

PAGE 9

LISTOFTABLES Table page 3{1ModelNotation .............................. 29 3{2DescriptionofNumericalExamples .................... 40 3{3ResultsforNumericalExamples ..................... 41 4{1LinearDiscountHeuristicPerformance .................. 56 4{2IncrementalUnitsDiscountHeuristicPerformance ........... 58 4{3All-UnitsDiscountHeuristicPerformance ................ 61 4{4ProductABidInformationData ..................... 62 4{5ProductBBidInformationData ..................... 63 4{6CPOProductASolutionsComparison .................. 64 4{7CPOProductBSolutionsComparison .................. 64 4{8NumberofSubproblemsSolvedforeachTestProblem ......... 68 5{1SensitivityAnalysisoftheKeyParameters ............... 88 5{2ImpactofMinimumOrderQuantityontheSourcingStrategy ..... 89 5{3InteractionsbetweenMinimumOrderQuantitiesandReliabilities ... 90 5{4ParameterValuesforRankedSupplierCharacteristics ......... 91 5{5SelectedResultsforCapacityAdjustedModelA ............ 91 B{1LinearDiscountPricingTestProblemsData-12 ........... 109 B{2LinearDiscountPricingTestProblemsData-24 .......... 110 B{3LinearDiscountPricingTestProblemsData-30 .......... 111 B{4IncrementalandAll-UnitsPricingTestProblemsData-3 ...... 112 B{5IncrementalandAll-UnitsPricingTestProblemsData-6 ...... 115 B{6IncrementalandAll-UnitsPricingTestProblemsData-9 ...... 118 B{7IncrementalandAll-UnitsPricingTestProblemsData0-12 .... 121 ix

PAGE 10

B{8IncrementalandAll-UnitsPricingTestProblemsData3-15 .... 124 B{9IncrementalandAll-UnitsPricingTestProblemsData6-18 .... 127 B{10IncrementalandAll-UnitsPricingTestProblemsData9-21 .... 130 B{11IncrementalandAll-UnitsPricingTestProblemsData2-24 .... 133 B{12IncrementalandAll-UnitsPricingTestProblemsData5-27 .... 136 B{13IncrementalandAll-UnitsPricingTestProblemsData8-30 .... 139 x

PAGE 11

AbstractofDissertationPresentedtotheGraduateSchooloftheUniversityofFloridainPartialFulllmentoftheRequirementsfortheDegreeofDoctorofPhilosophySOURCINGSTRATEGIESINASUPPLYCHAINByGerardJosephBurkeJr.August2005Chair:AsooJ.VakhariaMajorDepartment:DecisionandInformationSciencesThefocusofthisdissertationisonsupplychainmanagementSCM,andmorespecicallyontheupstreamconnectionbetweenarmanditssuppliers.Myresearchexaminessingleversusmultiplesuppliersourcingstrategiesunderthreespecicscenarios.Ingeneral,mydissertationseekstocharacterizewhenabuyingrmshouldsinglesourceitsrequirementsinsteadofemployingadiversiedpurchasingpolicyundervariouscommonlyencounteredoperatingscenarios.First,theeectsofupstreamanddownstreamuncertaintyonarmssourcingstrategyareexamined.Ourresultsshowthatorder-splittingi.e.,choosingamultiplesupplierstrategyisanoptimalchoiceforthermevenwhensuppliersarecompletelyheterogeneousintermsoftheirreliabilityandcosts.Additionally,thechoiceofsingleversusmultiplesourcingdependstosomeextentonsupplierprices.Thislatterresultmotivatesinvestigatingoursecondscenariotogaininsightintohowalternatesupplierpricingschemesmayimpactsourcingdecisions.Thesecondscenarioweexamineismotivatednotonlythroughtheresultsobtainedunderthestochasticsupplysettingdescribedearlierbutalsothroughanunderstandingofthesupplierselectionandquantityallocationdecisionsmadeby xi

PAGE 12

amajoroceproductsretailerlocatedinFlorida.Thepricingschemesquotedbyitssupplierstendtobeeitheraconstantprice,alinearlydiscountedprice,orastagedquantitydiscounti.e.,all-unitsand/orincrementaldiscountedpriceprice.Foreachtypeofpricingscheme,wedevelopauniqueoptimizationmodelwheretheobjectiveistominimizethesumofconcavecostfunctionswhilesatisfyingthermstotalrequirements.Weadaptexistingbranchandboundalgorithmsinordertoidentifytheoptimalnumberofsupplierswhoshouldreceiveanorder.Finally,weincorporateexplicitdiversicationbenetsduetochannelpowerleverageandpricecompetitionintoanewsvendorframeworktoanalyzearmssourcingdecisionswhensuppliersareunreliable.Analysisrevealsthatatrade-obetweenthemarginalbenetofdiversicationandthemarginalcostofshiftingallocatedorderquantitiesawayfromlowercostsuppliersneedtobeassessed.Managerially,thismodelstressestheimportanceofconsistencybetweenarmssourcingstrategyanditscorporatestrategy. xii

PAGE 13

CHAPTER1INTRODUCTION 1.1 SupplyChainManagement-AnOverviewAsupplychaincanbevisualizedasanetworkofrmsservicingandbeingservicedbyseveralotherrms.However,itisconceptuallyeasiertoimagineachainasariver,originatingfromasource,movingdownstreamandterminatingatasink.Thesupplychainextendsupstreamtothesourcingofrawmaterialsanddownstreamtotheafterlifeactivitiesoftheproduct,suchasdisposal,recyclingandremanufacturing.Regardlessofmagnitude,everysupplychaincanbevisualizedasconsistingofsourcingstages,manufacturingstagesanddistributionstages.Eachofthesestagesplaysbothaprimaryusuallyphysicaltransformationorservicecreationandadualmarketmediatorrole.Theapproachtakentoexecuteactivitiesinsupportofbothrolesdependsonthestrategyofthesupplychain,whichinturn,isafunctionoftheservicedproducts'demandpatternFisher,1997.Dependinguponthestructureofthechainintermsofproductsandprocessesemployed,channelpowercanresidewiththesourcinge.g.,monopolistsupplierofkeycommoditiessuchasoil,manufacturinge.g.,dominantproducerofauniqueproductsuchassemiconductors,ordistributione.g.,keydistributorofconsumeritemsstagesinthesupplychain.Relativepowerinthesupplychaininuencesstrategicpositioningofeachlinkinthechain.Thus,managingsupplychainsisanegotiationbetweentheobjectivesofconstituent'sbenetateachstageandtheimpactofeachconstituent'sobjectivetotheoverallobjectiveofmaximizingthebenetoftheentirechain.Thecontributioncapturedateachstagedependsonthenatureofthedealingsbetweenthebuyerandsupplier.Thetraditionalmodelischaracterizedbyan 1

PAGE 14

2 adversarialrelationshipwhereabuyerawardscontractstooneormorecompetingsuppliersbasedonpriceandotherpertinentcriteria.Thisparadigmhasbeenwidelycriticizedasshortsightedbyproponentsofpartneredbuyer-supplierrelationships.Thepartneredapproachfavorsasmallerorevensinglesupplierbaseformoresuppliermanagementinitiativestominimizeinventoryinvestmentsandencouragecollaborationin,amongotherthings,newproductdevelopment.SupplychainmanagementSCMistheartandscienceofcreatingandaccentuatingsynergisticrelationshipsamongthetradingmembersthatconstitutesupplyanddistributionchannels.Supplychainmanagersstrivetodeliverdesiredgoodsorservicesontimetotheappropriateplaceintheorderedquantityinthemosteectiveandecientmanner.Usuallythisisachievedbynegotiatingabalancebetweenconictingobjectivesofcustomersatisfactionandcosteciencies.Eachlinkineachsupplychainrepresentsanintersectionwheresupplymeetsdemand,anddirectingtheproductandinformationowsatthesecrossroadsisatthecoreofSCM.Theintegralvaluepropositionofanintegratedsupplychainisasfollows.Totalperformanceoftheentirechainisenhancedwhenalllinksinthechainaresimultaneouslyoptimizedascomparedtotheresultingtotalperformancewheneachindividuallinkisseparatelyoptimized.Supplychainperformanceasawholehingesonachievingtbetweenthenatureoftheproductsitsupplies,thecompetitivestrategiesoftheinteractingrms,andtheoverallsupplychainstrategy.Coordinationoftheindividuallinksinthechainisessentialtoachievethisobjective.Theabilityoftradingpartnerstojointlycommunicateinrealtimeandthetransactionaleaseofdigitaldealingsallowweb-connectedrmstovirtuallyintegrate.TheInternetandinformationtechnologyingeneralfacilitatetheintegrationofmultitudesofchannelenterprises.On-linecollaborationenables

PAGE 15

3 betterinformedeconomicdecisionmaking,reducesthecostsoforderplacement,trackingandreceipt,andenhancescustomersatisfaction.Informationtechnologiesareakeydriverofmodernoperationaleciency,andecientoperationalexecutionisadriverofeectiveSCM.Selectionoftradingpartners,locationoffacilities,manufacturingschedules,transportationroutesandmodes,andinventorylevelsandlocationarethefundamentaloperationsdecisionsthatrunsupplychains.Theseoperationaldimensionsarethetributariesthatpilotthechaindownstreamthroughitschanneltoenddemand.Accurateandtimelyintegratedinformationnavigatesthechainfromsourcetosink.Asupplychainisacollectionofmultiplesuppliers',manufacturers'anddistributors'processes.Eachprocessemploysadistinctfocusandarelateddimensionofexcellence.Keyissuesinmanaginganentiresupplychainrelatetotacticalandstrategicanalysisofcoordinateddecisionsinlogistics,manufacturing,distribution,andaftersalesactivitiesofserviceanddisposalorrecycling;analyzingproductstrategies;andnetworkdesigndecisions.Themotivationforthisresearchisderivedfromthedebateastothebestnum-berofsupplierstoemployforsatisfyingabuyer'srequirements.Further,thebuyerconsideredisanintermediaryinthesupplychainandthereforemustincorporatedownstreamdemandintoitssourcingdecision.Essentially,thedecisionsanalyzedaddressthequestionofwhetherasinglesourcingstrategyisoptimalornot.Tounderstandtherelevanceofstrategicsourcingdecisions,itmustbeunderstoodhowarm'ssupplychainstrategyisanchoredtoitssourcingstrategy. 1.2 StrategicIssuesinSCMAsupplychainisonlyasstrongasitsweakestlink.Howthechaindenesstrengthisatthecoreofasupplychain'sstrategy,andthereforedesign.Isstrengthanchoredineciencyorresponsiveness?Regardlessofwhichstrategicpositionischosen,arm'sabilitytomaintainacompetitiveadvantagewilldependonhow

PAGE 16

4 wellitreinforcesitsrmlevelvaluepropositionwithfunctionalanddepartmentalstrategicdecision-making.Byanalyzingproductdemandcharacteristicsandthesupplychain'scapabilities,andcraftingatbetweenthem,anindividualsupplychainmanagercanbeassuredthatthespecicproductandprocessstrategyemployeddoesnotcreatedissonancewithinhisrmandfurtherthroughouttheentiresupplychain. 1.2.1 ProductStrategyAchievingatighttbetweenthecompetitivestrategiesofsupplychainmembersandthesupplychainitselfisgainedbyevaluatingthecharacteristicsoftheproductsservicedbythechain.Therootcauseoftheproblemsplaguingmanysupplychainsisamismatchbetweenthetypeofproductandthetypeofsupplychain"Fisher,1997,p.106.Criticalproductattributesareathedemandpattern;bthelife-cycle;cvarietyofoerings;anddtheproductdeliverystrategy.Fisher997categorizesaproductasbeingeitherfunctionalbasic,predictable,long-lived,lowprotmarginorinnovativedierentiated,volatile,short-lived,highprotmargin.Further,usingtheproductlifecycleargument,innovativeproductsifsuccessfulwilleventuallyevolvetobecomefunctionalproducts.Thetypesofsupplychainsneededtoeectivelyservicethesetwocategoriesofproductsarequitedistinct.Anecientorlowcostsupplychainismoreappropriateforafunctionalproductwhilearesponsiveorcustomerattunedsupplychainbetterservicesaninnovativeproduct.Obviously,aspectrumofchainvarietiesexistsbetweentheendpointsofresponsivenessandeciency,andhence,mostsupplychainsarehybridswhichtargetresponsivenessrequirementsforeachproductservicedwhileexploitingcommonalitiesinservicingallproductstogaineconomiesofscope.Thus,thestrategicpositionofasupplychainbalancescustomersatisfactiondemandsandtherm'sneedforcostminimization.

PAGE 17

5 Informationtechnologiesenablebothecientandresponsivesupplychainssincetheyhavethepotentialtoprovideimmediateandaccuratedemandandorderstatusinformation.Eciencygainsviainformationtechnologiesaregleanedfromdecreasedtransactionalcostsresultingfromorderautomationandeasieraccesstoinformationneededbychainmembers.Likewise,responsivenessgainscanbeobtainedbyaquickerresponsetocustomerorders.Hence,inpractice,itseemstohavebecomestandardpracticeforallsupplychainstoutilizesomeformofinformationtechnologytoenablenotonlyamoreecientphysicalowoftheirproductsbutalsotosimultaneouslyimprovetheirmarketmediationcapability.However,theeciencyofphysicalowprimarilydependsonasupplychain'sinfrastructure. 1.2.2 NetworkDesignIngeneral,networkdesigndeterminesthesupplychain'sstructure.Thesignicantcapitalinvestmentsrequiredinbuildingsuchastructureindicatetherelativelongrunorstrategicimportanceofnetworkdecisions.Networkdecisionsinasupplychaininvolvefacilityfocus,facilitylocation,capacityplanning,andsourcing/distributionchannelsChopraandMeindl,2001.Eachnetworkdesigndecisionimpactstherm'sabilitytoprovidevalue.Therefore,thesedecisionsmustincorporatetheirstrategicinuenceintotheanalysis.Facilityfocusrelatestohownetworkinvestmentsfacilitatethesupplychainstrategy.Ifthefacilityinquestionisamanufacturingplantandtheplantissetuptoproduceonlyaspecicproducttype,thechainwillbemoreecientbutlessexiblethanitwouldbeiftheplantproducedmultipleproducttypes.Facilitylocationdecisionsareessentialtoarm'sstrategy.Thecostrami-cationsofasub-optimallocationdecisioncouldbesubstantial.Further,shuttingdownormovingafacilityissignicantnotonlyintermsofnancialresources,butalsointermsoftheimpactonemployeesandcommunities.Otherfactorswhich

PAGE 18

6 shouldbeconsideredaretheavailableinfrastructureforphysicalandinforma-tiontransportation,exibilityofproductiontechnologiesemployed,externalormacroeconomicinuences,politicalstability,locationofcompetitors,availabilityofrequiredlaborandmaterials,andthelogisticscostscontingentonsiteselection.Dependingontheexpectedlevelofoutputforafacility,capacityallocationsshouldbemadesothatidletimeisminimal.Under-utilizationresultsinlowerreturnoninvestmentandissuretogettheattentionofcompanyexecutives.Ontheotherhand,underallocatingcapacityorlargeutilizationswillcreateabottleneckorconstrictedlinkinthesupplychain.Thiswillresultinunsatiseddemandandlostsalesorincreasedcostsasaresultofsatisfyingdemandfromanon-optimallocation.Thecapacityallocationdecisionisarelativelylong-termcommitment,whichbecomesmoresignicantassophisticationandpriceoftheproductiontechnologyincrease.Themostbasicquestionofanenterpriseis:Whowillserveourneedsandwhoseneedswillweserve?Thisisarecurringquestion.Decisionsregardingthesupplierstoafacilityandthedemandtobesatisedbyafacilitydeterminethecostsofmaterialinputs,inventory,anddelivery.Therefore,asforcesdrivingsupplyand/ordemandchange,thisdecisionmustbereconsidered.Theobjectivehereistypicallytomatchsuppliersandmarketstofacilitiesinordertominimizenotonlythesystem-widecostsbutalsothecustomerresponsivenessofthesupplychain.Eachofthesenetworkdesigndecisionsisnotmadeinisolationsincethereisaneedtoprioritizeandcoordinatetheircombinedimpactonthermanditssupplychain.Ingeneral,networkcongurationisthestructureofthesupplychainanditiswithinthisstructurethatoperationsstrategiesandtacticsareimplementedtoreinforcetheoverallstrategyoftheentirechain.Ofparticularrelevanceforthisresearchisarm'ssourcingstrategy.

PAGE 19

7 1.3 OperationalIssuesinSCMSCMhasevolvedfromprocessreengineeringeortstocoordinateandintegrateproductionplanningatthefactorylevelinordertoexpandthescopeofstrategictChopraandMeindl,2001.Positiveresultsfromtheseintra-functionaleortshaveextendedtheSCMphilosophythroughouttheenterprise.Further,processimprovementsatthermlevelhighlightedtheneedforsuppliersandcustomersofsupplychainmanagedrmstoadoptanintegratedSCMphilosophy.Makingasupplychain'slinkagesasfrictionlessaspossibleisthetacticalgoalofsuchanintegratedphilosophy.KeytacticalcoordinationdecisionsforSCMrelatetotransportation,transformation,andinformationtransmission. 1.3.1 TransportationTransportationdecisionsimpactproductownotonlybetweensupplychainmembersbutalsotothemarketplace.Inmanysupplynetworks,transportationcostsaccountforasignicantportionoftotalsupplychaincost.Indeterminingthemodesandroutestoemploythroughthesupplychain,transportationdecisionsseektostrikeabalancebetweeneciencyandresponsivenesssoastoreinforcethestrategicpositionofthesupplychain.Forexample,aninnovativeproduct'stypicallyshortlife-cyclemaywarrantexpensiveairfreightspeedforaportionorallofitsmovementthroughthechain,whileacommodityisgenerallytransportedbyslowbutrelativelyeconomicalwaterorrailfreight.Shippingviatruckisalsousedfrequently.Truckingismoreresponsiveandmoreexpensivethanrail,andlessresponsiveandlessexpensivethanair.Mostsupplychainsemployanintermodalstrategye.g.,rawmaterialsaretransportedbyrailorship,componentsbytruck,andnishedgoodsbyair.Asupplychain'stransportationnetworkdecisionsareinextricablylinkedtostrategicnetworkdesigndecisions.Transportationnetworkdesignchoicesdriveroutingdecisionsinthesupplynetwork.Themajordecisionsarewhethertoship

PAGE 20

8 directlytobuyersortoadistributioncenter,andwhetheraroutingschemeisneeded.Asconsumers'expectationsregardingmerchandiseavailabilityanddeliverybecomemoreinstantaneous,theroleofasupplychain'stransportationnetworkismorecritical. 1.3.2 TransformationAtransformationnetworklinksproductionfacilitiesconductingwork-inprocessinventoriesthroughthesupplychain"Erenguc,Simpson,andVakharia,1999,p.224.Supplierslinkedtomanufacturerslinkedtodistributionsystemscanbeviewedasatransformationnetworkhingingonthemanufacturer.Transformingsuppliesbeginsatthereceivingstationsofmanufacturers.Thecongurationofmanufacturingfacilitiesandlocationsoftransformationprocessesaredeterminedbyplantleveldesigndecisions.Themanufacturingprocessstrategyemployedataspecicplantlargelydrivesthedecisions.Whileanassemble-to-orderATOplantmayhaveverylittleinvestmentinproduction,itrequireslargerinvestmentinsub-assemblyinventories.Ontheotherhand,amake-to-stockMTSfacilitymayhavelittleornoinvestmentinprocessinventories,ittypicallyrequireslargerinvestmentsinrawmaterialsandnishedgoodsinventories.Amake-to-orderMTOfacilitymayhavesignicantinvestmentincomponentsandproductionfacilities,withfewrawmaterialsandnishedgoodsinventories.Aproduct'snalformcanalsotakeshapeclosertotheendconsumer.Tokeepnishedgoodsinventorycostsaslowaspossible,andbettermatchenddemand,asupplychainmayemploypostponementtodelaycustomizingendproducts.Majordesigndecisionssuchasfacilitycongurationandtransformationprocessesareconsideredlongertermdecisions.Thesedecisionsconstraintheshorttomid-termdecisionsaddressedinaplant'saggregateplan.Anaggregateplanisageneralproductionplanthatencompassesaspecicplanninghorizon.Informationrequiredtodevelopaneectiveaggregateplanincludeaccuratedemandforecasts,

PAGE 21

9 reliablesupplydeliveryschedules,andthecosttrade-osbetweenproductionandinventory.Eachsupplychainmemberdevelopsanaggregateplantoguidemedium-termtacticaldecisions.Toensurethattheseindividualplanssupporteachother,theplanningprocessmustbecoordinated.Thedegreeandscopeofcoordinationwilldependontheeconomicsofcollaborativeplanningversusthecostsofunder-supplyandover-supply.Ingeneral,amanufacturershoulddenitelyinvolvemajorsuppliersandbuyersinaggregateplanning.Whetherthisplanninginformationtricklestoothersupplychainmembersakeyforthesuccessofintegratedsupplychainmanagementwilldependonthecoordinationcapabilitiesofsuccessivelayersofmembersemanatingfromacollaborativeplanningcenter,whichisoftenthemajormanufacturer.Theexecutionoftheaggregateplanisafunctionoftheinformationinputsintotheaggregateplan.Itisvitalthattheseinputsbeasaccurateaspossiblethroughouttheentiresupplychain.Integratedplanninginasupplychainrequiresitsmemberstoshareinformation.Theinitiatorofintegratedplanningistypi-callyanintermediary.Tounderstandwhy,wemustunderstandthedynamicsofdistribution. 1.3.3 InformationSharingAdistributionchannelistypicallycomposedofamanufacturer,awholesaler,adistributor,andaretailer.Thebull-whipeect"isaclassicillustrationofdysfunctioninsuchachannelduetothelackofinformationsharing.Thiseectischaracterizedbyincreasingvariabilityinordersastheordersaretransferredfromtheretailerupstreamtothedistributor,thentothewholesaler,andnallytothemanufacturer.Distorteddemandinformationinducesamplicationsinordervarianceasordersowupstream.Therefore,themanufacturerbearsthegreatestdegreeofordervariability.Thisisamajorreasonmanufacturersinitiatecollaborativeeortswithdownstreamchannelmembers.Toanticipatethequantity

PAGE 22

10 ofproducttoproduceandwhen,amanufacturermustcompiledemandforecastsfromdownstreamsupplychainmembers.Forecastingaccuracyisparamountbecauseitisthebasisforeectiveandecientmanagementofsupplychains.AmajorchallengeofSCMistominimizecostsandmaintainexibilityinthefaceofuncertaindemand.Thisisaccomplishedthroughcapacityandinventorymanagement.Similarly,marketersattempttomaximizerevenuesthroughdemandmanagementpracticesofpricingandpromotion.Therefore,itisvitalthatmarketingandoperationsdepartmentscollaborateonforecastsandshareharmoniousincentivestructures.Thedegreeofcoordinationbetweenorderacquisition,supplyacquisition,andproductionprocessdirectlyaectshowsmoothlyarmoperates.Likewise,buyers',suppliers',andproducers'coordinationlevelsdirectlyaecthowsmoothlythesupplychainoperates.Summarily,accurateinformationowsbetweenchannelmembersareessentialforeectiveSCM. 1.4 FocusofthisResearch:StrategicSourcingThestrategicimportanceofsourcingactivitiesisinherentinpurchasing'ssupplychainposition.Purchasingactivitieslinkthermtotheupstreamvaluesys-temandallowabuyingrmtoobtainappropriateinputsfromexternalsuppliers.Procurementactivitiesinlargepartsupportarm'sinboundlogisticsandarevitaltovaluecreationPorter,1985.Arm'ssourcingstrategyisthereforeakeydriverofaneectivesupplychainvaluesystem.Innovationsintechnologyandincreasedglobalcompetitionprovideopportu-nitiesandchallengesthatdrivermstocontinuouslyevaluateandmodifytheirsourcingstrategies.Moreover,recentstudiesrevealthatthelong-termimplicationsofpoorsupplychainmanagementarefarreaching,ultimatelyimpactingbothrmperformanceandmarketvalue.Sinceatypicalmanufacturingrmspends55%ofearnedrevenueonpurchasedmaterialsLeendersandFearon,1998,disruptionsduetosupplyinadequaciescouldhaveamajorimpactonprotability.Hendricks

PAGE 23

11 andSinghal01reinforcethisbyshowingthatinthe90dayspriorandsub-sequenttoareportedsupplychainproblemstemmingfromsupplierglitches,thebuyingrm'saverageshareholderreturntypicallydecreasesby12%.Clearly,amanufacturer'soperationsstrategyandnanciallivelihoodrelyonitschosensup-plierpoolandthusdecisionswithregardtosuppliersarefundamentaltosuccessfulenterprise.Arm'ssourcingstrategyischaracterizedbythreekeyinterrelateddecisionsBurkeandVakharia,2002:acriteriaforestablishingasupplierbase;bcriteriaforselectingsuppliersasubsetofthebasewhowillreceiveanorderfromtherm;andcthequantityofgoodstoorderfromeachsupplierselected.Tostartwith,criteriafordevelopingasupplierbasearetypicallybasedupontherm'sperceptionofthesupplier'sabilitytofullltheobjectivesofquality,quantity,deliveryandprice.Whilethesupplier'spricemaybethemostimportantcriteriaforprotmaximization,theotherdimensionscanalsoaecttheoverallprotabilityoftherm.Scoringmodelsaregenerallyusedtoevaluatesuppliersforinclusioninthebase.Ingeneral,thisapproachrankseachsupplierintermsofobjectivesandthen,basedonarelativeweightingofeachoftheobjectives,atotalscoreforeachpotentialsupplierisderived.Next,byspecifyingathresholdscore,allsupplierswhoachievethisthresholdareincludedinthebase.Objectivesusedindevelopingrankingsvaryacrossrms.Forexample,SunMicrosystemsranksitssupplierswithascorecard"basedonquality,delivery,technology,andsuppliersupportHollowayetal.,1996.Fromtheapprovedsupplybase,thespecicsubsetofsupplierswhichwillactuallyreceiveanordermustbedetermined.Sinceallsuppliersinthebasemeetthequality,delivery,andotherobjectivesoftherm,dominantindustrypracticeappearstobasethisdecisionprimarilyoncostconsiderations.Whilethesupplier'spricemaybethemostimportantcriterionforprotmaximization,somebuying

PAGE 24

12 rmsimposealternatecriteriarelatedtorobustdeliveryreliabilitycapabilities.Oncetheselectedsetofsuppliersasubsetofthebaseisdetermined,thermmustallocateproductsrequirementsamongthem.Fortheallocationdecision,supplieryieldsintermsofpercentageofgood"units,orderquantitypolicies,leadtimes,andtransportationcostsaresalient.Armmaychooseeitheraspecializedi.e.,singlesupplierorageneralizedi.e.,multiplesuppliersstrategicsourcingposition,andallocationofrequirementswillseektooptimizethevaluetothermfromthisdecision-makingprocess.Sinceitisthecollectivesuppliers'capabilitiesthatcanenableorlimitsupplychainperformanceatitsinception,arm'ssourcingstrategyisvitaltosuccessfulenterprise.Single-sourcingstrategiesstriveforpartnershipsbetweenbuyersandsupplierstofostercooperationandachievesharedbenets.Thetightercoordinationbetweenbuyerandsuppliersrequiredforsuccessfuljust-in-timeJITinventoryinitiativesencouragessupplieralliancestostreamlinethesupplynetworkandtendstoshiftsupplyrelationstowardsinglesourcing.Managingmorethanonesourceisobviouslymorecumbersomethandealingwithasinglesource.However,web-basedSCMapplicationsenablecloserman-agementofdiversesuppliers,streamlinesupplychainprocessesanddrivedownprocurementcosts.Forexample,GMutilizedInternettoolstopurchasemoreofitstotalpurchasingbudgeton-line,whichresultedinastreamlinedprocurementprocessanddecreasedvehicledeliverytimesVeverka,2001.Otherdocumentedbenetsofeectivelyutilizingweb-basedprocurementtoolsincludeareductioninpriceofmaterials,administrativecosts,inventorycosts,andpurchaseandfulll-mentcycles.Consequently,rmsthatprefersingle-sourcingforeaseofmanagementcanembracemultiple-sourcingviainformationtechnology-basedSCMapplications

PAGE 25

13 asamoreviablestrategytocapturerisk-poolingbenets.AshortcomingofIn-ternetprocurementtoolsistheirlimitedcapabilitytoprovidecomplexdecisionsupportforstrategicsourcingdecisionsAberdeenGroup,1999.Nonetheless,thedierentialtacticalorlow-costcompetitiveadvantagessupportedbysingle-sourcingovermultiple-sourcingarediminishedbytheproliferationofInternetprocurementcapablerms.Moreover,single-sourcedependencyexposesthebuyingrmtoagreaterriskofsupplyinterruption.Toyota'sbrakevalvecrisisin1997providesarecentexampleofrealizedsupplyriskresultingfromasinglesourcingstrategyinaJITinventorysystem.In1997,Toyota'sassemblyplantswereforcedtoshutdownafterareatAisin'smainplant.Thissingle-supplier'sparticularfacilityprovided90%ofallbrakepartscomponentsandpracticallyallbrakevalvesforToyota,beforeitwasdestroyedNishiguchiandBeaudet,1998.ItisestimatedthattheimpactonToyota'snetincomefromthissingleeventatAisinwasadecreaseof$300million.Thereafter,ToyotasoughtatleasttwosuppliersforeachpartTreece,1997.Operationally,multiple-sourcingprovidesgreaterassuranceoftimelydelivery,andgreaterupsidevolumeexibilityduetothediversicationoftherm'stotalrequirementsRamasesh,et.al.,1991.Single-sourcingallofarm'srequirementsalsoexposesthebuyingrmtohold-uprisk.LandRover'scontractualproblemwithitsonlychassissupplierisanexampleoftheoperationaldicultiesthissituationcreatesLester,2002.Strategically,supplierpowerexertedonthebuyerisweakenedwhenthermsplitsitstotalrequirementsamongmultiplesources.MultiplesourcinghedgestherisksofcreatingamonopolisticsolesourcesupplybaseandsupplierforwardintegrationNewman,1989.Insum,thechoiceofmultiple-sourcingversussingle-sourcingdependsonthetrade-obetweenthebenetsofmultiple-sourcingversusthoseofsingle-sourcing.

PAGE 26

14 1.5 OrganizationofthisDissertationTopositionthisresearchinthebroaderareaofsupplychainmanagementresearch,thischapterhasprovidedanoverviewofsupplychainmanagementandstrategicsourcing.Theremainderofthisproposalisorganizedasfollows.Chaptertwoprovidesacomprehensiveliteraturereviewofresearchinsupplychainmanagementthatrelatestosupplierselectionandrequirementsallocationdecisions.InChapters3-5weseektocharacterizeconditionsunderwhicharmshouldchooseeitheraspecializedi.e.,singlesupplierorageneralizedi.e.,multiplesupplierssourcingstrategyunderdierentscenarios.Morespecically,chapter3incorporatestheimpactofvariabilityinbothsupplyanddemandinaddressingsourcingdecisions;chapter4examineshowthesourcingstategyismoderatedwhensuppliersoeravarietyofcommonquantitydiscountschemes;andchapter5integratestheimpactofexplicitdiversicationbenetsonthesourcingdecision.Chapter6providesasummaryofkeyresultsfromthisresearch,anddiscussionofopportunitiesforfutureresearch. 1.6 StatementofContributionResearchintheareaofsupplychainmanagementcontinuestobeactive.Thisresearchfocusesontheupstreamconnectionsofabuyingrmtoanapprovedsuppplierbaseundervariouscommonlyencounteredindustrialoperatingcondi-tions.Weaddtotheexistingbodyofknowledgeinsupplychainmanagementbysurveyingcloselyrelatedexistingresearch,anddevelopingandanalyzingrealisticmodelsforstrategicsourcingdecisionmaking.Ourfocusonoptimaldecisionmak-inginregardtothenumberofsupplierstorequisitionhighlightstheimportanceofriskmitigationthroughsupplierdiversication.Inchapter3weillustratethisvaluetothermbymodelingtherm'spurchasedecisionwithbothupstreamanddownstreamuncertainty.weareabletoanalyticallysolvethissourcingproblemforoptimalsolutionsandshowthatinsomescenarioscostdominatesreliability.

PAGE 27

15 Further,ouranalysisisuncommonlygeneralascomparedtoexistingworkforthisparticularproblem.Inchapter5weexplicitlyconsiderstrategicbenetsofdiversicationwithunreliablesupplyanddownstreamproductdemanduncertainty.Thisapproachallowsrmlevelexibilityinderivingthemoststrategicallydesiredsizeoftheselectedsupplierpool,andgaugesthemarginalcosttradeoofselectinghigherunitcostsupplyagainsttheimputedbenetsofsupplydiversication.Chapter4considersstrategicbehaviorofcapacitatedsuppliersviaquantitydis-countpricingquotes.Wedevelopconstant,lineardiscount,incrementaldiscountandall-unitsdiscountpricingschemessourcingmodels.Forthediscountpricingmodelswendthattheyareconcaveminimizationproblemsandcombinatorialinnature.Wedevelopanoptimalbranchandboundalgorithmandwell-performingheuristicstoaidinstrategicsourcingdecisionmakinginthisoperatingenvi-ronment.Inallthisresearchprovidesguidanceforpractitionersandacademicsconcernedwithsuppliermanagementinregardstothefrequentdecisionmakingtasksofchoosingtheappropriatenumberofsupplierstosourcefromandtheirrespectiveorderquantityallocations.

PAGE 28

CHAPTER2LITERATUREREVIEW 2.1 OverviewTherstsectioninthischapterdescribestheevolutionoftheroleofsourcingfromapurelytransactionalpursuittooneofstrategicinuence.Next,areviewofliteratureexaminingthebuyer-supplierrelationshiprevealsthedividebetweenproponentsofsinglesourcingandmultiplesourcing.Theliteraturereviewedinthenalsectionfocusesonsupplierqualication,supplierselection,andquantityallocationcriteria. 2.2 StrategicEvolutionofSourcingEllramandCarr1994provideahistoryandreviewofliteraturerelatedtosourcing'sstrategicimportance.Theynotethatevenastheoilcrisisof1973-1974highlightedtheperilsofrawmaterialsupplyshortages,researchonindustrialbuyingbehaviorlargelyviewedthepurchasingfunctionasadministrative.Itwasnotuntilthe1980swhenPorter'sFiveForcesmodelgainedpopularitythatthestrategicroleoftheinterfacebetweensuppliersandbuyerswasbetterunderstood.Thishasledtomorecontemporaryresearchinvestigatingthestrategicimpactofsourcingasanintegrativelinkbetweenthermanditssuppliers.Astheinterfacebetweensuppliersandtherm,purchasing'sinuenceonrmperformanceincreasesassuppliercontributiontothermincreases.Further,theinclusionofpurchasingsourcingstrategyincorporatestrategyismorevitalasglobalcompetitionandthepaceoftechnologicalchangequickens.Arm'ssuppliermanagementorientationisreectedinitscontractingpoliciesforexternalpurchases.CohenandAgrawal99modelthetrade-osbetweenshorttermandlongtermcontractualrelationships.Shorttermcontractsprovidemoreexibility 16

PAGE 29

17 andavoidxedinvestments,butalsoforgoimprovementandpricecertaintybenetsaordedfromlongtermcontracts.Theiranalysisrevealsthatshorttermcontractingisoptimalunderawiderangeofconditions.Inasurveyofsupplymanagerstheyndthatmanagementintendtodeveloplongtermrelationships,butoftenengageinshorttermcontracting.Thisconictovertheoptimalbuyer-supplierrelationshipisnotexclusivetotheranksofsupplymanagers. 2.3 Buyer-SupplierRelationshipsResearchonthenumberofsourcesforfulllmentofproductrequirementsissomewhatcontroversial.Atoneextreme,wehaveempiricalevidenceofmanyrmsshrinkingtheirsupplierbaseperitemandorderingthemajorityoftotalunitsrequiredfromasinglesourcee.g.,Spekman988andPillingandZhang992.Further,thedocumentedbenetsofsingle-sourcingsuchasquantitydiscountsfromorderconsolidation,reducedorderleadtimes,andlogisticalcostreductionsasaresultofascaleddownsupplierbase.Hahnetal.,86,andBozarthetal.,998reinforcethisevidence.Infact,MohrandSpekman94contendthatsingle-sourcingperformancebenetsoftenoutweighthebenetsofapricecentricmultiple-sourcingstrategy.ThesebenetsarealsoenhancedbyincorporatingperformanceimprovementcriteriainmanagingsupplierrelationshipsFawcettandBirou,1992.Incontrast,Bhote987observedthatrelationshipmanagementcosts,intermsoftimeandcapital,mayoutweighthebenetsofsingle-sourcing.Theprimaryrationaledrivingthisargumentisthatsingle-sourcingrequiresthermandthesuppliertodevelopapartnershipbasedontrust.Inlinewiththisreasoning,McCutcheonandStewart00assertthatthepartiesmustachievegoodwilltrusttohaveasuccessfulpartnership.Further,theyconjecturethatthisleveloftrustisrarelyattained.Adversarialhistorybetweenchannelmembersandcompetitionforlargersharesofaproduct'slimitedtotalmargindrivedistrust.A

PAGE 30

18 signicantamountofresearchexistsonprocurementcompetitionswheresupplybasesizeandorderallocationisofconcern.Elmaghraby00andMinner003bothoerexcellentoverviewsofthesupplychainliteraturerelevanttorequirementsallocationdecisions. 2.4 StrategicSourcingAspreviouslystated,arm'ssourcingstrategycanbecharacterizedbythreekeyinterrelateddecisionsBurkeandVakharia,2002:acriteriaforestablishingasupplierbase;bcriteriaforselectingsuppliersasubsetofthebasewhowillreceiveanorderfromtherm;andcthequantityofgoodstoorderfromeachsupplierselected.Literaturerelatedtothesekeydecisionsisnowreviewed. 2.4.1 QualicationCriteriaQualityanddeliveryarestrategicallyimportantsupplierqualicationcrite-ria.Inthecontextofthesecriteria,asarm'ssupplymanagementorientationincreases,supplierandbuyerperformanceincreasesShin,etal.,2000.Thisrein-forcesDickson's1966surveyofvendorselectioncriteria,whichranksqualityanddelivery,respectively,astherstandsecondmostimportantsupplierattributes.Similarly,inasurveyofcompaniesatdierentlevelsoftheU.Sautoindustry,ChoiandHartley1996foundthatpriceisoneoftheleastimportantfactorsforsupplierselectionacrossthesupplychain,whileconsistencydenedasthemar-riageofqualityanddeliverywasthemostimportantcriterion.Areviewofpriorresearchratesqualityasthesingleextremelyimportantcriterion,whiledeliveryisofconsiderableimportanceWeberetal.,1991.Measurementsofqualityarediculttogeneralizeacrossindustries.Garvin987outlinesseveraldimensionsofquality,includingperformance,features,reliability,conformance,durability,serviceability,aestheticsandperceivedquality.Inpractice,armmayusequalitymeasurestoscreenoutallsupplierswhofailtomeetsomepre-speciedminimalqualitylevel.Forexample,HillviewHospital

PAGE 31

19 utilizedsuchapolicyineliminatingpotentialsuppliersforhandsoapbaseduponthesoap'sabilitytoeradicateinfection-causinggermsMaurer,1997.Throughaqualicationprocess,anapprovedsupplybasecanbeassembledtoensureminimalcapabilitiesonkeydimensions. 2.4.2 SelectionandAllocationAsnotedearlier,supplierselectionisconcernedwithidentifyingthesubsetofqualiedsupplierswhowillbeconsideredfororderplacement,andallocationfocusesonsplittingtherequiredquantitybetweentheselectedsuppliers.Obviouslythesedecisionsareinterdependentandarealsodrivenbythetotaldeliveredcoststothermofanorderquantityfromeachsupplier.Pan1989proposesalinearprogrammingmodeltooptimallyidentifythenumberofsuppliersandtheirrespectivequantityallocationstomeetpre-speciedproductrequirements.Otherconstraintsincorporatedarerelatedtoaggregateincomingquality,leadtimes,andservicelevel.Theoverallobjectiveistominimizethepriceperunitasaweightedaverageofselectedsuppliers'prices.Itisassumedthatproductrequirementsaredeterministicandsupplyisreliableandunlimited.Inreality,however,itiscommonforsupplierstoquotealternativepricingschemesanduncertaintyexistsinbothsupplyanddemandmarkets.Sincethefocusofthisdissertationisonincorporatingtheseconsiderationsinmakingsupplierselectionandquantityallocationdecisions,priorresearchrelatingtothisareaarereviewednext.Analyticalstudiesonsupplierselectionandquantityallocationdecisionsshowthatincertaincases,multiple-sourcing,order-splitting,ordiversicationispreferabletosingle-sourcing.Horowitz986providesaneconomicanalysisofdualsourcingasingleinputatdieringcosts.Itisshownthatuncertaintyinsupplypriceandrisk-aversionofthebuyermotivatearmtoplacepositiveordersfromthehighcostseller.KelleandSilver990investigateacontinuousreviewinventorypolicyreplenishedbysupplierswithstochasticdeliverylead-times,and

PAGE 32

20 ndthatorder-splittingamongmultiplesourcesreducessafetystockwithoutincreasingstockoutprobability.Ramaseshetal.991alsoanalyzeareorderpointinventorymodelwithstochasticsupplylead-time,andndthatinthepresenceofloworderingcostsandhighlyvariablelead-times,dualsourcingcanbecostpreferable.GerchakandParlar990examinesecond-sourcinginanEOQcontexttoreducetheeectiveyieldrandomnessofabuyingrm'spurchasequantity.Thebenetsofdiversicationaretraded-oagainstthecostsofmanagingalargersupplybasetodeterminewhethersecond-sourcingisworthwhile.Theyalsoanalyzetheoptimalnumberofidenticalsources.Rosenthaletal.95introduceamixedintegerprogrammingmodelforsolvingasupplierselectionproblemwithbundling.Thesuppliersarecapacitated,oerdierentprices,dieringqualitylevels,anddiscountbundles.AgrawalandNahmias97examineasingleperiodsupplierselectionandallocationproblemwithnormallydistributedsupplyanddeterministicdemandforasingleproductwithxedorderingcosts.Theyareabletoshowthatfortwonon-identicalsuppliers,theexpectedprotfunctionisconcaveinthenumberofsuppliers.ParlarandWang993comparethecostsofsingleversusdual-sourcingforarmassumingthattheoverallobjectiveistominimizepurchasingandinventoryrelatedcosts.Intheirapproach,theyassumethatactualincomingquantitiesareafunctionofarandomvariablerepresentingtheyield.SeparatelyusinganEOQandnewsboybasedorderingpolicy,theyareabletoshowthatincertaincasesdual-sourcingdominatessingle-sourcing.Bothofthesestudiesignorethesuppliercapacityissueinmakingsupplierselectionandquantityallocationdecisions.Further,ParlarandWang93notethatsupplieryieldsanddemanduncertaintyplayacriticalroleintheanalysis.

PAGE 33

21 Otheranalyticalstudiessimilartothisresearchexaminesupplierselectionandorderallocationdecisionswithstochasticdemandfortheproductpurchased.GallegoandMoon1993employScarf'sorderingruleforadistributionfreeoptimalnewsboyorderquantity.Theymaximizeprotagainsttheworstpossibledistributionofdemandwithknownmeanandvariance.Separateextensionsincorporateasecondpurchasingopportunity,xedorderingcosts,randomyields,andmultipleitemsintotheanalysis.Inparticular,thecaseofrandomsupplieryieldsassumesthateachunitsuppliedhasthesameprobabilityofbeinggood,andthebuyerpaysforallunits.BassokandAkella991introducetheCombinedComponentOrderingandProductionProblemCCOPP.Theproblemisoneofselectingorderingandproductionlevelsofacomponentandanishedgoodforasingleperiodwithsupplyanddemanduncertainty.Intheirmodelthedistributionofsupplydependsontheorderquantitygiventoasinglesource.AnupindiandAkella993consideratwosupplier,singleproductprocure-mentproblemwithstochasticsupplyanddemand.Theysuggestthatminimumorderquantitypoliciesofsuppliersmayaecttheirndings.Gurnanietal.000simultaneouslydetermineorderingandproductiondecisionsforatwocomponentassemblysystemfacingrandomnishedproductdemandandrandomyieldfromtwosuppliers,eachprovidingadistinctcomponent.Theyalsoconsiderajointsup-plieroptionanddeterminethevaluetotheassemblerofreliablecomponentsupply.Diversicationoccurswhenpositiveordersareplacedwiththejointsupplierandindividualcomponentsupplier.Theyshowthatiftherearenotmismatchedinitialinventoriesofcomponents,eachcomponentwillbesinglesourced.Kim,etal.002modelacapacitatedmanufacturer'ssupplybasecongurationproblemformultiplepartsusedtoproducemultipleproductswithindependentstochasticdemands.Theydevelopaniterativealgorithmtodeterminethemanufacturer'sendproductmixandspecifythesupplierswhowillbeusedtosatisfytheparts

PAGE 34

22 requirementsforeachendproduct.Intheirmodelsuppliersarecapacitatedandreliable.YanoandLee95alsooerareviewofthenormativeliteraturewhichaddressestherandomyieldproblem.Orderquantityorlotsizingdecisionscanbelargelyinuencedbyalternativesupplierpricingschemes.Quantitydiscounts,especiallyall-unitsandincremental,arecommonpricingpractices.Inaquantitydiscountschedule,therangeofpotentialpurchasequantitiesissegmentedbyquantitybreakpoints.Orderquantitiesthatfallbetweentwobreakpointsqualifyforaspecicunitprice.Typicallythelargerthequantityordered,theloweristheunitprice.Thekeydierencebetweenanall-unitdiscountandincrementalscheduleisthatintheformer,allunitsorderedaresuppliedattheunitprice,whileinthelatter;onlythenumberofunitsinaspecicbreakpointaresuppliedattheunitprice.Suppliersoerthesetypesofschedulestoencouragebuyerstoprocurelargerquantitiesandtoreaptheoperatingadvantagesassociatedwiththeselargerquantitiessuchaseconomiesofscale.Boththebuyerandsuppliercanrealizehigheroverallprotsbymakingdecisionsjointly.SCMcoordinationresearchinpartresidesintheresearcheortsforjointoptimalsupplierpricingschemesandbuyerlotsizedecisionsthatconsiderquantitydiscounts.Operationsliteraturetakesacoordinatedcostminimizationapproachforanalyzingtheimpactofpricingschemesonsupplychains.MunsonandRosenblatt001analyzeathree-leveldistributionchannelforasingleproductwhereasupplierprovidesanorderquantitydiscountscheduletoamanufacturer.Themanufacturerthenproposesanorderquantitydiscountscheduledownstreamtoaretailer.Themanufacturersolicitsandproposesorderquantitydiscountsupstreamanddownstreamrespectively,andistermedthechannelcaptainforhisdictatorialrole.Assuch,themanufacturerretainsallchannelsavingsfromcoordinatedquantitydiscounts.Numericalexperimentsprovidea40%increaseinmanufacturer

PAGE 35

23 savingsfromtwowayorderquantitydiscountsoverretaileronlyquantitydiscounts.Further,optimizingoverbothsupplierandretailersimultaneouslyresultsinsavingsgreaterthanthesumofoptimizingovereachindividually.Weng1995developsanintegratedmodelofquantitydiscountsandchannelcoordination.Inhismodel,operatingcostsareafunctionofpurchasequantityanddemandisafunctionofsellingprice.Inthisscenario,protincreasesresultfromcostreductionsanddemandenhancement.However,aquantitydiscountschemealoneisnotenoughtoachievejointprotmaximization.Axedpaymentorfranchisefeeinconcertwithaquantitydiscountscheduleisnecessarytomotivatesystemoptimaldecisionmaking.Abad988a,1988balsoincorporatespricedependentdemandintolot-sizingmodelswithalternatesupplyacquisitionschemes.Optimallotsizeandsellingpricearesimultaneouslysolvedunderlinearandconstantprice-elasticdemand.Asupplieroersanall-unitquantitydiscountinAbad988aandanincrementalquantitydiscountin988b.Aniterativeprocedureisdevelopedtohandlethelot-sizeandsellingpriceinterdependency.Burwelletal.97extendtheworkofAbadbydevelopinganoptimallotsizingandsellingpricealgorithmforasingleitemgivensupplieroeredquantityandfreightdiscountsandpricedependentdownstreamdemand.Numericalresultsindicatethatthemaximizedprotwithall-unitsdominatesthemaximizedprotsunderincrementalandmixedquantityandfreightdiscounts.Itisclearfromtheabovestudiesthatorderquantitydiscountscaninuencedecisionmakingateachsupplychainlinkregardlessofanalyticalperspective.BentonandPark1996provideathoroughreviewofthelot-sizingliteraturewithquantitydiscountconsiderationsfortimephasedaswellasnon-timephaseddemands.Theirtaxonomyfurtherclassiesresearchbasedonthetypeofdiscountall-unitsorincrementalandtheperspectiveofthemodelingeortbuyeror

PAGE 36

24 supplier-buyer.Thescopeofthisdissertationisoneofnon-timephaseddemandfromabuyerperspective.Assuch,theremainderofthissectionreviewsliteraturewithsimilarscope.Benton91usesLagrangianrelaxationtoevaluateapurchasingmanager'sresourceconstrainedorderquantitydecisionsgivenalternativepricingschedulesfrommultiplesuppliers.Theexampledecisionmakerhasalimitedbudgetandstoragespacefortenitemsoeredbythreevendors,eachquotingthreeall-unitsdiscountlevelsforeachitem.Theobjectiveistominimizetotalacquisitionandinventorycosts.Themanagermustchooseasinglesupplierforallitems.However,ifmultiplesourcingisallowed,theoptimalobjectiveis8%lowerthanthesinglesourceoptimum.RubinandBenton1993useLagrangianrelaxationtoformulateaseparabledualproblem.Abranchandboundalgorithmwithabestboundbranchingruleisdevelopedtocloseanydualitygap.Toallowmultiplesourcingamongthemultipleitems,amergeddiscountscheduleisconstructedwhichquotesthelowestpriceamongallsuppliersforeachquantityofeachitem.Ashortcomingofthismodelisthattherearenoordercostsavingsfromconsolidateddeliveryororderplacement.Asanextensiontothiswork,RubinandBenton003analyzethesamepurchasingscenario,exceptthatsuppliersquoteincrementalquantitydiscountschedulesinsteadofall-units.Asimilarsolutionmethodologyisemployed.Theynotethebestfeasiblesolutionobtainedfromtheirrelaxationalgorithmshouldbeacceptableforuseunlessalargedualitygapexists.Further,theyobservethatinnumericalstudiestheprimalcostminimizationobjectiveisratherat.Thisimpliesthatgoodsolutionsarerathereasytond,butfathomingisslow.Thus,anoptimalsolutionsearchisprobablymorecostlythanbenecial.AnotherLagrangianrelaxationbasedheuristicisdevelopedbyGuderetal.994tosolveabuyer'smultipleitemmaterialcostminimizationproblemwithincrementaldiscountsoeredbyasinglesupplier.Thebuyerhasasingleresource

PAGE 37

25 constraintanddemandsforitemsareindependent.Thecomplexityoftheproblemliesinevaluatingallfeasiblepricelevelsequences.Forlargeproblemswithnitemsandmpricebreaks,theoptimalpricelevelsequencecanbeobtainedinOnm.For100itemswith8pricebreakseach,numericalexperimentsresultinprecisesolutionsquickly.TheirheuristicisadaptedfromPirkulandAras985whichanalyzedtheall-unitsversionofthisproblem.Themajorityofresearchonpurchasedecisionsinvolvingalternatepricingschemesassumesthatdemandforthepurchasedproductisknown.ModelingeortsthatconsiderstochasticdownstreamdemandandupstreamquantitydiscountsareJuckerandRosenblatt985,Khouja96a,b,andLinandKroll997.Alloftheseanalyzetheimplicationsofquantitydiscountsonthenewsboyproblem.JuckerandRosenblatt985developamarginalanalysisbasedalgorithmtominimizetotalcostsofpurchasingandtransportationwithseparatebutsimultaneouspurchasingandtransportationquantitydiscounts.Khouja996a,bimplementsabisectionmethodtosolveforaretailer'soptimalorderquantitywithsupplieroeredall-unitsdiscountsandprogressivemultipleretaildiscounts.LinandKroll997developasolutiontechniqueforsolvingariskconstrainednewsboyproblemwithall-unitsorincrementalsupplierpricingschemes.AsevidencedbytheproliferationofresearchonintegratingsupplychainsandadoptionofSCMpracticesinindustry,theimportanceofsuppliermanagementonarm'scompetitivestrategyisnowwidelyaccepted.Arm'ssuppliermanagementapproachischaracterizedbyitssourcingstrategy.Thisresearchprovidesmanage-rialinsightbyexaminingtheinuenceofalternatepricingschemesfromsuppliers,uncertaintyinproductsupplyanddemand,andexplicittreatmentofdiversi-cationbenetsonarm'sstrategicsourcingdecisions.Ofparticularinterestischaracterizingwhenitisoptimalforarmtosingle-sourceversusmultiple-source.

PAGE 38

26 Ourmodelingandanalyticalapproachestotheseproblemsofsupplierselectionandorderquantityallocationpositionthisresearchamongtheexistingliteraturereviewedinthischapter.

PAGE 39

CHAPTER3STRATEGICSOURCINGDECISIONSWITHSTOCHASTICSUPPLIERRELIABILITY 3.1 IntroductionSuppliersourcingstrategiesareacrucialfactordrivingsupplychainsuccess.Whilemanyrmsutilizeasinglesupplierforaparticularitem,othersdiversifytheirsupplyriskbysourcingfrommultiplesuppliers.Inparticular,arm'sallocationdecisiondetermininganappropriatesuppliersetandorderallotmentimpactsonallcompetitivedimensionsforthedeliveryofnishedgoodstoitscustomers,includingcost,quality,reliabilityandexibility.Benetsofasinglesupplierstrategyhavebeentoutedinthepopularpress,suchasJITreplenishmentandincreasedqualitylevels.Morerecently,thebenetsofsupplierdiversicationasapossibledefenseagainstsupplydisruptionhasgainedattention.Inthischapter,weinvestigatetheimplicationsofuncertainsupplierreliabilityonarm'ssourcingdecisionsinanenvironmentwithuncertaindemand.Inparticular,wecharacterizecircumstancesunderwhicharmshoulddiversifyitsordersamongstseveralsupplierstoincreaseitstotalprot,ratherthanutilizingasinglesuppliersourcingstrategy.ThemodelingframeworkdevelopedinthischapterissimilartoAnupindiandAkella993.Theyanalyzebothsingleandmulti-periodmodelswithuncertainsupplierreliabilityandstochasticdemand.Inparticular,inatwosuppliersetting,theycharacterizescenariosunderwhichitwouldbeoptimaltosourcefromonevs.twosuppliersbasedoninitialinventorylevels.Whiletheirgeneralresultsarederivedforthetwosuppliercasewithnospecicdistributionalassumptions,theyalsoanalyzespecicscenariosutilizinganexponentialdemanddistribution, 27

PAGE 40

28 andnormalandgammadistributionstoreectsupplierreliability.ParlarandWang993alsoconsideratwosuppliernewsboymodelwithrandomyields.Theyobtainrstorderoptimalityconditions,andestablishtheconcavityoftheobjective.Furthermore,theyutilizealinearapproximationtoobtainoptimalorderquantitiesforeachsupplier.AnothercloselyrelatedworkisAgrawalandNahmias1997.Theyanalyzeavariationofthesingleperiodnewsvendormodelwheredemandisactuallydeterministic,butsupplierreliabilitiesarenormallydistributed.Inaddition,theyexaminetheoptimalnumberofsuppliersandcorrespondingorderquantitiesassumingthataxedordercostisincurredforeachsupplierwithapositiveorder.OptimalpoliciesforNsuppliersarederivedforthecaseofhomogenoussuppliersi.e.,allsuppliershavesimilarreliabilitydistributionsandcosts.Forthecasewithheterogeneoussupplierreliabilityandsimilarsuppliercosts,theyshowoptimalityconditionsfortwosuppliersandconjecturethatthesealsoholdforcaseswithmorethantwosuppliers. 3.2 SourcingModelInthissection,weintroducethemodelnotationandvariables.First,weintroducethedecisionvariables.LetNreectthenumberofavailablesuppliers.Weassumethatthissethasbeenpre-qualiedsuchthattheyallmeetminimumsourcingstandardssetbytherm.Thekeydecisionvariableinthemodelistodeterminethenumberofunitstopurchasefromsupplieri,wherei=1;:::;N.Althoughwedonotincludeanexplicitvariablereectingtheoptimalnumberofsuppliers,itcanbedeterminedimplicitlybyidentifyingthenumberofsupplierswithanon-zeroorderquantity.Next,weintroducethesupplierspecicparameters.Weassumethateachsupplierquotesaunitcostofci.Inaddition,thermhassomeknowledgeofthehistoricalreliabilityintermsofthenumberofgoodunitsdeliveredforeach

PAGE 41

29 supplier.Wetreatthisqualityoryieldreliabilityforeachsupplierasarandomvariable,ri.Letgiridenotethecontinuousprobabilitydensityfunctionassociatedwithyieldforeachsupplieri.Wealsoassumethatthedensityfunctionistwicedierentiablewithriandirepresentingthemeanandstandarddeviation.Finally,severalrmspecicparameterstypicallyassociatedwithasingle-periodnewsboyframeworkarerelevanthere.WeassumethatthetotaldemandisunknownbutrepresentedbyastochasticparameterxwithaprobabilitydensityfunctionfxandacumulativedistributionfunctiondenotedbyFx.Also,theunitpricep,salvagevaluesandunderagecostsuarealsoassumedtofollowstandardassumptionsassociatedwithanewsboymodel,includingp>ci>s.AcompletelistofthemodelvariablesandparametersisincludedinTable3-1below. Table3{1:ModelNotation Variable Description N Numberofsuppliersavailable qi Numberofunitstopurchasefromsupplieridecisionvariable x Demandrandomvariable fx Probabilitydensityfunctionassociatedwithdemand Fx Cumulativedistributionfunctionassociatedwithdemand Meandemand a Minvalueparameterforuniformdemanddistribution b Maxvalueparameterforuniformdemanddistribution ci Unitcostforsupplieri ri Yieldforsupplierirandomvariable giri Probabilitydensityfunctionassociatedwithyieldforsupplieri Giri Cumulativedistributionfunctionassociatedwithyieldforsupplieri ri Meanyieldfactorforsupplieri i Standarddeviationoftheyieldfactorforsupplieri Vi Secondmomentoftheyieldfactorforsupplieri p Unitprice s Unitsalvagevalue u Unitunderagecost ci Unitcostforsupplieri Q Totalnumberofgoodunitsreceived Ki Criticalratioderivedfromcostforsupplieri bi Uniformdemandadjustedbycriticalrationforsupplieri b' Uniformdemandadjustedcriticalrationforhomogeneoussuppliers

PAGE 42

30 3.3 ModelDevelopmentTheobjectiveofthermistodeterminetheappropriateorderquantitiesforeachsuppliersuchthattheexpectedprotassociatedwithsatisfyingdemandismaximized.UtilizingtheframeworkfromthetraditionalnewsboyproblemSilver,PykeandPeterson1998,theobjectivefunctioninEquation 3.1 maximizesthesingleperiodseasonexpectedprotsfortherm.Notethatweassumethatthebuyingrmonlypayssuppliersfor"good"'unitsdelivered.Inaddition,anon-negativityconstraintEquation 3.2 isalsoincludedinourformulation.MaxE=1Z0g1r11Z0g2r2:::1Z0gNrN24QZ0px)]TJ/F23 7.97 Tf 16.805 14.944 Td[(NXi=1ciriqi+sQ)]TJ/F22 11.955 Tf 11.955 0 Td[(xfxdx35drN:::dr2dr1+1Z0g1r11Z0g2r2:::1Z0gNrN241ZQpQ)]TJ/F23 7.97 Tf 16.804 14.944 Td[(NXi=1ciriqi)]TJ/F22 11.955 Tf 11.955 0 Td[(ux)]TJ/F22 11.955 Tf 11.955 0 Td[(Qfxdx35drN:::dr2dr1.1subjecttoqi08i.2whereQ=NPi=1riqi 3.4 AnalysisInthissection,wedescribetheoptimalsuppliersourcingstrategyundercertainconditions.Tomakeouranalysismoretractable,weassumethatdemandisuniformlydistributedwithparameters[a;b]. 1 Tosupportsubsequentanalysis,werstensurethatakeyresultholds. 1Giventhatpriceperunitpisassumedtobexed,itisreasonabletoassumethatdemandaroundthisxedpriceisuniformlydistributed.

PAGE 43

31 Corollary3.1: TheexpectedprotfunctionshowninEquation 3.1 isconcaveintheorderquantitiesqiforNsupplierswhendemandisuniformlydistributedbetween[a;b].Proof :SeeAPPENDIXA.ItisinterestingtonotethatAgrawalandNahmias97wereunabletoproveconcavityoftheprotfunctionforNheterogeneoussupplierswithnormallydistributedreliabilityfunctionsanddeterministicdemand.However,byassumingthatdemandisuniformlydistributed,wecanderiveaclosedformexpressionforoptimalsupplierorderquantitiesmakingnodistributionalassumptionsforsupplierreliability.Westartouranalysisforthegeneralcaseofheterogeneoussuppliers.Thenweanalyzethecasewheresuppliersarehomogeneouswithrespecttothereliabilitydistributionsbuthavedieringcoststructures.Next,weconsiderthecasewherethereverseistruei.e.,allsuppliershavehomogeneouscoststructuresbutdierentreliabilitydistributions.Finally,weanalyzethecasewhereallsuppliersarecompletelyhomogeneous. 3.4.1 HeterogeneousSuppliersWhenallsuppliershaveheterogeneouscostandreliabilityfunctions,akeyresultisthatthereisnoonedominantsourcingstrategy.Thatis,undercertaincircumstances,itwillbeoptimaltoplaceasingleorderwiththelowestcostsup-plier;underothercircumstances,anorderisplacedwithasubsetofthesuppliers.However,itispossibletoanalyticallydeterminetheexactorderquantitiesforeachsupplierasshowninTheorem3.1below.

PAGE 44

32 Theorem3.1 :Whensuppliersareheterogeneouswithrespecttocostsandreliabil-ityparameters,thentheoptimalsourcingquantityforeachsupplieriis:qi=ri 2i"NPj=1bi)]TJ/F22 11.955 Tf 11.955 0 Td[(bjrj j2+bi# NPj=1rj j2+1.3wherebi=Kib)]TJ/F22 11.955 Tf 11.955 0 Td[(a+a,andKi=p+u)]TJ/F23 7.97 Tf 6.586 0 Td[(ci p+u)]TJ/F23 7.97 Tf 6.586 0 Td[(s.Proof :SeeAPPENDIXA.Ofcourse,theresultofTheorem3.1needstobemoderatedtoaccountforanegativeorderquantityi.e.,,qj*ascomputedaboveinEquation 3.3 islessthanzeroforsomesupplierj.Later,weanalyzethecaseincludingthenon-negativityconstraints.Wenowfocusonfurthercharacterizingthesituationwhereasinglesuppliersourcingstrategyisappropriate.AlthoughtheresultsofTheorem3.1donotprovideguidelinesintoageneralizableoptimalsourcingstrategy,therearecertainspecicinsightsfortheheterogeneoussuppliercasewhichcanbeobtainedasdiscussedbelow.Corollary3.2: Thermwillalwaysorderapositivequantityfromthelowestcostsupplier.Proof :SeeAPPENDIXA.Thekeyinsightstemmingfromthisresultisthatsuppliercoststructuresdominatethereliabilitydistributionsforselection.Basedonthis,wecanderiveadditionalinsightsintooptimalsourcingstrategiesfortherm.Werststartbyindexingsuppliersi=1;::;Ninincreasingorderofcostsi.e.,c1c2c3:::cN.Corollary3.3belowspecieswhenasinglesuppliersourcingstrategyisoptimalfortherm.

PAGE 45

33 Corollary3.3 :Thermwillsourceallitsrequirementsfromthelowestcostsupplieri.e.,useasinglesourcingstrategyifandonlyif:1 r12
PAGE 46

34 rstsupplierwilloptimallyreceivethecompleteorder.Notethatthereliabilitydistributionissuchthatmostvaluesarelessthan100%.Consequently,thisnecessitatesthatr<1.Forexample,AgrawalandNahmias97assumethat3
PAGE 47

35 costsupplierwillalwaysreceiveapositiveorder.Corollary3.4discussestheorderquantityforthelowestcostsupplierrelativetotheothersuppliersbysimplystatingthat,althoughthelowestcostsupplierwillreceiveanorder,theactualordersizemaybelowerthanotherhighercostsuppliers.Toillustrate,considerthecasewhereadualsupplierstrategyisoptimal.Inthissituation,ifthestandarddeviationoftherstlowestcostsupplierisfairlyhigh,thenthesecondhighercostsuppliermayreceiveahigherorder.Corollary3.4: Whenamultiplesuppliersourcingstrategyisoptimal,thelowestcostsupplierwillnotnecessarilyreceivethehighestorderquantity.Proof: SeeAPPENDIXA.Thelogicusedinthetwosupplierexamplefollowsthroughindetermininganoptimalsubsetofsupplierswhichreceiveapositiveorderquantity.Corollaries3.5and3.6summarizetheimportanceofsuppliercostindeterminingthisoptimalsubsetofsuppliers.Corollary3.5: Ahighercostsupplierwillneverreceiveapositiveorderwhenalowercostsupplier'sorderquantityisequaltozero.Proof: SeeAPPENDIXA.Corollary3.6: Theoptimalsubsetofsuppliersn*receivingapositiveorderquantityisthelowestcostsubsetofsupplierssuchthatthefollowingrelationshipshold:bn>n)]TJ/F21 7.97 Tf 10.821 0 Td[(1Pj=1bjrj j2 1+n)]TJ/F21 7.97 Tf 10.821 0 Td[(1Pj=1rj j2andbn+1nPj=1bjrj j2 1+nPj=1rj j2.5Proof :SeeAPPENDIXA. 3.4.2 HeterogeneousCostSuppliersToextendthebasicresultsderivedfromtheanalysisofheterogeneoussuppli-ers,wenowconsiderthesituationwheresupplierreliabilityishomogeneous,butthecostsarenot.Inparticular,leteachsupplierhaveacostofciandreliability

PAGE 48

36 parametersri=r,gi=g,Gi=G,andi=,8i.Theorem3.2characterizestheoptimalsolutionundertheseconditions.Theorem3.2: Whensuppliersahaveheterogeneouscoststructuresandbhaveidenticalreliabilitydistributions,thenthereisnoonedominantsuppliersourcingstrategy.Theoptimalorderquantityforeachsupplieriis:qi=r"r2NPj=1bi)]TJ/F22 11.955 Tf 11.955 0 Td[(bj+bi2# Nr22+48i.6Proof :SeeAPPENDIXA.Whilethermoptimallyplacesanorderwiththelowestcostsupplier,theordersfortheremainingsuppliersaredeterminedviacostdierentialsandvariancereduction.TheresultsofTheorem3.2aresimilartothoseofTheorem3.1,inthatthelowestcostsupplierwillreceiveapositiveorder,whiletheothersmaynot.AnupindiandAkella993analyzethetwosuppliercasewithstochasticdemandandsupplierreliabilitywhereonesupplierhasarelativecostadvantage.Theyshowthatwhentheinitialon-handinventoryfallsbelowsomeminimallevel,itisoptimalforthermtosourcefrombothsuppliers.Incontrast,Theorem3.2andEquation 3.4 foraNsuppliersettingwithuniformdemandshowsthat,intheabsenceofcapacityconstraints,therearecircumstancesunderwhichitisneveroptimaltoorderfrommorethanonesupplierwhenreliabilitydistributionsofsuppliersareidentical. 3.4.3 HeterogeneousReliabilitySuppliersInthiscase,wefocusonasituationwheresuppliercostsareroughlyequiv-alenti.e.,ci=c8i,buteachsupplierhasauniquehistoricalreliabilityfunctionassociatedwiththenumberofgoodsactuallydelivered.Theorem3.3characterizestherstorderconditionsofoptimalsourcingstrategyforthiscase.

PAGE 49

37 Theorem3.3: Whensuppliersahaveidenticalcoststructuresandbhaveuniquereliabilitydistributions,thenitisoptimaltoorderfromallsuppliers.Theoptimalorderquantityforeachsupplieriis:qi=ri 2ib0 NPj=1rj j2+18i.7whereb0=Kb)]TJ/F22 11.955 Tf 11.955 0 Td[(a+a,andK=p+u)]TJ/F23 7.97 Tf 6.587 0 Td[(c p+u)]TJ/F23 7.97 Tf 6.587 0 Td[(s.Proof :SeeAPPENDIXA.Fromtheexpressionsfortheoptimalorderquantity,thermoptimallyordersdierentnon-zeroorderquantitiesfromallsuppliers.Furthermore,itseemsthatthesupplierreliabilitiesdirectlyimpactontheorderquantitiessuchthatthebuyingrmrealizesthediversicationbenets.Theorderquantityforaparticularsupplierdependsnotonlyuponitsuniquereliabilityfunction,butalsoonthereliabilityoftheotheravailablesuppliers.Thisoptimalorderquantityforanindividualsupplierincreasesinresponsetothefollowing:aanincreaseinthemeanreliabilityofthatsupplier,badecreaseinthestandarddeviationofreliabilityofthatsupplier,cadecreaseinthemeanreliabilityofothersuppliers,anddanincreaseinthestandarddeviationofreliabilityofothersuppliers.Consequently,theorderquantityfortherstsupplierisadjustedfortheuncertaintyinthereliabilityofothersuppliers.Further,armoptimallysourcesfromtheentiresupplierpoolforthiscase. 3.4.4 HomogeneousSuppliersFinally,considerthescenariowherethesuppliersareroughlyequivalentincostsandreliabilityexpectations.Letci=c,gi=g,Gi=G,ri=r,andi=,8i.Theorem3.4characterizestheoptimalsolutionundertheseconditions.Theorem3.4: Whensuppliersahaveidenticalcoststructuresandbhaveidenticalreliabilitydistributions,thenitisoptimaltoorderthesameamountfrom

PAGE 50

38 allsuppliers.Theoptimalorderquantityforeachsupplieris:qi=rb0 2+Nr28i:.8Proof :SeeAPPENDIXA.Thisresultindicatesthatitisoptimalforthermtodiversifyitssupplybaseandplaceanequalorderfromeveryqualiedsupplier.Furthermore,itisobviousthatthetotalquantitysourcedbythermincreasesinthenumberofavailablesuppliersN.ThisresultconcurswiththosederivedbyAgrawalandNahmias997forasettingwithnormallydistributedreliabilitiesandconstantdemand.Corollary3.7: SupposemoftheNsupplierswithm
PAGE 51

39 3.5 NumericalAnalysis 3.5.1 ExperimentalDesignWeutilizeasetofparametervalueswhichsatisfytheassumptionsofthenewsvendormodel.Forthebuyingrm,weletp=19,s=2,andu=6.Demandisuniformlydistributedover[300,700]andthusthecorrespondingmeandemandi.e.,a+b=2is500whilethespreadofdemandi.e.,b)]TJ/F22 11.955 Tf 12.375 0 Td[(ais400.Fortheseexamples,weconsiderapossiblesupplierseteachmeetingminimalqualicationcriteriathatconsistsofthreesuppliers.Inaddition,weassumethatthesupplierreliabilitiesarealsouniformlydistributedovertherange[Mi-Li/2,Mi+Li/2]withameanofMiandaspreadofLi.Furthermore,weassumethatallthreesuppliershavesimilarreliabilitydistributionswithameanof0.7andaspreadof0.1.Forthebasecaseexamples,weassumethatthesuppliershavenotspeciedanyminimalorderquantities.Later,weinvestigatetheimpactofsuchconstraintsontheoptimalsolutions.Threesetsofexamplesareincludedwhichinvestigatetheimpactofchangesinvariousparametersgiventhatthesuppliersethascertaincoststructures.Thersttwosetsofexamplesillustratethecasewherethesuppliershaveheterogeneouscosts,whilethethirdsetofexamplesillustratesthecasewherethesuppliershavehomogeneouscosts.Notethattheheterogeneoussuppliersarealwaysrankorderedsuchthattherstsupplierhasthelowestcostandthelastsupplierhasthehighestcost.Fortherstsetofexamples,thesuppliercostsaresetatc1=6.75,c2=7.00,andc3=7.25.Forthesecondsetofexamples,thesuppliercostsaresetatc1=6.95,c2=7.00,andc3=7.05.Forthethirdsetofexamples,thesuppliercostsareequalwithc1=c2=c3=7.00.TheexperimentaldesignforthenumericalexamplesiscontainedinTable3-2.Werefertotherstexampleforeachseti.e.,Examples1,2and3asthebasecaseexamplewhichreectstheparametervaluesdiscussedhere.Within

PAGE 52

40 Table3{2:DescriptionofNumericalExamples Example CostType c1,c2,c3 ParameterChanged ParameterValue 1 Heterogeneous [6.75,7,7.25] BaseCase1 1A Heterogeneous [6.75,7,7.25] MeanDemand a+b/2=5200 1B Heterogeneous [6.75,7,7.25] SpreadDemand b-a=800 1C Heterogeneous [6.75,7,7.25] MeanReliabilitySupplier1 M1=.5 1D Heterogeneous [6.75,7,7.25] SpreadReliabilitySupplier1 L1=.5 1E Heterogeneous [6.75,7,7.25] MinOrderQuantitySupplier1 minq1=1000 2 Heterogeneous [6.95,7,7.05] BaseCase2 2A Heterogeneous [6.95,7,7.05] MeanDemand a+b/2=5200 2B Heterogeneous [6.95,7,7.05] SpreadDemand b-a=800 2C Heterogeneous [6.95,7,7.05] MeanReliabilitySupplier1 M1=.5 2D Heterogeneous [6.95,7,7.05] SpreadReliabilitySupplier1 L1=.5 2E Heterogeneous [6.95,7,7.05] MinOrderQuantitySupplier1 minq1=1000 3 Homogeneous [7,7,7] BaseCase3 3A Homogeneous [7,7,7] MeanDemand a+b/2=5200 3B Homogeneous [7,7,7] SpreadDemand b-a=800 3C Homogeneous [7,7,7] MeanReliabilitySupplier1 M1=.5 3D Homogeneous [7,7,7] SpreadReliabilitySupplier1 L1=.5 3E Homogeneous [7,7,7] MinOrderQuantitySupplier1 minq1=300 3F Homogeneous [7,7,7] MinOrderQuantityAll minq=300 eachsetofexamples,wevaryonespecicparameterrelativetothoseusedinthebasecase.TherstcolumninTable3-2containstheexamplereferencewiththecorrespondingnumberi.e.,1-3denotingtheparticularsettheexamplebelongsto,andthecorrespondingletteri.e.,A-Fdenotingaparametervariationrelativetothebasecaseexampleforthatset.Thesecondcolumnspeciesthetypeofcoststructureforthesuppliersi.e.,heterogeneousvs.homogeneous,whilethethirdcolumnidentiestheparticularunitcostsusedforeachexample.Finally,thefourthandfthcolumnsdescribetheparametersandtheircorrespondingvaluesthatarechangingrelativetothebasecaseexample.Toillustrate,inExample1Athethreesuppliershaveheterogeneouscostsc1=6.75,c2=7,andc3=7.25andexpecteddemandi.e.,a+b=2=5200muchhigherthanthatofthebasecase. 3.5.2 ResultsTable3-3containsasummaryoftheresultsofthenumericalexamples.Theparticularperformancemetricsincludedaretheoptimalorderquantityforeachsupplier,thetotalunitsordered,thenumberofsupplierswhichreceiveapositive

PAGE 53

41 order,andthermprot.Intheremainderofthissection,wediscusshighlightsfromthesenumericalexamples. Table3{3:ResultsforNumericalExamples Example ParameterChanged q1* q2* q3* Total n* Prot 1 BaseCase1 880 0 0 880 1 $5,353 1A MeanDemand 5619 1968 0 7587 2 $61,751 1B SpreadDemand 1048 0 0 1048 1 $4,604 1C MeanReliabilitySupplier1 1231 0 0 1231 1 $5,335 1D SpreadReliabilitySupplier1 174 700 0 875 2 $5,218 1E MinOrderQuantitySupplier1 0 874 0 874 1 $5,199 2 BaseCase2 803 73 0 876 2 $5,230 2A MeanDemand 3259 2529 1798 7586 3 $61,183 2B SpreadDemand 1038 0 0 1038 1 $4,458 2C MeanReliabilitySupplier1 759 333 0 1092 2 $5,220 2D SpreadReliabilitySupplier1 60 772 42 874 3 $5,202 2E MinOrderQuantitySupplier1 0 802 72 874 2 $5,199 3 BaseCase3 292 292 292 876 3 $5,211 3A MeanDemand 2529 2529 2529 7586 3 $61,158 3B SpreadDemand 346 346 346 1037 3 $4,430 3C MeanReliabilitySupplier1 249 349 349 946 3 $5,210 3D SpreadReliabilitySupplier1 17 429 429 875 3 $5,208 3E MinOrderQuantitySupplier1 300 288 288 875 3 $5,211 3F MinOrderQuantityAllSuppliers 438 0 438 875 2 $5,208 First,considertheimpactofthesuppliercoststructuresasshowninthebasecaseexamples.Examples1and2conrmTheorem3.2forthecasewithhetero-geneouscostsandhomogeneousreliabilities.InExample1,thecostdierentialsaresignicantenoughsuchthatthermoptimallyplacesasingleorderwiththelowestcostsupplier.Incontrast,Example2showsthesituationwherethecostdierentialsaresmallenoughsuchthatthermoptimallyplacesanorderwiththetwolowestcostsuppliers.Example3essentiallyillustratestheresultsshowninTheorem3.4forhomogeneoussuppliers.Specically,thetotalorderisequallysplitamongallsuppliers.Acomparisonoftheoptimalprotlevelsfortheseexamplesshowsthatthecostbenetsassociatedwithhavingthelowestcostsupplierc1=6.75availableoutweighthediversicationbenets.Specically,theprotforExample1ishigherthanthatofExamples2or3.

PAGE 54

42 Second,theimpactofchangesindemandontheoptimalsuppliersourcingstrategiesconrmstheanalyticresultsshowninCorollary3.3.InExamples1Aand2A,theoptimalnumberofsupplierswithapositiveorderincreasesinresponsetohigherlevelsofmeandemand.Therefore,ifarmanticipatesasignicantincreaseindemand,itshouldconsiderenlargingitssupplierbase.InExample1B,thermstillsourcesfromasinglesupplier,butthetotalorderquantityisincreasedtobueragainstdemanduncertainty.InExample2B,thermdecreasesthenumberofsupplierswhichreceiveapositiveordertoonly1,butitincreasesitstotalordersizetothatsinglesupplier.Forthecasewithhomogeneoussuppliers,theoptimalnumberofsuppliersremainsthesamebutthetotalorderquantityincreasesinresponsetoeitheranincreaseinthemeandemandoranincreaseindemandvariability.Tosummarize,thermoptimallydecreasesthenumberofsuppliersreceivingapositiveorderandincreasesthetotalorderquantityinresponsetohigherlevelsofdemanduncertainty.Third,considertheimpactofsupplierreliabilityontheoptimalsourcingstrategy.Ingeneral,itappearsthatthemeansupplierreliabilityimpactsthecorrespondingorderquantity,butdoesnotaecttheoptimalnumberofsuppliers.Thespreadinsupplierreliability,however,impactsonboththecorrespondingorderquantitiesandtheoptimalnumberofsupplierswhenthesuppliercostsareheterogeneous.InExamples1C,2C,and3C,whenthemeanreliabilityoftherstsupplierisreduced,thermoptimallylowersitsorderquantitytotherstsupplierbutsourcesfromthesamenumberofsuppliers.Forthecasewithheterogeneouscostsi.e.,Examples1Dand2D,inresponsetoanincreaseinthespreadassociatedwiththereliabilityforsupplier1,thermoptimallylowersitsorderquantitytotherstsupplierandincreasesthenumberofsuppliersthatreceivepositiveorders.Forthecasewithhomogeneouscostsi.e.,Example3D,theorderquantitiesareadjusteddownwardforthesupplierwithincreased

PAGE 55

43 reliabilityuncertaintywhiletheorderquantitiesforothersuppliersareadjustedupward.Recallthatforthecasewithhomogeneoussuppliercosts,itisalwaysoptimaltosourcefromthefullpoolofsuppliers.Therefore,anincreaseinsupplierreliabilityuncertaintydoesnotimpacttheoptimalnumberofsupplierswhichreceivepositiveorders.Next,considertheimpactofsupplierdictatedminimumorderquantitiesontherm'soptimalsourcingstrategy.Notethataminimumorderquantityisessentiallyanalogoustoaxedordercostassociatedwithaparticularsupplier.Therefore,theminimumorderquantitycanbeconsideredaproxyforcostsassociatedwithmaintainingthebuyer/supplierrelationship.Examples1Eand2Eillustratethesituationwherethelowestcostsupplieralsohasafairlyhighminimalorderquantityi.e.,minq1=1000.Inthissituation,thelowestcostsuppliernolongerreceivesanorderfromtherm.Instead,thermoptimallysourcesfromanalternativesetofsupplierswhichhaveslightlyhighercosts,thusforgoingthestringentminimumorderquantityrestrictionsassociatedwiththelowcostsupplier.Interestingly,theoptimalnumberofsuppliersremainsthesameasthebasecasefortheseexamples.InExample1E,thermoptimallysourcestheentireorderfromsupplier2insteadofsupplier1.InExample2E,thermoptimallysourcesfromsuppliers2and3insteadofsuppliers1and2.Thetotalprotforbothoftheseexamplesdecreasesrelativetothecorrespondingbasecasesbecausethermnolongersourcesfromthelowestcostsupplier.Theeectoftheminimumorderquantityisslightlydierentforthecasewithhomogeneoussuppliers.Whenoneofthesuppliershasasignicantminimumorderquantity,thermstillsourcesfromallthreesuppliersforatotalorderquantityandprotwhichareessentiallythesameasthebasecase,seeExample3E.However,thedistributionofordersischangedsuchthattheordertotherestrictedsupplierisincreasedandtheorderstotheremainingsuppliersaredecreased.When

PAGE 56

44 allsuppliershaveasimilarminimumorderquantitywhichissignicant,thentheoptimalnumberofsuppliersreceivingapositiveorderisreduced,andtheresultsfromTheorem3.3nolongerhold.InExample3F,theoptimalnumberofsuppliersisdecreasedsuchthattheoptimalorderquantityforeachsupplierslightlyexceedstheminimalrestriction.Finally,considertheperformanceofthehomogeneoussetofsuppliersasaresultofchangesinsupplierrelatedfactors.InExamples3C-3F,theindividualorderquantitiesforeachsupplierchangesinresponsetochangesinthemeansupplierreliability,thespreadinsupplierreliability,andsupplierdictatedminimumorderquantities.However,theresultingprotremainsveryclosetothatofthebasecaseshowninExample3.ForExamples3D-3F,thetotalorderquantityisalsosimilartothebasecase.Therefore,itappearsthatthescenariowithhomogeneoussuppliersisfairlyrobusttochangesinanindividualsupplier'sreliabilityorminimumorderquantities.Thisresultmayhavesomebearingonriskaversedecisionmakeroperatinginanenvironmentwhereconsistencyinperformanceandoutputisdesirable. 3.6 ConclusionsThischapterprovidesstructuralandnumericalresultsfordetermininganappropriatesuppliersourcingstrategyinthepresenceofupstreamanddownstreamuncertainty.Arm'ssourcingstrategyischaracterizedbythreeinterrelateddecisions:1thecriteriaforqualifyingasanapprovedvendor;thesuppliersselectionfromtheapprovedbasefororderplacement;andtheorderquantitiestoplacewitheachselectedsupplier.Ouranalyticalresultsdirectlyaddressthesecondandthirddecisions,whileournumericalresultsoermanagerialguidelinesformakingtherstdecision.Inthecontextofthesupplierselectiondecision,ourresultsareinlinewithobservedpractice.Forexample,VermaandPullman96ndthatwhilesupply

PAGE 57

45 managersrecognizetheimportanceofquality,costprimarilydrivestheirsupplierselectiondecisions.Inourmodel,asupplier'scostandnotitsreliabilityisthekeyfactorwhichcomesintoplaywhenarmisdecidingwhetherornottoplaceanorderwiththatsupplier.Consequently,thelowestcostsupplierinthepre-qualiedpoolwillalwaysreceiveapositiveorder.Anexceptiontothisruleshownthroughnumericalexamplesiswhenthelowestcostsupplierhasarestrictivelyhighminimumorderquantity.Itfollowsthatifallpre-qualiedsuppliershavesimilarcosts,thenit'soptimaltoplaceanorderwithallsuppliersinthepool.Whenaddressingthesecondsourcingdecision,armmaydecidetotakeanextremeapproachandlimititssupplierselectiontoonlyasinglesupplier.Inthechapter,wederiveasimpleratiotoanalyticallydeterminewhetherornotasinglesupplierstrategyisappropriate.Thisratioreectsatrade-obetweentherstsupplier'sreliabilityand itscostadvantagerelativetoothersuppliers.Essentially,ifthelowestcostsupplierhasareliabilitydistributionwithahighmeanandalowstandarddeviation,andhasalargecostadvantage,thenasinglesupplierstrategyiswarranted.Anotherkeyfactorinuencingthesinglesuppliersourcingdecisionistherm'santicipateddemand.Bothanalyticandnumericalresultsconrmthatasinglesupplierstrategyisfavorablewhenthemeandemandislow.However,ifarmanticipatesasignicantincreaseindemand,itshouldconsiderenlargingitssupplierbaseevenwhenthelowcostsuppliercouldprovidethefullorderquantity.Surprisingly,anincreaseinthevariabilityindemandalsofavorsasinglesourcingstrategy.Inthiscase,thermlimitsitsnancialriskbysourcingonlywiththesinglelowestcostsupplierwhenthermanticipatesgreatuncertaintyindemand.Otherfactorscontributingtoasinglesuppliersourcingstrategyarealsodiscussedinthepaper.

PAGE 58

46 Thethirdsourcingdecisionconcernshowmuchtoorderfromeachselectedsupplier.Analyticexpressionsaredevelopedwhichdeterminetheoptimalorderquantityforeachselectedsupplierunderavarietyofcircumstances.ThemostgeneralcaseisshowninTheorem3.1whichaddressesthesituationwhereallsuppliershavedierentcostsandreliabilityfunctions.Inthiscase,eachsupplierwillreceiveanorderamountbasedonitsunitcost,meanreliability,andvarianceinreliability.Notethatwhilethelowestcostsupplierisguaranteedtoreceiveapositiveorder,he/shewon'tnecessarilyreceivethelargestorder.Incontrast,wealsoanalyzethesituationwherethesuppliersarehomogeneousintheircostsandreliabilityfunctionsinTheorem3.4.Inthissituation,allpre-qualiedsuppliersreceiveanequivalentorderquantity.Finally,therstsourcingdecisionaddresseswhatcriteriaarmshouldusetopre-qualifyasetofsuppliers.Whiletherearemanyfactorsimpactingthisdecisionthatarenotaddressedhere,wecanextrapolatesomeinsightsconcerningtheappropriatenumberofsuppliersthatshouldbepre-qualied.Thexedcostsofqualifyingasuppliertoensurethatitmeetsaminimalsetofcriteriabasedonquality,costs,anddeliverycanbeexorbitant.Itfollowsthatarmshouldjudiciouslytargetasetofsupplierswhichsupportsitsanticipatedsourcingstrategy.Forthesetofnumericalexamplesthatweconsidered,sourcingonlyfromasinglesupplierwithverylowcostsresultedinahigherprotthanothermulti-suppliersourcingstrategies.Therefore,ifarmanticipatesthatasinglesupplierhasmuchlowercosts,thenitmightbebetterotofocusonqualifyingthatsinglesuppliertoensurethatthesupplier'sreliability,qualityanddeliveryaresucienttomeetdemand.Essentially,additionaleortscanbemadetoensureconsistentreliabilityofthelowestcostsupplier.However,ifseveralsuppliersareverycloseincost,thecompanyshouldcon-siderqualifyingthemallsuchthattheymeetspecicreliabilitycriteria.Indeed,the

PAGE 59

47 numericalexamplesshowthatourresultsforthecasewithhomogeneoussuppliersarefairlyrobustwithrespecttochangesindemandandsupplierreliability.Specif-ically,whenallsuppliersinthepre-qualiedpoolhavesimilarcostsandreliability,thenthetotalquantityorderedandthetotalrmprotisfairlyconstantevenwhendemandandsupplierreliabilityfactorschange.Thereareseveralfutureareasofresearchrelatedtothismodelwhichwarrantfurtherinvestigation.First,amoredetailedmodelcouldbedevelopedwhichaddressestherstsuppliersourcingdecision,orappropriatecriteriaforqualifyingsuppliers.Second,weassumethatdemandisuniformlydistributedtofacilitatethedevelopmentofsimpliedexpressions.Othertypesofdemanddistributionscouldbeexploredtoenhancethegeneralizabilityoftheresultsderivedhere.Lastly,thefocusofthismodelisondecisionmakingatthebuyingrm;futureresearchcouldincorporatethesupplier'sdecisionmakingprocessaswell.

PAGE 60

CHAPTER4IMPACTOFSUPPLIERPRICINGSCHEMESANDCAPACITYONSOURCINGSTRATEGIES 4.1 IntroductionConsiderthefollowingprocurementprocessfollowedbyamajoroceproductsretailerwithheadquartersinFlorida,USA.ThecentralizedpurchasingorganizationCPOfortheretailerisresponsiblefortheprocurementofallcommoditytypeproductswhicharesoldthroughtheretailoutlets.OrdersforthesetypesofproductsareplacedbytheCPOonaperiodicbasisandingeneral,afourstepprocesscharacterizestheprocurementprocess.First,foreachtimeperiod,theCPOaggregatesthetotalestimatedrequirementsforeachproductbasedoninputfromtheretailoutlets.Second,usingaweb-basedinterfacewhichisset-uptoallowaccessonlytopre-qualiedsuppliers,theCPOpoststimingandquantityrequirementsforthecommodityproducts.Third,suppliersrespondbyquotingpricesandquantitylimitations,ifany,fordeliveringtheproductstothecentralwarehousemaintainedbytheCPO.Finally,aCPOanalystanalyzesthesuppliersubmittedinformationandallocatesrequirementstosuppliers.Thenalstepofthisprocessforsupplierselectionandorderplacementwasbiasedtowardsselectingasinglesupplierwhocouldsupplytheentiresetofrequirementsratherthananexplicitfocusoncostsquotedbythesuppliers.Theprimarymotivationforsuchastrategyi.e.,singlesourcingwasthatitwaseasiertomanageorderreceiptsfromonesupplieratthecentralwarehouse.InamorerecentanalysisofsupplierresponsestotheCPOpostingofrequirementsandtimeinginformation,itwasfoundthatsupplierswerestartingtooerquote 48

PAGE 61

49 pricingschemesandthus,thesupplierselectionandorderplacementdecisionwasbecomingincreasinglydicult.Thisscenariomotivatesthischapter'sanalysisofsupplierselectionandorderplacementdecisionsinthepresenceofalternativesupplierpricingschemes.Itiswellknownthatsourcingdecisionsinthissettingareextremelycomplexandalsorequireafrequentreassessment.Forexample,suppliersoftenoerdiscountschedulestoinducelargerpurchasesbyoeringprogressivelylowerunitpricesforprogessivelylargerpurchasequantities.Evenforasingleproduct'spurchasingdecision,ifitisavailablefrommanyvendors,eachwithvariousqualifyingordersizes,identifyingtheoptimalselectedsuppliersetandcorrespondingquantityallocationsisadicultdecision.FurthercomplicatingmattersisthatdecisionsmustoftenbemadequicklyandwithlimitedinformationduetotimepressuresRubinandBenton,1993.Thisfastpaceddecisionenvironmentisnotconducivetooptimallysolvingcombinatorialproblemsbycompleteenumeration.Therefore,heuristicproceduresthatproduceoptimalornearoptimalfeasiblesolutionsareofsignicantvaluefordecisionmakers.Morespecically,thischapterexaminessupplierselectionandquantityallocationdecisionsforacquisitionofasingleproduct'stotalrequirementsfromapoolofsuppliersoeringquantitydiscountschedules.Weexaminethesedecisionsforenvironmentswheresuppliersoerconstant,lineardiscount,incrementaldiscount,andall-unitsdiscountpricingschemes.Theremainderofthispaperisorganizedasfollows.Inthenextsection,wereviewtherelevantpriorliterature.Section3developsourgeneralsourcingmodelforeachofthepricingschemesandduetothecomplexitiesassociatedwiththismodel,weproposeheuristicsforobtainingfeasiblesolutionsveryquickly.InSection4,weanalyzetheperformanceoftheseheuristicsthroughanextensivenumericalanalysisandthisisfollowedbyanapplicationofourapproachtodataobtainedfromtheoceproductsretailerin

PAGE 62

50 Section5.Abranchandboundalgorithmforndingaminimumcostsolutiontothesupplierselectionandquantityallocationproblemwithincrementaldiscountsisdevelopedandstatedinsection6.Finally,Section7discussesimplicationsandconclusionsofourresearch. 4.2 SourcingModel 4.2.1 PreliminariesOuranalysisofthesuppliersourcingdecisionsfocusesonasingleproduct,singleperiodanalysisofasystemconsistingofNsuppliersi=1;:::;Nandasinglebuyingrm.Ingeneral,weassumethefollowingthree-stagesequentialdecisionframeworkforouranalysis.Attherststage,thermFcommunicatesthetotalquantityofthesingleproductQwhichitwillprocurefromthesup-pliers.Followingthis,inthenextstage,eachsupplieridisclosesapricingschemefiandarelatedmaximumquantitywhichitcanprovidetothermyi.Afterreceivingthisinformation,thermmakesthesuppliersourcingdecisionqiforeachsupplierinthethirdstage.WemodeltheproblemforabuyingrmwhichiseitherachannelintermediarywithaxedquantitycontractfromasetofdownstreamrmsoramanufacturermakingprocurementquantitydecisionsusingautomatedmaterialsplanningsystemssuchasMRP.InbothcasesitisreasonabletoassumethatthebuyingrmcandeclarewithreasonablecertaintythetotalquantityQtobeprocuredfromthesuppliers.Asnotedearlier,zeroxedorderingcostsforthebuyingrmarebeingassumedinlinewithourmotivatingexampleoftheoceproductsretailer.Therearethreeadditionalfactorsthatneedtobeclariedinthecontextofouranalysis.First,eachsupplierisanindependentoperatorandhence,therearenoopportunitiesforsuppliercollusion/collaborationinoursetting.Second,thepricingschemedisclosedbyeachsupplieri.e.,fiisallinclusiveandincludesthelogistics/transportationcost.Finally,supplyleadtimesareassumedtobe

PAGE 63

51 relativelyconstantandthus,arenotincorporatedinouranalysis.Inthenextsection,wedescribethealternativesupplierpricingschemesandinsomecases,therelatedsuppliercapacityparameterswhichareinvestigatedinthispaper. 4.2.2 SupplierPricingSchemesandCapacityThereareseveraltypesofsupplierpricingschemeswhichareusedinthemar-ketplace.Theseschemesareprimarydriversforanalyzingthesourcingdecisionssincetheyhaveadirectimpactonrm-levelprots.Further,suppliercapacityalsocandrivesomeofthesepricingschemes.Onreviewingpriorliteratureintheeld,thefourdistinctsupplierpricingschemeswhichweincorporatealongwithcapacityconsiderationare: 1. ConstantPriceUnderthisschemeeachsupplieridisclosesaconstantpriceperuniti.e.,fiqi=ciitcanprovidethermwithinthecapacityrange[0;yi]. 2. LinearDiscountPriceUnderthisschemeeachsupplieridisclosesalinearlydecliningpriceinthequantityqipurchasedfromsupplieri.Thus,fiqi=ai)]TJ/F22 11.955 Tf 12.551 0 Td[(biqiai,bi>0withinthecapacityrange[0;yi].Wealsoassumethatthelinearlydiscountedpricingschemedisclosedissuchthatai)]TJ/F22 11.955 Tf 11.955 0 Td[(biyi>0foreachsupplieri. 3. IncrementalUnitsDiscountedPriceUnderthisschemeeachsupplieridisclosesthetraditionalincrementalunitsdiscountingschemeNahmias,2001whichisdependentonthequantityqipurchasedfromsupplieri.Tospecifysuchascheme,werstdenek=1;:::;Kiastheindexfordiscountclassesoeredbysup-plieri.Correspondingtoeachdiscountclasskforasupplieri,dene[lik;uik]astheminimumandmaximumquantitiesfortheclasssuchthatuiKi=yi.Basedonthesedenitions,thepricingschemecanbespeciedas

PAGE 64

52 fiqi=Pk)]TJ/F21 7.97 Tf 6.586 0 Td[(1j=1wijuij)]TJ/F22 11.955 Tf 12.157 0 Td[(uij)]TJ/F21 7.97 Tf 6.587 0 Td[(1+wikqiiflikqiuikfork=1;:::;Ki.Itisassumedthatwi1>wi2>:::>wikforeachsupplieri. 4. AllUnitsDiscountPriceUnderthisschemeeachsupplieridisclosesthetraditionalall-unitsdiscount-ingschemeNahmias,2001whichisdependentonthequantityqipurchasedfromsupplieri.Aswiththeincrementalunitsscheme,letk=1;:::;Kiastheindexfordiscountclassesoeredbysupplieri,andcorrespondingtoeachdiscountclasskforasupplieri,dene[lik;uik]astheminimumandmaximumquantitiesfortheclass,respectively,suchthatuiKi=yi.Basedonthesedenitions,thepricingschemecanbespeciedasfiqi=vikqiiflikqiuikfork=1;:::;Ki.Itisassumedthatvi1>vi2>:::>viKiforeachsupplieri.Next,weformulatethesourcingmodelforeachpricingscenarioandprovideanalytical/experimentalinsightsintothesourcingstrategyundereachcase. 4.3 AnalysisandInsights 4.3.1 ConstantPriceGiventhatfiqi=ciforeachsupplieri,oursourcingmodelforthiscaseisasfollows:MinimizeZC=nXi=1ciqi.1subjectto:nXi=1qi=Q .2 0qiyi8i .3 Analysisofthismodelleadsustothefollowingtheoremwhichcharacterizestheoptimalsourcingstrategyforthiscase.Theorem4.1 :Underconstantsupplierprices,theoptimalsourcingpolicyforthermis:

PAGE 65

53 Indexsuppliersinnon-decreasingorderofpricesi.e.,c1c2:::cn. Ify1Q,thensourcethecompleterequirementQfromsupplier,i.e.,qi=Qandqj=08j=2;:::;n. Ify1
PAGE 66

54 Ontheotherhand,ifthereisatleastonesupplierisuchthatyi
PAGE 67

55 5. Setq[j]=minfQ;y[j]g.Removesupplierjfrom. 6. SetQ=Q)]TJ/F22 11.955 Tf 11.955 0 Td[(q[j].IfQ>0goto2,elsegoto7. 7. Forallsuppliersk,withqk>0,exploreallpossibleimprovementsinthesolutionbyswitchingthepartialorderquantitywithinthisselectedsupplierset. 8. StorethebestsolutionasSolutionA 9. Repeattheaboveprocess,exceptatstep2,calculateri=minfQ;yigai)]TJ/F22 11.955 Tf -384.491 -23.908 Td[(biminfQ;yig,andinstep6,storethebestsolutionasSolutionB. 10. ChoosesolutionAorSolutionBbasedonthebetterobjectivefunctionvalue.Toevaluatethesolutionqualityofthisheuristic,wecarriedoutnumericalex-perimentsbyrandomlygenerating30testproblemsAPPENDIXB.Foreveryproblem,thesupplybasesize,N,wassetto10,andthetotalrequirements,Q,equals2000units.Further,foreachsupplieri,thepricingandcapacityparameterswererandomlygeneratedfromauniformdistributionasfollows:baseprice,ai~U[0,200];priceelasticity,bi~U[0,1];andcapacity,yi~U[0,1000].TheresultsofthisevaluationarepresentedinTable4-1.ThegapreportedinthistableisthepercentagedeviationoftheheuristicsolutionfromtheoptimalsolutionobtainedusingLINGO.Ascanbeseen,in20ofthe30problems,theheuristicsolutionwasoptimal.Fortheremaining10problems,theworstcaseheuristicsolutiongapwas5.78%,whiletheaveragegapwas1.81%.Furtheralltheheuristicsolutionswereobtainedislessthan1secondusingaPCwith256KBRAM,aPentiumIIIProcessor,and550MhzclockspeedwhiletoobtaintheoptimalsolutionsusingLINGOonasimilarmachines,theruntimeswerebetweenseveralminutesandaover2hours.Thisleadsustoconcludethatourheuristicprovidesgoodqualitysolutionsinreasonablecomputationtimes.

PAGE 68

56 Table4{1:LinearDiscountHeuristicPerformance Instance Heuristic Optimal Gap% Instance Heuristic Optimal Gap% 1 88282.77 88282.77 0.00% 16 168636.10 168636.10 0.00% 2 103315.00 103315.00 0.00% 17 120547.20 120457.20 0.00% 3 128455.30 128455.30 0.00% 18 98735.40 98583.63 0.15% 4 89650.40 88850.70 0.90% 19 94898.40 94898.40 0.00% 5 129274.70 127915.70 1.06% 20 174835.10 174835.10 0.00% 6 130886.20 128455.30 1.89% 21 39921.43 39921.43 0.00% 7 58198.44 58198.44 0.00% 22 45553.20 43856.09 3.87% 8 79593.48 79593.48 0.00% 23 89040.30 88585.22 0.51% 9 79593.48 79593.48 0.00% 24 111166.30 111166.30 0.00% 10 119205.40 119205.40 0.00% 25 66280.70 66051.12 0.35% 11 79593.48 79593.48 0.00% 26 86099.20 81393.94 5.78% 12 42926.90 41538.80 3.34% 27 53897.25 53897.25 0.00% 13 110474.80 110474.80 0.00% 28 119360.00 119360.00 0.00% 14 69594.40 69444.00 0.22% 29 55034.56 55034.56 0.00% 15 174675.70 174675.70 0.00% 30 195287.90 195287.90 0.00% MeanGap 0.60% WorstGap 5.78% 4.3.3 IncrementalUnitsDiscountPriceInthiscase,fikqik=Pk)]TJ/F21 7.97 Tf 6.586 0 Td[(1j=1wijuij)]TJ/F22 11.955 Tf 11.814 0 Td[(uij)]TJ/F21 7.97 Tf 6.586 0 Td[(1+wikqik)]TJ/F22 11.955 Tf 11.815 0 Td[(likiflikqikuikfork=1;:::;Ki.Thesourcingmodelforthiscaseis:MinimizeZIU=nXi=1KiXk=1fikqikyik.7subjectto:nXi=1KiXk=1qik=Q .8 KiXk=1yik18i .9 likyikqikuikyik8i;k .10 yik2[0;1] .11 ForthecasewhereeachsupplierdisclosesapricingschemesuchthatuiKiQ8i,tomeettheaggregaterequirementQ,itistrivialtoshowthatthermwillchoosetosourcethecompleterequirementQfromsupplierjsuchthatwjk=

PAGE 69

57 min1in;1kKifwikjlikQuikg.Inlinewiththepriorpricingschemes,thesinglesourcingstrategyisanoptimalchoiceforthissituation.Aswiththelineardiscountpricingscheme,thedominanceofthesinglesourcingstrategyisquestionablewhenevery supplierdoesnothaveadequatecapacitytomeettheaggregaterequirementsQ.Inthiscase,again,wenotethatresult4.1statedaboveforthelineardiscountpricingschemestillholdssinceourobjectivefunctionispiecewiselinearconcaveinqi.Basedonthis,weproposethefollowingheuristicalgorithmforobtainingsolutionstothesourcingmodelunderthispricingscenario. 1. Denetheactivesupplierset,,asconsistingofallsuppliers. 2. Foreachsupplieriincludedin, IfuiKi0goto2,elsegoto5. 5. Forallsuppliersk,withqk>0,exploreallpossibleimprovementsinthesolutionbyswitchingthepartialorderquantitywithinthisselectedsupplierset. 6. StorethebestsolutionasSolutionA 7. Repeattheaboveprocess,exceptatstep2,calculateintherstifstatementri=fiuiKi,andintheDOloopri=fikQ;andinStep6,storethebestsolutionasSolutionB. 8. ChoosesolutionAorSolutionBbasedonthebetterobjectivefunctionvalue.

PAGE 70

58 Toevaluatethesolutionqualityofthisheuristic,wecarriedoutnumericalexper-imentsbyrandomlygenerating30testproblemsAPPENDIXB.Aswiththelineardiscountpricingcase,foreveryproblem,thesupplybasesize,N,wassetto10,andthetotalrequirements,Q,equals2000units.Further,foreachsupplieri,thenumberofpricebreaks,upperandlowerboundsofquantitiesforeachpricebreak,andunitpricesforeachpricebreakwererandomlygeneratedfromauni-formdistributionasfollows:baseprice,Ki~U[0,10];li1=0,andlik=uik+1fork=2;:::;Ki;ui1~U[0,100],anduik=lik+U[0,100]fork=2;:::;Ki;andwi1~U[0,1]andwik=wik)]TJ/F21 7.97 Tf 6.586 0 Td[(1)]TJ/F15 11.955 Tf 13.2 0 Td[(0.05fork=2;:::;Ki. Table4{2:IncrementalUnitsDiscountHeuristicPerformance Instance Heuristic Optimal Gap% Instance Heuristic Optimal Gap% 1 2944.16 2944.16 0.00% 16 2435.61 2435.61 0.00% 2 2620.90 2613.21 0.29% 17 3046.59 3046.59 0.00% 3 2937.55 2937.55 0.00% 18 3202.77 3202.77 0.00% 4 2680.84 2680.84 0.00% 19 2993.12 2993.12 0.00% 5 2477.86 2477.86 0.00% 20 2465.22 2465.22 0.00% 6 2592.35 2592.35 0.00% 21 2888.12 2888.12 0.00% 7 3306.59 3306.59 0.00% 22 2841.58 2841.58 0.00% 8 2964.60 2962.86 0.06% 23 3178.13 3178.13 0.00% 9 2696.40 2685.76 0.40% 24 3178.13 3178.13 0.00% 10 2574.28 2574.28 0.00% 25 2853.53 2853.53 0.00% 11 2670.19 2670.19 0.00% 26 2992.17 2992.17 0.00% 12 2365.78 2365.78 0.00% 27 2753.51 2753.51 0.00% 13 3017.81 3017.81 0.00% 28 2850.40 2850.40 0.00% 14 2964.46 2964.46 0.00% 29 2699.23 2699.23 0.00% 15 2546.91 2546.91 0.00% 30 2704.60 2704.60 0.00% MeanGap 0.03% WorstGap 0.40% TheresultsofthisevaluationarepresentedinTable4-2.ThegapreportedinthistableisthepercentagedeviationoftheheuristicsolutionfromtheoptimalsolutionobtainedusingLINGO.Ascanbeseen,in27ofthe30problems,theheuristicsolutionwasoptimal.Fortheremaining3problems,theworstcaseheuristicsolutiongapwas0.40%,whiletheaveragegapwas0.24%.Aswiththelineardiscountcase,alltheheuristicsolutionswereobtainedislessthan1secondusingaPCwith256KBRAM,aPentiumIIIProcessor,and550MhzclockspeedwhiletoobtaintheoptimalsolutionsusingLINGOonasimilarmachine,

PAGE 71

59 theruntimeswerebetweenseveralminutesandaover2hours.Thisleadsustoconcludethatourheuristicprovidesverygoodqualitysolutionsfairlyquickly. 4.3.4 AllUnitsDiscountPriceInthiscase,fikqik=vikqikiflikqikuikfork=1;:::;Kiforeachsupplieri.Thesourcingmodelisnowformulatedasfollows.MinimizeZAU=nXi=1KiXk=1vikqikyik.12subjectto:nXi=1KiXk=1qik=Q .13 KiXk=1yik18i .14 likyikqikuikyik8i;k .15 yik2[0;1] .16 Forthecasewhereeachsupplierhasadequatecapacityi.e.,uiKiQ8itomeettheaggregaterequirementQ,itistrivialtoshowthatthermwillchoosetosourcethecompleterequirementQfromasinglesupplierjsuchthatVQj=min1infVQig.Inthiscase,VQiisthepriceperunitoeredbysupplieriforquantityQ.Again,thedominanceofthesinglesourcingstrategyisquestionablewhenevery supplierdoesnothaveadequatecapacitytomeettheaggregaterequirementsQ.Further,inthiscase,ourobjectiveisdiscontinuousandthus,theresultstatedforthepriortwocasesforacontinuousconcaveobjectivefunctiondonotnecessarilyapply.Ontheotherhand,thepropertystatedintheresultisquiteeasytoincorporateinaheuristicandhence,thefollowingheuristicwasproposedtogeneratefeasiblesolutionstoourproblem. 1. Denetheactivesupplierset,,asconsistingofallsuppliers.

PAGE 72

60 2. Foreachsupplieriincludedin, IfuiKi0goto2,elsegoto5. 5. Forallsuppliersk,withqk>0,exploreallpossibleimprovementsinthesolutionbyswitchingthepartialorderquantitywithinthisselectedsupplierset. 6. StorethebestsolutionasSolutionAitemRepeattheaboveprocess,exceptatstep2,calculateintherstifstatementri=viKiuiKi,andintheDOloopri=vikQ;andinStep6,storethebestsolutionasSolutionB. 7. ChoosesolutionAorSolutionBbasedonthebetterobjectivefunctionvalue.Toevaluatethesolutionqualityofthisheuristic,wecarriedoutnumericalexper-imentsbyrandomlygenerating30testproblemsAPPENDIXB.Theparametersettingforeachproblemareidenticaltothosegeneratedfortheincrementalunitsdiscountpricecaseexceptthatforeachsupplieri,theunitpricevik=wikfork=1;:::;Ki.TheresultsofthisevaluationarepresentedinTable4-3.ThegapreportedinthistableisthepercentagedeviationoftheheuristicsolutionfromtheoptimalsolutionobtainedusingLINGO.Ascanbeseen,in17of30problems,theheuristicsolutionwasoptimal.Fortheremaining13problems,theworstcaseheuristicsolutiongapwas1.58%,whiletheaveragegapwas0.27%.Alltheheuristic

PAGE 73

61 solutionswereobtainedislessthan1secondusingaPCwith256KBRAM,aPentiumIIIProcessor,and550MhzclockspeedwhiletoobtaintheoptimalsolutionsusingLINGOonasimilarmachine,theruntimeswerebetweenseveralminutesandaover2hours.Thisleadsustoconcludethatourheuristicprovidesfairlygoodqualitysolutionsfairlyquickly. Table4{3:All-UnitsDiscountHeuristicPerformance Instance Heuristic Optimal Gap% Instance Heuristic Optimal Gap% 1 2675.01 2675.01 0.00% 16 2116.36 2116.36 0.00% 2 2313.10 2308.95 0.18% 17 2753.34 2753.34 0.00% 3 2746.50 2736.80 0.35% 18 3023.82 3023.82 0.00% 4 2465.49 2465.49 0.00% 19 2756.77 2756.77 0.00% 5 2185.60 2181.90 0.17% 20 2245.90 2240.93 0.22% 6 2303.10 2267.30 1.58% 21 2529.90 2527.17 0.11% 7 3117.94 3117.94 0.00% 22 2497.80 2494.66 0.13% 8 2694.41 2694.41 0.00% 23 2931.00 2920.13 0.37% 9 2381.40 2380.86 0.02% 24 2161.84 2161.84 0.00% 10 2319.18 2319.18 0.00% 25 2578.03 2578.03 0.00% 11 2433.14 2433.14 0.00% 26 2650.97 2650.97 0.00% 12 2101.40 2099.66 0.08% 27 2530.91 2530.91 0.00% 13 2777.31 2777.31 0.00% 28 2523.10 2523.10 0.00% 14 2727.40 2721.61 0.21% 29 2493.43 2493.43 0.00% 15 2262.40 2259.67 0.12% 30 2461.70 2449.75 0.49% MeanGap 0.13% WorstGap 1.58% 4.3.5 SummaryofInsightsfromAnalysisBasedonouranalysis,wecanoerthefollowinggeneralconclusions.Inregardtotherm'ssourcingstrategy,foranypricingscheme,theonlytimeasinglesuppliersourcingstrategyispreferrediswhenthelowestcostsupplierhasadequatecapacitytomeettheentiredemandfortherm.Inallothercases,amultiplesourcingstrategyisthegeneralchoice.Inregardtosupplierquantityallocations,forthecasewhereamultiplesuppliersourcingstrategyispreferred,wecandeterminetheoptimalsupplierquantityallocationsonlywheneachsupplierquotesaconstantprice.Whensuppliersquotequantitydiscountpricingschemes,theoptimalquantityallocationsarediculttodetermine.However,wehavedevelopedecientheuristicswhichcangeneratesupplierquantityallocationswhich

PAGE 74

62 areclosetooptimal.Wenowturntoanapplicationofourheuristicapproachesfordataobtainedfromanoceproductsmanufacturer. 4.4 ApplicationTheCentralPurchasingOrganizationCPOofamajoroceproductsretailerisfrequentlyfacedwithcomplexsourcingdecisionsduetothepresenceofquantitydiscountschemesembeddedwithinvarioussuppliers'bids.GiventheCPO'stotalquantitycommitmentsandsupplierbiddatafortwodistinctcommodityproductsweapplyourmodelingapproach.ThedatasharedbytheCPOisfortwocommodityproductsamonghundredsofcommoditystockkeepingunitsSKUthatmustbefrequentlyreplenishedforretailsale.TotalquantityrequirementsforeachproductareaggregatedfromretaillocationsbytheCPO.TheCPOthenpostsarequestforquotationRFQforeachproduct.TwodistinctsetsofsuppliersquoteonproductAandproductBrespectively.TheportfolioofbidsforproductAconsistsofsixsuppliers,fourwithconstantpricequotesandtwowithdiscountpricing.ProductB'sbidportfolioconsistsofeightsuppliers,fourwithconstantpricequotesandfourwithdiscountpricing.Allsuppliersinbothportfoliosalsoprovidelimitationsontheirrespectiveunitcapacities.Tables4-4and4-5containbidinformationforproductsAandBrespectively. Table4{4:ProductABidInformationData TQC=9855 PriceBreakMin PriceBreakMax UnitPrice SupplierA1 0 1000 623 1001 2100 534 2101 3200 465 SupplierA2 0 2100 452 SupplierA3 0 2650 457 SupplierA4 0 1000 449 SupplierA5 0 700 654 701 1920 494 SupplierA6 0 2200 453

PAGE 75

63 Fromthebidinformationwegenerateadatasetforeachproduct'ssuppliers.Eachdatasetisimplementedtwice.Firstitisassumedthatquantitydiscountquotesareoftheincrementalvariety.Secondly,anall-unitsquantitydiscountstructureisassumed.Further,constantorxedpricingcanbemodeledasanincrementaloranall-unitsdiscountwithonlyonediscountlevel.Therefore,ourheuristiccanhandleportfoliosofsupplierbidsthatconsistofconstantpricequotesaswellasmoresophisticatedbidswithdiscountpricing. Table4{5:ProductBBidInformationData TQC=7680 PriceBreakMin PriceBreakMax UnitPrice SupplierB1 0 1200 634 SupplerB2 0 600 875 601 1300 800 1301 2500 725 2501 2900 634 SupplierB3 0 700 790 701 1600 710 1601 3000 620 SupplierB4 0 1460 621 SupplierB5 0 1275 625 SupplierB6 0 2600 632 SupplierB7 0 400 922 401 1400 822 1401 2000 722 2001 2600 622 SupplierB8 0 800 821 801 1750 700 1751 2400 610 Usingourheuristicswedeterminesupplierselectionandquantityallocationforthisindustryproblem.Tables4-6and4-7provideasummaryofheuristicsolutionperformanceversusLINGO'sGlobalSolveroptimalsolutionsforproductsAandBrespectively.Fortheincrementalinstances,theheuristicsolutionisoptimalandtheCPOshouldsourcefromvesuppliersforeachproduct.Intheall-unitsinstancestheheuristicarrivesatnearoptimalsolutionsthatvaryfromtheoptimalsolutionsinboththepoolofselectedsuppliersandallocationsamongthem.Forexample,itisoptimaltosourcefromvesuppliersforproductAandtheheuristicselects

PAGE 76

64 Table4{6:CPOProductASolutionsComparison Incremental qA1 qA2 qA3 qA4 qA5 qA6 Cost Gap Optimal 0 2100 2650 1000 1905 2200 4658920 Heuristic 0 2100 2650 1000 1905 2200 4658920 0% AllUnits qA1 qA2 qA3 qA4 qA5 qA6 Cost Gap Optimal 2101 2100 2454 1000 0 2200 4493243 Heuristic 2905 2100 2650 0 0 2200 4507675 0.32% Table4{7:CPOProductBSolutionsComparison Incremental qB1 qB2 qB3 qB4 qB5 qB6 qB7 qB8 Cost Gap Optimal 1200 0 1145 1460 1275 2600 0 0 4976485 Heuristic 1200 0 1145 1460 1275 2600 0 0 4976485 0% AllUnits qB1 qB2 qB3 qB4 qB5 qB6 qB7 qB8 Cost Gap Optimal 0 0 3000 279 0 0 2001 2400 4741881 Heuristic 0 0 3000 1460 820 0 0 2400 4743160 0.03% onlyfour.Optimally,supplierA4isselectedandallocated1000unitsbydecreasingallocationtosuppliersA1andA3.ForproductBtheheuristicandoptimalsolutionconsistsoffourselectedsuppliers.However,theoptimalsolutionselectssupplierB7andnotsupplierB5andourheuristicselectssupplierB5insteadofB7.Also,theoptimalallocationtosupplierB4is279unitsandourheuristicallocates1460units.Therefore,whiletheoptimalitygapisslightfortheall-unitsinstances,thesourcingstrategydecisionsofsupplierselectionandquantityallocationarequitedierent. 4.5 OptimalAlgorithmforIncrementalQuantityDiscountedSourcingInthissectionwepresentabranchandboundalgorithmforoptimallysolvingoursourcingproblemwhenallsuppliersquotepriceswithincrementalquantitydiscounts.Thesourcingproblemisexactlyasdescribedinsection4.3.3.Assuchweneedtosolvefortheminimumofthesumofpiecewiselinearconcavefunctions.Again,byleveragingResult4.1,wesearchforasolutionsuchthatatmostonesupplier'sallocationisnotataboundaryofzeroorhiscapacity.Sinceoursourcingproblemwithincrementalquantitydiscountsisaspecialcaseoftheseparablepiecewiselinearconcavecostallocationproblem,wedevelopandoutlinean

PAGE 77

65 optimalalgorithmforsolvingtheseparablepiecewiselinearconcavecostallocationproblem.Themathematicalformulationoftheproblemisasfollows:SPLCCAP:Minimizefxsubjecttox2DTCwherex=x1;:::;xn2Rn.Givenarefx=Pni=1fixiandforeachi,ficoncaveandboundedon[li;ui];CQni=1Ci;Ci=[li;ui],andli;ui2R+;Dx:Pni=1xi=Q;Q2R+;andDTCassumedtobebounded. 4.5.1 AlgorithmDescriptionInthisalgorithmwebranchifthesolutionatthecurrentnode,pallocatesanyxisuchthatlpi
PAGE 78

66 isexecutedtodeterminewhetherUBfcanbeupdated.IfLBpUBf,nodeNpispruned.Further,ifLBp
PAGE 79

67 solutiontoSPLCCAP.Thevalidityoftheupperboundonf,UBf,followsfromthealgorithmsettingUBf=minfxp:p=0;1;:::.AlowerboundonfoverDTC,LBfisnowexpressed.LetNBkdenotethesetofnodesnotbranchedfromatstagekofthealgorithm.AnodeNp2NBkduetofathomingorcurrentinclusioninthecandidatelist.Itcanbeshownthat:LBf=minNp2NBkLBp.Thereforeateachstagek;k=0;1;:::ofthealgorithmLBffUBf. 4.5.3 TheBranchingProcessAtstagekofthealgorithm,anodeNpisselectedfromthecandidatelist.BranchingisinvokedandnodeNpispartitionedintotwonewnodesN2k)]TJ/F21 7.97 Tf 6.586 0 Td[(1andN2k.Thisisaccomplishedbyrstchoosingaseparationvariable,xpswhichsatisesfsxps)]TJ/F22 11.955 Tf 12.495 0 Td[(gpsxps>0i.e.,lps
PAGE 80

68 3. RemoveanynodeNpfromthecandidatelist.IfLBpUBfgotostep2.Otherwise,generatenodesN2k)]TJ/F21 7.97 Tf 6.586 0 Td[(1andN2kandndLB2k)]TJ/F21 7.97 Tf 6.586 0 Td[(1andLB2k.Ifnecessary,updateUBfandx.IfLBj
PAGE 81

69 atQunits.However,ifevenonesupplierinthebaseisincapableofprovidingQunits,thensinglesourcingmaynotbetheoptimalsourcingstrategy.Incaseswherequantitydiscountquotingsuppliers'capacityisindividuallyinadequateforabuyingrm'sproductrequirements,arm'ssourcingproblemcanbecomeextremelycomplex.Therefore,identifyinganoptimalsolutionmayberesourcetimeprohibitiveforsupplychainsourcingprofessionals.Insuchcases,theheuristicsdevelopedinthischaptercanbeexpectedtoecientlyprovidegoodqualitysolutions.Insum,theheuristicsolutionstorandomlygeneratedtestproblemsarrivedattheoptimalsolutionin62%oftheinstances.Furthermore,inthe38%ofnon-optimalsolutioninstances,arespectable:64%averageoptimalitygapexists.

PAGE 82

CHAPTER5STRATEGICSOURCINGWITHDIVERSIFICATIONCONSIDERATION 5.1 IntroductionTobesuccessfulinthelongterm,arm'ssourcingstrategytodaymustbeconsistentwiththevisionofthermfortomorrow.Thisvisionisatthecoreofarm'scorporatestrategy.Corporatestrategyisnotonlyacompany'spatternofpurposesandgoalsthatdenesthetypeofcompanythatitis,itisalsothepoliciesthatdirectgoalachievementAndrews,1980.Theintentofthegeneralmodeldevelopedinthischapteristoprovideastrategicsourcingframeworktocapturesubjectiveelementsofarm'scorporatestrategythatdirectlyinuencesourcingdecisions.Thegeneralapproachinthischapterforexaminingsupplierselectionandquantityallocationdecisionsistwofold.First,anintegratedmodelisdevelopedtoaddressbothofthesedecisions.Givenuncertainproductdemand,wesimultaneouslyconsidersuppliercosts,supplierreliabilities,suppliercapacities,manufacturerinventorycosts,andmanufacturerdiversicationbenetsinmakingtheseintegrateddecisions.Second,usingourmodel,wecharacterizeconditionsunderwhichsingle-sourcingandmultiple-sourcingstrategiesareoptimal.Akeyfeatureofthemodelfacilitatingthisanalysisisanexplicittreatmentofsupplydiversicationbenets.Theseexecutivelevelvaluationsofalternatesupplybasesizesareassumedtobeconsistentwiththerm'scompetitivestrategy.Thisstrategicdirectionmaybepresentlyinconvenient,inthatitmayindicatethatcurrentprotsshouldbeforegoneinlieuofpotentialfuturegains.ThisworkismostcloselyrelatedtothoseofPan89andParlarandWang993.Pan989proposesalinearprogrammingmodeltooptimallyidentifythenumberofsuppliersandtheirrespectivequantityallocationstomeetpre-specied 70

PAGE 83

71 productrequirements.Otherconstraintsincorporatedarerelatedtoaggregateincomingquality,leadtimes,andservicelevel.Theoverallobjectiveistominimizethepriceperunitasaweightedaverageofselectedsuppliers'prices.Itisassumedthatproductrequirementsaredeterministicandsupplyisreliableandunlimited.ParlarandWang993comparethecostsofsingleversusdual-sourcingforarmassumingthattheoverallobjectiveistominimizepurchasingandinventoryrelatedcosts.Intheirapproach,theyassumethatactualincomingquantitiesareafunctionofarandomvariablerepresentingtheyield.SeparatelyusinganEOQandnewsboybasedorderingpolicy,theyareabletoshowthatincertaincasesdual-sourcingdominatessingle-sourcing.Bothofthesestudiesignorethesuppliercapacityissueinmakingsupplierselectionandquantityallocationdecisions.Further,ParlarandWang93notethatsupplieryieldsanddemanduncertaintyplayacriticalroleintheanalysis.Webuilduponbothofthesestudiesbyanalyzingthesimultaneoussupplierselectionandquantityallocationdecisionsforasinglermfacingsupplyunreliabil-ityanddemanduncertainty.Further,weincorporateanexplicitbenetrelatedtorequirementsdiversicationamongthesupplierbaseinidentifyingoptimalsourc-ingstrategies.Inthenextsection,theintegratedsupplierselectionandquantityallocationmodelwhichformsthebasisofouranalysisisdescribed.InSection5.3,wecharacterizetheoptimalsourcingstrategiesundervariousscenarios.InSection5.4,weproceedtodiscussextensionstothismodelingeortandthisisfollowedbyanextensivenumericalanalysisinSection5.5.Finally,conclusionsandmanagerialimplicationsarediscussedinSection5.6. 5.2 IntegratedSelection/AllocationModel 5.2.1 PreliminariesOurexaminationofthesupplierselectionandquantityallocationdecisionsfocusesonasingle-periodanalysisofatwostagesupplychainconsistingofN

PAGE 84

72 suppliersi=1;:::;Nandonebuyingrm.AllNsuppliersareassumedtohavebeenpre-screenedbythermandarethus,includedinthesupplierbase.Thermfacesanuncertainsingle-perioddemandwwithfwandFwrepresentingdensityanddistributionfunctions,respectivelyfortheproductrequirementswhichitsatisesthroughprocurementfromtheNsuppliers.Weassumethatthisproductisbeingsuppliedtothenextstageofthesupplychainataunitpricep.Excessinventoryoftheproductisdisposedofattheendofthesingleperiodbythermwhichreceivesapriceofsperunitwhileunsatiseddemandcosts"thermuperunit.Foreachsupplieriweassumethatthermhasinformationonathecostperunitci;bthecapacityinunitsyi;andcthereliabilityindexrirepresentingthehistoricalpercentageofgood"unitsi.e.,0ri18ireceivedfromthesupplier 1 .Inlinewiththeassumptionsinsingle-periodinventorymodels,weassumethatp>ci>sforalliSilver,Pyke,andPeterson,1998.Weassumezeroxedorderplacementcostsforthermsinceincurrentindustrialsettingswhereordersareissuedonlinethrough,forexample,B2Bexchanges,thesecostsarenegligibleNahmias,2001.Analcomponentofourmodelcharacterizesthediversicationbenetfunc-tiond.Themotivationforincorporatingthisfunctionstemsfromobservedindustrypractices.Forexample,considerHP'ssophisticatedProcurementRiskManagementprogramwhichwasinitiallyaimedatbettermanagingthepro-curementriskofcriticalmemorycomponents.Incorporatedinthisprogramisaportfolioapproachtoassessandmitigatepricingandavailabilityexposuretoinsure 1IncontrasttoParlarandWang1993,weassumethatthermhasexactknowledgeonthereliabilityofeachsupplier.ThisisprimarilyduetothefactthattheuncertaintyinthisparameterrequiredParlarandWang1993toimposesomeveryrestrictiveconditionsonotherparametersinordertoshowthatdual-sourcingispreferredtosingle-sourcing.Wedo,however,analyzetheimpactofchangesinthisparameterontheresultingselection/allocationpolicies.

PAGE 85

73 marginstability.HPexpectstherevenuecontributionofthisprogramtoapproach$1billionasitexpandsitsusethroughoutthecompany.Thisnancialbenetstemsfromreducedunitcostsandavoidingcoststhatarisewhenproductdeliveryisdelayedduetosourcingallocation.Additionally,asaresultofthisprogram,HPhaslearnedthatbysegmentingitsexpectedrequirementsitcantargetcontractstotakeadvantageofasupplier'sparticularstrengths.ThisresultsinmoreecientsupplychainpracticesthatcreatesharedsavingsforHPanditssuppliersShah,2002.However,thesesupplychainsavingsaresometimesdiculttoquantify.Theissueofpositivediversicationbenetsiswelldocumentedthroughanec-dotaland/orcaseexamples.InadditiontoHP,UnineRichardson'sdecisiontochangefromasinglehoneysuppliertomultiplehoneysupplierstomeetmorestrin-gentregulationrequirements;andWendy'sdecisiontondasecondhigh-volumesupplierofchickengiventheincreasingdemandforitschickenproductsillustratestheimportanceofthesediversicationbenetsPrahinski,2002;LambertandKnemeyer,2004.ToyotaandHondaalsohaveapolicyofsourcingallcomponentsfromaminimumoftwoorthreesuppliersLikerandChoi,2004.Thediversicationfunctiondreectsbuyerspecicsupplychaineciencysavingsandstrategicpositioningbenets.Ingeneral,thisfunctioncapturesthenetbenetsofchoosingtosourceproductrequirementsfrommultiplesuppliersandisanalogoustotherisk-averseexpectedutilityfunctionmaximizedintheportfolioselectionproblemGerchakandParlar,1990.BychoosingmultiplesuppliersthermcanreducetheriskassociatedwithselectingasinglesupplierRamaseshetal.,1991.Consequently,thediversicationfunctionisessentiallyinsuranceagainstsupplydisruptionsattributabletothesizeofthesupplybaseforaspeciedpart.Wealsorecognizethatthethereisapotentialdeclineindiversicationbenetsifthenumberofselectedsuppliersistoolargeduetoexcessiveindividualorderrelatedcosts.WeincorporatethisbyassumingthatthatdXisstrictly

PAGE 86

74 concaveinXwhereXrepresentsthenumberofsuppliersselectedbytherm.Tosupportthisspecicfunctionalform,considerthendingsofAgarwalandNahmias997.Theseauthorsshowthatexpectedrmprotsareconcaveinthenumberofsuppliersselectedinasettingcharacterizedbystochasticsupplierquantityreliability.Further,thenotionofsuchbenetsbeingpiece-wiseconcaveisdocumentedinpriorworke.g.,GerchakandParlar,1990;andRamaseshetal.,1991.Thus,ourdXfunctioncouldberegardedasaproxyfordiversicationbenetsassociatedwithuncertaintyinthequantitiesdeliveredbyeachsupplier.Whilethediversicationfunctionmaybediculttoquantify,avalue/utilityfunctionprocedureoutlinedbelowcanbeusedtogaugediversicationvaluationsofarm'sdecisionmakers. 1. Tostartwithassumethatdiversicationbenetsarenotincluded.ThenusingtheresultofTheorem5.1,determinethetotalquantityorderedbythermMQnc=q[1]andthecorrespondingexpectedrmlevelprotsbeMZnc. 2. Setj=2 3. Posethefollowingquestiontothermlevelsourcingmanager:Selectingj)]TJ/F15 11.955 Tf 12.727 0 Td[(1suppliers,youwillorderMQncunitsandexpectprotsofMZnc.AssumingthatyoudecidetoselectjsuppliersfororderingMQncunits,yourexpectedprotswilldeclineforcertainby$cj-cj)]TJ/F21 7.97 Tf 6.587 0 Td[(1perunitorderedfromsupplierj.What$amountwouldcompensateyouforthedeclineinexpectedprotsassumingthatyoudecideto:asourceexactlyMQncunitsfromalljsuppliers?;andborderatleast1unitfromsupplierj.Recordthis$amountasdj. 4. Ifj=Nstopelsesetj=j+1andrepeatStep.Theprocessoutlinedisbasedonestablishedworkinutilitytheoryandde-cisionmakingunderuncertainty,andisanadaptationofthereferencegamble

PAGE 87

75 procedure.Furthermore,itutilizesresultsfromTheorem5.1wherenodiversica-tionbenetsarerealized.Theoutcomeofthisprocessistodetermineavalue"fordifori=2;:::;nd=0sincenodiversicationbenetsarerealizedbythermwithasinglesupplier.However,elicitationofthesevaluesfromprocurementpolicysettersordecisionmakersisnothefocusofouranalysis.Thekeydecisionvariablesforourselection/allocationmodelareboththenumberofsuppliersandtheorderquantityforeachsupplier.Wedenethebinarydecisionvariablexitobe1ifwechoosetosourcefromsupplieri,and0otherwise;andtherelatedallocationquantityqiinunitsprocuredfromsupplieri. 5.2.2 ModelDevelopmentLetusstartbyexpressingtheprotfunctionwithoutdiversicationbenetsas 2 :=8><>:pw)]TJ/F28 11.955 Tf 11.955 8.966 Td[(PNi=1ciriqi+s[PNi=1riqi)]TJ/F22 11.955 Tf 11.956 0 Td[(w]ifw
PAGE 88

76 where=meandemandES=expectednumberofunitsshort=Z1PNi=1qiri[w)]TJ/F15 11.955 Tf 11.955 0 Td[(NXi=1qiri]fwdwanddependinguponthedistributionofdemand,wecanspecifytheexpectedshortageES 3 .Basedonthis,therm'sexpectedprotincludingthediversicationbenetmaximizationsourcingmodelcanbedenedasfollows:MaximizeZqi;xi=E+dX=p)]TJ/F22 11.955 Tf 11.955 0 Td[(s)]TJ/F23 7.97 Tf 16.804 14.944 Td[(NXi=1ciqiri+sNXi=1qiri)]TJ/F15 11.955 Tf 11.955 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uES+dX.1subjectto:qiyixi8i .2 X=NXi=1xi .3 qi08i .4 xi=f0;1g8i .5 whereconstraintset 5.2 integratescapacitylimitationswhenxi=1orsupplieriisselectedorforcesthequantityallocationdecisionqitobe0,ifxi=0orwhensupplieriisnotselected,constraint 5.3 determinesthetotalnumberofsupplierschosenforsourcingtotalproductrequirements,andequations 5.4 and 5.5 3Forexample,ifdemandisnormallydistributedwithmeanandstandardde-viation,thenES=GZwhereGisthelossfunctionandZisthestandardnormaldeviatei.e.,Z=PNi=1qiri)]TJ/F23 7.97 Tf 6.586 0 Td[(

PAGE 89

77 arethenon-negativityandbinaryrestrictionsonthedecisionvariablesqiandxi,respectively.Inthenextsection,weproceedtoanalyzethismodelandcharacterizetheoptimalsolutionsandsourcingstrategiesforseveralcases. 5.3 AnalysisThefocusofourpaperisoninvestigatingsourcingstrategiesforthesupplychain.Primarilyweareinterestedinidentifyingwhenitisoptimalforthemanu-facturertousemultiplesuppliersversusasinglesupplier.Westartouranalysisbyrstassumingazerodiversicationbenet.Theseresultswillserveasabasecaseinouranalysis. 5.3.1 NoDiversicationBenetAssumingthatthermdoesnotobtainanyexplicitdiversicationbenetforchoosingtosourcefrommorethanonesupplier,letusexaminethestructureoftheoptimalsourcingpolicies.WhenthediversicationbenetdX=0,ourintegratedsourcingmodelcanbeformulatedasfollows.MaximizeZqi=p)]TJ/F22 11.955 Tf 11.955 0 Td[(s)]TJ/F23 7.97 Tf 16.804 14.944 Td[(NXi=1ciqiri+sNXi=1qiri)]TJ/F15 11.955 Tf 11.955 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uES.6subjectto:qiyi8i .7 qi08i .8 Letusrstanalyzethisproblemassumingthatsuppliercapacityisnotasignicantissue.Notethatthisscenariomayberelevanttosmallermanufacturingrmswithlargersuppliers.Inthiscase,itisrelativelyeasytoshowthatthermcommitsallitsrequirementstoasinglesupplierandthissupplieris,aswouldbeexpected,theonewhichoersthelowestperunitcosttotherm.Thetheoremgivenbelowformalizesthisresult.

PAGE 90

78 Theorem5.1 : 4 Whenthesuppliersareuncapacitatedandtherearenodiversica-tionbenetsandthereisauniqueleastcostsupplier,thenitisoptimalforthermtoorderitstotalrequirementsfromtheleastcostsupplier.Underthisscenario,thetotalusablequantityorderedfromtheleastcostsupplieri.e.,q[1]r[1]isdeterminedsuchthat:Fq[1]r[1]=p)]TJ/F22 11.955 Tf 11.955 0 Td[(c[1]+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uwherec[1]isthecostperunitchargedbythelowestcostsupplier.Proof :SeeAPPENDIXC.Onesurprisingresultofthissinglesourcingstrategyisthatsupplierreliabili-tiesdonotimpactthesupplierchoicei.e.,thesupplierchoiceisbasedstrictlyoncostconsiderationsregardlessofthequantityreliabilityparameterri.Onfurtherinvestigation,wendthatthisresultissolelyduetothefactthatthemanufactureronlyincursthepurchasingcostforgood"unitsi.e.,incurscostciperunitforriqiunits.Incertainsituations,thecostofdefectiveunitsinadeliverymayneedtobeabsorbedbythemanufacturer.Toreectthisscenario,theuncapacitatedsuppliermodelwithoutdiversicationbenetscanbereformulatedas: 4Whilethisresultisanalyticallytrivial,itisdrivenbythefactthatforallsup-pliersj=1;:::;N,cjrjisaconstant.However,itpointstothefactthatwhenthermuseshistoricalinformationofsupplierreliabilitiesandonlycompensatesasupplierforthegood"unitsreceived,thenonlycostconsiderationsplayaroleindeterminingwhichsupplierwillbechosentoreceivedthecompleteorder.Thisresultalsoformsthebasisofhowtherm'ssinglesupplierselectiondecisionismoderatedshoulditdecidetocompensatesuppliersforallunitsorderedi.e.,thermabsorbsthecompletecostsofdefectiveunitswhichcouldoccurinsituationswheresuppliersholdmorepower"inthechannel.Inthiscase,weshowthatratioofcoststoreliabilitydrivesthechoiceofthesinglesupplierwhowillreceivetheentireorderfromtherm.

PAGE 91

79 MaximizeEe=p)]TJ/F22 11.955 Tf 11.24 0 Td[(s)]TJ/F23 7.97 Tf 16.089 14.944 Td[(NXi=1ciqi+sNXi=1qiri)]TJ/F15 11.955 Tf 11.24 0 Td[(p)]TJ/F22 11.955 Tf 11.24 0 Td[(s+uZ1PNi=1qiri[w)]TJ/F15 11.955 Tf 11.24 0 Td[(NXi=1qiri]fwdw.9subjectto:qi08i.10ItiseasytoshowthatEeisstrictlyconcaveinqi,andthus,theFOCarenec-essaryandsucienttoidentifyaglobaloptimalsolutiontothismodel.Essentially,themanufacturerdeterminesthetotalgood"quantityreceivedfromallsuppliersi.e.,PNi=1qirisuchthat:FNXi=1qiri=p)]TJ/F15 11.955 Tf 11.955 0 Td[(ci=ri+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+u.11Theissue,ofcourse,iswhichsupplier'sreliabilityadjustedunitcosti:e:;ci=riisrelevantindeterminingthistotalquantity.Theoptimalpolicy,whichcanbeeasilyveried,isforthemanufacturertoplaceanorderforthetotalquantityfromasinglesupplierwiththelowestcost/reliabilityratio.Hence,ifsuppliersareindexedinorderofdecreasingcost/reliabilityratiosuchthat:c[1] r[1]c[2] r[2]:::c[N] r[N].12thenthemanufacturerdeterminesthequantitytopurchasefromsupplier[1]suchthat:Fq[1]r[1]=p)]TJ/F15 11.955 Tf 11.956 0 Td[(c[1]=r[1]+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+u.13andorderszerounitsfromallothersuppliers.Thus,eveninthiscase,itisoptimalforthemanufacturertoadoptasinglesourcingstrategyexceptthatthechoiceofthesupplierisbasedonthelowestcost/reliableunit.Giventhateliminatingcapacityconstraintsresultsinsinglesourcing,letusnowproceedtoexaminehowthissolutionchangesifsuppliercapacityconstraints

PAGE 92

80 donotnecessarilypermitthermtoplaceordersforallrequirementswiththeleastcostsupplier 5 .Thetheorembelowcharacterizestheoptimalsupplierselectionandquantityallocationpolicywithcapacitatedsuppliers.Theorem5.2 :Whensuppliersarecapacitatedandtherearenodiversicationbenets,thentheoptimalnumberofsuppliersselectedandthecorrespondingquantityallocatedtoeachsuppliercanbedeterminedasfollows.Step1:Indexallsuppliersinincreasingorderofcostperuniti.e.,c[1]c[2]c[3]:::c[N].Step2:Foreachsupplier[i]i=1;:::;N,determineQ[i]suchthat:FQ[i]=p)]TJ/F23 7.97 Tf 6.587 0 Td[(c[i]+u p)]TJ/F23 7.97 Tf 6.586 0 Td[(s+uandbasedonthisdetermine:Fori=1,t[i]=Q[i].Otherwise,lett[i]=Q[i])]TJ/F28 11.955 Tf 11.955 8.966 Td[(Pi)]TJ/F21 7.97 Tf 6.587 0 Td[(1j=1y[j]r[j]Step3:Theoptimalnumberofsuppliersselectedkismaxf1kNjt[k]0g.Step4:Thequantitiesallocatedtosupplierj=1;:::;k)]TJ/F15 11.955 Tf 12.432 0 Td[(1aredeterminedsuchthatq[j]=y[j],andthequantityallocatedtosupplierkisq[k]=minft[k];y[k]g.ThetotalquantityorderedbythermfromallsupplierscanbedeterminedasminfQ[k];Pkj=1y[j]g.Proof :SeeAPPENDIXC.Aninterestingobservationbasedontheseresultsisthattherm'soptimaltotalorderquantitywhensuppliersarecapacityconstrainedisalwayslowerthantheoptimaltotalorderquantitywhenthelowestcostsupplier'scapacityisnotbinding.Thisleadstothegeneralresultthatexpectedprotsforthermdealing 5Alltheanalysisintheremainderofthischapterreectstheoriginalassump-tionthatsuppliersarecompensatedbythermforonlythegood"unitsreceivedbythermratherthanalltheunitssupplied.

PAGE 93

81 withcapacitatedsuppliersareneverhigherthantheprotsrealizedbyarmdealingwithuncapacitatedsuppliers.Thisservesasarationalefortheobservedindustrypracticeofrmsexpendingresourcestoencouragelowercostsupplierstoincreasecapacity. 5.3.2 DiversicationBenetInthissection,weanalyzeourcompletemodelwhichincludesexplicitbenetsderivedfromthesizeoftheselectedsupplierpool.Tostartwithassumethatsuppliersareuncapacitated.Inthiscase,oursourcingmodelis:MaximizeZqi;xi=p)]TJ/F22 11.955 Tf 11.955 0 Td[(s)]TJ/F23 7.97 Tf 16.805 14.944 Td[(NXi=1ciqiri+sNXi=1qiri)]TJ -90.739 -27.984 Td[()]TJ/F15 11.955 Tf 9.299 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uES+dX .14 subjectto:X=NXi=1xi .15 qi08i .16 xi=f0;1g8i .17 TheoptimalsolutiontothisproblemischaracterizedinTheorem5.3below.Theorem5.3 :Ifthediversicationbenetsarepositiveandsuppliersareuncapaci-tated,thentheoptimalnumberofsupplierstosourcefromisv,where1vn,andisdeterminedsuchthatvmaximizesthediversicationbenetfunctiondX.TheproofofthistheoremisadirectextensionofTheorem5.1.NotethatbasedonTheorem5.1,wechoosetosourcetheentirequantityfrom

PAGE 94

82 thelowestcostsupplier.Ifthediversicationbenetfunctionismaximizedwhenwechoosevsupplierswhere10tosuppliersk+1;:::;vassumingthatsuppliersareindexedinincreasingorderofunitcosts. 6Obviouslyifv=1,thenwewouldsimplysourcetheentirequantityfromasinglelowestcostsupplier

PAGE 95

83 TheimplicationsofTheorem5.3and5.4fordecisionmakingconcerninganappropriatesupplierbaseareclear.RecallfromtheintroductionthattherearethreeinterrelateddecisionswithregardstoarmssourcingstrategyBurkeandVakharia,2002acriteriaforestablishingasupplierbase;bcriteriaforselectingsuppliersasubsetofthebasewhowillreceiveanorderfromtherm;andcthequantityofgoodstoorderfromeachsupplierselected.Theorem5.3oersdirectmanagerialguidanceconcerningtheseconddecision,orthenumberofsupplierswhowillreceiveanorderfromtherm.Furthermore,asaconsequenceofthepositivediversicationbenets,thetotalorderquantitywillslightlyexceedthatoftheoriginalsolutionforthecapacitatedsuppliers.Finally,theremayalsobebenetsbeyondthosedirectlycapturedbythemodelfororderingatanegligiblelevelfromsomeofthesuppliers.Dependingupontheparticularcontractsnegotiatedwiththesesuppliersconcerningupsideorderexibility,themanufacturingrmcouldpotentiallyplacelargerorderswiththesesuppliersintheeventthatlowercostsupplierscannotdelivergoodunits.Notethatif1v
PAGE 96

84 constraintwhichreectstheminimumorderquantityforeachsupplierasshownbelow.qizixi8i.18Whilethecompletesolutionalgorithmforthismodelisfairlycomplicated,wecaneasilyobtainboundariesfortheoptimalnumberofsuppliers.AscomparedtoTheorems5.3and5.4intheprevioussection,wenowhavenon-negligiblecostsassociatedwithincludingadditionalsuppliersinoursupplierbase.Inparticular,thecostsareassociatedwithshiftingenoughunitsfromalowercostsuppliertoahighercostsuppliertomeetthathighercostsupplier'sminimumorderquantity.TheintuitiondevelopedinTheorems5.3and5.4astotheappropriatenumberofsuppliersisstillrelevant.Inparticular,wewouldwanttoconsiderallcandidatesolutionsforthenumberofsupplierswherethemarginaldiversicationbenetsarenon-negative.Therefore,itslikelythattheoptimalnumberofsupplierswillbelessthanorequaltothenumberdeterminedinTheorem5.3.Whilethesediversicationbenetsmaybediculttoquantifyprecisely,thermcananalyzethemarginalcostsassociatedwithare-allocationstrategytoevaluatethediversicationbenets.Toillustrate,iftheorderquantityforthen+1thsupplierisfairlylow,andthecostperunitofthen+1thsupplierisonlyslightlyhigherthanforthenthsupplieri.e.,whenregretismorelikely,thenitmaybeworthwhiletosourcefromn+1suppliers.RecalltheToyotadilemmadiscussedintheintroductionwhereToyotawasforcedtoshutdownanassemblyplantbecauseofaproblemwithitssolesupplierNishiguchiandBeaudet,1998.Inthiscase,asecondarysupplierhedgesagainstthepotentialcostsincurredfromproblemsassociatedwithasinglesupplierstrategy.Themarginalcostsofincludingasecondarysupplierataminimumrequiredlevelcanbeutilizedasaproxyfortheinsurance"premiumnecessarytoreapthebenetsofalargerpoolofsuppliers.Likewise,theimplicitriskpremiumforsinglesourcingcanbe

PAGE 97

85 determinedbycomparingprocurementcostsforsingleversusmultiplesupplierallocationstrategies.Ifweassumethatthermhasalreadydeterminedthesubsetofsuppliersthatwillreceiveorders,thenTheorem5.5belowspeciesthestructureofasimplealgorithmwhichcanbeusedtodetermineoptimalorderquantities.Theorem5.5 :Wheneachsupplierhasbothmaximumandminimumlimitationsplacedonthesizeoftheorder,thentheoptimalquantityallocatedtoeachsuppliercanbedeterminedasfollows.Step1:Indexallchosensuppliersinincreasingorderofcostperuniti.e.,c[1]c[2]c[3]:::c[X].Step2:Foreachsupplier[i]i=1;:::;X,determineQ[i]suchthat:FQ[i]=p)]TJ/F23 7.97 Tf 6.587 0 Td[(c[i]+u p)]TJ/F23 7.97 Tf 6.586 0 Td[(s+uandbasedonthisdetermine:t[i]=Q[i])]TJ/F28 11.955 Tf 11.955 8.966 Td[(Pi)]TJ/F21 7.97 Tf 6.587 0 Td[(1j=1y[j]r[j])]TJ/F28 11.955 Tf 11.955 8.966 Td[(PXj=i+1z[j]r[j]Step3:Thequantityallocatedtosupplieriisq[i]=minfmaxft[i];z[i]r[i]g;y[i]r[i]gandthetotalquantityorderedbythermfromallsupplierscanbedeterminedasPXi=1q[i].Proof :SeeAPPENDIXC.FromTheorem5.5,weknowthatatmostoneofthechosensupplierswillbeunconstrained.Supposethatsupplier[i]isunconstrainedi.e.,q[i]=t[i].Then,theoptimalorderquantityforthelowercostsuppliersj=1;:::;i)]TJ/F15 11.955 Tf 12.631 0 Td[(1isdeterminedbythecapacityconstraintforeachsupplier.Similarly,theoptimalorderquantityforthehighercostsuppliersj=i+1;:::;Xisdeterminedbytheminimumorderquantitydictatedbyeachsupplier.Interestingly,thetotalorderquantityinthissituationie.sumofallordersplacedtothesubsetofsuppliersisdeterminedbythecostoftheunconstrainedsupplier[i]andissuchthatFQ[i]=p)]TJ/F23 7.97 Tf 6.587 0 Td[(c[i]+u p)]TJ/F23 7.97 Tf 6.586 0 Td[(s+u.

PAGE 98

86 However,becausethegeneralizedstructureofthisproblemhasawiderangeofsupplieroptionscorrespondingtoalternatecostlevelswithdieringcombinationsofminimumandmaximumorderquantities,asimplesolutionalgorithmcannotbeeasilyderived.Thisleadsustotheproblemofdeterminingtheoptimalsubsetofpotentialsuppliersthatwillreceiveanorderfromtherm,inadditiontoallocatingappropriateorderquantities.BranchandboundmethodologiescanbeutilizedbasedonTheorem5.5toenumerateallpossiblesubsetsofsuppliersforanoptimalsolution.Inadditiontotheoptimalmathematicalsolution,rmsmaywanttoconsiderdierentqualitativeevaluationmeasuresindetermininganappropriatesubsetofsupplierstosourcefrom.Forexample,inaninternationalsourcingcontext,rmsmaywishtopickasubsetofsuppliersinavarietyofcountries,therebyhedgingagainstcountryspecicriskssuchaschangingpoliticalclimateand/orexchangerates.Wenowturntoanextensivenumericalanalysisinordertoillustratesomeofourresultsandexplorethesensitivityoftheseresultsforkeyparametersinouranalysis. 5.5 NumericalAnalysisAnalyticresultshavebeenpresentedoeringinsightsconcerningtheoptimalchoiceofsuppliersandappropriateorderquantitiesforamanufacturer.Inthissection,wepresenttheresultsofanumericalstudytoillustrateseveralkeycasesforwhichtheanalyticalinsightscannotbeobtained.Wealsoexaminethesensitivityofourresultsbasedonchangesinthekeyinputparameters.Ourintentionistoshowanoverviewoftheseexampleswhichoerinsightsconcerningtherelativeimpactofthesefactorsonarm'ssourcingstrategy. 5.5.1 ExperimentalDesignTheparametersandfunctionswerechosentocapturetheunderlyingassump-tionsoutlinedinSection3.Theexplicitnumericalparametersselectedforthebase

PAGE 99

87 caseexamplereectthoseshowninJuckerandRosenblatt1985.Forthemanu-facturer:aprice/unitp=$19;bsalvagevalues=$2/unit;clostsalescostu=$6/unit;andddemandisassumedtobeuniformlydistributedwithparameters[300,700].Oursupplierbaseconsistsof5suppliersi.e.,i=1;:::;5withidenticalreliabilitiesri=0:9,minimumorderquantitieszi=2008i,andcapacitiesyi=3008i.Suppliersareassumedtobeheterogeneouswithrespecttocostsi.e,ciandtheseparametersettingsarec1=6:5;c2=7;c3=8;c4=9;andc5=10.Whilethediversicationbenetfunctionisdiscrete,weassumethatitisroughlyquadraticinthenumberofsuppliersXandwhenX=1;:::;5,thisfunctionisdenedasdX=d1)]TJ/F22 11.955 Tf 12.453 0 Td[(d2d3)]TJ/F22 11.955 Tf 11.955 0 Td[(X2andwhenX=0,d=0.ThisfunctionalformofdXwaschosensincethesingleparameterd3representstheoptimalnumberofsupplierswhichmaximizesthisfunction.AlltheresultsdiscussednextwereobtainedusingLINGOoptimizationsoftware. 5.5.2 ResultsTable5.1summarizetheresultsofasetofnumericalexampleswhichshowsensitivityoftheoptimalsupplierstrategytochangesinparametervalues.ModelArepresentsthecasewherebothdiversicationbenetsandsupplierminimumorderquantitiesareincluded.FortheremainingexamplesinTable5.1,theparameterchangesarespeciedinthevariablerangecolumn.Notsurprisingly,modelsBandCinTable1showthatanincreaseinthepriceorsalvagevalueoftheitemsincreasesthetotalquantityordered,thetotalnumberofsupplierssourcedfromandthetotalprot.Similarly,anincreaseintheunderagecostsincreasesthetotalquantityordered,thetotalnumberofsuppliers,butdecreasesthetotalprotearned.ModelEisintendedtoillustratetheimpactofchangesinthediversicationbenetfunctionontheoptimalsourcingstrategy.Inthismodel,thepeakinthemagnitudeofdiversicationbenetsearnedi.e.,d1isvariedbetween$250and

PAGE 100

88 $2000itissetto$1000forthebasecaseexample.Inresponsetoanincreaseinthepeakvalueofthediversicationfunction,thetotalorderquantityandtheoptimalnumberofsuppliersremainsthesame,whiletheprotincreases.Thiswouldindicatethattheoptimalsourcingpolicyisfairlyrobustinthatitisnotsensitivetolargeincreasesinthepeakdiversicationvalueforthisexample.ModelsFandGillustratehowtheoptimalsourcingpolicychangeswithalterationsintherstsupplier'scostandreliability.Whilesmallincreasesintherstsupplier'scostdoesnotchangetheoptimalnumberofsuppliers,itdoesdecreasethetotalquantityorderedandthetotalprot.Similarly,anincreaseintherstsupplier'sreliabilitydecreasesthetotalnumberofunitsorderedandincreasesprot.Ingeneral,thermsimplycompensatesforsmallchangesinreliabilitybyorderingproportionatelymoreitemssinceitdoesnotpayforthebadunits. Table5{1:SensitivityAnalysisoftheKeyParameters Model Parameter FirmProt n* Q* NonzeroAllocations Range $ q1 q2 q3 A NA 6103.59 3 691 291 200 200 B p=[10,25] [1737,9065] [2,3] [600,701] [300,300] [300,201] [0,200] C s=[-6,6] [5375,6703] [2,3] [591,704] [300,300] [291,254] [0,200] D u=[0,12] [6166,6065] [3,3] [660,701] [260,300] [200,201] [200,200] E d1=[250;2000] [5354,7103] [3,3] [691,691] [291,291] [200,200] [200,200] F c1=[6:25;6:75] [6170,6039] [3,3] [667,658] [296,286] [200,200] [200,200] G r1=[0:5;1] [6045,6104] [3,3] [700,662] [300,262] [200,200] [200,200] NextweshowtheimpactoftheminimumorderquantityconstraintsontheoptimalsourcingstrategyseemodelextensionsdescribedinSection5.5.Inthiscase,wechoosethreescenariostoillustrateourresults.Inallthreescenarios,theresultsaregeneratedassumingcapacitatedsuppliers.Table5.2containstheresultsforthesethreescenariosinthefollowingmanner:ModelAisthesameasinTable5.1;modelHrepresentsthecasewherediversicationbenetsareincludedwithoutsupplierminimumorderquantities;andmodelIrepresentsthecasewheresupplierminimumorderquantitiesareincorporatedintheabsenceofdiversication

PAGE 101

89 benets.FormodelsAandH,thediversicationbenetfunctionparameterswered1=1000,d2=62:5andd3=4.ComparingmodelsAandH,theoptimalnumberofsuppliersisreducedinthepresenceoftheminimumorderquantityconstraints.Inthiscase,themarginalbenetsofdiversicationfromincludinganadditionalsupplierdonotoutweighthemarginalcostsofplacingaminimalorderof200unitsfromthe4thsupplier.InmodelH,notethattheoptimalnumberofsuppliersis4withthefourthsupplieractuallyreceivinganorderofzero.Thisoccursduetotheabsenceofaminimumorderquantity,andreectsthesituationwherethebuyingrmwouldoptimallyqualifythefourthsupplier.Ideally,acontractualarrangementwouldbenegotiatedwiththissupplierfacilitatinganagreementwherebythebuyingrmcouldplaceanactualorderinanemergencysituation.ComparingmodelsAandI,theoptimalnumberofsuppliersisreducedwhenthediversicationbenetsareequaltozero.Morespecically,whennodiversicationbenetsexist,thenthemarginalcostofplacinganorderwiththethirdsupplierattheminimallevelof200unitsisnoteconomicalgreat. Table5{2:ImpactofMinimumOrderQuantityontheSourcingStrategy Model FirmProt n* Q* QuantityAllocations $ q1 q2 q3 q4 q5 Sales Diversication Total A 5166.09 937.50 6103.59 3 691 291 200 200 0 0 H 5288.04 1000.00 6288.04 4 662 300 300 62 0 0 I 5199.00 0.00 5199.00 2 600 300 300 0 0 0 AthirdsetofnumericalexamplesshowninTable5.3illustratesfurtherinterestinginteractionsthatcanoccurbetweentheminimumorderquantitiesandthereliabilityfactorsforthesuppliers.Diversicationbenets,minimumorderquantities,andcapacityconstraintsareincludedforbothofthesemodels.Inbothoftheseexamples,thesuppliercostsarec1=6:5;c2=7;c3=8;c4=9;andc5=10.InmodelA,thereliabilitiesoftheindividualsuppliersarehomogenous

PAGE 102

90 andequalto0.90.InmodelJ,thereliabilityfactorofthe4thsupplierissettoequalhalfofthatofothersuppliers,i.e.,0.45.Theimpactofthelowerreliabilityfactorontheoptimalsolutionisthatsupplier4receivesanorderwhilesupplier3doesnot.Moreover,thebuyingrmordersmoreunitsandmakesmoreprot.Thisresultissomewhatcounter-intuitive,inthatlowerreliabilityleadstohigherprotandtotalorderquantities.Moreover,protincreasesbysourcingfromahighercostsupplier.RecallrstthatintheoriginalmodelfromSection5.3.2,weassumethatthebuyingrmpaysonlyforgoodunitsdelivered.Theneteectofthisassumptionisthattheminimumorderquantityintermsofthegoodunitsdeliveredismuchlowerforthelowerreliabilitysupplier.Also,notethatthetotalquantitydeliveredisactuallylowerformodelJthanformodelAduetothelowerreliabilityfactor.Therefore,theremaybesituationswhereitisoptimaltosourcemoreunitsfromalowerreliabilityyethighercostsupplier. Table5{3:InteractionsbetweenMinimumOrderQuantitiesandReliabilities Model FirmProt n* Q* QuantityAllocations $ q1 q2 q3 q4 q5 Sales Diversication Total A 5166.09 937.50 6103.59 3 691 291 200 200 0 0 J 5142.39 937.50 6079.89 3 781 300 281 0 200 0 Finally,basedonmodelAandadjustedcapacitiesof700foreachsupplier,weexaminecaseswherethethreelowestcostsuppliersalsohavedieringreliabilitiesandminimumorderquantities.EachofthethreelowestcostsuppliersisrankedasbestB,middleM,orworstWforcost,reliabilityandminimumorderquantity.Notationally,instanceBWMsigniesthatthelowestcostsupplierisworstinregardtoreliabilityandmiddleinregardtominimumorderquantity.Table5.4providesasummaryofparametervaluesusedforthisexperimentforeachrankorder.

PAGE 103

91 Table5{4:ParameterValuesforRankedSupplierCharacteristics Cost Reliability MinimumOrderQuantity Best 6.50 .9 100 Middle 7.00 .8 150 Worst 8.00 .7 200 Table5.5summarizestheoptimalsourcingstrategiesforselectedinstances.AgeneralinsightfromthisexperimentisthatthetotalusablequantityisdeterminedbythelowestcostsupplierwhoisnotoptimallyallocatedanamountequaltoitsminimumorderquantityoritscapacityconrmingtheintuitiveresultdevelopedinTheorem5.5.Additionally,itistypicallypreferabletohavehighercostsupplierswithlowerreliabilities.Thissituationeectivelylowerstheminimumorderquantitiesofhighercostsuppliersandrequiresshiftingfewerunitsfromlowercostsupplierstogainincrementaldiversicationbenets. Table5{5:SelectedResultsforCapacityAdjustedModelA Supplier Firm Optimal Good Characteristic Prot Quantity Units Vectorc,r,z $ Allocations Supp.1 Supp.2 Supp.3 Sales Divers. Total q1 q2 q3 Q* BBB MMM WWW 5466.09 750.00 6216.09 557 150 0 622 BBB MMW WWM 5288.59 937.50 6226.09 396 200 150 622 BBW MMM WWB 5361.09 937.50 6298.59 480 150 100 622 BMB MBW WWM 5278.59 937.50 6216.09 421 200 150 622 BMB MWW WBM 5456.09 750.00 6206.09 602 200 0 622 BWB MBW WMM 5256.09 937.50 6193.59 460 200 150 622 BWB MMM WBW 5462.39 750.00 6212.39 700 154 0 613 BWM MBB WMW 5462.39 750.00 6212.39 700 137 0 613 BWM MBW WMB 5316.09 937.50 6253.59 517 200 100 622 5.6 ConclusionsandImplicationsAnalyticandnumericalanalysisofourmodelprovideseveralmanagerialinsights.Tostartwith,considerthesituationwheresupplierminimumorderquantitiesarenotconsideredandwhenthermdoesnotobtainanyexplicitbenetsbydiversifyingitssupplierbase.First,theindustrypracticeofsinglesourcingisonlyoptimalwhensuppliercapacitiesarerelativelylargeascomparedtoproductdemand.Insuchacase,therm'soptimalchoiceistosourceallits

PAGE 104

92 requirementsfromtheleastcostsupplier.Interestingly,supplierreliabilitiesdonotmoderatethechoiceofthesupplierunlessthermisrequiredtocompensatesuppliersforallunitsorderedratherthansimplythegood"unitsreceived.Inthelattercase,theratioofcoststoreliabilitiesisrelevantindeterminingthesupplierfromwhichalldemandissourced.Second,weshowthatwhensuppliercapacitiesarerelevant,theoptimalstrategyforthermistosourcefrommultiplesuppliers.Underthisscenariowendthattherm'stotalorderquantityacrossallsuppliersandexpectedprotsarebothlowerthanthatcomparedtothescenariowheresuppliersareuncapacitated.Thedierenceinprotscouldberegardedasthevaluetothermwhichcouldberealizedifthelowestcostsuppliercouldbemotivatedtoincreasehis/hercapacity.Whenpositivenetdiversicationbenetsareincorporatedwithoutsupplierminimumorderquantities,thekeyresultsareasfollows.Ifsuppliersareunca-pacitated,thenmultiplesuppliersourcingstrategiesarealwaysoptimalwherethenumberofsuppliersisdeterminedbythediversicationbenetfunction.Manage-rially,thisimpliesthatthermshoulddeterminethetotalorderquantitybasedontheleastcostsupplier.However,inplacingorders,itshouldordertherequiredamountfromtheleastcostsupplierandordermarginalquantitiesfromalltheotherselectedsuppliers.Whensuppliersarecapacitated,asimilarsimpledecisionrulecanbeusedbythermwhenthenumberofsupplierswhichoptimizesthediversicationbenetsislargerthanthenumberwhichareselectedwithoutsuchabenet.Throughanextensivenumericalanalysiswealsoexaminetherobustnessofourresultswhensupplierminimumorderquantitiesarerelevantinmakingarm'ssourcingdecisions.Acounter-intuitiveinsightweobtainforthiscaseisthatthereisaninteractionbetweenreliabilities,costs,andminimumorderquantities.Forexample,weshowthatincertaincases,itmaybeoptimaltosourcefroma

PAGE 105

93 highercost,lowerreliabilitysupplierascomparedtoalowercost,higherreliabilitysupplier.Thisisgenerallythecasewhenalowereectiveminimumorderquantityiseconomicallypreferable.Theinsighthereisthattheexibilityofasuppliermayhavegreaterbearingonselectionthanunitcost.Thegeneralmodelinthischapterincorporatesstrategicdiversicationcon-siderationswithinthetraditionalnewsvendorframeworktodeterminetheoptimalnumberofsupplierstoplaceanorderwithandthecorrespondingquantitiesofthoseorders.Throughtheintroductionofthediversicationbenetfunction,weexplicitlyaccountforbuyerspecicsupplybasemanagementbenetsbasedonthesizeofaselectedsupplierpool.Inessence,thisfunctioncouldbeconstruedascapturingthemonetizedutilityofinsurancenaturallyprovidedbythesizeoftheselectedsupplierpool.Thisconsiderationhighlightstheneedforapurchasingmanagertoexecutesourcingstrategiesconsistentlywiththerm'soperationsandbroadercorporatestrategy.

PAGE 106

CHAPTER6SUMMARYThestrategicimportanceofsourcingactivitiesisinherentintheirvitalitytoarm'smeansofvaluecreation.Sourcingactivitieslinkthermtoitschainofsup-ply.Arm'ssourcingstrategyisthereforeakeydriverofaneectivesupplychainstrategy.Arm'ssourcingstrategycanbecharacterizedbythreekeydecision:acriteriaforestablishingasupplybase;bcriteriaforselectingsuppliersfromtheapprovedbasetoreceiveorders;andcthequantityofgoodstoorderfromeachsupplierselected.Decisionsbandcaredecisionsofselectionandallocationrespectively.Inthisdissertationtheselectionandallocationdecisionsofarm'ssourcingstrategyarestudiedundervariousoperatingconditionscommonlyencoun-teredinpractice.Weareespeciallyfocusedondeterminingwhenitisoptimalforarmtochooseaspecializedsinglesourcingstrategyversusageneralizedmultiplesourcingstrategy.Inparticular,thesupplychainweexamineconsistsofasingleintermediarybuyingrmservicedbymultipleupstreamsuppliers.Byseparatelymodelingandexaminingtheinuenceofuncertaintyinproductsupplyandde-mand,alternatequantitypricingschemesquotedbysuppliers,andexplicitbenetsofadiversiedsourcingstrategy,weprovidemanagerialinsightsforthisbuyingrm'soptimalsourcingstrategy.Theremainderofthissummarychapterreviewsthekeyresultsofouranalysisandnumericalexperimentsineachofchapters3,4and5,anddiscussesvariousavenuesforfutureresearchrelatedtothisdissertation. 6.1 KeyResultsandDirectionsforFutureResearchfromChapter3Businessoperationsareoftenfacedwithuncertaintiesinproductsupplyanddemand.Chapter3providesstructuralandnumericalresultsfordetermininganappropriatesuppliersourcingstrategyinthepresenceofupstreamanddownstream 94

PAGE 107

95 uncertainty.Inthecontextofthesupplierselectiondecision,ourresultsareinlinewithobservedpractice.Forexample,VermaandPullman996ndthatwhilesupplymanagersrecognizetheimportanceofquality,costprimarilydrivestheirsupplierselectiondecisions.Inourmodel,asupplier'scostandnotitsreliabilityisthekeyfactorwhichcomesintoplaywhenarmisdecidingwhetherornottoplaceanorderwiththatsupplier.Consequently,thelowestcostsupplierinthepre-qualiedpoolwillalwaysreceiveapositiveorder.Anexceptiontothisruleshownthroughnumericalexamplesiswhenthelowestcostsupplierhasarestrictivelyhighminimumorderquantity.Itfollowsthatifallpre-qualiedsuppliershavesimilarcosts,thenit'soptimaltoplaceanorderwithallsuppliersinthepool.Also,forconvenientevaluationoftherm'ssupplierselectiondecision,wederiveasimpleratiotoanalyticallydeterminewhetherornotasinglesupplierstrategyisappropriate.Thisratioreectsatrade-obetweentherstsupplier'sreliabilityand itscostadvantagerelativetoothersuppliers.Essentially,ifthelowestcostsupplierhasareliabilitydistributionwithahighmeanandalowstandarddeviation,andhasalargecostadvantage,thenasinglesupplierstrategyiswarranted.Anotherkeyfactorinuencingtherm'ssupplierselectiondecisionistherm'santicipateddemand.Bothanalyticandnumericalresultsconrmthatasinglesupplierstrategyisfavorablewhenthemeandemandislow.However,ifarmanticipatesasignicantincreaseindemand,itshouldconsiderenlargingitssupplierbaseevenwhenthelowcostsuppliercouldprovidethefullorderquantity.Surprisingly,anincreaseinthevariabilityindemandalsofavorsasinglesourcingstrategy.Inthiscase,thermlimitsitsnancialriskbysourcingonlywiththesinglelowestcostsupplierwhenthermanticipatesgreatuncertaintyindemand.Therm'sorderquantityallocationdecisionconcernshowmuchtoorderfromeachselectedsupplier.Analyticexpressionsaredevelopedwhichdeterminethe

PAGE 108

96 optimalorderquantityforeachselectedsupplierunderavarietyofcircumstances.ThemostgeneralcaseisshowninTheorem3.1whichaddressesthesituationwhereallsuppliershavedierentcostsandreliabilityfunctions.Inthiscase,eachsupplierwillreceiveanorderamountbasedonitsunitcost,meanreliability,andvarianceinreliability.Notethatwhilethelowestcostsupplierisguaranteedtoreceiveapositiveorder,he/shewon'tnecessarilyreceivethelargestorder.Incontrast,wealsoanalyzethesituationwherethesuppliersarehomogeneousintheircostsandreliabilityfunctionsinTheorem3.4.Inthissituation,allpre-qualiedsuppliersreceiveanequivalentorderquantity.Thereareseveralareasofresearchrelatedtochapter3whichwarrantfurtherinvestigation.First,themodelmaybeextendedtoencompassmultipleperiodsandmultipleproducts.Also,amoredetailedmodelcouldbedevelopedwhichaddressesthexedandpotentiallycumbersomecostsoftheinitialsupplierqualicationprocess.Anothermodicationofthemodelforfutureconsiderationistoincorpo-ratenotonlyqualityreliability,butalsotimelinessofdeliveryinamultipleperiodmodel.Further,weassumethatdemandisuniformlydistributedtofacilitatethedevelopmentofsimpliedexpressions.Othertypesofdemanddistributionscouldbeexploredtoenhancethegeneralizabilityofchapter3'sresults.Anothermodicationofthemodelforfutureconsiderationistoincorporatenotonlyqualityreliability,butalsotimelinessofdeliveryinamultipleperiodmodel.Lastly,thefocusofthismodelisondecisionmakingatthebuyingrm;futureresearchcouldincorporatethesupplier'sdecisionmakingprocessaswell. 6.2 KeyResultsandDirectionsforFutureResearchfromChapter4Evaluatingalternativepricingschedulesoeredasquantitydiscountsisaquantitativeproblemcommonlypresentedtosourcingdecisionmakers.Inchapter4,wemodelacentralpurchasingorganization'ssupplierselectionandorderquantityallocationdecisionswithsupplierquotedquantitydiscountpricing.We

PAGE 109

97 examinethesedecisionsforenvironmentswheresuppliersoerconstant,lineardiscount,incrementaldiscount,andall-unitdiscountpricingschemes.Ouranalysisofalternatesupplierbasepricingschemesprovidesguidanceforabuyingrm'soptimalsourcingstrategy.Fortheconstantcostcase,ifthelowestcostsupplierhasenoughcapacitytosatisfyallofthebuyingrm'sproductrequirements,singlesourcingisoptimal,otherwisemultiplesourcingisbest.Forsourcingscenarioswheresupplierbasesoerquantitydiscountpricingschemes,ifallsuppliersinthebasepossessenoughcapacitytoindividuallyprovideQunits,thenitisoptimaltosinglesourcefromtheleastcostsupplierevaluatedatQunits.However,ifevenonesupplierinthebaseisincapableofprovidingQunits,thensinglesourcingmaynotbetheoptimalsourcingstrategy.Incaseswherequantitydiscountquotingsuppliers'capacityisindividuallyinadequateforabuyingrm'sproductrequirements,arm'ssourcingproblemcanbecomeextremelycomplex.Therefore,identifyinganoptimalsolutionmayberesourcetimeprohibitiveforsupplychainsourcingprofessionals.Insuchcases,theheuristicsdevelopedinthischaptercanbeexpectedtoecientlyprovidegoodqualitysolutions.Insum,theheuristicsolutionstorandomlygeneratedtestproblemsarrivedattheoptimalsolutionin62%oftheinstances.Furthermore,inthe38%ofnon-optimalsolutioninstances,arespectable:64%averageoptimalitygapexists.LeveragingResult4.1,wehavedevelopedecientandwell-performingheuristicstogeneratesupplierquantityallocationswheretheselectedsupplierscapacitybinds.Additionally,wevalidateLINGO'sGlobalSolveroptimalsolutionsforourincrementaldiscountstestproblemsbydevelopingabranchandboundalgorithm.Animmediateextensionofthemodelsdevelopedinchapter4istoconsidermultipleproductsrequiredbythebuyingrmandsuppliedbymultiplesuppliers.

PAGE 110

98 Additionally,quantitydiscountscanbeappliedbasedontotalpurchasevalueinsteadofnumberofunits.Anothernoteworthymodicationisforeachsuppliersaggregatecapacitytodependontheproductmixorbundleorderedbytherm.Finally,incorporatingdownstreampricedependencyordemanduncertaintywithupstreamquantitydiscountpricingwarrantsfurtherexamination. 6.3 KeyResultsandDirectionsforFutureResearchfromChapter5Chaptervemodelsabuyingrmssourceselectionandorderquantityallocationdecisionsinanoperatingenvironmentcharacterizedbyunreliableupstreamsupplyanduncertaindownstreamdemand.Analyticandnumericalanalysisofthismodelprovideseveralmanagerialinsights.First,ifsupplierminimumorderquantitiesarenotconsideredandthermobtainsnoexplicitbenetsbydiversifyingitssupplierbase,thensinglesourcingisoptimalwhenthelowestcostsupplierscapacityisrelativelylargeascomparedtoproductdemand.Interestingly,supplierreliabilitiesdonotmoderatethechoiceofthesupplierunlessthermisrequiredtocompensatesuppliersforallunitsorderedratherthansimplythegood"unitsreceived.Inthelattercase,theratioofcoststoreliabilitiesisrelevantindeterminingthesupplierfromwhichalldemandissourced.Second,weshowthatwhensuppliercapacitiesarerelevant,theoptimalstrategyforthermistosourcefrommultiplesuppliers.Underthisscenariowendthattherm'stotalorderquantityacrossallsuppliersandexpectedprotsarebothlowerthanthatcomparedtothescenariowheresuppliersareuncapacitated.Thedierenceinprotscouldberegardedasthevaluetothermwhichcouldberealizedifthelowestcostsuppliercouldbemotivatedtoincreasehis/hercapacity.Whenpositivenetdiversicationbenetsareincorporatedwithoutsupplierminimumorderquantities,thekeyresultsareasfollows.Ifsuppliersareunca-pacitated,thenmultiplesuppliersourcingstrategiesarealwaysoptimalwherethe

PAGE 111

99 numberofsuppliersisdeterminedbythediversicationbenetfunction.Manage-rially,thisimpliesthatthermshoulddeterminethetotalorderquantitybasedontheleastcostsupplier.However,inplacingorders,itshouldordertherequiredamountfromtheleastcostsupplierandordermarginalquantitiesfromalltheotherselectedsuppliers.Whensuppliersarecapacitated,asimilarsimpledecisionrulecanbeusedbythermwhenthenumberofsupplierswhichoptimizesthediversicationbenetsislargerthanthenumberselectedwithoutsuchabenet.Throughextensivenumericalanalysiswithsupplierminimumorderquantityconstraints,aninsightweobtainisthatthereisacounter-intuitiveinteractionbetweenreliabilities,costs,andminimumorderquantities.Forexample,weshowthatincertaincases,itmaybeoptimaltosourcefromahighercost,lowerreliabilitysupplierascomparedtoalowercost,higherreliabilitysupplier.Thisisgenerallythecasewhenalowereectiveminimumorderquantityiseconomicallypreferable.Theinsighthereisthattheexibilityofasuppliermayhavegreaterbearingonselectionthanunitcost.Thisisalsosimilartoatotalcostofownershipapproachtosupplierselection.Theexplicitdiversicationbenetsmodeldevelopedinchaptervecanbeextendedforfutureresearchbyexaminingmultipleproductsourcingscenariosaswellasmultipleperiodproblems.Aninterestingandgrowingareaofpracticalapplicationisincombinatorialprocurementauctions.Also,operationalizingthediversicationbenetfunctionisanareaofempiricalresearchinterest.

PAGE 112

APPENDIXAPROOFSFORCHAPTER3 A.1 ProofofCorollary3.1Corollary3.1: TheexpectedprotfunctionshowninEquation 3.1 isconcaveintheorderquantitiesqiforNsupplierswhendemandisuniformlydistributedbetween[a;b].Proof :Giventhatdemandisuniformlydistributed,theobjectivefunctioninEquation 3.1 canbesimplied:E[]=)]TJ/F23 7.97 Tf 6.586 0 Td[(ua+b 2)]TJ/F21 7.97 Tf 13.151 5.698 Td[(p+u)]TJ/F23 7.97 Tf 6.587 0 Td[(sa2 2b)]TJ/F23 7.97 Tf 6.586 0 Td[(a+NPi=1p+u)]TJ/F22 11.955 Tf 11.955 0 Td[(ciriqi+p+u)]TJ/F23 7.97 Tf 6.587 0 Td[(sa b)]TJ/F23 7.97 Tf 6.587 0 Td[(aNPi=1riqi)]TJ/F21 7.97 Tf 10.494 5.699 Td[(p+u)]TJ/F23 7.97 Tf 6.586 0 Td[(s 2b)]TJ/F23 7.97 Tf 6.586 0 Td[(aNPi=1Viq2i+2N)]TJ/F21 7.97 Tf 6.587 0 Td[(1Pi=1NPj=irirjqiqj!TheHessianfortheobjectivefunctionwithNsuppliersisasfollows:HN=L0H0=L00BBBBBBBBBBBBBBBBB@)]TJ/F22 11.955 Tf 9.298 0 Td[(V1)]TJ/F15 11.955 Tf 9.823 0 Td[(r1r2:)]TJ/F15 11.955 Tf 9.823 0 Td[(r1ri::)]TJ/F15 11.955 Tf 9.823 0 Td[(r1rN:::::::::::::)]TJ/F15 11.955 Tf 9.822 0 Td[(rir1)]TJ/F15 11.955 Tf 9.823 0 Td[(rir2:)]TJ/F22 11.955 Tf 9.298 0 Td[(Vi::)]TJ/F15 11.955 Tf 9.823 0 Td[(rirN::::::::::::::)]TJ/F15 11.955 Tf 9.822 0 Td[(rNr1)]TJ/F15 11.955 Tf 9.823 0 Td[(rNr2:)]TJ/F15 11.955 Tf 9.823 0 Td[(rNri::)]TJ/F22 11.955 Tf 9.299 0 Td[(VN1CCCCCCCCCCCCCCCCCAwhereL0=p+u)]TJ/F23 7.97 Tf 6.586 0 Td[(s b)]TJ/F23 7.97 Tf 6.586 0 Td[(aandVi=2i+r2i.Then,thedeterminantoftheHessianforNsuppliersis:jHNj=)]TJ/F22 11.955 Tf 9.298 0 Td[(L0N"NXi=1r2iYj6=i2j+NYi=12i# 100

PAGE 113

101 Sincethemeanandvarianceforallreliabilitydistributionsarepositive,thenthesignofthedeterminantisdeterminedby)]TJ/F22 11.955 Tf 9.298 0 Td[(L0N.Also,utilizingstandardassumptionsfornewsboyproblems,wehavep>ci>s.Sinceb>afortheuniformdistribution,thenL0>0andthesignofthedeterminantoftheHessianmatrixissimply)]TJ/F15 11.955 Tf 9.299 0 Td[(1N.Inordertoestablishconcavityoftheobjective,allofthekprincipalminorsfortheHessianmatrixmusthavethesamesignas)]TJ/F15 11.955 Tf 9.299 0 Td[(1k.BysubstitutingkforNintheexpressionforthedeterminantoftheHessian,itfollowsthattheprincipalminorshavetheappropriatesignsuchthattheobjectiveisnegativedeniteandconsequentlyconcaveforanyN. A.2 ProofofTheorem3.1Theorem3.1 :Whensuppliersareheterogeneouswithrespecttocostsandreliabil-ityparameters,thentheoptimalsourcingquantityforeachsupplieriis:qi=ri 2i"NPj=1bi)]TJ/F22 11.955 Tf 11.955 0 Td[(bjrj j2+bi# NPj=1rj j2+1A.1wherebi=Kib)]TJ/F22 11.955 Tf 11.955 0 Td[(a+a,andKi=p+u)]TJ/F23 7.97 Tf 6.586 0 Td[(ci p+u)]TJ/F23 7.97 Tf 6.586 0 Td[(s.Proof :Theprooffollowsdirectlyfromtherstorderconditionsofoptimality.ConsiderrsttheoptimalorderquantityfortherstsupplierwhenN=1-5:q1=r1b1 r21+21;forN=1q1=r1[r22b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b2+b122] DwhereD=2122+r2221+r2122;forN=2q1=r1[b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b2r2223+b1)]TJ/F23 7.97 Tf 6.587 0 Td[(b3r2322+b12223] r212223+r222123+r232122+212223forN=3q1=r1[b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b2r222324+b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b3r232224+b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b4r242223+b1222324] r21222324+r22212324+r23212224+r24212223+21222324forN=4q1=r1[b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b2r22232425+b1)]TJ/F23 7.97 Tf 6.587 0 Td[(b3r23222425+b1)]TJ/F23 7.97 Tf 6.587 0 Td[(b4r24222325+b1)]TJ/F23 7.97 Tf 6.586 0 Td[(b5r25222324+b122232425] r2122232425+r2221232425+r2321222425+r2421222325+r2521222324+2122232425forN=5ItfollowsthatforN>2:

PAGE 114

102 qi=ri2666666666666666666666666664NPj=1j6=ibi)]TJ/F23 7.97 Tf 6.586 0 Td[(bjr2jNQk=1k6=ik6=j2k+biNQj=1j6=i2j3777777777777777777777777775 NPi=1r2iNQj=1j6=i2j+NQi=12iorqi=ri 2i"NPj=1bi)]TJ/F23 7.97 Tf 6.586 0 Td[(bjrj j2+bi# NPj=1rj j2+1. A.3 ProofofCorollary3.2Corollary3.2: Thermwillalwaysorderapositivequantityfromthelowestcostsupplier.Proof :SupposethatciKj,andbi>bk.FromEquation A.1 ,itisclearthatqi>0. A.4 ProofforCorollary3.3Corollary3.3 :Thermwillsourceallitsrequirementsfromthelowestcostsupplieri.e.,useasinglesourcingstrategyifandonlyif:1 r121receiveanorder.Then,ifthefollowingEquation

PAGE 115

103 holds,supplierjj=2;::;nwillreceiveahigherquantityorderthansupplier1:b1)]TJ/F22 11.955 Tf 11.955 0 Td[(bj=b)]TJ/F22 11.955 Tf 11.955 0 Td[(acj)]TJ/F22 11.955 Tf 11.955 0 Td[(c1 p+u)]TJ/F22 11.955 Tf 11.955 0 Td[(srjbj21)]TJ/F15 11.955 Tf 12.479 0 Td[(r1b12j r1rjr1)]TJ/F15 11.955 Tf 12.479 0 Td[(rj A.6 ProofofCorollary3.5Corollary3.5: Ahighercostsupplierwillneverreceiveapositiveorderwhenalowercostsupplier'sorderquantityisequaltozero.Proof: Considerathreesupplierexample,andorderthesupplierssuchthatc10i.e.,q2=0and3=0i.e.,q3>0.From2>0andsolvingforc1,wehave:2=r2[c2)]TJ/F22 11.955 Tf 11.955 0 Td[(c3r2321+c2)]TJ/F22 11.955 Tf 11.955 0 Td[(c1r2123+c2)]TJ/F22 11.955 Tf 11.955 0 Td[(ah)]TJ/F22 11.955 Tf 11.956 0 Td[(p)]TJ/F22 11.955 Tf 11.956 0 Td[(u2123] r2321+r2123+2123>0c10andsolvingforc1,wehave:q3=c1r21)]TJ/F22 11.955 Tf 11.955 0 Td[(c3r21)]TJ/F22 11.955 Tf 11.956 0 Td[(c321+21ah+p+u>0c1>K2=c3r21+c321)]TJ/F15 11.955 Tf 11.956 0 Td[(ah+p+u21 r21SinceK1n)]TJ/F18 5.978 Tf 9.576 0 Td[(1Pj=1bjrj j2 1+n)]TJ/F18 5.978 Tf 9.575 0 Td[(1Pj=1rj j2andbn+1nPj=1bjrj j2 1+nPj=1rj j2Proof:TheprooffollowsdirectlyfromtherstorderconditionsofoptimalityshowninEquation A.2 assumingthatqn>0andqn+10.

PAGE 116

104 A.8 ProofofTheorem3.2Theorem3.2: Whensuppliersahaveheterogeneouscoststructuresandbhaveidenticalreliabilitydistributions,thenthereisnoonedominantsuppliersourcingstrategy.Theoptimalorderquantityforeachsupplieriis:qi=r"r2NPj=1bi)]TJ/F23 7.97 Tf 6.586 0 Td[(bj+bi2# Nr22+48i.Proof :FollowsdirectlyfromEquation A.1 A.9 ProofofTheorem3.3Theorem3.3: Whensuppliersahaveidenticalcoststructuresandbhaveuniquereliabilitydistributions,thenitisoptimaltoorderfromallsuppliers.Theoptimalorderquantityforeachsupplieriis:qi=ri 2ib0 NPj=1rj j2+18iwhereb0=Kb)]TJ/F22 11.955 Tf 11.955 0 Td[(a+a,andK=p+u)]TJ/F23 7.97 Tf 6.587 0 Td[(c p+u)]TJ/F23 7.97 Tf 6.587 0 Td[(s.Proof :FollowsdirectlyfromEquation A.1 A.10 ProofofTheorem3.4Theorem3.4: Whensuppliersahaveidenticalcoststructuresandbhaveidenticalreliabilitydistributions,thenitisoptimaltoorderthesameamountfromallsuppliers.Theoptimalorderquantityforeachsupplieris:qi=rb0 2+Nr28i:A.3Proof :Letci=c,ri=r,gi=g,Gi=G,andi=inEquation A.1 .Then,theoptimalorderquantityforallsuppliersis:qi=r[0+b02N)]TJ/F18 5.978 Tf 5.756 0 Td[(1] 2N+Nr22N)]TJ/F18 5.978 Tf 5.757 0 Td[(18iorqi=rb0 2+Nr28i. A.11 ProofofCorollary3.7Corollary3.7: SupposemoftheNsupplierswithm
PAGE 117

105 Proof :TheprooffollowsdirectlyfromEquation A.1 .Itmayalsobethecasewherethemhomogeneoussuppliersdonotreceiveapositiveorderquantityi.e.,qi=0.

PAGE 118

APPENDIXBPROOFSFORCHAPTER4 B.1 ProofofTheorem4.1Theorem4.1 :Underconstantsupplierprices,theoptimalsourcingpolicyforthermis: Indexsuppliersinnon-decreasingorderofpricesi.e.,c1c2:::cn. Ify1Q,thensourcethecompleterequirementQfromsupplier,i.e.,qi=Qandqj=08j=2;:::;n. Ify11,suchthatcjc1.InthiscaseourtotalcostofnotsinglesourcingfromSupplier1is:TC2=c1Q)]TJ/F22 11.955 Tf 11.956 0 Td[(+cj=c1Q)]TJ/F22 11.955 Tf 11.955 0 Td[(c1+cj=c1Q+cj)]TJ/F22 11.955 Tf 11.955 0 Td[(c1Andsincecjc1,TC2)]TJ/F22 11.955 Tf 11.955 0 Td[(TC10.Thisconcludeourproof. 106

PAGE 119

107 B.2 ProofofResult4.1Result4.1 :Thereexistsatleastoneoptimalsolutiontooursourcingmodelsuchthatqi=0orqi=yiforallisuppliersexceptthattheremaybeatmostonesupplierjforwhich0
PAGE 120

108 Obtainanewfeasiblesolution,S2,byreplacingq1iwithq2iandq1jwithq2j.Adding B.2 and B.3 :q2i+q2j=q1i+q1jThistogetherwiththedenitionof,insurethatS2isfeasible.Now,sincerelation B.1 holdsforallfeasiblesolutionsweknowthat:@fi @qiqi@fj @qjqjB.4andforanyqiq1iandqjq1j,weknowthat:fiqi)]TJ/F22 11.955 Tf 11.955 0 Td[(fiq1ifjq1j)]TJ/F22 11.955 Tf 11.955 0 Td[(fjqj.Sofiq2i)]TJ/F22 11.955 Tf 11.955 0 Td[(fiq1ifjq1j)]TJ/F22 11.955 Tf 11.955 0 Td[(fjq2j,andtherefore,Pni=1fiq2iPni=1fiq1i.RepeatingthisprocesswillleadtoasolutionS2thatveriesResult3.1'sconditions.

PAGE 121

109 B.3 LinearDiscountPricingTestProblemsData TableB{1:LinearDiscountPricingTestProblemsData-12 1 2 3 4 Supplier y a b y a b y a b y a b 1 945 71 0.02 164 121 0.55 12 15 0.33 138 108 0.54 2 91 109 0.71 83 57 0.49 126 56 0.26 130 27 0.11 3 250 165 0.5 860 198 0.07 296 116 0.05 201 197 0.64 4 35 21 0.47 21 56 0.83 45 83 0.79 72 162 0.68 5 367 142 0.13 219 116 0.02 468 88 0.06 746 170 0.14 6 777 175 0.16 249 108 0.38 7 22 0.13 18 17 0.84 7 978 165 0.01 42 30 0.33 74 189 0.2 193 184 0.7 8 50 171 0.13 74 70 0.58 766 95 0.05 981 87 0.05 9 295 91 0.29 762 17 0.02 80 176 0.97 77 154 0.81 10 35 125 0.38 12 8 0.4 143 103 0.54 128 81 0.57 5 6 7 8 Supplier y a b y a b y a b y a b 1 968 185 0.08 78 166 0.99 190 67 0.19 139 125 0.6 2 674 195 0.23 240 47 0.09 956 158 0.13 58 177 0.81 3 89 195 0.42 468 169 0.29 4 7 0.68 597 180 0.29 4 15 16 0.43 810 138 0.03 49 36 0.09 129 156 0.4 5 156 120 0.68 9 21 0.69 169 99 0.53 57 84 0.54 6 239 38 0.1 17 75 0.76 46 125 0.86 87 147 0.35 7 74 60 0.71 179 196 0.75 479 167 0.33 901 133 0.08 8 70 62 0.71 28 102 0.92 439 137 0.15 259 27 0.02 9 29 149 0.98 290 189 0.15 278 159 0.35 25 180 0.41 10 62 103 0.92 133 95 0.67 91 135 0.69 264 199 0.37 9 10 11 12 Supplier y a b y a b y a b y a b 1 139 125 0.6 120 110 0.27 139 125 0.6 422 75 0.17 2 58 177 0.81 169 181 0.85 58 177 0.81 26 36 0.39 3 597 180 0.29 70 56 0.77 597 180 0.29 268 106 0.23 4 129 156 0.4 125 109 0.36 129 156 0.4 124 99 0.59 5 57 84 0.54 93 93 0.78 57 84 0.54 274 136 0.33 6 87 147 0.35 525 90 0.13 87 147 0.35 817 53 0.05 7 901 133 0.08 397 24 0.02 901 133 0.08 22 103 0.7 8 259 27 0.02 83 180 0.9 259 27 0.02 53 73 0.87 9 25 180 0.41 855 177 0.04 25 180 0.41 3 65 0.45 10 264 199 0.37 47 178 0.25 264 199 0.37 35 35 0.01

PAGE 122

110 TableB{2:LinearDiscountPricingTestProblemsData3-24 13 14 15 16 Supplier y a b y a b y a b y a b 1 140 153 0.96 542 110 0.09 6 4 0.54 251 150 0.55 2 8 4 0.46 574 175 0.06 65 167 0.68 8 4 0.48 3 32 160 0.56 90 152 0.76 157 165 0.6 763 158 0.11 4 24 162 0.65 130 140 0.92 612 199 0.06 134 21 0.12 5 247 172 0.17 0 10 0.99 42 69 0.96 67 143 0.44 6 543 190 0.19 6 31 0.09 53 144 0.93 15 66 0.73 7 577 185 0.28 62 83 0.37 241 132 0.35 160 135 0.2 8 520 171 0.23 308 96 0.18 476 182 0.06 151 177 0.55 9 68 44 0.61 166 62 0.09 150 184 0.99 556 188 0.07 10 37 137 0.85 704 158 0.22 545 91 0.13 12 33 0.52 17 18 19 20 Supplier y a b y a b y a b y a b 1 32 98 0.64 168 79 0.36 17 4 0.09 252 193 0.21 2 329 84 0.14 732 135 0.09 179 127 0.37 54 21 0.39 3 382 98 0.01 12 17 0.24 703 182 0.24 39 102 0.74 4 73 65 0.15 129 177 0.82 31 127 0.81 407 152 0.14 5 754 186 0.17 22 115 0.43 321 152 0.21 185 178 0.3 6 28 43 0.65 88 13 0.11 296 155 0.4 252 40 0.07 7 287 86 0.11 907 105 0.06 41 20 0.01 15 18 0.93 8 10 93 0.93 258 189 0.62 428 197 0.27 160 71 0.03 9 106 79 0.3 197 98 0.32 45 155 0.68 652 133 0.06 10 95 157 0.17 28 53 0.75 7 5 0.06 12 48 0.82 21 22 23 24 Supplier y a b y a b y a b y a b 1 78 57 0.25 831 190 0.07 249 188 0.33 578 170 0.21 2 233 176 0.6 93 26 0.23 10 28 0.67 940 189 0.06 3 57 83 0.79 873 141 0.15 112 97 0.16 121 145 0.71 4 64 67 0.61 206 67 0.3 159 182 0.94 250 172 0.39 5 2 173 0.69 120 189 0.05 204 174 0.85 162 172 0.72 6 25 54 0.17 235 94 0.35 95 119 0.57 291 141 0.45 7 38 174 0.85 93 112 0.74 985 139 0.11 207 119 0.17 8 47 100 0.97 957 160 0.13 36 107 0.99 134 129 0.56 9 486 41 0.01 3 2 0.62 67 144 0.95 21 161 0.44 10 998 141 0.14 94 27 0.22 130 111 0.45 219 90 0.11

PAGE 123

111 TableB{3:LinearDiscountPricingTestProblemsData5-30 25 26 27 28 Supplier y a b y a b y a b y a b 1 379 142 0.21 34 73 0.9 60 155 0.75 196 135 0.43 2 165 156 0.8 350 67 0.08 163 176 0.77 16 28 0.4 3 697 116 0.09 600 161 0.22 125 94 0.72 171 68 0.16 4 159 91 0.28 40 91 0.82 283 160 0.29 176 82 0.38 5 637 121 0.18 161 92 0.44 910 199 0.14 809 154 0.06 6 132 105 0.69 40 11 0.12 303 100 0.33 306 147 0.46 7 47 29 0.15 63 152 0.77 120 34 0.15 139 141 0.97 8 28 153 0.91 652 179 0.18 241 149 0.06 106 56 0.24 9 18 20 0.55 170 141 0.65 483 137 0.11 95 121 0.02 10 77 68 0.58 14 47 0.61 966 152 0.14 23 68 0.71 29 30 Supplier y a b y a b 1 18 193 0.45 185 73 0.35 2 19 60 0.99 45 50 0.84 3 200 49 0.01 471 34 0.01 4 55 138 0.59 841 181 0.01 5 154 78 0.35 10 92 0.89 6 36 86 0.94 112 140 0.14 7 373 91 0.13 10 24 0.84 8 373 108 0.26 112 165 0.82 9 837 72 0.06 150 166 0.19 10 93 176 0.81 133 41 0.2

PAGE 124

112 B.4 IncrementalandAll-UnitDiscountPricingTestProblemsData TableB{4:IncrementalandAll-UnitsPricingTestProblemsData-3 Dataset1 Dataset2 Dataset3 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 65 1.66 0 0 96 1.96 0 0 58 1.59 0 0 47 1.48 0 97 167 1.91 188.16 59 101 1.54 92.22 48 83 1.43 69.56 0 12 1.12 0 102 141 1.49 158.44 84 87 1.38 121.04 13 22 1.07 13.44 0 80 1.81 0 88 148 1.33 126.56 23 80 1.02 24.14 81 139 1.76 144.8 149 190 1.28 207.69 81 165 0.97 83.3 140 154 1.71 248.64 191 202 1.23 261.45 166 214 0.92 165.75 155 210 1.66 274.29 0 92 1.93 0 215 244 0.87 210.83 211 262 1.61 367.25 93 160 1.88 177.56 245 255 0.82 236.93 0 86 1.86 0 161 230 1.83 305.4 256 339 0.77 245.95 87 150 1.81 159.96 231 327 1.78 433.5 0 36 1.37 0 151 238 1.76 275.8 328 383 1.73 606.16 37 63 1.32 49.32 239 243 1.71 430.68 384 440 1.68 703.04 64 123 1.27 84.96 0 79 1.79 0 441 520 1.63 798.8 124 150 1.22 161.16 80 137 1.74 141.41 521 590 1.58 929.2 151 199 1.17 194.1 138 236 1.69 242.33 591 682 1.53 1039.8 0 39 1.39 0 237 323 1.64 409.64 0 82 1.83 0 40 68 1.34 54.21 324 412 1.59 552.32 83 143 1.78 150.06 69 163 1.29 93.07 413 441 1.54 693.83 144 188 1.73 258.64 164 183 1.24 215.62 0 85 1.86 0 189 252 1.68 336.49 184 224 1.19 240.42 86 148 1.81 158.1 253 339 1.63 444.01 225 317 1.14 289.21 149 228 1.76 272.13

PAGE 125

113 TableB-4.Continued Dataset1 Dataset2 Dataset3 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 340 422 1.58 585.82 318 413 1.09 395.23 229 299 1.71 412.93 423 439 1.53 716.96 0 50 1.5 0 0 7 1.07 0 0 75 1.76 0 51 87 1.45 75 0 5 1.05 0 76 131 1.71 132 88 120 1.4 128.65 6 9 1 5.25 132 184 1.66 227.76 0 61 1.62 0 10 76 0.95 9.25 185 187 1.61 315.74 62 107 1.57 98.82 77 127 0.9 72.9 188 230 1.56 320.57 108 184 1.52 171.04 128 202 0.85 118.8 231 255 1.51 387.65 185 200 1.47 288.08 203 277 0.8 182.55 256 275 1.46 425.4 201 269 1.42 311.6 278 355 0.75 242.55 0 38 1.39 0 270 344 1.37 409.58 0 33 1.33 0 39 67 1.34 52.82 345 424 1.32 512.33 34 58 1.28 43.89 68 157 1.29 91.68 425 489 1.27 617.93 59 76 1.23 75.89 158 191 1.24 207.78 490 499 1.22 700.48 77 78 1.18 98.03 192 264 1.19 249.94 0 71 1.72 0 0 33 1.33 0 265 305 1.14 336.81 72 124 1.67 122.12 34 58 1.28 43.89 0 1 1.02 0 125 129 1.62 210.63 59 75 1.23 75.89 2 3 0.97 1.02 130 215 1.57 218.73 76 153 1.18 96.8 4 27 0.92 2.96 216 291 1.52 353.75 154 195 1.13 188.84 28 35 0.87 25.04 292 297 1.47 469.27 0 8 1.08 0 36 73 0.82 32 298 389 1.42 478.09 9 15 1.03 8.64 0 45 1.45 0 390 465 1.37 608.73 16 21 0.98 15.85 46 79 1.4 65.25 0 99 1.99 0 22 24 0.93 21.73 80 148 1.35 112.85 0 72 1.73 0 25 63 0.88 24.52 149 228 1.3 206 0 53 1.53 0

PAGE 126

114 TableB-4.Continued Dataset1 Dataset2 Dataset3 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 69 1.7 0 70 120 1.65 117.3 121 198 1.6 201.45 199 227 1.55 326.25 228 309 1.5 371.2 310 318 1.45 494.2 319 371 1.4 507.25 0 1 1.01 0 2 3 0.96 1.01 4 16 0.91 2.93 17 24 0.86 14.76 25 56 0.81 21.64

PAGE 127

115 TableB{5:IncrementalandAll-UnitsPricingTestProblemsData-6 Dataset4 Dataset5 Dataset6 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 14 1.14 0 0 13 1.14 0 0 66 1.66 0 15 25 1.09 15.96 14 24 1.09 14.82 67 115 1.61 109.56 26 103 1.04 27.95 25 97 1.04 26.81 116 148 1.56 188.45 104 142 0.99 109.07 98 143 0.99 102.73 149 216 1.51 239.93 143 201 0.94 147.68 144 231 0.94 148.27 217 274 1.46 342.61 202 209 0.89 203.14 232 233 0.89 230.99 275 359 1.41 427.29 210 242 0.84 210.26 234 258 0.84 232.77 360 422 1.36 547.14 243 302 0.79 237.98 259 274 0.79 253.77 423 496 1.31 632.82 303 330 0.74 285.38 0 73 1.74 0 0 61 1.61 0 0 78 1.78 0 74 128 1.69 127.02 62 106 1.56 98.21 79 136 1.73 138.84 129 159 1.64 219.97 107 174 1.51 168.41 137 218 1.68 239.18 160 191 1.59 270.81 175 242 1.46 271.09 219 229 1.63 376.94 0 34 1.35 0 243 298 1.41 370.37 230 307 1.58 394.87 35 60 1.3 45.9 299 361 1.36 449.33 308 340 1.53 518.11 0 39 1.39 0 362 441 1.31 535.01 0 55 1.55 0 40 68 1.34 54.21 0 71 1.71 0 56 96 1.5 85.25 69 161 1.29 93.07 72 123 1.66 121.41 97 194 1.45 146.75 162 245 1.24 213.04 124 218 1.61 207.73 0 75 1.75 0 246 277 1.19 317.2 219 240 1.56 360.68 76 131 1.7 131.25 278 317 1.14 355.28 241 311 1.51 395 0 46 1.47 0 318 404 1.09 400.88 312 328 1.46 502.21 47 81 1.42 67.62 0 86 1.87 0 329 414 1.41 527.03

PAGE 128

116 TableB-5.Continued Dataset4 Dataset5 Dataset6 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 82 172 1.37 117.32 87 150 1.82 160.82 415 482 1.36 648.29 173 238 1.32 241.99 151 241 1.77 277.3 483 551 1.31 740.77 239 268 1.27 329.11 242 297 1.72 438.37 0 83 1.84 0 269 282 1.22 367.21 298 359 1.67 534.69 84 145 1.79 152.72 0 27 1.28 0 0 52 1.53 0 0 51 1.52 0 28 48 1.23 34.56 53 91 1.48 79.56 0 37 1.38 0 49 100 1.18 60.39 0 64 1.65 0 0 27 1.28 0 0 93 1.94 0 65 112 1.6 105.6 28 48 1.23 34.56 94 162 1.89 180.42 113 130 1.55 182.4 49 97 1.18 60.39 163 244 1.84 310.83 131 138 1.5 210.3 98 139 1.13 118.21 245 340 1.79 461.71 139 171 1.45 222.3 0 10 1.1 0 341 390 1.74 633.55 172 223 1.4 270.15 11 18 1.05 11 0 57 1.58 0 224 308 1.35 342.95 19 47 1 19.4 58 100 1.53 90.06 309 355 1.3 457.7 48 138 0.95 48.4 101 126 1.48 155.85 0 95 1.96 0 139 204 0.9 134.85 127 173 1.43 194.33 0 70 1.7 0 205 226 0.85 194.25 174 271 1.38 261.54 71 122 1.65 119 0 64 1.65 0 272 348 1.33 396.78 123 204 1.6 204.8 65 112 1.6 105.6 0 21 1.22 0 205 214 1.55 336 113 129 1.55 182.4 0 15 1.16 0 215 285 1.5 351.5 130 205 1.5 208.75 16 27 1.11 17.4 286 292 1.45 458 0 3 1.03 0 28 29 1.06 30.72 293 304 1.4 468.15 4 6 0.98 3.09 30 52 1.01 32.84 0 1 1.01 0 7 45 0.93 6.03 53 145 0.96 56.07 2 3 0.96 1.01 46 102 0.88 42.3

PAGE 129

117 TableB-5.Continued Dataset4 Dataset5 Dataset6 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 146 239 0.91 145.35 4 14 0.91 2.93 103 170 0.83 92.46 15 100 0.86 12.94 171 236 0.78 148.9 101 175 0.81 86.9 237 256 0.73 200.38 176 249 0.76 147.65 257 299 0.68 214.98 250 323 0.71 203.89 300 329 0.63 244.22 324 383 0.66 256.43 384 404 0.61 296.03

PAGE 130

118 TableB{6:IncrementalandAll-UnitsPricingTestProblemsData-9 Dataset7 Dataset8 Dataset9 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 52 1.53 0 0 81 1.81 0 0 40 1.4 0 53 91 1.48 79.56 82 141 1.76 146.61 41 70 1.35 56 92 155 1.43 137.28 142 167 1.71 252.21 71 77 1.3 96.5 0 99 1.99 0 168 199 1.66 296.67 78 91 1.25 105.6 100 172 1.94 197.01 200 245 1.61 349.79 92 121 1.2 123.1 173 225 1.89 338.63 246 321 1.56 423.85 122 123 1.15 159.1 226 325 1.84 438.8 322 419 1.51 542.41 124 146 1.1 161.4 0 5 1.06 0 0 75 1.76 0 147 237 1.05 186.7 6 10 1.01 5.3 0 55 1.55 0 0 56 1.56 0 11 80 0.96 10.35 56 96 1.5 85.25 57 98 1.51 87.36 0 94 1.94 0 97 192 1.45 146.75 99 105 1.46 150.78 0 68 1.69 0 193 234 1.4 285.95 106 113 1.41 161 69 119 1.64 114.92 235 245 1.35 344.75 114 143 1.36 172.28 120 185 1.59 198.56 246 322 1.3 359.6 144 158 1.31 213.08 186 217 1.54 303.5 323 330 1.25 459.7 159 203 1.26 232.73 218 262 1.49 352.78 0 37 1.37 0 204 268 1.21 289.43 0 72 1.72 0 0 27 1.27 0 0 2 1.03 0 73 125 1.67 123.84 28 48 1.22 34.29 3 4 0.98 2.06 126 137 1.62 212.35 49 93 1.17 59.91 5 36 0.93 4.02 138 235 1.57 231.79 94 155 1.12 112.56 0 41 1.42 0 0 79 1.79 0 156 212 1.07 182 42 72 1.37 58.22 80 137 1.74 141.41 213 285 1.02 242.99 73 98 1.32 100.69

PAGE 131

119 TableB-6.Continued Dataset7 Dataset8 Dataset9 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 138 235 1.69 242.33 0 46 1.47 0 99 145 1.27 135.01 0 82 1.83 0 47 81 1.42 67.62 146 239 1.22 194.7 83 143 1.78 150.06 82 170 1.37 117.32 240 249 1.17 309.38 144 182 1.73 258.64 171 197 1.32 239.25 250 315 1.12 321.08 183 252 1.68 326.11 198 251 1.27 274.89 316 335 1.07 395 253 345 1.63 443.71 252 274 1.22 343.47 336 377 1.02 416.4 346 432 1.58 595.3 0 81 1.82 0 0 14 1.15 0 433 523 1.53 732.76 82 141 1.77 147.42 15 25 1.1 16.1 524 577 1.48 871.99 142 171 1.72 253.62 26 114 1.05 28.2 0 21 1.22 0 172 177 1.67 305.22 115 130 1 121.65 22 37 1.17 25.62 0 95 1.96 0 131 205 0.95 137.65 0 73 1.74 0 0 69 1.7 0 0 86 1.86 0 70 121 1.65 117.3 0 62 1.63 0 122 201 1.6 203.1 63 109 1.58 101.06 202 275 1.55 331.1 110 202 1.53 175.32 276 347 1.5 445.8 203 296 1.48 317.61 348 385 1.45 553.8 297 311 1.43 456.73 386 427 1.4 608.9 0 45 1.46 0 0 16 1.16 0 46 79 1.41 65.7 80 157 1.36 113.64 158 191 1.31 219.72 192 265 1.26 264.26 0 70 1.71 0 71 122 1.66 113.64

PAGE 132

120 TableB-6.Continued Dataset7 Dataset8 Dataset9 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 123 214 1.61 185.02 0 77 1.78 0 78 134 1.73 137.06 135 211 1.68 235.67 212 232 1.63 365.03 233 285 1.58 399.26 286 381 1.53 483 382 420 1.48 629.88 421 487 1.43 687.60 488 527 1.38 783.41

PAGE 133

121 TableB{7:IncrementalandAll-UnitsPricingTestProblemsData0-12 Dataset10 Dataset11 Dataset12 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 84 1.85 0 0 21 1.22 0 0 30 1.3 0 85 146 1.8 155.4 22 37 1.17 25.62 31 52 1.25 39 147 213 1.75 267 38 111 1.12 44.34 53 131 1.2 66.5 214 255 1.7 384.25 112 174 1.07 127.22 132 183 1.15 161.3 256 275 1.65 455.65 175 255 1.02 194.63 184 263 1.1 221.1 0 41 1.41 0 256 344 0.97 277.25 0 67 1.67 0 42 72 1.36 57.81 345 374 0.92 363.58 68 117 1.62 111.89 73 93 1.31 99.97 0 5 1.06 0 118 165 1.57 192.89 94 140 1.26 127.48 6 10 1.01 5.3 166 184 1.52 268.25 141 148 1.21 186.7 11 84 0.96 10.35 185 208 1.47 297.13 149 182 1.16 196.38 85 142 0.91 81.39 0 11 1.12 0 183 252 1.11 235.82 0 96 1.97 0 0 8 1.08 0 253 254 1.06 313.52 0 70 1.71 0 9 15 1.03 8.64 255 269 1.01 315.64 71 122 1.66 119.7 16 22 0.98 15.85 0 46 1.46 0 0 92 1.93 0 23 36 0.93 22.71 47 80 1.41 67.16 93 160 1.88 177.56 37 70 0.88 35.73 0 83 1.83 0 161 229 1.83 305.4 71 144 0.83 65.65 84 144 1.78 151.89 230 301 1.78 431.67 145 205 0.78 127.07 145 194 1.73 260.47 302 333 1.73 559.83 206 239 0.73 174.65 195 243 1.68 346.97 334 378 1.68 615.19 0 73 1.74 0 0 31 1.32 0 379 444 1.63 690.79 74 127 1.69 127.02 32 55 1.27 40.92 445 475 1.58 798.37 128 153 1.64 218.28

PAGE 134

122 TableB-7.Continued Dataset10 Dataset11 Dataset12 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 56 154 1.22 71.4 0 21 1.22 0 154 197 1.59 260.92 155 250 1.17 192.18 0 15 1.16 0 198 252 1.54 330.88 251 301 1.12 304.5 16 27 1.11 17.4 253 295 1.49 415.58 302 365 1.07 361.62 28 127 1.06 30.72 296 319 1.44 479.65 366 458 1.02 430.1 0 7 1.08 0 320 417 1.39 514.21 459 460 0.97 524.96 8 13 1.03 7.56 0 74 1.74 0 0 1 1.01 0 14 110 0.98 13.74 75 129 1.69 128.76 2 3 0.96 1.01 0 59 1.59 0 130 164 1.64 221.71 4 19 0.91 2.93 0 43 1.43 0 165 166 1.59 279.11 20 78 0.86 17.49 44 75 1.38 61.49 167 171 1.54 282.29 79 80 0.81 68.23 76 122 1.33 105.65 172 255 1.49 289.99 81 113 0.76 69.85 123 215 1.28 168.16 256 293 1.44 415.15 114 179 0.71 94.93 216 314 1.23 287.2 0 47 1.48 0 0 34 1.34 0 315 405 1.18 408.97 48 82 1.43 69.56 35 60 1.29 45.56 406 457 1.13 516.35 83 84 1.38 119.61 61 93 1.24 79.1 458 537 1.08 575.11 85 97 1.33 122.37 0 54 1.55 0 98 114 1.28 139.66 0 40 1.4 0 115 202 1.23 161.42 41 70 1.35 56 203 301 1.18 269.66 0 4 1.05 0 302 393 1.13 386.48 5 8 1 4.2 394 476 1.08 490.44 9 71 0.95 8.2 0 18 1.19 0 72 146 0.9 68.05 19 32 1.14 21.42 147 224 0.85 135.55 33 70 1.09 37.38

PAGE 135

123 TableB-7.Continued Dataset10 Dataset11 Dataset12 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 225 260 0.8 201.85 71 110 1.04 78.8 261 279 0.75 230.65 111 185 0.99 120.4 186 260 0.94 194.65 261 350 0.89 265.15 351 391 0.84 345.25 392 394 0.79 379.69 0 48 1.49 0 49 84 1.44 71.52 85 101 1.39 123.36 102 178 1.34 146.99 179 200 1.29 250.17 201 275 1.24 278.55 276 354 1.19 371.55 0 52 1.53 0

PAGE 136

124 TableB{8:IncrementalandAll-UnitsPricingTestProblemsData3-15 Dataset13 Dataset14 Dataset15 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 88 1.89 0 0 13 1.14 0 0 57 1.58 0 89 154 1.84 166.32 14 24 1.09 14.82 0 42 1.42 0 155 173 1.79 287.76 25 98 1.04 26.81 43 73 1.37 59.64 174 195 1.74 321.77 99 155 0.99 103.77 74 105 1.32 102.11 196 268 1.69 360.05 156 226 0.94 160.2 106 154 1.27 144.35 269 310 1.64 483.42 227 236 0.89 226.94 155 191 1.22 206.58 0 19 1.19 0 237 298 0.84 235.84 192 221 1.17 251.72 20 34 1.14 22.61 0 64 1.65 0 222 230 1.12 286.82 35 77 1.09 39.71 65 112 1.6 105.6 0 50 1.5 0 78 105 1.04 86.58 0 15 1.16 0 51 87 1.45 75 106 182 0.99 115.7 16 27 1.11 17.4 88 121 1.4 128.65 183 193 0.94 191.93 28 125 1.06 30.72 122 190 1.35 176.25 0 82 1.83 0 126 208 1.01 134.6 191 267 1.3 269.4 83 143 1.78 150.06 209 226 0.96 218.43 268 275 1.25 369.5 144 182 1.73 258.64 0 98 1.98 0 276 310 1.2 379.5 183 250 1.68 326.11 0 71 1.72 0 311 400 1.15 421.5 251 319 1.63 440.35 72 124 1.67 122.12 401 441 1.1 525 320 405 1.58 552.82 125 129 1.62 210.63 0 93 1.93 0 0 78 1.79 0 0 73 1.74 0 94 162 1.88 179.49 79 136 1.74 139.62 74 127 1.69 127.02 163 236 1.83 309.21 137 225 1.69 240.54 128 157 1.64 218.28 237 297 1.78 444.63

PAGE 137

125 TableB-8.Continued Dataset13 Dataset14 Dataset15 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 226 257 1.64 390.95 158 159 1.59 267.48 298 343 1.73 553.21 258 299 1.59 443.43 160 167 1.54 270.66 344 430 1.68 632.79 0 9 1.1 0 168 200 1.49 282.98 431 519 1.63 778.95 10 17 1.05 9.9 201 266 1.44 332.15 0 28 1.29 0 18 39 1 18.3 267 293 1.39 427.19 29 49 1.24 36.12 40 106 0.95 40.3 0 63 1.64 0 50 110 1.19 62.16 0 44 1.44 0 64 110 1.59 103.32 111 154 1.14 134.75 0 32 1.32 0 111 114 1.54 178.05 155 205 1.09 184.91 0 23 1.24 0 115 170 1.49 184.21 206 272 1.04 240.5 24 41 1.19 28.52 171 224 1.44 267.65 273 317 0.99 310.18 42 138 1.14 49.94 225 244 1.39 345.41 318 385 0.94 354.73 0 63 1.63 0 245 273 1.34 373.21 386 453 0.89 418.65 64 110 1.58 102.69 274 372 1.29 412.07 0 58 1.59 0 111 210 1.53 176.95 0 98 1.99 0 59 101 1.54 92.22 211 215 1.48 329.95 99 171 1.94 195.02 102 139 1.49 158.44 216 301 1.43 337.35 172 216 1.89 336.64 140 189 1.44 215.06 302 375 1.38 460.33 217 274 1.84 421.69 0 60 1.6 0 376 440 1.33 562.45 275 361 1.79 528.41 61 105 1.55 96 441 443 1.28 648.9 0 95 1.95 0 106 163 1.5 165.75 0 48 1.48 0 96 165 1.9 185.25 164 260 1.45 252.75 166 263 1.85 318.25 0 64 1.65 0 0 84 1.85 0 65 112 1.6 105.6 85 146 1.8 155.4 113 129 1.55 182.4 147 213 1.75 267 130 220 1.5 208.75

PAGE 138

126 TableB-8.Continued Dataset13 Dataset14 Dataset15 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 214 259 1.7 384.25 221 275 1.45 345.25 260 337 1.65 462.45 0 40 1.4 0 41 70 1.35 56 71 77 1.3 96.5 78 93 1.25 105.6 94 169 1.20 125.6 170 262 1.15 216.8 263 346 1.10 323.75 347 391 1.05 416.15 0 62 1.63 0

PAGE 139

127 TableB{9:IncrementalandAll-UnitsPricingTestProblemsData6-18 Dataset16 Dataset17 Dataset18 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 60 1.6 0 0 46 1.47 0 0 70 1.71 0 61 105 1.55 96 0 34 1.34 0 0 51 1.52 0 106 166 1.5 165.75 35 59 1.29 45.56 52 89 1.47 77.52 167 207 1.45 257.25 60 89 1.24 77.81 90 141 1.42 133.38 208 296 1.4 316.7 90 96 1.19 115.01 142 232 1.37 207.22 297 327 1.35 441.3 97 104 1.14 123.34 233 290 1.32 331.89 328 350 1.3 483.15 105 139 1.09 132.46 0 87 1.87 0 351 431 1.25 513.05 0 96 1.96 0 88 151 1.82 162.69 0 79 1.79 0 97 167 1.91 188.16 152 250 1.77 279.17 80 138 1.74 141.41 168 177 1.86 323.77 251 341 1.72 454.4 139 140 1.69 244.07 178 240 1.81 342.37 0 65 1.65 0 141 151 1.64 247.45 241 313 1.76 456.4 66 113 1.6 107.25 152 241 1.59 265.49 314 354 1.71 584.88 114 136 1.55 184.05 0 35 1.35 0 355 359 1.66 654.99 137 217 1.5 219.7 36 61 1.3 47.25 360 445 1.61 663.29 218 309 1.45 341.2 62 105 1.25 81.05 446 520 1.56 801.75 310 381 1.4 474.6 106 161 1.2 136.05 0 84 1.84 0 382 415 1.35 575.4 162 211 1.15 203.25 85 146 1.79 154.56 0 72 1.73 0 0 48 1.49 0 147 204 1.74 265.54 0 53 1.53 0 49 84 1.44 71.52 0 96 1.96 0 54 93 1.48 81.09 85 100 1.39 123.36 97 167 1.91 188.16 94 165 1.43 140.29 101 161 1.34 145.6 168 177 1.86 323.77 166 194 1.38 243.25

PAGE 140

128 TableB-9.Continued Dataset16 Dataset17 Dataset18 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 162 208 1.29 227.34 178 240 1.81 342.37 195 287 1.33 283.27 209 215 1.24 287.97 0 72 1.72 0 0 97 1.97 0 216 236 1.19 296.65 73 125 1.67 123.84 98 169 1.92 191.09 237 283 1.14 321.64 126 133 1.62 212.35 0 24 1.24 0 284 289 1.09 375.22 134 167 1.57 225.31 0 17 1.18 0 0 90 1.91 0 168 236 1.52 278.69 18 30 1.13 20.06 91 157 1.86 171.9 237 313 1.47 383.57 31 53 1.08 34.75 158 197 1.81 296.52 314 337 1.42 496.76 0 92 1.93 0 198 276 1.76 368.92 338 435 1.37 530.84 93 160 1.88 177.56 277 325 1.71 507.96 436 508 1.32 665.1 161 230 1.83 305.4 326 362 1.66 591.75 0 50 1.5 0 231 323 1.78 433.5 363 383 1.61 653.17 51 87 1.45 75 324 422 1.73 599.04 384 437 1.56 686.98 88 119 1.4 128.65 0 23 1.24 0 120 159 1.35 173.45 24 41 1.19 28.52 160 232 1.3 227.45 42 137 1.14 49.94 0 41 1.41 0 138 186 1.09 159.38 42 72 1.36 57.81 187 227 1.04 212.79 73 89 1.31 99.97 228 231 0.99 255.43 90 170 1.26 122.24 0 61 1.62 0 171 254 1.21 224.3 62 107 1.57 98.82 255 284 1.16 325.94 108 187 1.52 171.04 285 289 1.11 360.74 0 62 1.63 0 0 69 1.69 0 63 108 1.58 101.06 70 120 1.64 116.61

PAGE 141

129 TableB-9.Continued Dataset16 Dataset17 Dataset18 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 109 200 1.53 173.74 121 194 1.59 200.25 201 269 1.48 314.5 0 68 1.68 0 270 348 1.43 416.62 349 408 1.38 529.59 0 19 1.2 0 20 34 1.15 22.8 35 80 1.1 40.05 81 166 1.05 90.65 167 239 1 180.95 240 289 0.95 253.95 290 335 0.9 301.45 336 414 0.85 342.85 415 462 0.8 410 0 22 1.23 0 23 39 1.18 27.06 40 125 1.13 47.12 126 196 1.08 144.3 197 209 1.03 220.98 210 228 0.98 234.37 229 254 0.93 252.99

PAGE 142

130 TableB{10:IncrementalandAll-UnitsPricingTestProblemsData9-21 Dataset19 Dataset20 Dataset21 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 43 1.43 0 0 36 1.37 0 0 67 1.67 0 44 75 1.38 61.49 37 64 1.32 49.32 68 117 1.62 111.89 76 121 1.33 105.65 65 129 1.27 86.28 118 165 1.57 192.89 0 85 1.86 0 130 145 1.22 168.83 166 181 1.52 268.25 86 148 1.81 158.1 146 220 1.17 188.35 182 248 1.47 292.57 149 226 1.76 272.13 221 297 1.12 276.1 249 296 1.42 391.06 227 252 1.71 409.41 0 12 1.12 0 297 311 1.37 459.22 253 292 1.66 453.87 13 22 1.07 13.44 312 353 1.32 479.77 293 378 1.61 520.27 23 78 1.02 24.14 354 375 1.27 535.21 379 447 1.56 658.73 79 135 0.97 81.26 0 67 1.67 0 0 82 1.83 0 136 214 0.92 136.55 68 117 1.62 111.89 0 60 1.6 0 215 262 0.87 209.23 118 161 1.57 192.89 61 105 1.55 96 263 277 0.82 250.99 162 209 1.52 261.97 106 166 1.5 165.75 278 329 0.77 263.29 210 229 1.47 334.93 0 39 1.39 0 0 79 1.8 0 230 262 1.42 364.33 40 68 1.34 54.21 80 138 1.75 142.2 263 324 1.37 411.19 69 164 1.29 93.07 139 142 1.7 245.45 325 380 1.32 496.13 165 210 1.24 216.91 143 201 1.65 252.25 381 438 1.27 570.05 211 297 1.19 273.95 202 215 1.6 349.6 0 96 1.97 0 298 392 1.14 377.48 0 38 1.39 0 97 167 1.92 189.12 393 420 1.09 485.78 39 67 1.34 52.82 168 182 1.87 325.44 0 67 1.68 0 68 154 1.29 91.68 183 237 1.82 353.49

PAGE 143

131 TableB-10.Continued Dataset19 Dataset20 Dataset21 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 49 1.49 0 155 237 1.24 203.91 238 281 1.77 453.59 50 86 1.44 73.01 238 260 1.19 306.83 282 329 1.72 531.47 87 109 1.39 126.29 261 344 1.14 334.2 330 343 1.67 614.03 110 193 1.34 158.26 0 36 1.37 0 344 384 1.62 637.41 194 228 1.29 270.82 37 64 1.32 49.32 385 386 1.57 703.83 229 314 1.24 315.97 65 130 1.27 86.28 0 22 1.23 0 315 389 1.19 422.61 131 156 1.22 170.1 0 16 1.17 0 0 81 1.81 0 157 191 1.17 201.82 0 12 1.12 0 82 141 1.76 146.61 192 286 1.12 242.77 13 21 1.07 13.44 142 163 1.71 252.21 287 320 1.07 349.17 22 76 1.02 23.07 164 231 1.66 289.83 0 73 1.74 0 77 111 0.97 79.17 232 298 1.61 402.71 74 128 1.69 127.02 112 200 0.92 113.12 299 339 1.56 510.58 129 158 1.64 219.97 201 230 0.87 195 0 91 1.91 0 0 7 1.07 0 231 246 0.82 221.1 92 158 1.86 173.81 8 13 1.02 7.49 0 66 1.66 0 159 206 1.81 298.43 14 103 0.97 13.61 67 115 1.61 109.56 207 225 1.76 385.31 104 142 0.92 100.91 116 151 1.56 188.45 226 249 1.71 418.75 0 56 1.57 0 152 161 1.51 244.61 0 8 1.09 0 57 98 1.52 87.92 162 219 1.46 259.71 99 114 1.47 151.76 220 317 1.41 344.39 115 174 1.42 175.28 318 401 1.36 482.57 175 206 1.37 260.48 402 444 1.31 596.81 207 249 1.32 304.32 445 471 1.26 653.14 0 35 1.35 0 0 62 1.62 0

PAGE 144

132 TableB-10.Continued Dataset19 Dataset20 Dataset21 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 36 61 1.3 47.25 0 45 1.46 0 62 104 1.25 81.05 0 33 1.33 0 105 138 1.2 134.8 34 58 1.28 43.89 0 83 1.83 0 59 78 1.23 75.89 84 144 1.78 151.89 79 116 1.18 100.49 145 193 1.73 260.47 194 221 1.68 345.24 222 298 1.63 392.28 299 310 1.58 517.79

PAGE 145

133 TableB{11:IncrementalandAll-UnitsPricingTestProblemsData2-24 Dataset22 Dataset23 Dataset24 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 69 1.7 0 0 91 1.91 0 0 12 1.13 0 70 121 1.65 117.3 92 158 1.86 173.81 13 22 1.08 13.56 122 201 1.6 203.1 159 208 1.81 298.43 23 84 1.03 24.36 202 264 1.55 331.1 209 253 1.76 388.93 85 148 0.98 88.22 265 342 1.5 428.75 254 324 1.71 468.13 149 239 0.93 150.94 343 377 1.45 545.75 325 336 1.66 589.54 240 298 0.88 235.57 378 379 1.4 596.5 337 432 1.61 609.46 299 397 0.83 287.49 380 395 1.35 599.3 433 483 1.56 764.02 398 492 0.78 369.66 0 63 1.63 0 484 553 1.51 843.58 0 25 1.26 0 64 110 1.58 102.69 0 96 1.97 0 26 44 1.21 31.5 111 207 1.53 176.95 0 70 1.71 0 45 71 1.16 54.49 208 271 1.48 325.36 71 122 1.66 119.7 72 125 1.11 85.81 272 356 1.43 420.08 123 214 1.61 206.02 126 153 1.06 145.75 357 406 1.38 541.63 0 70 1.7 0 154 225 1.01 175.43 407 465 1.33 610.63 71 122 1.65 119 226 258 0.96 248.15 466 479 1.28 689.1 123 207 1.6 204.8 259 321 0.91 279.83 480 518 1.23 707.02 208 265 1.55 340.8 0 69 1.69 0 0 58 1.59 0 266 360 1.5 430.7 0 50 1.51 0 59 101 1.54 92.22 361 392 1.45 573.2 51 88 1.46 75.5 102 140 1.49 158.44 393 431 1.4 619.6 89 126 1.41 130.98 141 202 1.44 216.55 0 60 1.6 0 127 165 1.36 184.56 203 258 1.39 305.83 0 44 1.44 0 166 235 1.31 237.6

PAGE 146

134 TableB-11.Continued Dataset22 Dataset23 Dataset24 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 259 309 1.34 383.67 0 32 1.32 0 236 327 1.26 329.3 310 380 1.29 452.01 0 23 1.24 0 328 406 1.21 445.22 381 386 1.24 543.6 24 41 1.19 28.52 407 460 1.16 540.81 0 88 1.88 0 42 139 1.14 49.94 461 475 1.11 603.45 89 153 1.83 165.44 140 214 1.09 161.66 0 54 1.55 0 154 162 1.78 284.39 215 302 1.04 243.41 55 95 1.5 83.7 163 202 1.73 300.41 303 308 0.99 334.93 96 187 1.45 145.2 203 289 1.68 369.61 0 92 1.93 0 188 256 1.4 278.6 290 383 1.63 515.77 93 160 1.88 177.56 0 83 1.83 0 384 392 1.58 668.99 161 225 1.83 305.4 0 60 1.61 0 393 440 1.53 683.21 226 236 1.78 424.35 61 105 1.56 96.6 441 457 1.48 756.65 0 78 1.78 0 0 64 1.64 0 0 80 1.81 0 79 136 1.73 138.84 65 112 1.59 104.96 81 140 1.76 144.8 113 121 1.54 181.28 141 160 1.71 250.4 122 175 1.49 195.14 161 194 1.66 284.6 176 203 1.44 275.6 195 265 1.61 341.04 204 283 1.39 315.92 0 14 1.14 0 284 356 1.34 427.12 15 25 1.09 15.96 357 403 1.29 524.94 26 102 1.04 27.95 0 94 1.94 0 103 127 0.99 108.03 95 163 1.89 182.36 128 145 0.94 132.78 164 247 1.84 312.77 146 149 0.89 149.7 248 283 1.79 467.33 150 207 0.84 153.26 0 4 1.05 0

PAGE 147

135 TableB-11.Continued Dataset22 Dataset23 Dataset24 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 80 1.81 0 5 8 1 4.2 0 59 1.59 0 0 43 1.43 0 44 75 1.38 61.49 76 120 1.33 105.65 121 193 1.28 165.5 0 43 1.43 0 44 75 1.38 61.49 76 122 1.33 105.65 123 129 1.28 168.16 130 148 1.23 177.12 149 165 1.18 200.49

PAGE 148

136 TableB{12:IncrementalandAll-UnitsPricingTestProblemsData5-27 Dataset25 Dataset26 Dataset27 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 47 1.48 0 0 69 1.7 0 0 49 1.5 0 48 82 1.43 69.56 70 121 1.65 117.3 50 86 1.45 73.5 83 84 1.38 119.61 0 78 1.79 0 87 113 1.4 127.15 85 92 1.33 122.37 79 136 1.74 139.62 114 169 1.35 164.95 93 129 1.28 133.01 137 231 1.69 240.54 0 49 1.5 0 130 152 1.23 180.37 232 253 1.64 401.09 50 86 1.45 73.5 0 79 1.8 0 254 331 1.59 437.17 87 115 1.4 127.15 0 58 1.58 0 332 373 1.54 561.19 116 203 1.35 167.75 59 101 1.53 91.64 374 391 1.49 625.87 204 303 1.3 286.55 102 137 1.48 157.43 392 396 1.44 652.69 304 319 1.25 416.55 0 18 1.18 0 0 79 1.8 0 0 72 1.72 0 19 32 1.13 21.24 0 58 1.58 0 73 125 1.67 123.84 33 60 1.08 37.06 0 42 1.43 0 126 136 1.62 212.35 61 127 1.03 67.3 43 74 1.38 60.06 137 216 1.57 230.17 128 165 0.98 136.31 75 110 1.33 104.22 217 287 1.52 355.77 166 216 0.93 173.55 111 120 1.28 152.1 0 12 1.12 0 217 281 0.88 220.98 121 190 1.23 164.9 13 22 1.07 13.44 282 284 0.83 278.18 191 285 1.18 251 23 78 1.02 24.14 0 35 1.36 0 286 318 1.13 363.1 79 129 0.97 81.26 36 62 1.31 47.6 0 64 1.64 0 130 205 0.92 130.73 63 112 1.26 82.97 65 111 1.59 104.96 206 210 0.87 200.65 113 164 1.21 145.97 112 120 1.54 179.69 211 284 0.82 205

PAGE 149

137 TableB-12.Continued Dataset25 Dataset26 Dataset27 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 165 242 1.16 208.89 121 162 1.49 193.55 0 70 1.71 0 243 283 1.11 299.37 163 184 1.44 256.13 0 51 1.52 0 0 92 1.93 0 0 70 1.7 0 52 89 1.47 77.52 93 160 1.88 177.56 71 122 1.65 119 90 140 1.42 133.38 161 229 1.83 305.4 123 209 1.6 204.8 141 210 1.37 205.8 230 307 1.78 431.67 210 295 1.55 344 211 213 1.32 301.7 308 342 1.73 570.51 296 360 1.5 477.3 214 258 1.27 305.66 343 441 1.68 631.06 361 362 1.45 574.8 259 329 1.22 362.81 442 443 1.63 797.38 363 391 1.4 577.7 330 349 1.17 449.43 444 457 1.58 800.64 392 486 1.35 618.3 0 28 1.28 0 458 496 1.53 822.76 487 516 1.3 746.55 29 49 1.23 35.84 0 56 1.56 0 0 1 1.01 0 50 107 1.18 61.67 57 98 1.51 87.36 2 3 0.96 1.01 0 87 1.87 0 99 109 1.46 150.78 4 12 0.91 2.93 0 63 1.64 0 0 84 1.84 0 13 56 0.86 11.12 64 110 1.59 103.32 85 146 1.79 154.56 57 98 0.81 48.96 111 112 1.54 178.05 147 205 1.74 265.54 99 108 0.76 82.98 113 133 1.49 181.13 206 212 1.69 368.2 109 166 0.71 90.58 134 180 1.44 212.42 0 4 1.04 0 0 94 1.94 0 181 278 1.39 280.1 5 8 0.99 4.16 95 164 1.89 182.36 279 365 1.34 416.32 9 64 0.94 8.12 0 89 1.89 0 0 82 1.83 0 65 118 0.89 60.76 90 155 1.84 168.21 119 144 0.84 108.82 156 179 1.79 289.65 0 32 1.32 0 180 188 1.74 332.61

PAGE 150

138 TableB-12.Continued Dataset25 Dataset26 Dataset27 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 33 56 1.27 42.24 189 240 1.69 348.27 57 61 1.22 72.72 241 329 1.64 436.15 62 140 1.17 78.82 330 357 1.59 582.11

PAGE 151

139 TableB{13:IncrementalandAll-UnitsPricingTestProblemsData8-30 Dataset28 Dataset29 Dataset30 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 0 95 1.96 0 0 27 1.27 0 0 53 1.53 0 96 165 1.91 186.2 28 48 1.22 34.29 54 93 1.48 81.09 166 168 1.86 319.9 49 94 1.17 59.91 94 165 1.43 140.29 169 210 1.81 325.48 0 90 1.91 0 166 191 1.38 243.25 211 219 1.76 401.5 0 66 1.66 0 192 238 1.33 279.13 220 262 1.71 417.34 67 115 1.61 109.56 0 7 1.07 0 263 288 1.66 490.87 116 150 1.56 188.45 8 13 1.02 7.49 289 329 1.61 534.03 151 250 1.51 243.05 14 104 0.97 13.61 330 422 1.56 600.04 251 269 1.46 394.05 105 162 0.92 101.88 0 99 1.99 0 270 286 1.41 421.79 163 251 0.87 155.24 100 172 1.94 197.01 0 75 1.75 0 252 282 0.82 232.67 173 221 1.89 338.63 0 55 1.55 0 283 312 0.77 258.09 222 248 1.84 431.24 56 96 1.5 85.25 313 316 0.72 281.19 249 305 1.79 480.92 97 191 1.45 146.75 317 385 0.67 284.07 0 78 1.78 0 192 210 1.4 284.5 0 78 1.78 0 79 136 1.73 138.84 0 20 1.21 0 79 136 1.73 138.84 137 224 1.68 239.18 21 36 1.16 24.2 137 224 1.68 239.18 225 324 1.63 387.02 37 99 1.11 42.76 225 237 1.63 387.02 325 337 1.58 550.02 100 172 1.06 112.69 238 258 1.58 408.21 338 349 1.53 570.56 173 219 1.01 190.07 259 310 1.53 441.39 350 441 1.48 588.92 220 227 0.96 237.54 0 87 1.88 0 442 509 1.43 725.08 228 256 0.91 245.22 88 152 1.83 163.56

PAGE 152

140 TableB-13.Continued Dataset28 Dataset29 Dataset30 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 510 565 1.38 822.32 257 350 0.86 271.61 153 157 1.78 282.51 0 51 1.52 0 0 15 1.16 0 158 238 1.73 291.41 0 37 1.38 0 16 27 1.11 17.4 239 326 1.68 431.54 38 65 1.33 51.06 28 126 1.06 30.72 327 339 1.63 579.38 66 143 1.28 88.3 127 217 1.01 135.66 0 16 1.16 0 144 178 1.23 188.14 0 54 1.54 0 17 28 1.11 18.56 179 180 1.18 231.19 55 94 1.49 83.16 29 33 1.06 31.88 181 189 1.13 233.55 95 180 1.44 142.76 34 112 1.01 37.18 190 238 1.08 243.72 181 249 1.39 266.6 113 168 0.96 116.97 239 278 1.03 296.64 250 335 1.34 362.51 0 46 1.47 0 279 362 0.98 337.84 336 414 1.29 477.75 47 81 1.42 67.62 0 39 1.4 0 415 464 1.24 579.66 82 171 1.37 117.32 40 69 1.35 54.6 0 50 1.5 0 172 205 1.32 240.62 70 72 1.3 95.1 51 87 1.45 75 206 274 1.27 285.5 0 39 1.39 0 88 121 1.4 128.65 0 71 1.72 0 40 68 1.34 54.21 0 69 1.69 0 72 124 1.67 122.12 69 165 1.29 93.07 70 120 1.64 116.61 125 126 1.62 210.63 166 224 1.24 218.2 121 189 1.59 200.25 127 145 1.57 213.87 225 229 1.19 291.36 146 165 1.52 243.7 230 307 1.14 297.31 166 205 1.47 274.1 308 341 1.09 386.23 206 284 1.42 332.9 0 74 1.74 0 285 327 1.37 445.08 75 129 1.69 128.76 0 39 1.4 0 130 162 1.64 221.71 0 28 1.29 0

PAGE 153

141 TableB-13.Continued Dataset28 Dataset29 Dataset30 LB UB Cost Incost LB UB Cost Incost LB UB Cost Incost 163 222 1.59 275.83 29 50 1.24 36.12 223 238 1.54 371.23 51 114 1.19 63.4 239 304 1.49 395.87 0 98 1.98 0 305 335 1.44 494.21 99 170 1.93 194.04 336 361 1.39 538.85 171 206 1.88 333 0 44 1.45 0 207 223 1.83 400.68 45 77 1.4 63.8 224 303 1.78 431.79 78 143 1.35 110 304 373 1.73 574.19 0 24 1.24 0 374 472 1.68 695.29 25 42 1.19 29.76 43 49 1.14 51.18 50 63 1.09 59.16

PAGE 154

APPENDIXCPROOFSFORCHAPTER5 C.1 ProofofTheorem5.1Theorem5.1 :Whenthesuppliersareuncapacitatedandtherearenodiversicationbenetsandthereisauniqueleastcostsupplier,thenthermwillchoosetoorderitstotalrequirementsfromtheleastcostsupplier.Underthisscenario,thetotalusablequantityorderedfromtheleastcostsupplierisdeterminedsuchthat:Fr[1]q[1]=p)]TJ/F22 11.955 Tf 11.955 0 Td[(c[1]+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uwherec[1]isthecostperunitchargedbythelowestcostsupplier.Ifmultiplesuppliershavethesamelowestcost,thenthetotalorderwillbesplitamongstallofthelowestcostsupplierssuchthatthetotalusablequantityorderedstillsatisestheabovecriticalratio.Proof :Beforeprovingtheresultinthistheorem,werstcharacterizetheoptimalsolutiontotheuncapacitatedsuppliersproblemwithnodiversicationbenet.Thisproblemcanbeformalizedasfollows:MaximizeZqi0=p)]TJ/F22 11.955 Tf 11.955 0 Td[(s)]TJ/F23 7.97 Tf 16.804 14.944 Td[(NXi=1ciriqi+sNXi=1riqi)]TJ/F15 11.955 Tf 11.955 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uESC.1Addingthenon-negativityconstraintstothisobjective,thecorrespondingKKTconditionsare:)]TJ/F22 11.955 Tf 11.955 0 Td[(ciri+sri+p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uri)]TJ/F22 11.955 Tf 11.956 0 Td[(FNXi=1riqi+i=08iC.2qii=08iC.3i08iC.4 142

PAGE 155

143 Wealsonotethat@2ZU @qi2=)]TJ/F15 11.955 Tf 9.298 0 Td[(p)]TJ/F22 11.955 Tf 12.303 0 Td[(s+uri@F @qi<0foralli.Thustheobjectiveisconcavei.e.,theHessianisnegativesemidenite,andtheKKTconditionsarenecessaryandsucienttoobtainaglobaloptimumsolutiontoequation C.1 .As-sumingthatqi>0foranysupplieri,theFOCresultsinthefollowingrelationship:FNXi=1riqi=p)]TJ/F22 11.955 Tf 11.955 0 Td[(ci+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uC.5Ifthecostsperunitciarenotequalamongsuppliers,thenthisrelationshipcanonlyholdforanyonesupplieri.Further,forallothersuppliersjj6=i,itisobviousthatqj=0.Thus,equation C.5 providesthefollowingresultforsupplieri:Friqi=p)]TJ/F22 11.955 Tf 11.956 0 Td[(ci+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uC.6Usingthisresult,supposethatqi>0forsomesupplieri,andqj=08j6=i.Then,forallsuppliersjwehavethefollowingrelationships:Friqi=p)]TJ/F22 11.955 Tf 11.955 0 Td[(cj+u p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+u+j rjp)]TJ/F22 11.955 Tf 11.956 0 Td[(s+uC.7j=rjp)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uFriqi)]TJ/F22 11.955 Tf 11.955 0 Td[(rjp)]TJ/F22 11.955 Tf 11.955 0 Td[(cj+uC.8j=rjcj)]TJ/F22 11.955 Tf 11.955 0 Td[(ciC.9InordertosatisfytheKKTconditions,thenjmustbenon-negative8j6=i.Thiscanonlyoccurwhentheithsupplieristhelowestcostsuppliersuchthatcj>c[1],wherec[1]isthecostperunitchargedbythelowestcostsupplier.Forthesituationwheremorethanonesupplierhasthelowestcost,thenitisobviousfromtheKKTconditionsofoptimalitythatthetotalusablequantityorderedmuststillsatisfytheabovecriticalratioinequation C.2 .Thisconcludesourproof.

PAGE 156

144 C.2 ProofofTheorem5.2Theorem5.2 :Whensuppliersarecapacitatedandtherearenodiversicationbenets,thentheoptimalnumberofsuppliersselectedandthecorrespondingquantityallocatedtoeachsuppliercanbedeterminedasfollows.Step1:Indexallsuppliersinincreasingorderofcostperuniti.e.,c[1]c[2]c[3]:::c[N].Step2:Foreachsupplier[i]i=1;:::;N,computeQ[i]suchthat:FQ[i]=p)]TJ/F23 7.97 Tf 6.587 0 Td[(c[i]+u p)]TJ/F23 7.97 Tf 6.586 0 Td[(s+uandbasedonthisdetermine:t[i]=Q[i])]TJ/F28 11.955 Tf 11.955 8.966 Td[(Pi)]TJ/F21 7.97 Tf 6.587 0 Td[(1j=1y[j]r[j]Step3:Theoptimalnumberofsuppliersselectedkcanbeidentiedasmaxf1kNjt[k]0g.Step4:Thequantitiesallocatedtosupplierj=1;:::;k)]TJ/F15 11.955 Tf 12.575 0 Td[(1areq[j]=y[j],thequantityallocatedtosupplierkisq[k]=maxft[k];y[k]g,andthetotalquantityorderedbythermfromallsupplierscanbedeterminedasminfQ[k];Pkj=1y[j]g.Proof :Recallthecapacitatedsuppliersproblemwithnodiversicationbenetisasfollows:MaximizeZqi=p)]TJ/F22 11.955 Tf 11.955 0 Td[(s)]TJ/F23 7.97 Tf 16.804 14.944 Td[(NXi=1ciriqi+sNXi=1riqi)]TJ/F15 11.955 Tf 11.955 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uESsubjectto:qiyi8i C.10 qi08i C.11 BasedontheproofofTheorem5.1,weknowthatZisstrictlyconcaveinqiandtheconstraintsarealllinear.Thus,bynotingtheconcavityofthelagrangeanfunctionLL=Z+Pni=1iyi)]TJ/F22 11.955 Tf 12.163 0 Td[(qi,weknowthatKarush-Kuhn-TuckerKKTconditionsarenecessaryandsucienttoidentifytheoptimalsolutiontothis

PAGE 157

145 problem.TheKKTconditionsforthelagrangeanare:qi[@L @qi]=riqi[)]TJ/F22 11.955 Tf 9.298 0 Td[(ci+s)]TJ/F15 11.955 Tf 11.955 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+u)]TJ/F22 11.955 Tf 11.956 0 Td[(FNXi=1riqi)]TJ/F22 11.955 Tf 11.955 0 Td[(i]8i C.12 i[@L @i]=i[yi)]TJ/F22 11.955 Tf 11.955 0 Td[(qi]8i C.13 TostartwithassumethatthereissomeamountQwhichwillbesourcedfromallthesuppliers.ThenitisobviousthatusingtheunderlyinglogicofTheorem5.1,wewouldchoosetosourcethemaximumamountpossiblefromthelowestcostsuppliers.Thus,ouroptimalalgorithmgivenQhasthefollowingstructure: 1. Indexsuppliersinincreasingorderofthecostscisuchthatc[1]c[2]:::c[N].Sets=0andqi=08i. 2. s=s+1.Determinet[s]=maxf0;Q)]TJ/F28 11.955 Tf 11.594 8.966 Td[(Ps)]TJ/F21 7.97 Tf 6.587 0 Td[(1i=1r[i]q[i]gandbasedonthis,q[s]=minfy[s];t[s]g. 3. Ifs=nstop,elserepeat2.BeforeproceedingtodeterminetheoptimalquantityQ,notethatforeachsupplierwhereqi=yi>0,thelagrangemultiplierishowsthemarginalprotwhichcouldbeobtainedbyincreasingthecapacityofsupplieriandthiscanbedeterminedfromequation C.12 .TodeterminetheoptimalquantityQwhichshouldbesourcedfromallthesuppliers,wenotethatforanyonesupplierk,)]TJ/F22 11.955 Tf 9.299 0 Td[(ck+s+p)]TJ/F22 11.955 Tf 12.005 0 Td[(s+u)]TJ/F22 11.955 Tf 12.005 0 Td[(FQ=0musthold.Obviously,sinceprotsaremaximizedbyorderingrstfromthelowercostsuppliers,determiningkiterativelyasinourTheoremmusthold.Thisconcludesourproof.

PAGE 158

146 C.3 ProofofTheorem5.5Theorem5.5 :Wheneachchosensupplierhasbothmaximumandminimimumlimitationsplacedonthesizeoftheorder,thentheoptimalquantityallocatedtoeachsuppliercanbedeterminedasfollows.Step1:Indexallchosensuppliersinincreasingorderofcostperuniti.e.,c[1]c[2]c[3]:::c[X].Step2:Foreachsupplier[i]i=1;:::;X,determineQ[i]suchthat:FQ[i]=p)]TJ/F23 7.97 Tf 6.586 0 Td[(c[i]+u p)]TJ/F23 7.97 Tf 6.586 0 Td[(s+uandbasedonthisdetermine:t[i]=Q[i])]TJ/F28 11.955 Tf 11.955 8.966 Td[(Pi)]TJ/F21 7.97 Tf 6.587 0 Td[(1j=1y[j])]TJ/F28 11.955 Tf 11.955 8.966 Td[(PXj=i+1z[j]Step3:Thequantityallocatedtosupplieriisq[i]=minfmaxft[i];z[i]g;y[i]gandthetotalquantityorderedbythermfromallsupplierscanbedeterminedasPXi=1q[i].Proof :Recallthecapacitatedsuppliersproblemwithnodiversicationbenetisasfollows:MaximizeZqi=p)]TJ/F22 11.955 Tf 11.955 0 Td[(sb+a 2)]TJ/F23 7.97 Tf 16.804 14.944 Td[(NXi=1ciriqi+sNXi=1riqi)]TJ/F15 11.955 Tf 11.956 0 Td[(p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+uESsubjectto:qiyi8i C.14 qi08i C.15 qizi8i C.16 SinceZisstrictlyconcaveinqiandtheconstraintsarealllinear,theKKTconditionsforthelagrangeanfunctionL=Z+PNi=1iyi)]TJ/F22 11.955 Tf 11.391 0 Td[(qi+PNi=1iqi)]TJ/F22 11.955 Tf 11.39 0 Td[(zi,arenecessaryandsucienttoidentifytheoptimalsolutiontothisproblem.These

PAGE 159

147 conditionsare:qi[@L @qi]=ri[)]TJ/F22 11.955 Tf 9.298 0 Td[(ci+s+p)]TJ/F22 11.955 Tf 11.955 0 Td[(s+u)]TJ/F22 11.955 Tf 11.955 0 Td[(FQ])]TJ/F22 11.955 Tf 11.955 0 Td[(i+i8i C.17 i[@L @i]=i[yi)]TJ/F22 11.955 Tf 11.955 0 Td[(qi]8i C.18 i[@L @i]=i[qi)]TJ/F22 11.955 Tf 11.955 0 Td[(zi]8i C.19 TostartwithassumethatthereissomeamountQwhichwillbesourcedfromallthesuppliers.Then,similartotheunderlyinglogicofTheorems5.1and5.2,weknowthatatmostonesuppliercanbeunconstrainedintheoptimalsolution.

PAGE 160

REFERENCES [1] Abad,P.L.,1988a.DeterminingOptimalSellingPriceandLotSizewhentheSupplierOersAll-UnitQuantityDiscounts,DecisionSciences,19,3,622-634. [2] Abad,P.L.,1988b.JointPriceandLotSizeDeterminationwhenSupplierOersIncrementalQuantityDiscounts.JournaloftheOperationalResearchSociety,39,6,603-607. [3] AberdeenGroup,Inc.,1999.StrategicProcurement:TheNextWaveofProcurementAutomation;AnExecutiveWhitePaper,July. [4] Agrawal,N.andNahmias,S.,1997.RationalizationoftheSupplierBaseinthePresenceofYieldUncertainty,ProductionandOperationsManagement,6,3,291-308. [5] Andrews,K.R.,1980.Directors'ResponsibilityforCorporateStrategy,HarvardBusinessReview,Nov/Dec80,58,6. [6] Anupindi,R.andAkella,R.,1993.Diversicationundersupplyuncertainty,ManagementScience,39,8,944-963. [7] Bassok,Y.andAkella,R.,1991.OrderingandProductionDecisionswithSupplyQualityandDemandUncertainty,ManagementScience,37,12,1556-1574. [8] Benton,W.C.,1991.QuantityDiscountDecisionsUnderConditionsofMultipleitems,MultipleSuppliers,andResourceLimitations,InternationalJournalofProductionResearch,29,10,1953-1961. [9] Benton,W.C.andPark,S.,1996.AClassicationofLiteratureonDetermin-ingtheLotSizeUnderQuantityDiscounts,EuropeanJournalofOperationalResearch92,219-238. [10] Bhote,K.,1987.SupplyManagement:HowtoMakeU.S.SuppliersCompeti-tive,AmericanManagementAssociation,NewYork. [11] Bozarth,C.,Handeld,R.,andDas,A.,1998.StagesofGlobalSourcingStrategyEvolution:AnExploratoryStudy,JournalofOperationsManage-ment16,2-3,241-255. [12] Burke,G.,andVakharia,A.,2002.SupplyChainManagement,TheInternetEncyclopedia,Wiley,NewYork,NY. 148

PAGE 161

149 [13] Burwell,T.H.,Dave,D.S.,Fitzpatrick,K.E.andRoy,M.R.,1997.EconomicLotSizeModelforPrice-DependentDemandUnderQuantityandFreightDiscounts,InternationalJournalofProductionEconomics48,141-155. [14] Chauhan,S.S.andProth,J.M.,2003.TheConcaveCostSupplyProblem,EuropeanJournalofOperationalResearch,148,374-383. [15] Choi,T.Y.andHartley,J.L.,1996.AnExplorationofSupplierSelectionPracticesAcrosstheSupplyChain,JournalofOperationsManagement14,333-343. [16] Chopra,S.andMeindl,P.,2001.SupplychainManagement:Strategy,PlanningandOperation,Prentice-Hall,Inc.,UpperSaddleRiver,NJ. [17] Cohen,M.A.andAgrawal,N.,1999.AnAnalyticalComparisonofLongandShortTermContracts,IIETransactions31,783-796. [18] Dickson,G.W.,1966.AnAnalysisofVendorSelectionSystemsandDecisions,JournalofPurchasing2,257-267. [19] Ellram,L.M.andCarr,A.,1994.StrategicPurchasing:AHistoryandReviewoftheLiterature,InternationalJournalofPurchasingandMaterialsManagement30,2,10-18. [20] Elmaghraby,W.,2000.SupplyContractCompetitionandSourcingPolicies,Manufacturing&ServiceOperationsManagement,2,4350-371. [21] Erenguc,S.S.,Simpson,N.C.,andVakharia,A.J.,999.IntegratedProduc-tion/DistributionPlanninginSupplyChains:AnInvitedReview,EuropeanJournalofOperationalResearch,115,219-236. [22] Fawcett,S.E.,andBirou,L.M.,1992.ExploringtheLogisticsInterfaceBetweenGlobalandJITSourcing,InternationalJournalofPhysicalDistribu-tionLogisticsManagement,22,1,3-14. [23] Fisher,M.L.997.WhatistheRightSupplyChainforYourProduct?HarvardBusinessReview,75,2,105. [24] Gallego,G.,andMoon,I.,1993.TheDistributionFreeNewsboyProblem:ReviewandExtensions,JournalofOperationalResearchSociety,44,8,825-834. [25] Garvin,D.A.,1987.CompetingontheEightDimensionsofQuality,HarvardBusinessReview,65,6,101-109. [26] Gerchak,Y.,andParlar,M.,1990.YieldRandomness,CostTradeos,andDiversicationintheEOQModel,NavalResearchLogistics,37,341-354.

PAGE 162

150 [27] Guder,F,Zydiak,J.andChaudrey,S.,CapacitatedMultipleItemOrderingwithIncrementalQuantityDiscounts,JournaloftheOperationalResearchSociety,45,10,1197-1205. [28] Gurnani,H.,Akella,R.,Lehoczky,J.,2000.SupplyManagementinAssemblySystemswithRandomYieldandRandomDemand,IIETransactions,32,701-714. [29] Hahn,C.K.,Kim,K.H.,andKim,J.S.,1986.CostsofCompetition:Im-plicationsforPurchasingStrategy,JournalofPurchasingandMaterialsManagement,22,3,2-7. [30] Hendricks,K.,Singhal,V.,2001.WhatisSCMWorth?,IIESolutions,33,2,12. [31] Holloway,C.,Farlow,D.,Schmidt,G.,andTsay,A.,1996.SupplierManage-mentatSunMicrosystemsA:ManagingtheSupplierRelationship,HarvardBusinessSchoolCase. [32] Horowitz,I.,1986.OnTwo-SourceFactorPurchasing,DecisionSciences,17,274-279. [33] Jucker,J.,andRosenblatt,M.,1985.Single-PeriodInventoryModelswithDemandUncertaintyandQuantityDiscounts:BehavioralImplicationsandaNewSolutionProcedure,NavalResearchLogistics,32,537-550. [34] Kelle,P.,andSilver,E.A.,1990.DecreasingExpectedShortagesThroughOrderSplitting,EngineeringCostsandProductionEconomics,19,351-357. [35] Kim,B.,Leung,J.M.Y.,Park,K.T.,Zhang,G.,andLee,S.,2002.Cong-uringaManufacturingFirm'sSupplyNetworkwithMultipleSuppliers,IIETransactions,34,663-677. [36] Khouja,M.,1996a.ANoteontheNewsboyProblemwithanEmergencySupplyOption,TheJournaloftheOperationalResearchSociety47,12,1530-1534. [37] Khouja,M.,1996b.TheNewsboyProblemwithMultipleDiscountsOeredbySuppliersandRetailers,DecisionSciences27,3,589-599. [38] Lambert,D.M.,andKnemeyer,A.M,2004.We'reinthisTogether,HarvardBusinessReview,82,12,114-122. [39] Leenders,M.R.,andFearon,H.E.,1998.PurchasingandSupplyManagement11thed.,Irwin,Chicago,IL. [40] Lester,T.,2002.MakingitSafetoRelyonaSinglePartner,FinancialTimes,April1,London,UK.

PAGE 163

151 [41] Liker,J.K.,andChoi,T.Y,2004.BuildingDeepSupplierRelationships,HarvardBusinessReview,82,12,104-113. [42] Lin,C.andKroll,D.E.,1997.TheSingle-itemNewsboyProblemwithDualPerformanceMeasuresandQuantityDiscounts,EuropeanJournalofOperationalResearch,100,562-565. [43] Maurer,C.,1997.HillviewHospital:TheHandsoapDecision,IveyManage-mentServicesCase9A97J006. [44] McCutcheon,D.,andStuart,F.I.,2000.IssuesintheChoiceofSupplierAlliancePartners,JournalofOperationsManagement,18,3,279-301. [45] Minner,S.,2003.Multiple-SupplierInventoryModelsinSupplyChainManagement:AReview,InternationalJournalofProductionEconomics,81-82,265-279. [46] Mohr,J.,andSpekman,R.,1994.CharacteristicsofPartnershipSuccess:PartnershipAttributes,CommunicationBehavior,andConictResolutionTechniques,StrategicManagementJournal,15,2,135-152. [47] Munson,C.L.andRosenblatt,M.J.,2001.CoordinatingaThreeLevelSupplyChainwithQuantityDiscounts,IIETransactions,33,371-384. [48] Nahmias,S.,2001.ProductionandOperationsAnalysis4thed.,Mc-Graw/Irwin,NewYork,NY. [49] Newman,R.G.,1989.SingleSourcing:Short-termSavingsVersusLong-TermProblems,JournalofPurchasingandMaterialsManagement,Summer,20-25. [50] Nishiguchi,T.,andBeaudet,A.,1998.CaseStudy:TheToyotaGroupandtheAisinFire,SloanManagementReview,40,1,49-59. [51] Pan,A.1989.AllocationofOrderQuantityAmongSuppliers,JournalofPurchasingandMaterialsManagement,Fall,36-39. [52] Parlar,M.,andWang,D.,1993.DiversicationUnderYieldRandomnessinInventoryModels,EuropeanJournalofOperationalResearch,66,52-64. [53] Pilling,B.K.,andZhang,L.,1992.CooperativeExchange:RewardsandRisks,InternationalJournalofPurchasingMaterialsManagement,28,Spring,2-9. [54] Pirkul,HandArasO.A.,1985.CapacitatedMulti-ItemOrderingProblemwithQuantityDiscounts,IIETansactions,17,206-211. [55] Porter,M.,1985.CompetitiveAdvantage,TheFreePress,NewYork,NY,pp.38-40.

PAGE 164

152 [56] Prahinski,C.,2002.UnineRichardson,Case902D20,RichardIveySchoolofBusiness,UniversityofWesternOntario,Canada. [57] Ramasesh,R.V.,Ord,J.K.,Hayya,J.C.andPan,A.,1991.SoleVersusDualSourcinginStochasticLead-Times,QInventoryModels,ManagementScience,37,4,428-443. [58] Rosenthal,E.C.,Zydiak,J.L.,andChaudhry,1995.VendorSelectionwithBundling,DecisionSciences,26,1,35-48. [59] Rubin,P.A.andBenton,W.C.,1993.JointlyConstrainedOrderQuantitieswithAll-UnitsDiscounts,NavalResearchLogistics,40,255-278. [60] Rubin,P.A.andBenton,W.C.,2003.EvaluatingJointlyConstrainedOrderQuantityComplexitiesforIncrementalDiscounts,EuropeanJournalofOperationalResearch,149,557-570. [61] Shah,J.02.HPWrestlesRisk.Downloadablefrom:http://www.ebnonline.com/story/OEG20020919S0063Dateofac-cess:January28,2003. [62] Shin,H.,Collier,D.A.,andWilson,D.D.,2000.SupplyManagementOrien-tationandSupplier/BuyerPerformance,JournalofOperationsManagement,18,317-333. [63] Silver,E.A.,Pyke,D.F.,andPeterson,R.,1998.InventoryManagementandProductionScheduling,JohnWiley&Sons,NewYork,NY. [64] Spekman,R.E.,1988.StrategicSupplierSelection:UnderstandingLong-TermBuyerRelationships,BusinessHorizons,July-August,75-81. [65] Treece,J.,1997.Just-Too-MuchSingle-SourcingSpursToyotaPurchasingReview:MakerSeeksatLeast2SuppliersforEachPart,AutomotiveNews,March3,p.3. [66] Verma,R.andPullman,M.E.,1998.AnAnalysisoftheSupplierSelectionProcess,Omega:TheInternationalJournalofManagementScience,26,6,739-750. [67] Veverka,M.,2001.TheRealDot.Coms:TheCompaniesoftheDow30areDeliveringontheVastPromiseoftheInternet,Barrons,81,52,December24,19-21. [68] Weber,C.,Current,J.,andBenton,W.C.,1991.VendorSelectionCriteriaandMethods,EuropeanJournalofOperationalResearch,50,2-18. [69] Weng,Z.K.,1995.ChannelCoordinationandQuantityDiscounts,Manage-mentScience,41,9,1509-1522.

PAGE 165

153 [70] Yano,C.A.,andH.L.Lee,1995.LotSizingwithRandomYields:AReview,OperationsResearch,43,311-334.

PAGE 166

BIOGRAPHICALSKETCHIn1967,GerardJosephBurkeJr.wasborninJacksonvilleBeach,Florida.HegraduatedfromtheUniversityofFloridawithaBachelorofSciencedegreeinbusinessadministrationin1991.From1991through1999,Gerardworkedasarestaurantmanagerforcasualdiningcompanies.Duringthistimehegainedvaluableexperienceintheeldasageneralmanagerofproductionandserviceoperations.In1999,hewasawardedagraduateassistantshipfromtheUniversityofFlorida'sMBAprogramandbecameaPoeEthicsFellow.AftercompletingtheMBAprogramin2001withaspecializationinsupplychainmanagement,GerarddecidedtopursuehisnewfoundcallinginacademicsandremainedattheUniversityofFloridaintheDepartmentofDecisionandInformationSciences'doctoralprogramforoperationsmanagement.Duringhisdoctoralstudiesfrom2001to2005,hewasawardedaUniversityofFloridaAlumniGraduateFellowshipandbecameamemberofthedecisionscienceshonorarysociety,AlphaIotaDelta.HeisalsoamemberoftheInstituteforOperationsResearchandtheManagementSciencesINFORMS,DecisionSciencesInstituteDSI,andProductionandOperationsManagementSocietyPOMS.Hehaspresented,sessionchaired,orrefereedforannualmeetingsofINFORMSandDSI.UponcompletionofhisdoctorateattheUniversityofFlorida,GerardwillbeonfacultyatGeorgiaSouthernUniversity'sCollegeofBusinessAdministration.


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 E20101124_AAAADF INGEST_TIME 2010-11-24T22:26:03Z PACKAGE UFE0011378_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES
FILE SIZE 12323 DFID F20101124_AACKHH ORIGIN DEPOSITOR PATH burke_g_Page_165.jpg GLOBAL false PRESERVATION BIT MESSAGE_DIGEST ALGORITHM MD5
f4105d26212a12d7382ad2f10a279e15
SHA-1
dce8ea4a96baeefa937a427b0b4216fc3ff8bb5e
30890 F20101124_AACKGT burke_g_Page_144.jpg
f1196a453d656a6c9ce7152327a643c2
131ba0a4ad058a38ea3938c73493182458533f4e
22032 F20101124_AACKHI burke_g_Page_001.jp2
cbd6759c544320c98adae64f2e079e5b
8198ac8a6b1efea3c345d2fe6a87593c0650530c
83638 F20101124_AACKGU burke_g_Page_145.jpg
2edbce99a60a1503aaacd72ca200da15
12e17190dba013a2aa09bb2d2dae66be5180ae97
6135 F20101124_AACKHJ burke_g_Page_002.jp2
388b89720241988b4c495560c44e61e6
fc31fc572ce34f2419cd7caff99fe0fb6c8bbe37
82444 F20101124_AACKGV burke_g_Page_149.jpg
66d21d3013d074d9853e4d9e0a7bd3ae
da3e422134dd52f471fa0db0a4ef4af1f23074ee
9111 F20101124_AACKHK burke_g_Page_003.jp2
9dccd92079a9c052b9bc7640be174393
1d63cd6ee6fb0aae18bc773c187ac689c461a18c
84030 F20101124_AACKGW burke_g_Page_151.jpg
e2f4651862ed4a48e4aaa7d62164b822
f6c890094fb67326aa1dfb598dfa5e77695784ec
109358 F20101124_AACKIA burke_g_Page_032.jp2
ec957a9286d4f1c46b6485b6fbb7e003
9cdb2c60a07cabc5329427c4c7e46d8a8fb18213
97139 F20101124_AACKHL burke_g_Page_004.jp2
b9ef692792f1f0fdc5e40c6ba56941a5
5e443fa5609b74b5bfc68d936deb35e47cf9e3f0
37337 F20101124_AACKGX burke_g_Page_153.jpg
2abe7d83ebf1acf4205913ff72ba47af
096d24290912e0d151955c7e99feed3c92806e60
119957 F20101124_AACKIB burke_g_Page_033.jp2
ba6b86c03fc16f819ffdf8a0a719c275
b7ffb14befc64de753fab36e9e155815d9a31ce9
26868 F20101124_AACKHM burke_g_Page_005.jp2
8b21e3834a842831dd44e510758e19b2
4921790bfb8fcaace74101e6c06c9fe97557bb30
45314 F20101124_AACKGY burke_g_Page_154.jpg
09e3949c1fd22fa7ae388bbec7b1cc06
876ddb6ad1229b7b8364af57bff91efe884819c5
114805 F20101124_AACKIC burke_g_Page_035.jp2
1b02367b629bae7985a18fd250149a45
a5575bd376e583a5db9fb1031595efb09b0924b3
1051986 F20101124_AACKHN burke_g_Page_007.jp2
d91b2137ced4653303ce2cd4fbcf4a53
f0142653e35bd290fd2555a58ba1bd089563c472
56210 F20101124_AACKGZ burke_g_Page_155.jpg
2d87d149455d77861d344570d43be478
312af2d12822f1bda906688f86fa0ce07ccd060e
117025 F20101124_AACKID burke_g_Page_036.jp2
14c25392a80574c19d8d50ec13c49fa8
4689a1b76e18e293a5e6e0ccea6b3e66d600c05c
1051918 F20101124_AACKHO burke_g_Page_008.jp2
bb37a576bca2ae8a5fd52cefb6ce6539
520e9804326925d0515bb31933ca654b7cbb03bb
115349 F20101124_AACKIE burke_g_Page_037.jp2
80809cd7eb62c453dc1d26b29ce328a9
d98078a9a79c1b28bf0537ccb700e1d9024e2527
1051983 F20101124_AACKHP burke_g_Page_009.jp2
b6548c53971997a05d7877f9dceec782
2777018d650e1353bd0972c25f39b02d7705850a
14508 F20101124_AACKIF burke_g_Page_038.jp2
769a12e32946a7e53ac53ba193a9b9fc
c2209304569630956240f3586befe7dcd7157904
101169 F20101124_AACKHQ burke_g_Page_013.jp2
52e6c49d2147ee3aa41c7df4f955df20
e63be59878347b18a322044cdc95e05437222dd5
111523 F20101124_AACKHR burke_g_Page_017.jp2
f6a5e91fe14e2d38bd7bdc2cdd846bae
e25d509218d3e0164e89f7ac8af49eda74833b5f
98309 F20101124_AACKIG burke_g_Page_039.jp2
f05716da821d5150c92386a017434acd
b9b32c75381f8ccd5daa07ac31305e504021b2e5
111007 F20101124_AACKHS burke_g_Page_021.jp2
aa3420368314c78dc402fd5031783f5a
462abf207fc6b98787011e757007d6704bd2b13d
107631 F20101124_AACKIH burke_g_Page_040.jp2
02a831a9f53814db7ab597e35c2d9886
9d3615a1d7241d0be6c57b3f21f2fd94cfe595e1
117263 F20101124_AACKHT burke_g_Page_022.jp2
7727c05604a932c3e8ccb7efbdea82bb
c5e3c30e52935280d8b6768458ae36e912955a65
769119 F20101124_AACKII burke_g_Page_042.jp2
892b74f749ca9ff978992887520cb752
205c31c1fe580eb54612dfb0f6a63d9d1b425ecc
115951 F20101124_AACKHU burke_g_Page_023.jp2
a75c5c2d92bd363c8e1721d21e188c75
843dcbf49f7a7f98968194dca5ce2b2cdcfbe298
836304 F20101124_AACKIJ burke_g_Page_044.jp2
e5b224f9b102fb83409c848f4f46a6bd
811a5f2b82cb9eb40ff8be0dfbf40e8c7f796971
110887 F20101124_AACKHV burke_g_Page_024.jp2
04a1ceecbfeabae08bc8bd8c524dd8b5
d77f81bc6d7b24df344198bcbdc0c556416a4119
924409 F20101124_AACKIK burke_g_Page_045.jp2
229405f213e14e0988fe3a3082630797
6b38e7494a3c39489d0cadeb0b583b05f95604ff
114327 F20101124_AACKHW burke_g_Page_025.jp2
c921184bc873bb7fe669e7705b324f49
618e0b9b94326285c90103470566756abffc9007
1051971 F20101124_AACKIL burke_g_Page_046.jp2
30c6cf8bad3814fd36465cb329a061e2
21fd993ca89e0ad8aa99ab1b7021356645550c07
72135 F20101124_AACKHX burke_g_Page_027.jp2
24e5a2c75da81b122c7cc3a6caba6d16
d68d4684e3cda247e5361a78f88b8497815f4d5d
87291 F20101124_AACKJA burke_g_Page_072.jp2
b112163e3ae67c40a4118009382e5439
03014f0f70e97a5410757b634c62685039f0830c
97360 F20101124_AACKIM burke_g_Page_047.jp2
4e4480292d0e80db2f33d5b0b4bbb590
7936e31a934854928289f4fe6a40fd5bd1792c2a
113802 F20101124_AACKHY burke_g_Page_029.jp2
8ff89c60db34bf88e5370fde83f89c6f
e2a25b476b24c5d6282ab304b0d3747cd5d03ecf
120835 F20101124_AACKJB burke_g_Page_073.jp2
db0372be478c1bfb4917a31398c4f5b8
c2ddfbf3abb19052abafd7e0c89d98788e3bc40d
96182 F20101124_AACKIN burke_g_Page_049.jp2
dd9eaadeb436eef5b6e57f1d98ce926a
6192deb8d14ecb118da72a71a54cf544553eaeab
111933 F20101124_AACKHZ burke_g_Page_030.jp2
5155ffec81279b46889dcf5a20c6b5e8
8f318cae36e819288bca746b889914a34c66dd23
91948 F20101124_AACKJC burke_g_Page_074.jp2
e197859ea49b5b5c36cf6276318e5cbf
c3c90f2d377540e16e2b7abd7af2c88e0f912f4b
89808 F20101124_AACKIO burke_g_Page_050.jp2
ee31439684917413dd018c94d93a854b
4a99a81758d10e5c70a01805f75d8b3b08c11283
110507 F20101124_AACKJD burke_g_Page_076.jp2
83e0251406755869e8972dfee0f84bec
92cbd9ed144dfd11e53a9d8e09a4106aef7a6b61
107586 F20101124_AACKIP burke_g_Page_051.jp2
f61e2db0601f9630b932d357f2c3b944
4cd444eb7638a6002bd4b35d2176df8b38f713eb
105518 F20101124_AACKJE burke_g_Page_077.jp2
9b75341990853afd8298d58cd8d0ea74
4482374a2ba232cedce42ef718ee65f6a00120ca
124168 F20101124_AACKIQ burke_g_Page_052.jp2
3e31f3d6868f2bf28c8702a46372eba8
5d6067b954b777566a2be01db6a179da9a2d5988
96915 F20101124_AACKJF burke_g_Page_078.jp2
fc1d86e23987a173037cf6d142605fc4
711318491eb8f7544cdd86c2154d28ebd1e7e1f2
116308 F20101124_AACKIR burke_g_Page_053.jp2
2ed2d5effc4c6a421c5089d88bb575df
f13afe7a6910689bd296ea61d1c66f18d13fbb98
48216 F20101124_AACKJG burke_g_Page_081.jp2
3c299020d5f693636263e93dba6135d2
2d1f33f398bf0a4b7e693cca3e77d40d6d85ae80
108234 F20101124_AACKIS burke_g_Page_056.jp2
5411f3546076d350b0e488e1acad1dd3
3bbfe2e4be98a1a6fcd47f7fe6ded0d8224fc256
96430 F20101124_AACKIT burke_g_Page_060.jp2
ebb0be53d0005606bfc7ce8ec24bdbcd
c698939c04a9a626b37c36e852a509fb6d7829c0
104868 F20101124_AACKJH burke_g_Page_082.jp2
c5cf967f3cb855e60dfbde1f8b23805d
f1f489e69a5d9acd6393420f7b47674f6d3f23dd
116755 F20101124_AACKIU burke_g_Page_061.jp2
522b7d86f3dc7f4fa1f3c6c7c6402b16
bd8acde8ddadb68564db3db50c365838224c8c5c
112658 F20101124_AACKJI burke_g_Page_083.jp2
cb8c942d3ff11154c4dc77ee113a9b3f
c1aa6d4439957d6b0caa2fa487cfaade9389f1ad
78518 F20101124_AACKIV burke_g_Page_064.jp2
59e04d923535a4eec5ff63d18402ae55
dfe587e9a9712f3df26ea0a5f5df2b2d9b7c858a
121187 F20101124_AACKJJ burke_g_Page_085.jp2
c517df0cb00e8f3fe70aec6a696f2a1d
4522464176d4a92593ca85e92ca9721f8484c8cb
105794 F20101124_AACKJK burke_g_Page_086.jp2
81d6fd2d77095b272acec806ecb4deeb
1df18838f2d8a0543791cf4b153a363004b14d2d
72071 F20101124_AACKIW burke_g_Page_065.jp2
ffabe724c15701df7b021d4cf0073aa4
ce6326448b1b0a6ab7bd4885f8d4fd65f1db75e5
91365 F20101124_AACKKA burke_g_Page_111.jp2
ba5f91fcbb64f4bcb4b69db919190336
d74dc05af71748eea498264d5b06731a0e86a632
880368 F20101124_AACKJL burke_g_Page_087.jp2
c36c38882fa0a44b2efcc5decc48378e
6fb37b02a47c70107df3cb695be5b74f7c0a4a14
98661 F20101124_AACKIX burke_g_Page_066.jp2
9c4ba8b6fe23a56c96d7a75e13a240f2
85d723d9e0066a739fc6a9aa84e61c4840c2f389
88726 F20101124_AACKKB burke_g_Page_113.jp2
f0ed07386c4141240e2b2e19551124c0
8d2896f34bbcf5398b2d23e431f3af150973a4da
1051969 F20101124_AACKJM burke_g_Page_090.jp2
0238c9d33bb939941f2879bc4dec5e96
ddfa6146012d7c5a82d0adaa2752cee01adeff1e
98560 F20101124_AACKIY burke_g_Page_067.jp2
0e94e1e3aa721397394897703db37164
7c0e39eb70d8d9f5719e80948c41d7711c606303
674538 F20101124_AACKKC burke_g_Page_114.jp2
4a60fddc05cde4774f2c3f645146cd84
0ca282637e10c2b322d69b4a6dd47be78ff66e93
66187 F20101124_AACKJN burke_g_Page_093.jp2
501e441680a77291743c1bf37be8738c
b841ce981b8b25f568107ad6d09459837be49bb0
81430 F20101124_AACKIZ burke_g_Page_071.jp2
f4ccb83573397b6f564876afe513e0b5
6acdeb74f6edb58b0c8abef878ae5b557cd10eae
111633 F20101124_AACKKD burke_g_Page_117.jp2
dc505fc4d8c038bd5ba7bf68da75c3a8
ca33f7f40ca3fdaaa388649e7a6d2a9b4d5fd1f2
1020535 F20101124_AACKJO burke_g_Page_094.jp2
e3852d3370eb4e0d55e9f3498e73cb3e
03a6757836c9bf8fefdaee0bd047cd35bee75a9c
62045 F20101124_AACKKE burke_g_Page_118.jp2
0314df602bb5c37374e2c18680481d3f
500b153dca8dfda4134bca20d702ce0cf849bcde
112783 F20101124_AACKJP burke_g_Page_095.jp2
39d8651a99ecceb401de66cfbca0ce59
a3f55b2fb25bf5171e460f43412e12041a9cf2c5
47820 F20101124_AACKKF burke_g_Page_119.jp2
9562c800f667a412e3bca13074ad65b0
416975513c2f9ef6519eb462931d5688c463aad9
110707 F20101124_AACKJQ burke_g_Page_096.jp2
d4eea32f0b01dd2d8b7c4db7db82ade9
c642f353dde3716ec63ab55921a64a3ef9e307db
345645 F20101124_AACKKG burke_g_Page_120.jp2
90261e1c8a8014242060b367d0dc5242
b9d314cc7cc430d92402ab30a5b4e56de8c6e283
107207 F20101124_AACKJR burke_g_Page_098.jp2
04cba4252a36a97ac17cfd038211c61e
8f7626ea64cc8852ce6218dccd3464c506e16f6c
97527 F20101124_AACKKH burke_g_Page_121.jp2
78e9d499933a6bb49ef16f0f37dce274
42f5725bcdae78ba756ccaae69f9180b5ea0d1e7
110808 F20101124_AACKJS burke_g_Page_099.jp2
102129b88f1e97812417aba2c99e4dc9
792830ffa0b54461325d4d5c8fbc5c1c31102574
118692 F20101124_AACKJT burke_g_Page_100.jp2
9b21ecac8b5c53652a44e9dcebeb1367
014243d0a9567cb542a866c0b3ade81d2147ff81
53647 F20101124_AACKKI burke_g_Page_123.jp2
0e9656ab88acc27c68c52527ab3c1d64
2e0f1464497710c48301faf76eac7763732fe2f0
113374 F20101124_AACKJU burke_g_Page_103.jp2
91a76c46901776a1b8d98ab27dc94c1f
56820fbc2f56dd7b556fb37c0910c7c9f4b487b2
91647 F20101124_AACKKJ burke_g_Page_124.jp2
6e6bfba96ca13e044946c8d2600ebe9a
2a625cc39b587beb067f7359b6677cc378bd0dbf
113244 F20101124_AACKJV burke_g_Page_104.jp2
28f42ad1fc7063e3f53f9eb401af146a
c11dd8b05ac3a92e951f5a0d5de8e3596a6351cb
92240 F20101124_AACKKK burke_g_Page_125.jp2
a16b12a15eee80a1cd98b149e810abca
a5ab533f671c2d4e24948f27b1b5202e84f41aaf
107078 F20101124_AACKJW burke_g_Page_106.jp2
784d56d346e85326c194bcafe6eac7ce
4d0742bb1afe723d1581f224118274418e33ee8c
24827 F20101124_AACKKL burke_g_Page_126.jp2
54b7fc9f209d1bfcc2601680a72e61ee
e043ed551056e891d8dc465c627ab5fc7eaaf984
116207 F20101124_AACKJX burke_g_Page_107.jp2
e3804bc75f5c2a09fcd984bc25b50ecc
b5c82ecfb6234ae11e1b31e94604d66b889493ec
90818 F20101124_AACKLA burke_g_Page_149.jp2
c04b94a90eaeb7fcf28d64825032cb8e
4a09cc3dca7791e5e5df991b7272ef29a5ea4db5
90824 F20101124_AACKKM burke_g_Page_127.jp2
7b02099d87fa0b6b213c2c27e6738337
36ac73db12b7add18589edcabe463705a2ae4286
112350 F20101124_AACKJY burke_g_Page_109.jp2
88ed3cbe2abd7e8efeb4cf8c7ff0b051
c0ae815bfffa05f95a2a5e7248e07655c65df2d3
19358 F20101124_AACKLB burke_g_Page_150.jp2
987f59c4dba3df697949967787154e50
48751bc66dd8fe5433907f5ae657a782e8a18d4c
90336 F20101124_AACKKN burke_g_Page_128.jp2
98e51824d8b8a0711832b7b8ee85d652
8d7329ea0e17eeb33f1242ff84abc70b4f31e0f9
113441 F20101124_AACKJZ burke_g_Page_110.jp2
734b8dcee4dc39e60f8b8d612ce918f0
d4e50d9386dc7b941761e84c461c87873752f9b8
87485 F20101124_AACKLC burke_g_Page_152.jp2
f0ba38e1c35d8816548fc8cb46f38040
4e62762f556104fb69abacedf5a65f5179ec7978
29288 F20101124_AACKKO burke_g_Page_129.jp2
3818c8e02345121106f1bf84f26cf965
6c40f545bc640d9884a4174049021e20fb7afd05
67797 F20101124_AACKLD burke_g_Page_154.jp2
97253c5fbe81a2d2d9f3623030aedebf
a385adb697cec9514f760afecb3c37078607ad93
86254 F20101124_AACKKP burke_g_Page_130.jp2
ab38101e4217f2bc609956c170a31414
46b51019eb10a40a897ef841202e18b9d18379ca
811844 F20101124_AACKLE burke_g_Page_155.jp2
17d31ff82bad69033323cdf5b49df269
b59f487802af4ac5d6a54c61d76ce46a4440ec6d
73864 F20101124_AACKKQ burke_g_Page_131.jp2
c16e7a8ba96827c76bdaa41942b80c19
9aa681a1032a4f79f79b2e4c107868ba542bf8e9
76862 F20101124_AACKLF burke_g_Page_156.jp2
7178abb4684b9a0b66ab8ed502c16ed6
76cc10e51d846df72eb030c472b3c6508d9f428c
25447 F20101124_AACKKR burke_g_Page_132.jp2
25f9704690cce15a3125d848b764848d
91d359c0e1ef4e345389def7c70c6aa3adbd9282
757971 F20101124_AACKLG burke_g_Page_157.jp2
76ed42eda569aafb1bb4bb1966b6ec74
fa54d3fa07502d86e6ea219a70011d2bf4b2d6e4
83041 F20101124_AACKKS burke_g_Page_134.jp2
715bbf54e92f1e35617e134515b9508c
c02d1b2d83fdc13d63ec3b36c02206289569132f
66604 F20101124_AACKLH burke_g_Page_158.jp2
0e3e76b45ab82c5eec46bda122136b0d
2e4d1820b11dab3b4519e4ac9040d5396787fcd4
32406 F20101124_AACKKT burke_g_Page_135.jp2
f01af5079285bf09a71972bf9a60418b
eb47eb21cf9749938ddd85a88824dc6c2aa675d0
25461 F20101124_AACKLI burke_g_Page_159.jp2
411dbf2aec56af6b1de009e3bf7d5dc9
dab953a35c757e12231869b8ae73cda0fb4e8345
90198 F20101124_AACKKU burke_g_Page_137.jp2
d435ccbd4c8a7801addf74e5770e91f0
c0a5c205fced07562a738e731385e4af37fdf1fb
26284 F20101124_AACKKV burke_g_Page_138.jp2
cc18387de66521e1d24bc5d8a05461d0
d704fc4995da01586fd666766f12dd29ead80fec
10130 F20101124_AACKLJ burke_g_Page_165.jp2
6420b7991bff9f292efed5789a6bf75d
a23a110eef4acf6ac3046ec59ca2a51f24996c16
39669 F20101124_AACKKW burke_g_Page_141.jp2
a6932f042fd45311691eaa7db17f4c47
56409bb894c05097bf3d34b4a9eda4f538050e70
85573 F20101124_AACKLK burke_g_Page_166.jp2
c70e1b181273b85825c4e67e618426c3
2e2ef45c442acab40615c9abb24408f0f25503ec
88764 F20101124_AACKKX burke_g_Page_143.jp2
1176bfcc36598623a453f950586007cb
69fadc2b4d885bbcf65e3ee3084c612fdecd6f98
1053954 F20101124_AACKMA burke_g_Page_021.tif
d7a1bcf3ac204aa31837e1691f8ecbbf
9fe41dc6958a758d6abd626b0ecb0ad3cadd99db
F20101124_AACKLL burke_g_Page_001.tif
f5b27d7bc73be4b204a8d4f67d485629
2179f6fcfb026e8a9ead9b438d7c70506fc1046e
28383 F20101124_AACKKY burke_g_Page_144.jp2
44a71800065e77ea1efddef7e2fb9cce
f5d356fa9127088868be09a858d2b01600bbc2c3
F20101124_AACKMB burke_g_Page_022.tif
515a75a2e2e0c1909a168d259208b364
a82a30e3dfc35d37054a7114a517323b0e5a7850
F20101124_AACKLM burke_g_Page_002.tif
984c8341d5c0fd5aca0be3694df79a45
97624d9038ceb332fad6aeedde8e9f7a341a7510
92384 F20101124_AACKKZ burke_g_Page_145.jp2
09066482fd9fd5c8ec5738944b28ec41
ffaefc2c73d5fdd753590ddb10b4c8d083ff1fe5
F20101124_AACKMC burke_g_Page_023.tif
f9460b33038dff5b62e171dc49610306
aa0d48910e3031e68657f01d4a163a5ff7848522
F20101124_AACKLN burke_g_Page_004.tif
e0982f00c83915a9f3b8721670359ba0
0b2356935565fdce0a095bf661084f322e00908c
F20101124_AACKMD burke_g_Page_024.tif
5eb75fa8a7a228a03d39d799afba72cc
d83dea70ad5326b1ed64e0da3b081a0ac1560c2d
F20101124_AACKLO burke_g_Page_005.tif
8e1c5fc22ff5e9b58fac3151907d6909
794614442e9f5400efd3123b8d7f2fc15e8ed43f
F20101124_AACKME burke_g_Page_026.tif
2245ab38773fda0b0632bae8a96bdacd
c05d517c4515eefe4444585afe9e1a1e77e5f1bd
25271604 F20101124_AACKLP burke_g_Page_006.tif
f3f62e509f8516ae582590e46b419262
e1f8aeb9aa13ec21d9056f478a0b48b2434beb2a
F20101124_AACKMF burke_g_Page_027.tif
e378c8132285313cff99fcd095400f25
f04d7c37802519e0ca949ea5be5a94cb1185cf21
F20101124_AACKLQ burke_g_Page_007.tif
efaeba4dc6477f8a554f76ff547d2932
6833eae5bfc23110207d1c8c973fdf567c4f5124
F20101124_AACKMG burke_g_Page_028.tif
96acc57e5f5d236785d1087e711f4c0f
01e4ea79c11673ba7fdfca27176d1d41494960e8
F20101124_AACKLR burke_g_Page_008.tif
be7505e401fb339fc208cf5f79ad336e
6c40f1339a748ad2203b96bf46349ba7da31ff82
F20101124_AACKMH burke_g_Page_031.tif
129314e820fa8769a3565af151174bc1
0bae7e56c777b006b72aabee04bcd296b6977d4b
F20101124_AACKLS burke_g_Page_009.tif
5da625b64425f92484ef3b8670e4f6fa
09f655d2156b8470e56666a2e5594b4d3c50058e
F20101124_AACKMI burke_g_Page_032.tif
985e3c0c143fcd0e2481298db082f362
bfcd1d68077f1964606e8f39826abcb01e2758e2
F20101124_AACKLT burke_g_Page_012.tif
bf785b24eb1958703320ed16b85b933c
b89e0ba70cf49f876f52ffe94f7f7efb7586da9e
F20101124_AACKMJ burke_g_Page_034.tif
6b9e7ed2e8551f88f6c6084eb25ec029
4258f6ccd13457c8a5440004b6eac2e73882c3b8
F20101124_AACKLU burke_g_Page_014.tif
29ff9ccfc112d191782956173654938e
02832e5c876924aa1d959c6569228e958a9e23d6
F20101124_AACKLV burke_g_Page_015.tif
abe5132b8193a4dbee8285665b5ebe66
29bcff1eb7d19a781c78aec9fdbb50275e7bc3cd
F20101124_AACKMK burke_g_Page_036.tif
11bb2368cb01cb559fe36932629d0e9b
697d971a4016cded813fa359d6efd246b2e5176e
F20101124_AACKLW burke_g_Page_016.tif
62df889e872f17103efd41eca7a0a245
588797d3a014f5f7f816d82d1a31f382174da2d1
F20101124_AACKNA burke_g_Page_061.tif
7fec93914710e123a4e3e492aca3679b
c7f9d0ad089a48ce4123a567269b5d3b42a11fce
F20101124_AACKML burke_g_Page_037.tif
a648f7839149d19a9356492864fd09e9
5a25f735dd0996f719f4930dd3cda3f9b1eff525
F20101124_AACKLX burke_g_Page_017.tif
4f6c6907a24b55da56c8726361e9ba16
f7137c20e4c800d64974939a6a8eb31f0ee1059c
F20101124_AACKNB burke_g_Page_063.tif
157f700380f2eeda8a839d9b2f1526d7
4908536c5f4c380ccbb9c6656a73ce39236bad9f
F20101124_AACKMM burke_g_Page_039.tif
1333b39b14ffb9ae05304f509c68b39f
2db0ac2908c241c6e1eb9c976483e44f7d81e3e2
F20101124_AACKLY burke_g_Page_018.tif
0a53775b6dfe5acfde885a93d899ba4b
8f764234e6f0adc7aedd82ecc63dddd229908218
F20101124_AACKNC burke_g_Page_064.tif
9a82f162f2820c0ba9205a07ec060b77
e148b4351f79bf2f4052068128ba19fc94cfc8a4
F20101124_AACKMN burke_g_Page_040.tif
9629c7c39a4494736255aea24ddc8696
ac510120a5de4db893a25ac2ac3db05ac7e70f71
F20101124_AACKLZ burke_g_Page_020.tif
dcc2b1cc8fb4e9f099ae7a679eea4a7e
aa49d0532476371e054dc8d0bcedd3b9dfbe6121
F20101124_AACKND burke_g_Page_065.tif
7deecbc6ae70e8cc28c6a27e1e908c87
7dcfb9d3a481712a0085955a65bbd0486443a684
F20101124_AACKMO burke_g_Page_041.tif
db30778f9aa8c4ad74b25a59021f41da
a7ae3633c8ced3e3edd2adb9a50e9caa615fd2df
F20101124_AACKNE burke_g_Page_067.tif
6a1098e4788f714a9ab3bb2877d1287b
41f1fec9c1ea017fd1b3e5cb247de61c1b61c6e2
F20101124_AACKMP burke_g_Page_043.tif
bf69d3b6e6670d7e9c5e4744e5856bc6
0355515f25e53f3ec2af60f8f1af8a9b1a385f65
F20101124_AACKNF burke_g_Page_068.tif
258f1c42aec32d1b352c8de4161f7ff2
a49d8af861d10090da317bed043687bdef7dfaf3
F20101124_AACKMQ burke_g_Page_045.tif
99f4cc6bfa7ca2be027328ff18edfed2
c7b7e17646c4416284f17ddbe701814f6ab8d846
F20101124_AACKNG burke_g_Page_069.tif
c9d305747ce8c3cabfdd08b6f8b53122
24f414d204a531770acd3d8374ea492356a1c6d0
F20101124_AACKMR burke_g_Page_048.tif
a8bf89772f7f8a1d54168229381dd78e
1f911c5fdd324a52b55a22d5107262b908e0edce
F20101124_AACKNH burke_g_Page_070.tif
8b13d2f49fe1075307ed20cd1a7a7c9c
a22982108f9f7f58893e7e00cf240638dcf398a5
F20101124_AACKMS burke_g_Page_050.tif
f284d3367c2223b3a62e5fd6b63db49a
fe77fa5c8bc20c396d58b7f26bd4667c57961306
F20101124_AACKNI burke_g_Page_071.tif
c37bb4286a983ddf184e15a40d99d79d
34656134a9a21390d2ccda71e133dd839430de7d
F20101124_AACKMT burke_g_Page_052.tif
f1fbebc6c189b51fec93dfb00401e2f1
fe0916e1547953a1d9e5d7b0e09e6ff5238b4d5b
F20101124_AACKNJ burke_g_Page_073.tif
08e936dad74f0ea8bc4b6871611e8ed3
5830595beff7479114e20f6f2e8939301ff4d0ee
F20101124_AACKMU burke_g_Page_053.tif
86cb7c751b3d66391d21c1a16396abd6
c626e003e7b17a20806a318b3ec2b2edbb9fbc5b
F20101124_AACKNK burke_g_Page_074.tif
8bf62eb43f771fe600c160ed36616b6c
41ce302de9f213e422da8fc7681ebcbcd0397d2f
F20101124_AACKMV burke_g_Page_054.tif
58a9c79e83fd7bc6b0eab1fc1d178190
7fedcc0bd12efa073519c5cf65e6552f65451d43
F20101124_AACKMW burke_g_Page_055.tif
901c79dfcc796d97ca18d9336f8a0b89
74a87c7134e65f790fca060724cf7b93634a56c1
F20101124_AACKNL burke_g_Page_075.tif
4c6349331b26b074c8cc4a204779b049
3501a2d068f71fb65552083bb36f74a147b78045
F20101124_AACKMX burke_g_Page_057.tif
87f4cf0788050b882f610564e5e40ef5
f23a1328a1c65f077efa12b646ce0ac19c08fd55
F20101124_AACKOA burke_g_Page_104.tif
b2dbd5f7554c33799238f4af027a7031
732b9f67e1f0b251b685c5a5c502e97004cd15f0
F20101124_AACKNM burke_g_Page_077.tif
9ad9aa09d36a915ced7210ecb1adacea
8a362f1ac3a9478777fbac81582bb48252577ef7
F20101124_AACKMY burke_g_Page_058.tif
2532319bdcf533583d8b5707736ff0ac
44c502af86a33f82c507cc9e3ae09308e13d8b86
F20101124_AACKOB burke_g_Page_105.tif
d99c64b4e50b6ff85ac8a0e89427ecd8
db0a693653c590d679bba7a130494bc5b347318b
F20101124_AACKNN burke_g_Page_080.tif
5ba238425b298081253103fa3eec9a24
0552463c4e1a88a2c50bdd73e07a3d906e5da22d
F20101124_AACKMZ burke_g_Page_059.tif
00ccd2361d332d7b14be046e9998b4a7
9d448b96ba4a3e281967241d0c7d5ab269beba85
F20101124_AACKOC burke_g_Page_107.tif
1400252da45f38d0b725d896040fb821
4679c6a4c11350dde8fbf4ce57e55edcb136a605
F20101124_AACKNO burke_g_Page_081.tif
1d63cc46f4b65098183368f2236123c2
2244db23ebf418918150bbc1ac8ed6e3b9406d81
F20101124_AACKOD burke_g_Page_109.tif
f5fc8d51642928d296fc4dc4dab9fdcd
918bc87000eb4266f36f451e6d9d0c8fb0b7d2b5
F20101124_AACKNP burke_g_Page_087.tif
030fb95901edc1496e8617876bc43a32
a9751ca210238346e2cf8c413cafe3602db3f4ff
F20101124_AACKOE burke_g_Page_110.tif
7dc5934dbab3ad5827638ea210eac701
1d397930eef7c05567cfc14d5a96ec9fbda68ace
F20101124_AACKNQ burke_g_Page_090.tif
7ca2f8c1b064a3ba506d9ccde81b6040
39dc48c6ee35ce8f0fea91d6e429405e871786ab
F20101124_AACKOF burke_g_Page_111.tif
80bee2a1e6ae88640df64b8ed3ff7812
8b68be8fa32345028fd9fb2ac02333f8551123f9
F20101124_AACKNR burke_g_Page_091.tif
a42fa93e31039bf248f24ac7b3fb693e
8fd46c7e848c068a6f25ee55d8d48d87661c8d3e
F20101124_AACKOG burke_g_Page_112.tif
dd97e7509c47f18149f6acc3dc70de30
9d3f16af2f0008f14ae6a3d7823f0d55a2236ecf
F20101124_AACKNS burke_g_Page_093.tif
46a7bf5d7f45b267f448ded1d464dbb8
6de9cd28e3ca04366cd99c346d509c591c6c35a8
F20101124_AACKOH burke_g_Page_113.tif
c256bd1832c8d00533a2eafd337beb20
dad3eff78b59b4b16bd4f608746dcfe70a80027c
F20101124_AACKOI burke_g_Page_114.tif
71b96b14251d4f9b90b4a32bd0eefb5e
c40f0ba1c5fc3afa0564c128c1b8b6571adfee3e
F20101124_AACKNT burke_g_Page_094.tif
dae87e1ed3fc03cc69a409994619f84b
451bebb82c88e53545f10d4af5a60eecd9c10a38
F20101124_AACKOJ burke_g_Page_117.tif
51e904990cc9483537b830ad9fa8417e
40f87ea9e67549614bd39dffc1ae20cf722d52e2
F20101124_AACKNU burke_g_Page_095.tif
a2de53f062092734fa3126919d8bad71
da611d5884b85abf758564756f7072f10d473fea
F20101124_AACKOK burke_g_Page_120.tif
a179d62b58d63ea2eb487b32d4af19a5
86fe7c479d4fb0ad2ca9bbfea693c15ae491faa9
F20101124_AACKNV burke_g_Page_096.tif
347b1cc632b2900161181afe8bc56b89
6831cee26400a65f9cc1231132e1196f3386d418
F20101124_AACKOL burke_g_Page_123.tif
a70621f1b77f838ae895d5ea3d316343
672c7e3b40f1b825f548118f6e18fa49cad099dc
F20101124_AACKNW burke_g_Page_097.tif
7a614f81dd6d534c810b6cebfce75f8e
9c1c2848d420a30905310c741c176e6603ba16e4
F20101124_AACKPA burke_g_Page_142.tif
48d05883fe7a88070d971b0e65e8a4db
e8dae700ba2bf578d0422e7d6564af8c011b9aef
F20101124_AACKNX burke_g_Page_099.tif
77c525eb89482536f9cbc4cb30eb9323
355ec4166adec324e5f7e1da7d20ac5758b99b59
F20101124_AACKPB burke_g_Page_143.tif
b10a4a3e4dc50a3630e49c96cbc6a044
e8f59cdb037763155ccc43a47267f2c0a884128b
F20101124_AACKOM burke_g_Page_124.tif
d6dcd75c6fd203f5193c139d60313bc5
cd64636cc2b9c99b2c4d2019dc540e4a36ec7420
F20101124_AACKNY burke_g_Page_100.tif
52b8f103b8ca074c1e4c5ae0cf7306ca
955ff931af8e6fb10b3a25e9ab940233ae8be7a7
F20101124_AACKPC burke_g_Page_145.tif
7c643a5da8e4a1ece2df602a479659d8
045f6aacaa8b15fa7642e51a4b82a0d6a0269c6e
F20101124_AACKON burke_g_Page_125.tif
1341abb05e12c368a32bd4b4b16fc0c6
3ef51db58781200843095f618189f106765e6a1d
F20101124_AACKNZ burke_g_Page_103.tif
bf66f516b9323e46487c348839967a1b
be87669b415acc675c41fe3187fa42d922046a8f
F20101124_AACKPD burke_g_Page_147.tif
247b1f95a471df9f835fbf617893365c
a9168f1618bba0f340174635fa44ea50831f8799
F20101124_AACKOO burke_g_Page_126.tif
b129b8c773eb6f8ea7a1847d2f737959
838faf6afaae81bbbb0e6c3e657bf81998e30f19
F20101124_AACKPE burke_g_Page_148.tif
cb544e9110863555c7920537e71b101c
fd395ed7c58fbf34713691f2e33b764633d44d86
F20101124_AACKOP burke_g_Page_128.tif
4bf42dbf679a41c5e219ed94ab5c5ef2
51641499f00ca300807b6dfc5c34c00e4b70d1e5
F20101124_AACKPF burke_g_Page_150.tif
e84bbedf80f767281a5316b5dc90e184
c48c98462893cec0b170bae62dea1291a19f6e21
F20101124_AACKOQ burke_g_Page_129.tif
c148a64c32199c39bd7116b0020e79ae
8a3869b02bd9cbf29b7b48e236f86a0798c51a73
F20101124_AACKPG burke_g_Page_153.tif
7dfd0aa28f22fcdd82807749f3e2cae8
6eb19bc5fe7a43483ed8ca42cf3124613f6b1880
F20101124_AACKOR burke_g_Page_130.tif
41c6eb3618926c25514a3e367e401667
3edebd17ea74497c6c3cb840f1fda770b3683bab
F20101124_AACKPH burke_g_Page_154.tif
746a829d5740fb29d128888f1339cb29
f99e76680d1d32c501cbc394228fd024b132f7a8
F20101124_AACKOS burke_g_Page_131.tif
e5dce5779040ab602b253f4777718c12
e81d865cde7c01c51080079024ba311cd1523a14
F20101124_AACKPI burke_g_Page_156.tif
edec7b14e4a73e3adf517bff12b318ed
e9d9c8011c71f86121efb107715c28face7dbdd2
F20101124_AACKOT burke_g_Page_132.tif
4922a675fb2e05845630bcfea16e143f
06f9a8aa0db241488236d7ec40884353d63c881d
F20101124_AACKPJ burke_g_Page_157.tif
5786028d1fc50429a6389b28dc5c763a
09ab2a6bdbe73c9ea08c73168136af76fb3a81ac
F20101124_AACKOU burke_g_Page_133.tif
5558d9ba010b15c2b6f2d62f08c656a5
f05a7fdcbac9b7c537a120df9f9801fb43d33253
F20101124_AACKPK burke_g_Page_158.tif
a7f25618101f3ea39d0ab5268080434b
69570ee30aab30a7b99c25ade1c402d13df3b159
F20101124_AACKOV burke_g_Page_134.tif
b748740772a513485ebb07a86f6b8ff3
520b50c0ca5a395abd20cc0c770c13020763ead9
F20101124_AACKPL burke_g_Page_159.tif
d8ad93862d560a1e128ab5f606f94278
8d66613bd6329d7169aacd5273efba088543a7e2
F20101124_AACKOW burke_g_Page_135.tif
76ad116a9959056be4040a00ffe27522
859d7717b648435436c3a73463541744fdce55af
F20101124_AACKPM burke_g_Page_160.tif
23e3d5901fb22b18e2a7d392d20d5479
f2fe7bc07b769b2323c287544586c9874f7354e7
F20101124_AACKOX burke_g_Page_136.tif
f9c049c76caa19b18071ccedd9a704ae
c56e2d88ec2014d39336b173a32f2eb6fa3aeeb2
53671 F20101124_AACKQA burke_g_Page_014.pro
879a446c38f2ea41a8d30529e4dfb373
73f6fa184cdcc6f808fc309c8967fd469097cc27
F20101124_AACKOY burke_g_Page_138.tif
c42ccb053660c59953bf26b3e01d8ec0
a10d527ff44c5315f9ca112d5a568a3780a15e4c
55253 F20101124_AACKQB burke_g_Page_015.pro
225f6e08ad92159ec9c2f9e6d585dfa9
3084285476ca327a9afbc76303acb27d49ddfd59
F20101124_AACKPN burke_g_Page_161.tif
54dddbe5a75cb638ae3ce0820facd4fb
b973513d0d5577b71a07ec623ea1f861167b5d0f
F20101124_AACKOZ burke_g_Page_139.tif
fb15da3a91e41e5d317f1e7ea1eae434
920151c51925a9791ad2ec1a0f50b061790381dd
5763 F20101124_AACJNA burke_g_Page_029thm.jpg
286d166708b48f50123ade3382809170
0a8fdd6e355186e99c4c7dc191c56f57c3be447c
52650 F20101124_AACKQC burke_g_Page_016.pro
89040ab3154722a3a7ae72b4f0bbd381
23879b40eee7ba49a25a38c35562cdaa03b8865b
F20101124_AACKPO burke_g_Page_162.tif
c7a4dd1b137b62c5409a361832d066e6
86fdec6eb0b19602e6343fc13043b39282cbcfd3
48958 F20101124_AACJNB burke_g_Page_160.pro
0c78ed6bfa42b47dd329d2d58e25c0de
fde0efc1d64e03d0eab4ce9e2fc403e0a7cdc089
54181 F20101124_AACKQD burke_g_Page_017.pro
1bbed8124767d6ef56cb5fb004f7ca91
433d358943fa48912ba4d4ef95047be24e1a2a0c
F20101124_AACKPP burke_g_Page_163.tif
59a8c1208d7993ac9e65690cc59168e8
6384fc89fb12b09167007b196f3cbccaefea2acd
2106 F20101124_AACJNC burke_g_Page_106.txt
102a1074f96703cb71c97093d9d41621
c31fdc077716bcbca6a40c26a8d615654d6831ab
54184 F20101124_AACKQE burke_g_Page_018.pro
75144dd3041a3fd20cde00932717eb33
57035f5d7b92f24b9eb6c0361a02b4353780fa36
F20101124_AACKPQ burke_g_Page_164.tif
d96b7ee658d61c392e8d27de895b2e45
59a79da29aa50448303d2a081e954b7b0edd36c8
126145 F20101124_AACJND burke_g_Page_070.jp2
778a5fe31933e330af174358b3fe2e9c
2630ddca7e5fda6f9d74bf52ab0e66699a998a67
53601 F20101124_AACKQF burke_g_Page_019.pro
6b54e4b45cedfd20934eae342a4fc78e
9601d8eb5dd1e34fedcfbf3fed0160afc0a9f972
F20101124_AACKPR burke_g_Page_166.tif
f4badfd296a204a09227d04db17e76e2
802bfb6f69d96651a8dde29c7b651f73e7ee7785
9296 F20101124_AACJNE burke_g_Page_129.QC.jpg
35944e07f719ab480607db1252ead029
65d01a9d1b900e767732f9900ea94b88e1af64d3
54492 F20101124_AACKQG burke_g_Page_020.pro
757fe6ffca6c35cc1b24c09feaff101e
3fb4b286261af047bc349a4d0cbb9a00bf709f83
1381 F20101124_AACKPS burke_g_Page_002.pro
983a9f64a16c6333e493b61ad9ab805f
03081c1e5ee158af670473e691e93762ae14a6df
2143 F20101124_AACJNF burke_g_Page_018.txt
a99148bacec24e4534c77e7b534b2681
bf03ae42627e18dbb9ca0a825386785137c8689c
53935 F20101124_AACKQH burke_g_Page_021.pro
ee9465feb9b493534a5787045eed461b
be0ea2655188a7a255f005baacee628eaaa9dcc2
76498 F20101124_AACJMR burke_g_Page_033.jpg
200877de121a5eeb7949a7c2630e0694
7d606e878d413d5eccea2346adbcdecb48ffb9b2
45151 F20101124_AACKPT burke_g_Page_004.pro
218bf5aa0839acfb538921bdc5f7dfc1
2078753027c64ddd9c13e3e6d54fae230e117811
40190 F20101124_AACJNG burke_g_Page_151.pro
97200609535db3601a6b5c4fdb7be1d1
2a0fb6a42f92647c5ca297e27634f47f6dac4980
52786 F20101124_AACKQI burke_g_Page_024.pro
90b00ebdfed1a4391daac3c728a93111
5a281762f716f4c61eb477af969fa2b715a0e9d2
4930 F20101124_AACJMS burke_g_Page_071thm.jpg
e6acd0c902d8dccb2c441f1f5af70b51
1a8bd24bf74e8556b6652bd16ff226caa6df4637
11306 F20101124_AACKPU burke_g_Page_005.pro
78dc2e2e3fb6941a9205a4d824ab9401
f2c0b959c839565969f4ccf2c5d7015f92632dad
56139 F20101124_AACJNH burke_g_Page_113.jpg
ece3546495165bde878e5da1bb45cc46
703c99c58433663b3bc9565b40aa0cac6a566ffe
54495 F20101124_AACKQJ burke_g_Page_025.pro
fed5d100f23dd38c073ed7c8b92fbbe1
5c2bce1b73c70e0b28a1f00290a45f46e67d1bf9
62343 F20101124_AACJMT burke_g_Page_078.jpg
2cddc1c70e6aa1ce6750f3ded18b4c98
266d091c65d184a3fee7d15366930accb53cba90
49827 F20101124_AACKPV burke_g_Page_006.pro
a04545b8bd3023d348cb77108b5d0b54
e53aa3fdcaca51425b396ad6bf63e5fc7db8f891
109733 F20101124_AACJNI burke_g_Page_057.jp2
8d7d3ef12056349ca0813aa6b7d73037
f1d4d9cc9bcded50551a09e43beeacdf0a036778
54737 F20101124_AACKQK burke_g_Page_029.pro
3237844f213c70e69dfbac658d1132db
443690d99b95965e6dc6b33c5b5fa5d9422f7ad6
23363 F20101124_AACJMU burke_g_Page_058.QC.jpg
1a842670f50075bbe64b3160b1d3c23e
6b24d11916402fa6fe13e22c997d7a419d82e118
46636 F20101124_AACKPW burke_g_Page_007.pro
73a243e5a73185f16562cc8655c645ea
397668411289e49632635791bd534c60669ed17c
100669 F20101124_AACJNJ burke_g_Page_079.jp2
e10319dd195e4c21c6dc44ed376d27c4
574c2a3dfe40f540007d5b9253cc2721c6b5e3ae
56029 F20101124_AACKQL burke_g_Page_031.pro
a7d05b826e167f5a553f535096cb6007
aebf2cb17793f9a31c01c5f8d41e37b4873fc48d
56311 F20101124_AACKRA burke_g_Page_055.pro
99b8f91f0903ece72a0ef1ee5f2e6912
6b1a3c8a659435dc5349b1470ad8b07faab35ef2
F20101124_AACJMV burke_g_Page_115.tif
dd9b1ee77500649e7a4b580d9a0026b3
3ce4eae15a23fb07ccfc880e093a947a73f0c0e0
35398 F20101124_AACKPX burke_g_Page_008.pro
d631075689d91e339f718b4465282de1
9743e0055be2b1ee44e2a76cfb11471f3ed6816b
F20101124_AACJNK burke_g_Page_122.tif
3d815b2c4abbf8874c911e05048aa2da
8eed04469c5ea91649828d02ee0c4c6f3326e382
56908 F20101124_AACKQM burke_g_Page_033.pro
a08444f9f4fc83ee283bcb7a67b7bc4c
002948f414a9b97e393d115e70b8ba428978830f
52299 F20101124_AACKRB burke_g_Page_056.pro
d89e7064486c86e077b9fcb20273f6a6
8931c3dad6387dfdd8b6446ae7aa5a995f2233d9
1051916 F20101124_AACJMW burke_g_Page_084.jp2
45548fc6287886df881540ae2bfba1e7
62a30bcc2f003504328e624d1dbd9125e7ba9e64
47297 F20101124_AACKPY burke_g_Page_009.pro
729c62dd2a6420545061119c56467127
9744c1de51a29b9813b9f6d4af0047222f178080
20878 F20101124_AACJNL burke_g_Page_080.QC.jpg
4f7c7916a52097b22518b8080ad5960b
dac73395f35bf24a7b1ad14c182ad7240f5b9a96
55931 F20101124_AACKQN burke_g_Page_034.pro
a8f5ac780cde09f246b4e0b1e3cdacdd
981416944c17ab905a73b0a12caafe8906ded8a1
53560 F20101124_AACKRC burke_g_Page_057.pro
6bea70feee1bb99d7d986164d97e7c84
a1a07f77c2847032b60be4781fc7c021a951aa19
60608 F20101124_AACJMX burke_g_Page_068.jpg
48dd779c8ae51f4d12900fbccd928a2b
d4e24adb950fc447b1188254d15efd5556a4ca1d
43151 F20101124_AACKPZ burke_g_Page_011.pro
bb367bdb0400cbec632fa484d19a1f2f
caa1ddd68627fd4f710ab5eab8a102547cc25363
1855 F20101124_AACJOA burke_g_Page_149.txt
bc4a1b4986d83f199257886d8e04930b
72e593e0cb3fb8b1fa7dc9e4628c2b0976b3a4f9
56785 F20101124_AACKRD burke_g_Page_058.pro
3cc07370625f16b035723c19a62a1fbf
88db73411120699d3c7639826970e06dfb455d16
52848 F20101124_AACJMY burke_g_Page_089.jpg
769f3743dc0eecff85c875efeb929ece
a976430322e835b62aad936d1ad26e9cd0037636
70670 F20101124_AACJOB burke_g_Page_095.jpg
eb27b6fdcfa9ba4c173f7a024b5ecca1
091ad41b9e29cabe9eb44771241aa5bb9183aaba
F20101124_AACJNM burke_g_Page_030.tif
a7872ccdede2f56e2252f74576c889a3
62a812fc1c265e6f4f0939703c98954f0bdf97fc
54840 F20101124_AACKQO burke_g_Page_035.pro
d3fdbfbc21c0ecee356519cb4bc7859c
51e6c702fc551786dc22d1272b4b495c2144dd2a
47102 F20101124_AACKRE burke_g_Page_063.pro
11162ba777a3d2a579759602160ae9c4
74fa7a1c0e3dc2c29e9443cfbace4fc3f8b2be4a
2704 F20101124_AACJMZ burke_g_Page_138thm.jpg
d392487e17a349ffccd55ab0fc0738f7
9d0c2a911c766b0fab598dcbec3c0ac24be9f2fc
53989 F20101124_AACJOC burke_g_Page_030.pro
45eeb8573ffdc2d25091d61245a720fc
92d04fd985751750c40a555a5651c2d838f4b301
2175 F20101124_AACJNN burke_g_Page_029.txt
c68fed2d27e87a575935246da9fcf609
1d569990ff7142f868814db0e94df78817392bf9
56472 F20101124_AACKQP burke_g_Page_036.pro
2f95ab9754bb3a234283537482dee26d
46686b1d721b53bf435904bbcd15e8bbe9ba2c93
38089 F20101124_AACKRF burke_g_Page_064.pro
8e812a744641f70e6fca4a6fd6f340c4
eb2181d17ef92ddbac1bea116c2616b3ce213eb3
13547 F20101124_AACJOD burke_g_Page_123.QC.jpg
8033b3f08d9c045432b820ba9800f49d
10c1be2606dce5477e27ce39dcc8dfef25abb7db
76252 F20101124_AACJNO burke_g_Page_146.jpg
9135e3f221cc5fed3106beb5547057ff
3b7fd15d81750f6069acb1609074d4b2311f3960
54936 F20101124_AACKQQ burke_g_Page_037.pro
44cf7b19b63ae31e9ad32fa6ccaf5f0b
c48cf1bb70d8ad4a7c544c805796bebffbdbc70f
48553 F20101124_AACKRG burke_g_Page_066.pro
d824e1c8bb784698b64d7d4b29f7d6c4
6f59cd883f2ec0b831dc00b80f1a6f6219397ec7
59238 F20101124_AACJOE burke_g_Page_084.pro
f00e1b5dd24b72df5eb861504aa180d9
104bf6d6b8628f7931d3c82c0e974a72dbc5865b
24171 F20101124_AACJNP burke_g_Page_151.QC.jpg
d64f483a1ebdedb45e331cb9376b93e6
f4dc7d51bb16085ed707c5aeaf6670910cea8d2b
52193 F20101124_AACKQR burke_g_Page_040.pro
67e6085f855011f66c4f272a2595c223
767f5ed2ab72a2147de9a519093c5557a995a531
47639 F20101124_AACKRH burke_g_Page_067.pro
faee43d7e8c6cc36d99e5677a167b3ec
7d39351a88adf0a320a949793441bd1252d24a59
2404 F20101124_AACJOF burke_g_Page_159thm.jpg
d3180a722729748ea588cc2513760198
6518b36d0a3b5a887ff6f929a54d3146edaeb3d6
114774 F20101124_AACJNQ burke_g_Page_020.jp2
719e266e6d4ed3c29073e273f079d163
dc7063f6a11a5ad49d919f925ec5f0067a205837
34378 F20101124_AACKQS burke_g_Page_042.pro
fcb7b7b4e52c09161703fd8126b55b0a
7a0f088a83fb4cb0a3638c1eed63d3550ffb413f
42347 F20101124_AACKRI burke_g_Page_069.pro
b058497d20b5266d8d4ea06622cfcdc8
c6c0798fac06fab1f6708755e8ccba6dcead98cf
19551 F20101124_AACJOG burke_g_Page_049.QC.jpg
54fcf7e065040fc54a5e108488b185a8
0d566cbb77986059b2458ed1c39df4cd699311bc
97515 F20101124_AACJNR burke_g_Page_080.jp2
af7241207a69695b290ec91f3448905a
490329b73a7af898d556f49756b1b2ec27ab58f9
40610 F20101124_AACKQT burke_g_Page_043.pro
0c3e8f05b55190aec7dc8909afa74079
eb8d575fd38a4ea670066649aa076d3bdff8d09b
39500 F20101124_AACKRJ burke_g_Page_071.pro
278c46055135028639c41e9bde38b568
b62ffe897ced16390770ab78f04b71146025f1eb
35998 F20101124_AACJOH burke_g_Page_157.pro
64e8e83d4516f15b497810b2bc05628b
ebf29927a784a9df46f129a4001cfd28ba924bf4
F20101124_AACJNS burke_g_Page_108.tif
3027dfc679adc1ec5c7e8b67bdb83269
c73aad0c5a0960709a084b8f81dada3187eff886
44092 F20101124_AACKQU burke_g_Page_045.pro
ce6fa56b06723b90984fc0d02244573a
2873fdba7caa5a0fe4b91c383a555b62445ff2ea
62086 F20101124_AACKRK burke_g_Page_073.pro
1c79c292b60b719fc7df631391a0ea25
5ec732b0d42e82d244ce8bc3983c18c4f8f0867a
F20101124_AACJOI burke_g_Page_092.tif
6cf53d045522a6f2fe2e7c859e84b74a
df2917ec993f9080479ea9853b9fe217d6aad066
F20101124_AACJNT burke_g_Page_121.tif
86171f21eb053ece0156b0bf185b74c1
5ab6689e2de24dd4836ccf237f611203bc767509
46783 F20101124_AACKQV burke_g_Page_047.pro
df46a31f0416f624b1f7f5279b093621
7695c283e4f27dc8a88ed37315067e9a4862b0b0
50039 F20101124_AACKRL burke_g_Page_075.pro
8c8542c918c173f9623bb31cd15a686d
cf87848f57d285da217b1112552719f39c6f921f
37989 F20101124_AACJOJ burke_g_Page_116.pro
cabcd203573f5468b47154b94b95f866
833661ad6e33b972d166329307ef90ab21efd055
1160 F20101124_AACJNU burke_g_Page_012.txt
d7836d2cfcfcbf616201141dca756028
e53bcca19469238889d63606249218e9d34ea004
46331 F20101124_AACKQW burke_g_Page_048.pro
30e1b8cdf69ffb4545f726184555fc88
3adecbb3677b13343f339c05d2125c3a71c2d166
44424 F20101124_AACKSA burke_g_Page_097.pro
d63dbde79c032d2b0eba1b246957ad69
e1bfb4c9a8d4fefca83dc4c0b962cf5f748d04bf
55017 F20101124_AACKRM burke_g_Page_076.pro
6b88bc79b9da41d35743819726f0d16c
8f24abed9ea06cc44dc36ce4dcfa8e9a6cee6412
47681 F20101124_AACKQX burke_g_Page_049.pro
40fecf845c813f0e935962e4e14bac0a
9096d4c637babf689e113727f1ea28f80cf0dd05
799832 F20101124_AACJOK burke_g_Page_116.jp2
e6d4c93c64719930ba828b84bc6a0fe9
dd12d3d802ccaee6f2fb285c6166383b3b15facd
F20101124_AACJNV burke_g_Page_086.tif
fc6b6f7a6b3b83818b1400348b7041e1
06ec1e70ff6d933119bea7e13290ed57e18063a0
51271 F20101124_AACKSB burke_g_Page_098.pro
45f58c526944cde02bc677c214e065dc
6e790f59b00c9264d1f3a6838d130a8b1d5821ed
50875 F20101124_AACKRN burke_g_Page_077.pro
df72d51e9cd5f0ad3adba6c93bb096fc
5d659ad01008a16d98f741e46a7ecb3b6d16c958
43725 F20101124_AACKQY burke_g_Page_050.pro
bd001801382323ba81a5443c0cc3fe11
e0b8607654fa79f53d1774dd66c01efa2954dcca
9201 F20101124_AACJOL burke_g_Page_126.pro
6fa59aeae89638c65724c82bc4c91893
6115db163e87ec00be728d643c2a2b8e0b5f2ab5
33140 F20101124_AACJNW burke_g_Page_027.pro
3b4a198758966445d2b25881a4c142ef
ec749fba7db3f5ced56970afad523382ba72ac0a
54506 F20101124_AACKSC burke_g_Page_099.pro
6b17fe0c053e3891de253b418fb6baf3
a87cbaa7f9ed1ad99d484bd06f998413d8c7dd17
49499 F20101124_AACKRO burke_g_Page_079.pro
fdf45d2829320ed459dbee38bd65c07d
0299534da7e6a4ebc419df99bd35a23bb07d2f45
56863 F20101124_AACKQZ burke_g_Page_054.pro
fc13d82519138ecebd1c90d2ec155fa7
93d098f4ea42978c9c9ed139e715e6f95074c717
16564 F20101124_AACJPA burke_g_Page_064.QC.jpg
24ee154b9d21690528a26f51679933ca
609375026c07f62bbaa2774cc12ddd5673b80583
23646 F20101124_AACJOM burke_g_Page_107.QC.jpg
2836765a461c7a0a26e12c7296423e01
b182fb2ba24337cb633c2df59e43e5e7847cbc1d
5845 F20101124_AACJNX burke_g_Page_083thm.jpg
c3dc8fc9a7437ae250857be804ccced8
ced4284973591b0c902a7444ce151934283c3021
62130 F20101124_AACKSD burke_g_Page_100.pro
017e40d05c1c74a466c79719e47bddd1
b6613f2619445dab9b304c36d261d9937ee4cf2e
489535 F20101124_AACJPB burke_g_Page_112.jp2
7e2fb0240558376480a45ef89d59a8c4
10e903b49e9fcb99d1a3d55d28e4b184858bb8ef
1794 F20101124_AACJNY burke_g_Page_097.txt
d90f02472bf1d40366e5e802e00c7054
4481d7bbb00d199e5900eda74720aeccf2cc4a31
54913 F20101124_AACKSE burke_g_Page_101.pro
35e391a65794ec15497f0ba237be9a65
f85780f7d5f98e4a73fd3a9592d24ad379641625
21599 F20101124_AACKRP burke_g_Page_081.pro
f24312a42f1da9f8b4f98a426559aadb
b3f3fbbde49f79a70f19e24876aea8a7ac07a67a
55335 F20101124_AACJPC burke_g_Page_166.jpg
1ca0f9df483250f82cc8e80ee77605a2
68d28d2742287d827471230558137a1d15ce3614
2120 F20101124_AACJON burke_g_Page_021.txt
d9b985bd511bec94999d2729ed8d039b
a8cb192357e62be54a95b4ff38c29797d980f68a
5454 F20101124_AACJNZ burke_g_Page_086thm.jpg
2f1b618abb55757746a634628b4397c3
024c6680eb498110288b59c96b0d5abf2c88c6c9
52396 F20101124_AACKSF burke_g_Page_102.pro
8f582ab649b439d59d7815b97261abca
b28640200c2e9582f69e96ae6830772249dc6352
49716 F20101124_AACKRQ burke_g_Page_082.pro
f5bd83899673faebe2f737f5bb472a5c
dc49a3a204b3e37ddf6d2a414c830ad25171461f
F20101124_AACJPD burke_g_Page_078.tif
41ffe1031340a7b09b2b885fd4af55fa
fca128ade7b86cd335bfa0e0fed5f0aab55a0f42
55545 F20101124_AACJOO burke_g_Page_022.pro
91be2176160c47f78fdbea5bf33a660c
9c506a1d39dd39728d759625b286ab6fcb503dbc
56025 F20101124_AACKSG burke_g_Page_104.pro
0ea1a7509cf53ed5401fe9b987243079
28c285b48c8f001002422f8899e2740257b3e4da
53997 F20101124_AACKRR burke_g_Page_083.pro
47b0e7df6fc7f0d07952f28dc95aa559
73d678d93e2f9181d898465e5c8c5547638c6202
10705 F20101124_AACJPE burke_g_Page_129.pro
ecde1618ab99cd82bb1b29765997d224
657cfb77faaedc21d20e847376345591584a9fb9
6169 F20101124_AACJOP burke_g_Page_128thm.jpg
8f5274dea22129ca9a5f2e5d88e7136a
4e29858de7471c8054aa642932d073a249fca8d8
51325 F20101124_AACKSH burke_g_Page_106.pro
b14be00c87f408e6354e8b23718a25f3
5b150da64acf42a95d0670ca32f3d5ad74ef6d11
58992 F20101124_AACKRS burke_g_Page_085.pro
db1acdffe679ed919fba8c6b6339e05a
7ba722d96a374f596ce80c8c452308390bb2f791
F20101124_AACJPF burke_g_Page_089.tif
cc946eae9e26c712f1869309148cda4e
dfb63acc4454e516a97a2a795c5765529047be3e
39509 F20101124_AACJOQ burke_g_Page_134.pro
a1be362f9f5bc849db617aa228a95079
940578aed09380dd5bbc6dda64b95e08227aaf84
57449 F20101124_AACKSI burke_g_Page_107.pro
ba1ed3eec29b2192afc0080a1b13286d
080e01a05526063dd27e1abdb47e3a76125577a9
43666 F20101124_AACKRT burke_g_Page_087.pro
ac2e5b8865007532fcbb079b3bb88427
8d232faf1c51e227523d2e62ae2378543118e787
1910 F20101124_AACJOR burke_g_Page_011.txt
809bf3171a921e8c9e8dfde6d40bd940
136f9d237e603c6d635d4dcccdd2a6f04c9316e9
F20101124_AACJPG burke_g_Page_011.tif
b6ffb31e37c477657ed782293ac6bd98
202c96dc68b60d181224b63c20657f9c4fee9371
57147 F20101124_AACKSJ burke_g_Page_108.pro
16b96173f90f760b55f004589e59b8a4
058ec2d9ce9e90df06b86ce9e6f5a6ac98ae0ba7
31248 F20101124_AACKRU burke_g_Page_088.pro
773093bd4d11464deb7070959beb0b82
f1de1b3f20539fe4d727e2f2a754e04694a52ce7
6124 F20101124_AACJOS burke_g_Page_085thm.jpg
454166e08741997c1e33b00d5c85f22d
34e18e07555813a17c07e89f863047cd81eaaa70
39899 F20101124_AACJPH burke_g_Page_124.pro
14619ed9293a85e86a592fb4ab986516
5923efffff1aae9630db8ae433aa434d8d90920d
53642 F20101124_AACKSK burke_g_Page_109.pro
9db859fa24d7fd892967e482c20357ae
ac050a26cf8f35d7f41d15023f0d6cdd4cc613eb
41717 F20101124_AACKRV burke_g_Page_089.pro
69b85e239d7579b68a7248939ec80dcf
bdb344c0b37c42bd377f92b66fa9bbc8d34998cc
115408 F20101124_AACJOT burke_g_Page_034.jp2
4ae3087f99a36fda09c4e4795a9b00df
51694e17f9e34a61ab72ae7946a065ea56f66f6f
F20101124_AACJPI burke_g_Page_165.tif
58f6e81cd68b4c74cb757b34e7f669cf
016a588fdd467692c8dc58ccc2744e359da7d240
55202 F20101124_AACKSL burke_g_Page_110.pro
2d68bbaa188701f71bf146f41a9ca146
325a16a312d90733eef0ce071b0c2bf119d1f655
53886 F20101124_AACKRW burke_g_Page_090.pro
9b5c5a37d77b63d7caf1d08d69b1c756
86fa9bec946b0b9db2e55347739be7a6e08ebb82
38054 F20101124_AACJOU burke_g_Page_044.pro
075823c4525e2412f5073403149af958
92185d0e4a6469edfa6f2adf301ce19560ba4dee
F20101124_AACJPJ burke_g_Page_029.tif
ab92f84dbf10d3ab7baf52d7a516370d
dd46d949f2c8ed3a508917fe17fed03f1a4204fd
12104 F20101124_AACKTA burke_g_Page_135.pro
b47eb05a085fb4b31a835c3de4ab3cb6
4114f642b599a61e73097d2db82b32c7a8b4546a
23310 F20101124_AACKSM burke_g_Page_112.pro
320bb87643f337b20ffccebcb9924cfb
e2da0bb5996bcd61b97487fc0cdfaccc657a9640
43703 F20101124_AACKRX burke_g_Page_092.pro
ff4e55dfd0626dbe6e7ee8451b81f110
257d75771e094c17082a2d9161965f12b042c7a4
F20101124_AACJOV burke_g_Page_106.tif
16f95986f15cdc5c9da154f42c2ce181
e0a998409d0cbfa841e4139d633fa232a55aa0a6
21755 F20101124_AACJPK burke_g_Page_016.QC.jpg
1a0bc1f428ee4dce2539f6b34203ed82
b19db4bb87eb916c245d10d1bae13dc58bc7c6ca
38034 F20101124_AACKTB burke_g_Page_136.pro
9b6ad39a14d8d75e3832ec7a04eb5b18
7bc7d80a0277d176f6c97107f93ac76a2aba8804
34941 F20101124_AACKSN burke_g_Page_114.pro
bcc09a02d0c9227eef9a9ed033ed2e13
58914b6c52c8ca5c70491248f4dca2e5f270585e
32827 F20101124_AACKRY burke_g_Page_093.pro
7a0f7c4dfca4a6e1097e083b08194992
500e86992aad26948877984d6600c5ae3051478c
6126 F20101124_AACJOW burke_g_Page_052thm.jpg
1c1764e68b28a542ff48c76627740a8a
74deafa6141dc54c295ed9f48d946cf600166493
F20101124_AACJPL burke_g_Page_119.tif
ccae685be32dedf67169caba26178161
058e8b3b6bb2c2c9db5562d9340212a3abf759b2
39890 F20101124_AACKTC burke_g_Page_137.pro
f0d2b87b11ae48da0e902d3ee662a993
08c72539d3eb239a6402429c5673fc23eb49b331
42505 F20101124_AACKSO burke_g_Page_115.pro
36718db57a03c212664ded6ec281028b
93eb4f57365534c6e42e9a570e0ea75345424326
47927 F20101124_AACKRZ burke_g_Page_094.pro
a3da3fb6788954c3b8a81bfd6a759072
b3e41fcda1e0533b5de56601a4e106c0e8abf707
61682 F20101124_AACJOX burke_g_Page_105.jp2
7df29af8c33dd3b7396697024f78070b
a75d68bc1a99d518b40a613753cb13671a73c664
5397 F20101124_AACJQA burke_g_Page_106thm.jpg
ac8f731d13277896a4dcf9cb9f7868fd
a8c6187d0f48a8d8edf7675fbfa581a2e43dff76
70806 F20101124_AACJPM burke_g_Page_048.jpg
a22f511117b4c1295637bb73e2d639e8
b953234c426b55de8f8bca810e59b010698fda7c
9571 F20101124_AACKTD burke_g_Page_138.pro
b42056e8286f190405addd34cb0ad8ac
fcc6ae6334667d3a1bcddaca85bd0918b71fdb0f
4767 F20101124_AACKSP burke_g_Page_117.pro
3d69adcd7afc2edd334d5efde0f086a1
f09ca2cda2ed3f1a0ffec8fe516c2c87f158874b
52719 F20101124_AACJOY burke_g_Page_062.pro
5c26382250ff5b27ece5dc93460541c5
1aca79b80dc7d612703f18960bd1984710f6fcbe
108350 F20101124_AACJQB burke_g_Page_062.jp2
1f4b652fe09973236cdb12379cd8892d
4e37475707ad27c90017b3f8ab2be54b46b03058
207 F20101124_AACJPN burke_g_Page_165.txt
774012749a4fabd1c5afa2d534605eaf
5622816cda0bb12ff4b196cdf9a0ad0a8263f700
15760 F20101124_AACKTE burke_g_Page_141.pro
268cacb2bcb3d995d48829c11a301025
020d62c45743f9b21f1843f096ec64a9ca8665df
80160 F20101124_AACJOZ burke_g_Page_152.jpg
d975d0900a141d3a4a82137005db8bce
971e720d0ff1b8a4c5beb70b676e72514b757cdd
58217 F20101124_AACJQC burke_g_Page_103.pro
74b68a6ab3a4ef982032dc71a67965e7
63dc0f418eb43d110a9f69b9927edf1f886266d8
39260 F20101124_AACKTF burke_g_Page_143.pro
2a4b7f87a4740074b833bd005fbea83a
8ce8a94304d90b6bbb0bf0e3c8ad5da3c8915e05
29525 F20101124_AACKSQ burke_g_Page_118.pro
439d48aaf7596f3fdb6831b854c7e1eb
be1d450179504bcae69a64abfcd4dfbf8b80ac22
5912 F20101124_AACJQD burke_g_Page_073thm.jpg
ad0af0508e01218a087f5019a9f09b3f
1be294b5b28eb266ef592a4d03c74f33f4338c74
1358 F20101124_AACJPO burke_g_Page_119.txt
3462c27bf08ff8ebbc6024e01c7e12f0
72538767f0ffe51b276a4470376528c09e4c43f9
10870 F20101124_AACKTG burke_g_Page_144.pro
438de45e2d96888dc6673a85cf87d19e
03328a6b16dd9089f294832e556a136006649de9
22383 F20101124_AACKSR burke_g_Page_119.pro
1b4cb0c084007e8c7b8e4313e649ff1a
28449a1e1093edc9c966eb6a443e8d618ee43287
29446 F20101124_AACJQE burke_g_Page_120.jpg
f1df025caec199c94c3358dd455150aa
f1731affe964febbe1f369e76f81e465b24e4831
22316 F20101124_AACJPP burke_g_Page_025.QC.jpg
b5261b8aab5ada31bb20fb141e8d1150
8f7422722fe2f6033a21cda27af274aca459a0f8
39921 F20101124_AACKTH burke_g_Page_145.pro
59d5652ef15be150663a05bb7b8258d5
cf0df122dd4e74512b22a7aa2ea270c8e0bde452
55887 F20101124_AACKSS burke_g_Page_121.pro
7d4046ea0cf0f90082594fa6d4d1198f
a537288ecfc753bd9b68b4f846103bd420e08d25
F20101124_AACJQF burke_g_Page_084.tif
a11bbafef9c4a4e326411aaaeba5ea8d
f729bdc5a7c834d61d3d0ac913c1a4d1763871d9
F20101124_AACJPQ burke_g_Page_010.tif
cb2604bfef84f1b0352f8410639df0f5
751b585f57e4406e9fbb1221982efe8fa4b62b7e
39508 F20101124_AACKTI burke_g_Page_148.pro
2cdbecfa117ad74026b889a86a051154
220498fd48c50522dfd658938f31778f74237710
54848 F20101124_AACKST burke_g_Page_122.pro
8448b5c210571e20c56059a5334f6757
411057d7cb4b1c4a9fde18106d10e2bcf8537de3
65998 F20101124_AACJQG burke_g_Page_070.pro
cdab86aafe1969d95d70169d7b835aaf
fd376fd52d4a4dfacd2a8b3a38b95aadc9065d9b
F20101124_AACJPR burke_g_Page_127.tif
d9e5d0184486ba7a4f5333105924933f
5b62a64db4b746b8f9c6e2189fc40208563118df
39850 F20101124_AACKTJ burke_g_Page_149.pro
4b27d5b226986decfd7d6fc39c8f5c46
26bdebdad2826962d65c39c4909bbdc3b599800d
29145 F20101124_AACKSU burke_g_Page_123.pro
b9120b3725692a595f908cf1b9d404c4
d970caecc93b7361e127b38c8b33775d1c3ceb85
F20101124_AACJQH burke_g_Page_082.tif
09118f33ebe43c25c01519ba232ef53a
b7695c9ed90242e577892433d279f23874125ccf
7274 F20101124_AACJPS burke_g_Page_001.pro
9c3e636cf7fc5c03dfbada36c1b4abf2
ab4a898540188e17626fda7c9f56014ecc0a197c
40525 F20101124_AACKTK burke_g_Page_152.pro
102f4020864e934dcb9a1dd258dabcd0
8bc3fca26ef7eb54d0513fc2b2900f5b77893c58
40220 F20101124_AACKSV burke_g_Page_125.pro
26ecdd1fd8e5299510f83b8c65487e69
1ca70b16bc5f8067f1277519739ff3949d2e41e4
16423 F20101124_AACJQI burke_g_Page_091.QC.jpg
56a6d83530f7895dc2e716ecf6396c54
e0e71c951c4723f215ce2e1ce02a1c21980b94ba
43381 F20101124_AACJPT burke_g_Page_072.pro
6c2d07fd12b8a25a6bff6db1fe7e10be
e049babbbf0ce5a13cddd172915f9efecfca5eea
14644 F20101124_AACKTL burke_g_Page_153.pro
fca8829c48583f8ec62ce134945e6009
019d379e58d68245c85e69c2ddbd238167a5f190
40054 F20101124_AACKSW burke_g_Page_127.pro
fa0aaee7c359b5e5ae396faa9e0731fa
4f1e0e5507fe61e6758b025f048e01c17731867f
119280 F20101124_AACJQJ burke_g_Page_164.jp2
8d0e2711a9cd88a29cd7ce61a237a3b6
e0a7192b84f4ff2eb60568a317a4c2d162732754
F20101124_AACJPU burke_g_Page_025.tif
efcc2537c655e4edbb00f5f846b57bba
23fea74d23387ee0f05a5876204897ae10efc510
2121 F20101124_AACKUA burke_g_Page_014.txt
6a704d02d60bccf5a2ed833dbac15757
eb5603040021cd636fb7d822b5078288026c5e31
32947 F20101124_AACKTM burke_g_Page_154.pro
fb773ea4323cee2c318413732254b0cb
eaa45365c70d01bf9350a5e835ca7d8febfda879
39192 F20101124_AACKSX burke_g_Page_130.pro
0b005bd85be1a195d9024581ae4f68ac
8a94c265ecb1289e82229f6eff1d9028bcb567e7
27230 F20101124_AACJQK burke_g_Page_132.jpg
66945828ea1be41615d9d9b39e78a10f
5dec0c5925855d7855ddb006cdd1fcff4325c186
2116 F20101124_AACJPV burke_g_Page_057.txt
f4392a1c0bd296c9b64a4ae648f8a965
29ff29cfc8424299751bf3f93e902e15d1ff04fc
2214 F20101124_AACKUB burke_g_Page_015.txt
8c61f16f6e6ea14509d39bda7ec9dc91
6ecebaf2f16f4572cee7edd611b635e0258fbfb2
32356 F20101124_AACKTN burke_g_Page_158.pro
0479f111cd7fa49f803b3f445f5a2dbd
bf68c7843de93647c34f66fe291cd84d1d10c263
32249 F20101124_AACKSY burke_g_Page_131.pro
4dd9eaa8c19b9342655824fe20d08a3d
66f27cc87cb2d40314ad82a59376eddf28104085
F20101124_AACJQL burke_g_Page_083.tif
bd7cf7c40f287fff0b361268f366fda9
978604d4e4d89853a651b550d7ca026b09d37a04
38782 F20101124_AACJPW burke_g_Page_166.pro
38f9037bdfab398c275d84ca7db7e743
8094cfe80c95b5292e45fc1f793cea24f6539939
2072 F20101124_AACKUC burke_g_Page_016.txt
bb2a2972de7ab708154038efbb0de3e2
dac9b812a701a052a25048addbdd77d96c376005
57200 F20101124_AACKTO burke_g_Page_161.pro
498a0b0e6a8d13d6dc5d405a8b453ae1
c09fe2e14b9968ac2dfd57f37f1fe6d7c6e739cb
39266 F20101124_AACKSZ burke_g_Page_133.pro
1498f74d6ab349ab5b462d31d9895b3c
ab6e169d514ea403eb81d6bd11f953a84b7be4e0
1885 F20101124_AACJRA burke_g_Page_060.txt
e96f425282a337eaa25eca99a769a38a
504c47470284686e5ff2063f21e7472529991ce9
77847 F20101124_AACJQM burke_g_Page_090.jpg
b467b2b92c6dd454be5ee92ec2736f17
e4289b8325a36b51b057887830afe94be177351a
3296 F20101124_AACJPX burke_g_Page_165.pro
5bebb2831cca2ae2afd38ae959ba21da
799703e9f15d1d1dc82fd2dd45be55ed6786e16e
2142 F20101124_AACKUD burke_g_Page_017.txt
9d61473f5d4b4a42dd2bc6e94e315396
1c08d95a5354dc27d820f0eb661f448f0a6486a9
54337 F20101124_AACKTP burke_g_Page_162.pro
34e5bb8c073a1350f2b4feadfc3060f2
2cdbd68f3e298db9ef1f55135e9ba21256b9e831
51374 F20101124_AACJRB burke_g_Page_114.jpg
1d3c398b95a0dbd39be556a214e7f29c
ba900847751986dbe9f88fb4489f09ae01709696
F20101124_AACJQN burke_g_Page_137.tif
91be2eb2dc1fcfe69345b1f09fffe500
1b15761e86e7319303e45fad664bbcdd3ebe0b2c
23818 F20101124_AACJPY burke_g_Page_152.QC.jpg
6860380b83bf867b677208109350b042
8acc0a5727a00665e8f20547bdd38ff0bd117ac1
2141 F20101124_AACKUE burke_g_Page_019.txt
1e28374e28ea04d416d99918ddf0bc78
1a07987923e4c10f2652d157afab1b765bf0ce21
55533 F20101124_AACKTQ burke_g_Page_164.pro
aa46f982a5c6754340db1749062d9a9b
52b2f9e7c7f26ea11b2cee0eb74d0b34f1cb180e
5264 F20101124_AACJRC burke_g_Page_041thm.jpg
2f00902da8542bb3637f6b45ce4c9db3
18c5360a756eab93782bf8214c8145e815d25265
43616 F20101124_AACJQO burke_g_Page_111.pro
082e2f56fa155a3332badb84ffb8daad
18a37895ba0497f575dc18ee67f4ac371fb8c164
1349 F20101124_AACJPZ burke_g_Page_118.txt
5394ab7c0b9b98ac76e24891181c290d
d8b5670fa37bacb76e781d9aee73d25b834f69fd
F20101124_AACKUF burke_g_Page_020.txt
2a702e2c9ab3bb869f8369b8599fe15b
a1f9ebf665be7e9de95bfaab87190cc735d3b6cb
84057 F20101124_AACJRD burke_g_Page_142.jpg
ac47fc3dcf3c79b3bb2d022a0994c31d
ca5ee33ff902f83a350c31dd8b4d31b6de154c07
25243 F20101124_AACLAA burke_g_Page_046.QC.jpg
ee9bee9d799b8ad8201183693a2c728b
297df27155c6b09a7e6b0fdb140129c2cee4ab78
2204 F20101124_AACKUG burke_g_Page_022.txt
f29c0b6200dba4d3fe0665dd88688d7d
211ed577f000703b794b2798969563f3bbcfd01e
378 F20101124_AACKTR burke_g_Page_001.txt
94b5a692c95d7cb8e7fb20eed8de3578
6d3a4e109627efeb2ed9bc225bb546b93b45a194
22471 F20101124_AACJRE burke_g_Page_146.QC.jpg
5405fb4eaad9e8c47bb47af33590696a
a220be0ca4938abbe5c13ec9888f7686927995df
5601 F20101124_AACJQP burke_g_Page_056thm.jpg
315f749521b2bf86abcdf2e74d5656e8
18cc86ad74b3806340c0ccf61686fd1115f0df39
6419 F20101124_AACLAB burke_g_Page_046thm.jpg
a77fbb0a7cf8c0a52e03398afb17e4bd
8f5e608445be47aa098f81257eaebf452e956c31
2206 F20101124_AACKUH burke_g_Page_023.txt
3250df7f92bec9f5cf28c54c34ec5c59
7539c1bbcdba3779c6a4f270ab331a1a842ac7b5
126 F20101124_AACKTS burke_g_Page_002.txt
2a9d192b30649d2dedd19fd5ac4202d0
f2a655698eb1d8115fa2c2e29ac85067bf15b813
72521 F20101124_AACJRF burke_g_Page_031.jpg
c312c5994c8bb58c790631c78adfbf2b
e165c1e36ffe5abd8ed8e1ff6f27d36b6f319ac8
4882 F20101124_AACJQQ burke_g_Page_116thm.jpg
16616c1a727a0d2337921d453863ed3a
dcbeae7b79c9e1cdbc4961569f9eb5f96b01c481
19201 F20101124_AACLAC burke_g_Page_047.QC.jpg
21b494fbb03bf88b68e0c96d7ab781f3
4eada5187f0434c776ce2772afcd5e165b257280
2200 F20101124_AACKUI burke_g_Page_026.txt
5993d9c720f11b0eccb9844a32c4fb31
ea34075f5c9af5e5573e4db130bfa816abc101d4
130 F20101124_AACKTT burke_g_Page_003.txt
50efdaf966f7afd0a6f98b4216c500d2
86fe72d8cd39473d4d729c5a874aa80c8f7a4fd4
14082 F20101124_AACJRG burke_g_Page_158.QC.jpg
0d9e68a393da64f138ddc45313f8e2f1
e12a19ce25ec5b7f468e1cd10a6b72668090c3cb
F20101124_AACJQR burke_g_Page_038.tif
d770fe05f97b99bcb839839c5b5ff1c4
51b3a1e7885743453bdb842a359e6be008f5f674
5401 F20101124_AACLAD burke_g_Page_047thm.jpg
b1de6018c3564b957b7ef55c69eb5c3f
c425d9fe727861d8e0c0a0d68675d64276501695
1317 F20101124_AACKUJ burke_g_Page_027.txt
2f6916bd978e794ecbb618fbe7215337
5cfd9a8ceb4f057ad81a515df6b9d7701bb7c342
2284 F20101124_AACKTU burke_g_Page_006.txt
50c1b3672d8c32087b060229d241b436
3109c73fe94c3191bf1372a9681d84100666b598
2063 F20101124_AACJRH burke_g_Page_098.txt
2607f784e71d1adbebed18292fb2583d
9ae6e1de3f358887217690d50776b3a2d167dc5f
5867 F20101124_AACJQS burke_g_Page_025thm.jpg
047866f326bc9c8c3b38341bdc6af23b
ce3314c6fe9708a7b2f222cbde3ac3a43dfe367a
18413 F20101124_AACLAE burke_g_Page_050.QC.jpg
28cc66093ea68f4b948c986afd821a18
3999467cee6570967968ba3246bf8830b1683206
1942 F20101124_AACKUK burke_g_Page_028.txt
1c693ced3a8c972275e5041e39360492
fc86e06b18ffb4849181a4335a69d06ef9d07eab
2083 F20101124_AACKTV burke_g_Page_007.txt
ee765a269b40c2516a471349975a0e3c
fc6018f5246e3fc0d8d71ba2a5243280dc9cf6ab
20927 F20101124_AACJRI burke_g_Page_086.QC.jpg
09614cdd582132c61753dea7e0405de8
922366bec6c6d7cf76a9239b7b1d518553b8c634
118459 F20101124_AACJQT burke_g_Page_163.jp2
2e176e16fbe4bf3685866163570ed76d
9e88e704e5ef28e244add6564a130c1d317555a5
4997 F20101124_AACLAF burke_g_Page_050thm.jpg
3c87bf4a44f9badc22df4ece852854e5
577f5ba00fe66d2279c197440ed535c91c4d9c53
2202 F20101124_AACKUL burke_g_Page_031.txt
e22cbc0fcba73cb1b79882d554796c09
460f95dcc34b14c6f6e75527f2a855911a88263f
1470 F20101124_AACKTW burke_g_Page_008.txt
fe63b3d4f29445c39f33fcf1d5fd76c9
cc608815c43cef95339ac84c0f4ad7adfc906a3e
22940 F20101124_AACJRJ burke_g_Page_110.QC.jpg
93faf73a57127fb1a02a5f61a64b6095
698d197007665cdaa216e004c8b15a08c5d7e917
39560 F20101124_AACJQU burke_g_Page_139.pro
7317f210211b5420afcaa3d8b690dcd9
5937b27c6ed915d1e4a8b42fe145e79a833e18f0
21956 F20101124_AACLAG burke_g_Page_051.QC.jpg
9bbf8436d1008a42fc38150de1601d3c
a09eb25259efd900d63bcaa67e017069c0a5c711
2860 F20101124_AACKVA burke_g_Page_053.txt
b2b33eb487e96e30626ab63db7be6bab
b15fd14018314999b67fa443dd59173c62ddb537
2236 F20101124_AACKUM burke_g_Page_033.txt
81255205a499384d91d4c26e7acd279e
e8270f16cf467c1daf5acd87ff752a531de3e35f
2000 F20101124_AACKTX burke_g_Page_009.txt
5140fa8bb205c20fbee4625e25053aa6
796f61a7ce1d2cd6162a32536cff2d24dee68b71
92512 F20101124_AACJRK burke_g_Page_142.jp2
3fb8ff8b409cc04315e5a425e4c4047b
75089ffa55b194420ab28d94e07454a870d90043
F20101124_AACJQV burke_g_Page_079.txt
e5cd6ede7bb608ff228be43086bba019
00f8728cb8e76e3e3047924d010f245ddd67b18b
24565 F20101124_AACLAH burke_g_Page_052.QC.jpg
140282d7b825fd390a436647763e8a03
ba8eb2bb43995cbd8c844ce56d93ea5d7733781e
2233 F20101124_AACKVB burke_g_Page_054.txt
50b05731bf8d24abad0e07b4bc4b6576
5fb37817391ca90a19751151367978ea96dac329
2197 F20101124_AACKUN burke_g_Page_034.txt
03f9ccb7fd1284d01a31d041eed0a20f
00fedb8b9ae1e2ae758180c27c12817a60f3e556
605 F20101124_AACKTY burke_g_Page_010.txt
5967789545df6241443f42d7f0168f81
54c7dbbb227481047fcc96205b500d82c5f29bc6
10516 F20101124_AACJRL burke_g_Page_135.QC.jpg
dd0cfbf5e5bb2e76591a5b9b536a10ab
8d1244de1016c9eb444bd0b422507974812f59e2
13517 F20101124_AACJQW burke_g_Page_012.QC.jpg
dceaa2386f7b41230c21a49c090950fd
2ae60bd28d870fe5f074a8993caa077cdfc37bde
23131 F20101124_AACLAI burke_g_Page_053.QC.jpg
9fe0d6cd78d56398f20a569ebe9bafd8
64a7f85192c229c4326225271dd3867acdce98bc
2101 F20101124_AACKVC burke_g_Page_056.txt
3bf8b773270ea9bf3d098a5e91ceba75
27c7870cc9526e12fc3c580e5e92c407a4bcc63b
2159 F20101124_AACKUO burke_g_Page_035.txt
b0c27b47c72a90aaec188d81e657f86a
ccc915bfe4d6657247c175439a60a5155f89f365
1991 F20101124_AACKTZ burke_g_Page_013.txt
76af16790959036c885e684cc024d626
a697979e0ff56956e9c850e9f193a7d1af472e7a
115114 F20101124_AACJRM burke_g_Page_015.jp2
6bfda54a4d4ddf989af352712dc28c51
323d1b17f44dc819c31786c10bfb218c82343e5d
79756 F20101124_AACJQX burke_g_Page_100.jpg
288ebcb788ba70ebc8294375cf12cebb
160d005b1ded82d897fc39641cefdb3eb3b33bad
4913 F20101124_AACJSA burke_g_Page_089thm.jpg
1027b8c735129be4630708f1234e51fa
e4aab56b32abda17c92d53c5cbcf3b1ebb2fb9b6
5819 F20101124_AACLAJ burke_g_Page_053thm.jpg
52640d0b5a7df0237a184570864efe73
c2e0603e2446e96d4807d1e0031232739f6cf5c9
1085 F20101124_AACKVD burke_g_Page_059.txt
1cf47b744b44c3926c34eba55a5c3628
dc6c9ffe864e381752bb3e921262574301f5106d
2219 F20101124_AACKUP burke_g_Page_036.txt
d72c22434ef6051150802983733b1068
ff5e999fc4ddfada00339884b28de4258a2b48f3
14185 F20101124_AACJRN burke_g_Page_088.QC.jpg
c09cc5c7fbfa25018aac6c372c7bdd00
2827f109141b18643e851bbcfc48409075ae830c
46029 F20101124_AACJQY burke_g_Page_028.pro
c9002952b49c3929c9a54a86ed574ae8
6eb50430e3434306adf9eda00f8347c4d376043b
74412 F20101124_AACJSB burke_g_Page_036.jpg
c81d18d296cdd9deaa0eccf21e41928a
d1ba4746098ab9d069c9b828c5374118834b8025
5944 F20101124_AACLAK burke_g_Page_054thm.jpg
ca2e3ab8568ce21625d568bb5f1ce60f
ff4a4e0539c8acad8010c0a541af34f50b662b82
2215 F20101124_AACKVE burke_g_Page_061.txt
dc100d3592f75ca79bc4e1f6ca562b5a
47315a43cdfd0c32c796321d4c1efbfa29c1095e
2158 F20101124_AACKUQ burke_g_Page_037.txt
6d54a079ca495c7b9674e8de26202715
2763fea746cadc941908fe6f064139398496c26d
51392 F20101124_AACJRO burke_g_Page_032.pro
a377f536d1434ca1593c437f4fe146ac
da0f50acb7c3a9cf5c59df4009c874e36bc3a3d9
82757 F20101124_AACJQZ burke_g_Page_140.jp2
be0adb920ca7769b78e0c4cd1850701e
fcb9a068b745fcb084acbc4050e1059486440627
5577 F20101124_AACJSC burke_g_Page_140thm.jpg
5ea63ae16f30a8aa925e4d25434c0606
0d5da201239c24f5099575ed34f96589d606fb7f
23376 F20101124_AACLAL burke_g_Page_055.QC.jpg
9ce29d8c6e8066a89ba4f74e572dc341
c90216e97352c56c4b2114e5146da3b457ed2bf2
2109 F20101124_AACKVF burke_g_Page_062.txt
82248fc9b0575e3a85d2395b52c03428
98a95f8b9321b85364964ae56d8ca4e46178f534
292 F20101124_AACKUR burke_g_Page_038.txt
c4aa046c7583a420b667d66263b98fa9
913d03e8adafeb529f6aee85c020775f45b34dce
1527 F20101124_AACJRP burke_g_Page_165thm.jpg
c870e7b172a94c60d3e098cbe05fa1e1
8abbc3e2e697d7dbe41b3628d0345302bb66b879
82495 F20101124_AACJSD burke_g_Page_124.jpg
2b9395904dce47faaa66d2ee2e1058cc
2e3b4ee2ca7f05fabe441c27dda35ec585d14f44
5213 F20101124_AACLBA burke_g_Page_067thm.jpg
17969c46f70f038613b153a0d2049786
c4ff10d7debe2be2310680a4cea52d5ce0777e9d
5904 F20101124_AACLAM burke_g_Page_055thm.jpg
3798076439a6a4b331de13d8250c0a35
c0a2ba72222e92150d3513c66dfb7637ebe57c57
1950 F20101124_AACKVG burke_g_Page_063.txt
616cca8416e65cdda87d58d062ee292c
07c6e9d6fc5896ca5879c4867e80a9694114adba
98697 F20101124_AACJSE burke_g_Page_063.jp2
129555e93e3c12001f4a36c990f6ee8f
05f29e746aaea4b1dd2ff2287e8e5bf2697559bd
17605 F20101124_AACLBB burke_g_Page_068.QC.jpg
369f9fa0ef10d6f1739c43432fde0419
ae806ce683c381f7e90e7dfd435b9de8b7f6f701
22033 F20101124_AACLAN burke_g_Page_056.QC.jpg
b4c892603feab0bf1d9f4edad12d05bf
52be7ccfddec85b40c2cbcf40b14a74cbd93d26f
1834 F20101124_AACKVH burke_g_Page_064.txt
a37f48f9f2da24422af330ba75c2880b
a4d943973226a5b826dd3442ee4f43fca3ab0996
2093 F20101124_AACKUS burke_g_Page_040.txt
1875c35a5ce21e6b3d820ff308abf6cb
17dbb703d0f8d03dff994ee1a3e2317f30b2864f
10485 F20101124_AACJRQ burke_g_Page_002.jpg
b3965b406414a1181eb8a5c03d9c1a98
8ddf69d7c8b7ef4c3841930cf70b16752f728015
57683 F20101124_AACJSF burke_g_Page_041.pro
2e42ff21d41bf3e67361eb5a2bcd00a0
d2f2c952e28495733a1692fd988f135d56105b19
4810 F20101124_AACLBC burke_g_Page_068thm.jpg
a6fdcb00c745aae8356ea4e84ed40d67
3b24a4ef96e5f359e1915108042383ea60feef9e
5905 F20101124_AACLAO burke_g_Page_058thm.jpg
a9234a7f2035edda706dd2250b56d381
6275d8c8c0b70a86c86971fe2d79628de272d39a
1694 F20101124_AACKVI burke_g_Page_065.txt
7c6958848e0343291ea383195c6b175c
1caaf8f243e69a56f64eea8a8cca1069842c5c06
2748 F20101124_AACKUT burke_g_Page_041.txt
f0bda0c661c4529045922651d3f28b45
112f393a8597ac8b0bbf92648154edc45341174c
13199 F20101124_AACJRR burke_g_Page_010.pro
94c0db9ea5aea5cabacaa12f9cbb68af
9c5cebb67107be53e4eb505ed1be964ff9601fe5
119885 F20101124_AACJSG burke_g_Page_161.jp2
68eb7c6feda962833d9c2a94a01af501
3965004da228f9abc15879c04d556233803883cf
18566 F20101124_AACLBD burke_g_Page_069.QC.jpg
0d27de225ed365a3393cf29ba9bba3ba
ad7cce9903856400c4d2266beb03ee7ca013f433
18870 F20101124_AACLAP burke_g_Page_060.QC.jpg
14ff06f015bb2045699313611b6dc0ba
a8964537f9243afc16ff42938b92cc84dcff3957
1948 F20101124_AACKVJ burke_g_Page_066.txt
1a6093eb4d38a2b883e459e46ee00f9c
0302117ab98388c2c0ce341bb9d7250403d751bc
1632 F20101124_AACKUU burke_g_Page_043.txt
ca41e956096957e2b8815394577e8c11
99c637912d3c33bd20c47b15c5d73e4e7ca665ce
1702 F20101124_AACJRS burke_g_Page_088.txt
2505fb511f6f019fd47c91a72f74facc
caa0ab5a51e992e882168e75d142998b3dc606d7
114377 F20101124_AACJSH burke_g_Page_162.jp2
0604af46d73140945dc75498d9b36265
645d032fdbac6df247352bcc85280776edc6043d
4934 F20101124_AACLBE burke_g_Page_069thm.jpg
8a77c25989bf7a33ebe9f9df9a8da3a0
bb05c784a6f9b860122b0863ed38fd6000de2aa4
23329 F20101124_AACLAQ burke_g_Page_061.QC.jpg
b49d1e44b9153317fa33f5774157bfbd
5fa378068b6589395170959d6591c9262006f80d
2198 F20101124_AACKVK burke_g_Page_068.txt
1b16bdc52746f33ff4e37e4c0013581c
768777ec2c0f7e9db8d72afe011c5a819472bf01
1835 F20101124_AACKUV burke_g_Page_045.txt
15b81d373ef6d541919741128f2fc341
d00f8cffc2eb362e500ceb39a77a8a8a10bba44a
5360 F20101124_AACJRT burke_g_Page_004thm.jpg
f366147bf2b6ea1fb9c017e1868e9ff4
030f5b097013827460797cded70e1e84ce5a3329
5789 F20101124_AACJSI burke_g_Page_136thm.jpg
91ec23b43bab32b1e4273c31f8483e77
9a2f00e6dea4405978aafef3f419a7c1b3d19354
24441 F20101124_AACLBF burke_g_Page_070.QC.jpg
060e164ebb1c157b4102f05608da5ca5
53ab2aa838330f731336ddb2737965a9d715b898
5862 F20101124_AACLAR burke_g_Page_061thm.jpg
1c57678b0abaaee0217f088a9a02d8ee
3ce671034f0e1e72398336d480b80e7a8ac0f9ca
2877 F20101124_AACKVL burke_g_Page_070.txt
050925f1b3f74d616d3ceab3edd88ba3
9529d281192de9cacdafdaa1069400fd0abc06e8
2176 F20101124_AACKUW burke_g_Page_046.txt
d0d68652f6aae2e57e66fa5190995bf2
fe798eef722e3bff64af7d1aae9874c6db92b734
23957 F20101124_AACJRU burke_g_Page_148.QC.jpg
d755511e57e68e6c876f607423baddb5
5cc9622d2a09d437257eadeea714291e5cc53ec8
40523 F20101124_AACJSJ burke_g_Page_155.pro
183e2809fb1df63a1cfe0de5f090cc0c
bc28861831f304675bea5d11334304e4ebd45598
6112 F20101124_AACLBG burke_g_Page_070thm.jpg
fd5ac26534856f2b6c5bcc8e014a0419
ca0de3a515f3f5b218ebcb82cbc579b820b5c7d1
21788 F20101124_AACLAS burke_g_Page_062.QC.jpg
c7ff42f1a5fe349efe915f5270a52b17
51f10287b5fec37f39b9bcf1914b021b98aea031
2565 F20101124_AACKWA burke_g_Page_100.txt
3f6d9655e203af8bedeaa69bc830effb
dfbad99f6579e474d74c1b4600c1c49ec1303407
1822 F20101124_AACKVM burke_g_Page_072.txt
72470dc2a8e620eb8e91b6e1d38240d9
2683e6914ecdc025522ba59bc3fc28fb97cb9b68
1992 F20101124_AACKUX burke_g_Page_047.txt
fbf1a7f8570bd2470b5b099ce10720e4
3761b736750133a4e48d47ee3632bd8d01588d9d
F20101124_AACJRV burke_g_Page_019.tif
058eade9c07cf0265b4df62715740d09
1c4eabccdd1a190efe58e77ccf202c7c53f05b36
4812 F20101124_AACJSK burke_g_Page_091thm.jpg
bba65cfd1ea35f593caa0aebac02a68f
0883edc9b04dd733478cfb0c4112e7634ededfd8
16846 F20101124_AACLBH burke_g_Page_071.QC.jpg
9c4e1e01dafbb281e162fd94b74fab8c
6c1a69108953e7e064f4f29ccb1ba5e4b1b5aec7
5663 F20101124_AACLAT burke_g_Page_062thm.jpg
1064e89d3723a7b11ec8cecea12aec52
f89d08f8cee99d5f6bc3a9a6a8c7ebb752fd707f
2203 F20101124_AACKWB burke_g_Page_104.txt
9db3ae80dcde2db3664731b827710e25
89358f1641ef4db464724cc117ea24b124696992
2171 F20101124_AACKVN burke_g_Page_074.txt
dda4e26a2b5295aadadda7c6c7361823
b1b8d6600e55c720ffb0d10a5fe37c5943d889a0
1882 F20101124_AACKUY burke_g_Page_050.txt
8e83d3d1deeb9241f5c894c27cd1b3cb
c25ca8ffc53713b10e240c6541e83d56287c0d87
F20101124_AACJRW burke_g_Page_144.tif
24ae759e03854bde2eaafec416af917f
7c2c2e3b9c65cb7a447468b0434972e529911bf4
5422 F20101124_AACJSL burke_g_Page_049thm.jpg
66cb9035a822d6c61609e2b6add5db8c
2e4a3fd13620ac6a2c79c2cbd51421f8b4c6ca17
19085 F20101124_AACLBI burke_g_Page_072.QC.jpg
b4923805f7a3fe3c6d0686a9b39297ff
a8b105b2b9c2bedc194f50f584616e75f3716dee
19532 F20101124_AACLAU burke_g_Page_063.QC.jpg
32b7eca76f9bd157a071c411e41c8292
8f9d1218cc126a4d3e5ba2883893dbfd657f8cda
1128 F20101124_AACKWC burke_g_Page_105.txt
33ae9046b6a1341dcf9975f79cd542ba
4169665483ed822128dca31f91c689d173262ff3
2417 F20101124_AACKVO burke_g_Page_076.txt
1f621c6ffc5e86ad006abfc6a6dea01c
4c7e799c216dec3872fd8fa3cc97444099443d12
2843 F20101124_AACKUZ burke_g_Page_052.txt
e3b210fb2b161a6b9758b7d1587996e6
14446a70f96833bfb6283b554c446ccb2c88c38c
F20101124_AACJRX burke_g_Page_072.tif
cce33ad7164c593572daeb086f0a8b33
4283f59a9e756f50f38ae917a2aebfa1d335cf72
5252 F20101124_AACJTA burke_g_Page_066thm.jpg
d7f2e413d00c811aaccb5c812a59f6fc
3854c795caa91e9c7a301c4635e0eede5cb0f4aa
F20101124_AACJSM burke_g_Page_102.tif
a4c9e2007115286f657364ee293d8ccb
bdc6b419b6496b3b0e34745613a69f716349175b
22962 F20101124_AACLBJ burke_g_Page_073.QC.jpg
46dc4709903bb4df7994b8a9691a2b31
411950841cca8721e6242a656d68a56c2f776ca0
5165 F20101124_AACLAV burke_g_Page_063thm.jpg
5b1fa8c3ce3fd3d2b8acf7e4c3e9323b
e6fad9b68d52ae866b97e49d26a9460b83542fd4
2127 F20101124_AACKWD burke_g_Page_109.txt
c5c2eb67750a96b515b1784f8f1b4355
0972c0991dbd8b0452603c6185d87d8447766ef2
2036 F20101124_AACKVP burke_g_Page_077.txt
4ea5a122bdf93bc9ef777f905b6c1f0e
3232e63d0e1e302116c599496be1c0224d4d5e57
F20101124_AACJRY burke_g_Page_141.tif
708aa5a1d5c081d458cb3633462b5289
b5ae7316582fc11e12a9b6af1512b586029c0ee0
4678 F20101124_AACJTB burke_g_Page_006thm.jpg
d1009beb640c668528011b2e263c64db
4490b5a12e03bc433dc6afa71798d0805de4940b
F20101124_AACJSN burke_g_Page_155.tif
8b53e990a6bb17e4c859176d1ce2ada6
829392eaa62798acbb7406164a7025773504145c
4965 F20101124_AACLBK burke_g_Page_074thm.jpg
43c84fb7b529568f3180485e76dc0373
18fb8fd7b89334ce63339880c60da9be74387680
4677 F20101124_AACLAW burke_g_Page_064thm.jpg
678bc89ead747a4dc59ea3c979d57857
4e2e7a9fdf0224b6c504df3787e539f252f9a1c6
1315 F20101124_AACKWE burke_g_Page_112.txt
9b692fad583a7b5214e9512c90056dc9
b88850569985a4d774d2dd2059edc17d2662c653
F20101124_AACKVQ burke_g_Page_080.txt
e56d581a1baf3eb312acf69f1cf8075b
e91f78dfcf6508bf847475bfed66ec052af745f0
28223 F20101124_AACJRZ burke_g_Page_105.pro
35790b91fb2a8ac9eae788d7a50f0adf
bdb5a620930a8112593d7f29295cc9f019af2fce
57252 F20101124_AACJTC burke_g_Page_042.jpg
641b9597b7974f672dae2da5e48dd5eb
db6fd6da7a2809a62b6be0bd22023724db1cd98f
F20101124_AACJSO burke_g_Page_066.tif
f43cfd8099ede0b59b7534e364b05706
460edb57ca77be807d3b48971a15c1928edb43c6
20468 F20101124_AACLBL burke_g_Page_075.QC.jpg
25ef400a77d4d8bd25b0a7b37282a1e0
7f11f618330c371fc38fbfd4857d87e1b491b5e0
14880 F20101124_AACLAX burke_g_Page_065.QC.jpg
9b8c086aa13aefebbacb1e6cfd90b242
07346a0532d1e8037b0ae353585e3524a38be9a0
1769 F20101124_AACKWF burke_g_Page_114.txt
8a5b5733cb2219d41640942c274395bf
7101d784afee06eefe9dadcff27d3bbd1ff0a857
2034 F20101124_AACKVR burke_g_Page_082.txt
2b62bff4ad62189da09dd6e30ea8b093
bfecb80dc350e2c1ee5ba7779db56a67948f6e96
1051928 F20101124_AACJTD burke_g_Page_006.jp2
fd1849b304d0587fedeb4a15a31e1295
d7cb5cc345331842e99f690a952ac5b09d281b9f
68200 F20101124_AACJSP burke_g_Page_160.jpg
91274d072f1a7286a5bd4aaffbc3c97a
45e6df485ee2efd461a0030519504beb79f0824d
F20101124_AACLCA burke_g_Page_092thm.jpg
59f7644df6c60f10b679c1c313581015
8e755e544f309791a25c67b9a6895300b369b351
5396 F20101124_AACLBM burke_g_Page_075thm.jpg
10383d9fc75556d68a2b0056eef60b2c
974356841d37807abea3a1733b54359139559823
20579 F20101124_AACLAY burke_g_Page_066.QC.jpg
c855c0dd05d22442667694febd3ecb07
9cb4a4f9842d74a3400e26799b0e7d3a5f21aa0b
269 F20101124_AACKWG burke_g_Page_117.txt
2b17c045f3d9df3ab46cf1942d91d47a
7516913e391a5d8493b0bea9ad0a7abd3b6ad310
2313 F20101124_AACKVS burke_g_Page_084.txt
38cdf0bb1270065803c9dcb25d635c8e
8a97a5875d59bb4f64f84ff35598968ad0e977b5
F20101124_AACJTE burke_g_Page_049.tif
c6ec7f1c5acea9f1449fa9c2c6267a4d
33e03080d78fcc57770de5f8ff63787f58132e00
110313 F20101124_AACJSQ burke_g_Page_101.jp2
0219d0c0e6f0bad6c3ea5d469012e121
6372dc028e8a2e14cc184630543eeb8769bed054
13569 F20101124_AACLCB burke_g_Page_093.QC.jpg
43a45852814465a78b58589eeebd2cdf
a9a1538686faea3fbde7ae5fbb87667573c5e85d
5895 F20101124_AACLBN burke_g_Page_076thm.jpg
ca80e82c2c418ce9383e4d9a71632e77
da99c8f7b00b197b703db4ae7ac87af61b60045c
20512 F20101124_AACLAZ burke_g_Page_067.QC.jpg
08855beb2365e757f6b5fc5eeaf5c84f
80bc9423a9341a96d72bfd47312c7caeb9434f65
2901 F20101124_AACKWH burke_g_Page_121.txt
d36919c1f8da445a3fb397084c54833d
7881b1f19ca19698b0fcddb43c7025f83cc76480
6051 F20101124_AACJTF burke_g_Page_100thm.jpg
5fb31300d33c4e2ec8c863362d7c1277
4e1a00946277275e3b2371e6e83fa284c23ff4be
22107 F20101124_AACLCC burke_g_Page_094.QC.jpg
0294f997ed1c63550f1fb41338f538ef
e9547b142fc2c307c7093fd158b6ab4a84a4a2e0
19974 F20101124_AACLBO burke_g_Page_078.QC.jpg
8728cfb929195f2f0be567532a67d2b0
bc73e3c9cef136fefefb1d451dd0e5786dc21683
1516 F20101124_AACKWI burke_g_Page_123.txt
e6d35232099c1bbd5fac10af936231c1
e30f74d248909b807705dc290c1e4ab58ee66a2e
2320 F20101124_AACKVT burke_g_Page_085.txt
4a6f80746a4af6db27ce7408953f7082
5d4b242bbbf72a985fa47b2cc91858b7f4372daa
5168 F20101124_AACJTG burke_g_Page_060thm.jpg
d8bab02b5139647a45a735a856446b2a
7b286869ea68a9cfb01ca8c027429b7ba43371fa
4405 F20101124_AACJSR burke_g_Page_154thm.jpg
593c525074e5ae2409c5e04112c21d26
2805e2a20830ab75e2b140c2c0d2c6a35e6fb327
5611 F20101124_AACLCD burke_g_Page_094thm.jpg
46c2a0029cbbc3bf2f76582794e64a2a
6ae7bb9dbccdbbeac1c45b63ba98f1e788d6be33
5214 F20101124_AACLBP burke_g_Page_078thm.jpg
18dd6477a8199be957b85d37445a3f38
9de20027e3a512edd0e94b24e15657221a256a01
1793 F20101124_AACKWJ burke_g_Page_124.txt
a76de74b4029d2b2b02168021c5ac3dd
06a85ef6a8fcebbcce69e643b3df3766364e2b2a
1883 F20101124_AACKVU burke_g_Page_089.txt
134e0090cbc6a0f2d53be419c19307a1
7d3944e6b8cdbe81d8ebe3bd05fd108d022dcae6
67884 F20101124_AACJTH burke_g_Page_062.jpg
fe493bf2bae27ff5daee34a3b4460e8a
86b75e848ba93740697494c7a4727db2b899abf5
1813 F20101124_AACJSS burke_g_Page_151.txt
0e53c7f893e0af68d01365bef7e37c7a
8d22057be7cb9e137e8300699aa739557dd4cd7d
22994 F20101124_AACLCE burke_g_Page_095.QC.jpg
c61fad0f8fd12b0e1b73cca064a1adb5
c5d2ccda08c09081aa26df10a401875310dd9031
5285 F20101124_AACLBQ burke_g_Page_080thm.jpg
776d0ad20a215bce5df5113e4db08a53
567de2fc08ebb47f8f06a51a13390d7157debef2
F20101124_AACKWK burke_g_Page_125.txt
0b4ee43299a6d5cb91f11d1a89e59a90
7eff2b298d9837e9c6048579768bb789405c563c
2232 F20101124_AACKVV burke_g_Page_090.txt
174c050d4d4c5acba418688c5fc2d280
fdd8cc3c69362d607bb1f6050d89e28d92006f03
94302 F20101124_AACJTI burke_g_Page_122.jp2
2073eace6fbf1f8f70f03e3da6115e2d
cb56224873e6514b3b4549b56cda8c5a5ab8c56c
5882 F20101124_AACJST burke_g_Page_101thm.jpg
6ce2a0597dc64bb51bdd5f9673964d3a
c212b757dc01ec906d6ebade86324815f5af2a35
5752 F20101124_AACLCF burke_g_Page_095thm.jpg
3a2a39a3ba36aa4e4e58a82337903b28
b5de6d03c58a56e2e647279079096e099a26201c
11007 F20101124_AACLBR burke_g_Page_081.QC.jpg
74030e75e6643ca5127d3b57f1aec2df
cb2d8a6e3638cf67e4608db27b341872c51c929a
441 F20101124_AACKWL burke_g_Page_126.txt
4475eb7baf1e200d300edfe1d469bbf5
6b4d7000c9efcb911a225c57120a878b87b2f9b2
1916 F20101124_AACKVW burke_g_Page_094.txt
aaec6d743ff90157c959afbc2a78c01d
5348f7fa58ad25ed9772478b6d6903b8f46df5ff
2608 F20101124_AACJTJ burke_g_Page_075.txt
1e00c41175f51edfd7c076dc3412cdfe
d0e8a25bb9a04c59b64479920a5bf59f154f80fc
24430 F20101124_AACJSU burke_g_Page_137.QC.jpg
a902dc824c512ca6a2f8ff13aaeaad24
fd3ac0edd199144e0ed7bcf0eecb5aa99bbdbe56
22626 F20101124_AACLCG burke_g_Page_096.QC.jpg
1bce04efb4b159f693d4f6750f08c790
4a8ac6f3810dae292d2879e2955a83ed81d45000
20746 F20101124_AACLBS burke_g_Page_082.QC.jpg
a29e2e9a02809416cb074cbcf3254f62
ea786a727545c5f67c812232384a088bc7273d80
487 F20101124_AACKXA burke_g_Page_147.txt
70d290feaade2e6a0f6b8ac97bb86a3f
b896991b44d149a40f65ee992caaf916cf39617e
1917 F20101124_AACKWM burke_g_Page_128.txt
31b8b0544c28263887e111cc78b45e64
505e77b6b4fca668c8163b81c073533c601424e0
2192 F20101124_AACKVX burke_g_Page_095.txt
e0fcdd0086993f50ef629467a274a2c3
ee4bed11daf346036d038ced4a7f306e37be8403
60937 F20101124_AACJTK burke_g_Page_115.jpg
a352a2c7c07bbe339530f178b6883a17
46e8f029dd2c936b06d8deef83d7873bc16c5bc4
115377 F20101124_AACJSV burke_g_Page_054.jp2
25e2275d7f30f2f157b1b205b4ab56b9
fd4d658ebfe4c03cbbcc992eb1c449339aa2d611
5849 F20101124_AACLCH burke_g_Page_096thm.jpg
55b179217758fb3c9ed463d89cebedc8
c23a4926ac7959c1e223a6b60d6ac17e760d12ac
22389 F20101124_AACLBT burke_g_Page_083.QC.jpg
c7774a23e59fcaf04642f7a89da68acb
926e792f7458a6b9d665e3ae0dd2466b084cde06
1833 F20101124_AACKXB burke_g_Page_148.txt
55e0216d117fa7217780ff5c73e0fa3e
5e407fd751472e087685b468e0fc18553becdd7a
649 F20101124_AACKWN burke_g_Page_129.txt
656aa727569dc936dfd9b25f92669711
aaae09a11ed6df33c557c557c14e10193240e017
2201 F20101124_AACKVY burke_g_Page_096.txt
5f8c86255c0dc73dcef5e23ee8ac2ccf
4114899ab3b9e983e5a88d686ca5b2bc92efe34e
F20101124_AACJTL burke_g_Page_152.tif
a236b7f597591a5a0c44042bd2a5750d
e4507e586a11913e02fc531cf8fbc7e0057de880
48742 F20101124_AACJSW burke_g_Page_123.jpg
9bef865417d196ac2d2c59d9ce7f9ccf
c7509c893f41729bd02598d3e04955013d39580e
18755 F20101124_AACLCI burke_g_Page_097.QC.jpg
66ccdfaf6a047b3eb95d182c7b36fb59
c87d899383304b047652eb7e1f04329139e3dbcf
6476 F20101124_AACLBU burke_g_Page_084thm.jpg
585812bb865627db55849fca5921a2b8
9b87c8830755d274a76543ebd52cbef4da5f6011
1944 F20101124_AACKXC burke_g_Page_152.txt
4f15eeaf418c67211f821c7b83d0ccaa
1dc7ae5d9d62bb52357e123a416f68d7255d3738
1927 F20101124_AACKWO burke_g_Page_131.txt
2545951b92c05c85a6a6ee6725d28a08
65903e3a709ba2b5a6501f0e2997e84d3e8db596
2148 F20101124_AACKVZ burke_g_Page_099.txt
8c13c1cda92ca0f6df8f4d36edaaf428
9530f9ff32f9d9e2b39221f45711a2fc27af9341
88208 F20101124_AACJUA burke_g_Page_139.jp2
93eefb3d5a3833614d60a056bd616d78
69eca26316b09f5e6ede47d00dcb309ef81326c8
42782 F20101124_AACJTM burke_g_Page_091.pro
610a6e001c80db7ffcec8426c349e0ef
9b9a074a0c360601d59c2131640b755d4bb7ba73
15627 F20101124_AACJSX burke_g_Page_156.QC.jpg
716a88c7f63c85b47852bf73ad38fe2a
44b8dcf31b3a7714d2344154d56f322272dde071
5051 F20101124_AACLCJ burke_g_Page_097thm.jpg
100fbff657e11fbbb3733021072e88fd
9b1bbc0a40cf9d393a3ddb4e99899fb6e12074dd
24363 F20101124_AACLBV burke_g_Page_085.QC.jpg
bdc111de6a48335e4e6729944efc1ff8
f9ae0120bf948e5d24cbc02e7ade0407a641bc85
690 F20101124_AACKXD burke_g_Page_153.txt
13a6993944b4c29c8c2ecb1cb0f09490
15105a2cff3c09bc3b159400e76ad58943d94e16
1953 F20101124_AACKWP burke_g_Page_134.txt
5b103fe783c847bc30bb5da158d44640
c4abd85a89ee499bd722871a8e836b1558e76997
70043 F20101124_AACJUB burke_g_Page_016.jpg
5a331eeac0fa8a61d981a9ef8810968f
4ee96ae19421034c455203428543b2904b8ef6f9
84066 F20101124_AACJTN burke_g_Page_127.jpg
fbd62e6629f4f5e444f61e77eea50ab2
b420b9801c83d1e9a06a91529ce8eb5dbef407a1
20100 F20101124_AACJSY burke_g_Page_092.QC.jpg
5edf8aefe09f8709b5e0d12d274c82b9
d9916a9bdaeebb47a7906fd0163bc7a4d014746c
21655 F20101124_AACLCK burke_g_Page_098.QC.jpg
b6ff9a0d95c1e43cbceed4f2ca1eba1b
d22624e650ef740b8ebde988f9c704a88dcf4ca6
19289 F20101124_AACLBW burke_g_Page_087.QC.jpg
9a7e74c6e2ace5a18db7d4433faa8dcb
3f6fb129074a0b8da43a4227918aba1787956836
1780 F20101124_AACKXE burke_g_Page_154.txt
a701242e5ef8c243bf4301e508672480
2e7548bcb960e1404692fb024cf8cc650499ecec
1755 F20101124_AACKWQ burke_g_Page_136.txt
9bcc4a77fcbcbc6efe1652fb63b1f734
dc178236bcb78176e316cc064a3c44e60e889578
65172 F20101124_AACJUC burke_g_Page_086.jpg
5b1ae9f77ed755d8700bac29da3565c1
8a060b7627f2c5c344d357dd87f4478293a4cf66
3626 F20101124_AACJTO burke_g_Page_012thm.jpg
2c3799da6b10d5b9afb0110327db0b8d
ab3973ae52e8a0a9f80fc1379bac26d5d17635c6
22796 F20101124_AACJSZ burke_g_Page_090.QC.jpg
8ddd35e6e90ff6d8879fb219d85408c6
83187e8fc27ed01350a1d05f4604da7c1fde7074
22616 F20101124_AACLCL burke_g_Page_099.QC.jpg
104d797a3873f428d0d1fa221567c2dc
cf7ed2398dd30abb27eb17d6f1ff72fcae598102
5120 F20101124_AACLBX burke_g_Page_087thm.jpg
55a523d973f95ab91deaa710691b381b
6285e2e5e1c6484134967f54e17a47629044cf97
2021 F20101124_AACKXF burke_g_Page_155.txt
76a44dc5b110569f83c1b5836ef10100
f58d5e22a131d51844a358e06a0f612d347c8b8d
1938 F20101124_AACKWR burke_g_Page_137.txt
1b8773f88a4d544e8bf658c2dedb2cf5
fa25deecfe55a42eeb11bd65c782d125c4bc638f
59923 F20101124_AACJTP burke_g_Page_011.jpg
56949b17aadb5ad39ad54c0203895212
590ebd583119e3039c7eac5fb9dff2f035dc97ea
1925 F20101124_AACJUD burke_g_Page_087.txt
d2a43f06a09a1e4795f7aac2c50a0f89
46f8d8563bbe1bf1fcd10526a45ae65ae1a7ec8c
17556 F20101124_AACLDA burke_g_Page_113.QC.jpg
244fdb2162878152f8afd1ea37a57f0e
342bdc8223747104310ca2731a32d29c540a2d2c
5718 F20101124_AACLCM burke_g_Page_099thm.jpg
8fc09f4dd2c64b22230249ddf7da73e5
ae9028c9dd2bfcd2cfe9d5f04573add268484ef1
4232 F20101124_AACLBY burke_g_Page_088thm.jpg
ff126761fe2f4b4e2a697bc21bfdc2e4
2171ecc74c9d183ff4480702cb4782922666615f
1741 F20101124_AACKXG burke_g_Page_156.txt
c63621e582568641d465cf9d7972a9af
7a46fc5753762880493079e0283957fd18440d4b
655 F20101124_AACKWS burke_g_Page_138.txt
2ed765ac4e5fab24dece439d5aa3fd7d
d9c269b5ec1f05eddff193f8cd240fbe46a32fd2
44455 F20101124_AACJTQ burke_g_Page_113.pro
214111e56b5d37fba39ffacfa3c8b279
8e73e4526c0482b145ee7b284fe17c54a46d9907
22941 F20101124_AACJUE burke_g_Page_076.QC.jpg
2f28db30e128027192ddab57392ac2e5
03017289965646b882d2c3fa4332a01945ebce6f
4860 F20101124_AACLDB burke_g_Page_113thm.jpg
594822a67d407d40f7d9288c8a68d5e8
b4a6ae13fb34e4637f784e3449b993bfde3d91ac
24084 F20101124_AACLCN burke_g_Page_100.QC.jpg
f8440c6c8aeced97d664e822120eedb2
f8a8b235f7e3fe577b97598262e6ea07f7b496b0
16823 F20101124_AACLBZ burke_g_Page_089.QC.jpg
538ca98e90069e6f7cb020e04d6f1f9e
c14693e00e259b5284e1c90b68b929845cc52df4
1608 F20101124_AACKXH burke_g_Page_157.txt
25d85bb371819c5dde8c86103c9f8603
80f94c1bef5a831f71632e1c5072cf9569189a8c
1819 F20101124_AACKWT burke_g_Page_139.txt
788e9a1dd645121164ab2a62aab7e1d7
12c60876652fbe5664e424639c4b3d09c91e774a
1791 F20101124_AACJTR burke_g_Page_078.txt
edda93f8f5421791a4d7be373ccc73cc
35f598cd8dddac8e870fa0fce98a227308c8107c
F20101124_AACJUF burke_g_Page_003.tif
428fb1bf5da96fd65baae1897de428c7
c560c935bea7e4a1db9406a50faa375d767253cc
15863 F20101124_AACLDC burke_g_Page_114.QC.jpg
a9b339dc91f1b913baa5c70827349a66
343748a9027bf521ef8047d1ece89cd002b2dede
21600 F20101124_AACLCO burke_g_Page_102.QC.jpg
744fd98ec5694892f00aab1c45db1804
77eab2c3383efbed37e025f31f045fa85b633679
1565 F20101124_AACKXI burke_g_Page_158.txt
ea356747465dd59de7b48da9f8bc8d0e
dfafaf3169cfe032faefe2a382946c50d099ee29
37260 F20101124_AACKAA burke_g_Page_153.jp2
c242fd7761efcd9765efa852a77ea3c2
c30949e6028df13bf2bea3784b5500ebc611c865
20798 F20101124_AACJUG burke_g_Page_079.QC.jpg
7336bdda0a4dae54e6991fc3cfa806da
3088a802a65aaf9e7f56777393e53d28ee642a17
4784 F20101124_AACLDD burke_g_Page_114thm.jpg
017c49d8a34bec517eb2b42f323cecfa
7065c7c494edebfdbc5dbd7fb38060d95c0af797
5805 F20101124_AACLCP burke_g_Page_102thm.jpg
2fce1cbe00c7fd80b72f75a162f875c0
f15ca4578c3290850484850e7836e36d56adcf51
F20101124_AACKXJ burke_g_Page_159.txt
58bcce94be77bc32bb8a8a2976c08afe
53184a5cf1330806ac1617ab1e8306bbdcc2c629
1626 F20101124_AACKWU burke_g_Page_140.txt
07dc4d56359f7eb3099d5af2288de0d5
a4074a9311e968944e9e1dfcddc3da89a221e749
F20101124_AACJTS burke_g_Page_062.tif
a95e53be6bd427c506c014e710c70e17
a31aed695a437cc289a8bd4f5f7723800e9015e1
86316 F20101124_AACKAB burke_g_Page_136.jp2
240cd1b9a0ecb30c4e9902f625a8852d
8893547ec296836b0b4e0a78eeca999e92a973a6
325 F20101124_AACJUH burke_g_Page_150.txt
c39f0df3791742247af8669fcd46c574
2de44d7960ddf6c8f3d8f29bb1d8480f6328b58d
19265 F20101124_AACLDE burke_g_Page_115.QC.jpg
5d85e8bf224116f4655806eb1aae2cdf
3318efd44a5e1dfe6647ca9256bc8a629b7b481b
23035 F20101124_AACLCQ burke_g_Page_103.QC.jpg
e343bfc5dbbb55e6e5abf1b4fcad5483
bdd1562aeed658b4714fec803037464d88ab9501
1990 F20101124_AACKXK burke_g_Page_160.txt
9e7c4ab41ca22307b417d24f7958c42b
21489cf2c0c55eec22b8da6831e5c59e3631f420
754 F20101124_AACKWV burke_g_Page_141.txt
e5ecb7b78110f7d26d4d2b912fc92e19
ac283b183c1c9e38f406fa2d53162dce1d8ad455
F20101124_AACJTT burke_g_Page_102.txt
58bf8a1d07e36a6b17c225c098cedb04
dc206b608be488dcbe73ee4af1daa6b6afae4214
F20101124_AACKAC burke_g_Page_118.tif
2a9d2165044053cb6edad410fc13acc1
0aa5b3f5c6eaa084cb910d922620dbb113d85648
39345 F20101124_AACJUI burke_g_Page_146.pro
47957dd8358f9b11ffb60fc120674f4c
7ec7fdb375dd1659e05122f682536d95e71e5282
17451 F20101124_AACLDF burke_g_Page_116.QC.jpg
a6d5097589657733037a1ff442a6feac
7db6e6e40e6edef3072ee933c48a4f3625acbaa1
5708 F20101124_AACLCR burke_g_Page_103thm.jpg
1149073acfb4a34b324b550d5316b67d
5a877d52bd836f0e70738c6283f8c80db05dcc59
2196 F20101124_AACKXL burke_g_Page_162.txt
f8b5d4ba4247c524923759eebcc913a9
cc8c53b7add9baab2635d7602a3278e2ece8181f
1832 F20101124_AACKWW burke_g_Page_142.txt
5585fc0cf2d97ed61efc9663e07b810b
26181ea2b647a170e3434b36092c7c37d6765651
F20101124_AACJTU burke_g_Page_013.tif
0234650eef15f9c5a534db4c3f4f624d
43be94876df9d40bfd2891cab707d753f08b692c
101598 F20101124_AACKAD burke_g_Page_041.jp2
68879127ce097758c7032bd37570d7b7
d24ca7a4916d281e8435c5e90df1b86fc1019faa
1844 F20101124_AACJUJ burke_g_Page_127.txt
233617d29c2857c1f1ace8dabd60f153
ddf4a82dfb6e070183c2ea663ffaedbac00ea0ee
4789 F20101124_AACLDG burke_g_Page_117.QC.jpg
0a1339912eabcc5fb22a61b13ef908b3
adf1982f074b6dc445b03b443d7d25b808377e1b
13193 F20101124_AACLCS burke_g_Page_105.QC.jpg
2e2bc4cf89a6d4add765941b63a9c080
c02521115bc8a3d821a82959563a70ddd24ecf53
21688 F20101124_AACKYA burke_g_Page_007.QC.jpg
5c700cced70ec6676759b5ce7faac177
9e705858a374cd7d8cd4283a73599b005d09b805
2241 F20101124_AACKXM burke_g_Page_164.txt
3d812c58837fce0037aa2eb3f0f31849
b14128cb15e73fb4badba4433cce257e8edc4e1e
1908 F20101124_AACKWX burke_g_Page_143.txt
5bd419ede383cc6f33a5d4bba2d7ead1
cbd7e03067af9a4ff36fc18d4911159d0cd73375
F20101124_AACJTV burke_g_Page_046.tif
cef467293ac84169cb4a543f9c853df3
4911af4aef58979b82288b31572c7e3d15c38729
18154 F20101124_AACKAE burke_g_Page_011.QC.jpg
1aedd3a2be351f90108d081fa46d3d73
9ec0d00b525309b8f1c04f6310bb6c961ee7f63c
F20101124_AACJUK burke_g_Page_088.tif
366c4fd6d24f5c1fd5cc5ae5dc21527a
2696d0036c187e9e9b4821cbb92e485337282e06
1779 F20101124_AACLDH burke_g_Page_117thm.jpg
c10f05985e2da3a3d9dd0f01a80a5ea1
958af9af32c9e8e9e9d63c6b9009048b19a298af
3579 F20101124_AACLCT burke_g_Page_105thm.jpg
33254e390de79bd3a2fc39c9971b320e
aee84078617c7dd564452255a2851a4e43ecab4d
5335 F20101124_AACKYB burke_g_Page_007thm.jpg
588efdb87f89b18a317ceeac0ed37e6d
1f2a78c1905c46a9f7a279926c0997b495f9db24
F20101124_AACKXN burke_g_Page_001thm.jpg
c6320e13ae41b04ba2158bd0af3ddf0d
66589198a16817d95722705508e21047ebb003e0
1829 F20101124_AACKWY burke_g_Page_145.txt
f8fd7e66d7bf6a756e5168eb98310ac2
f7d658451afefd48e3c445478df129545977d99c
14557 F20101124_AACJTW burke_g_Page_154.QC.jpg
265948db73c680bac4cca6f61673b6d8
db619862d4d3ec6ddf30fc9797b06374e197221c
17769 F20101124_AACKAF burke_g_Page_166.QC.jpg
bf1b9660dc62f96432503ff9403aef0b
37dd2b3a57de1e0d1f5f8b7027d070aba3939cd2
35286 F20101124_AACJUL burke_g_Page_065.pro
b651bc186102fb50d8b6b22d22434345
1a7ad934700a48513b59cddb23ead73451c877fa
13582 F20101124_AACLDI burke_g_Page_118.QC.jpg
56a2812731aa56dd131fa0049b54cdb0
4b1b21481936df8a382b2f03550fbc98d49ed6f1
21263 F20101124_AACLCU burke_g_Page_106.QC.jpg
36b0f1fb34740fd4448d0043805696fc
9a0d2dcf38b16cfc2bc3ffe91dad63b72ec9b161
4385 F20101124_AACKYC burke_g_Page_008thm.jpg
00a67b1aa7ccdaee8ac64626116b1146
e69cdddf8d00294219eab49eca256f62cacb01f7
707670 F20101124_AACKXO burke_g.pdf
164723047855e6b98b50c0ac21bbd5d4
8cac16777251943fcc1d31ca56c4c68681772772
1935 F20101124_AACKWZ burke_g_Page_146.txt
7f30324817b1035845bb4d40d3826704
c249839116dc87766fb221b964643b36a1f672fb
65623 F20101124_AACJTX burke_g_Page_006.jpg
378b01c16690f19fa4ea730b18a8eeed
df0ff28b590d533298efc0dbc945d67210fda562
F20101124_AACKAG burke_g_Page_079.tif
d228afc46df6dc7ffe9ad7b735fbbf1d
fcaeb4585d47fcb61876506c2de3bfff08bd74c1
68720 F20101124_AACJVA burke_g_Page_009.jpg
c881ad24b10703ec17f3058795a788c3
9c680a4ad638e3b5536042cebce985d65abd5847
1766 F20101124_AACJUM burke_g_Page_092.txt
4ab3b92bb18a8d26bc8ad34f67c1e843
87e972da5d7f1864dbeba204031978b813cc0747
10818 F20101124_AACLDJ burke_g_Page_119.QC.jpg
64582a42a65bac2507e8ad782f0bd3c3
b3ced5491d3a14de705caf85c81837b42a0fef77
5924 F20101124_AACLCV burke_g_Page_107thm.jpg
4a8f475c2a285583173729710f1087d6
00dde516e152552094487123388c8c7a115ec19c
20770 F20101124_AACKYD burke_g_Page_009.QC.jpg
ccd26689b411a4617ee82cc2508f267c
db8aa6da02d69e1f1421a8ec08802126bbdf2123
3082 F20101124_AACKXP burke_g_Page_081thm.jpg
302179125b225c40d1493b0fdc2835e0
997b1db1862c5c7833fe0aeb41dc6884022774e1
63813 F20101124_AACJTY burke_g_Page_053.pro
21e2815c7b3ff080f407cd7d63a1227e
3d695a51f3d3e79d1d4e25e3ae56bfa532934c1c
986807 F20101124_AACKAH burke_g_Page_048.jp2
017fe0eb0390c04819fdfb23e64159f9
6708ddf567cc41b3e7f2c1a6c25e40be52f1763e
29337 F20101124_AACJVB burke_g_Page_147.jpg
4e23c461767b3c1300ebbf9c0aa666ea
30c2b8216a225d4577d1d919b49799e4689ca947
14655 F20101124_AACJUN burke_g_Page_120.pro
37daed4a5fadc2ba10bbbcdc82539c90
244fd2f4c45d9410aa1ed571369451f0f815f754
3485 F20101124_AACLDK burke_g_Page_119thm.jpg
23ac322a0974b6119daf6fa2a56ccc31
54eda4a585355243088b3d62f8f8a53d82c2a56a
5943 F20101124_AACLCW burke_g_Page_108thm.jpg
c2a6918eda9797a8413879900f794f14
25b336833a7e7f37ca3bea8e4249816b1e43f695
5229 F20101124_AACKYE burke_g_Page_009thm.jpg
a05dfba1235fa6a2135d76ff45fc6835
3a4d94224d1377e09c745f89d053d2415edf9ea8
3499 F20101124_AACKXQ burke_g_Page_059thm.jpg
28174759b937a919626a4631e1502586
8c8442227d50e213b0e089fa0b02a2ba6c03f390
80545 F20101124_AACJTZ burke_g_Page_091.jp2
4a2ff6c22cd0a8754ccd252eb683deec
c28991200ee48b411d0ce3abe9451002fc255c80
2213 F20101124_AACKAI burke_g_Page_055.txt
fbe8ab261ab55eeccb035b3830708181
718fe13fb8da4b2d39eede380bfa97560798ba54
47252 F20101124_AACJVC burke_g_Page_068.pro
f5f3afd9fd8085440ff7930e3f8f0555
4f49310bea6706deeef010718f931d0a873168c5
2255 F20101124_AACJUO burke_g_Page_108.txt
1586a8da0d8d4076147ad5569cc6d387
3ac943fe9006995e0d1189f0b6cf20791b8a5d73
5878 F20101124_AACLEA burke_g_Page_130thm.jpg
e143b39dbb59fb14b840dd38d0af0f62
47e7f6bedb6e973cc404c6674e5d8a08ac1fc1ce
9013 F20101124_AACLDL burke_g_Page_120.QC.jpg
8a318d5dd9eb1884f71909415f7ffee0
dd73ea4ae5c5e714c2e068a6f260f6eb1aedfd03
F20101124_AACLCX burke_g_Page_109thm.jpg
735beb94bd5b5babc38580bdda368511
7df3dc54794a72983fb2631b9a57dcdfe0802603
8636 F20101124_AACKYF burke_g_Page_010.QC.jpg
82766abc2638028346a321c904326604
b0e18d4950beedab6e0ebe468625c61915cfde40
5889 F20101124_AACKXR burke_g_Page_145thm.jpg
adcdbda142dc9440cecc829a9a0fc883
a1b76f6f08b1ce216237594023b6123f71bd4ace
4005 F20101124_AACKAJ burke_g_Page_093thm.jpg
70c3f354ce8e45e0357ecd58bbc54a52
218e93a9f48c06f5aea36890b3b40cc262ec02cc
F20101124_AACJVD burke_g_Page_139thm.jpg
22e2176ded5acb37a81b9385650f4359
7b53ac580e60b7f909959c7af427d9b1fcd42d3c
56070 F20101124_AACJUP burke_g_Page_163.pro
24135226dea099b7eb0921db414fe540
0de4aa74850d8220cf1b376aa64d040ffe0563a7
3015 F20101124_AACLDM burke_g_Page_120thm.jpg
df9e5d8dfd82da3d83865f9a5188ffb0
e117ae19f14004905dc46d741adca5b72d1c6ec4
F20101124_AACLCY burke_g_Page_110thm.jpg
6b19e56d60e85e80e60f877ae666d009
3431b7d77b365a446cee1ebd5a584a86f0248267
2649 F20101124_AACKYG burke_g_Page_010thm.jpg
ac247003cb0e4bbcd67665b7c80af845
658fc7f1193474744072333a688256963e0b591d
246329 F20101124_AACKXS UFE0011378_00001.xml FULL
57f9e419f49a11210c9ed81d2e2cb585
c79f964e3fbd32b266c2cbb1b22ffd897420ba58
106022 F20101124_AACKAK burke_g_Page_102.jp2
8401aa92f2d8e4aa4f847233b1e727b2
560eddc8e9b17f21972894e59717e9c08f54861d
4097 F20101124_AACJVE burke_g_Page_118thm.jpg
115359771aaf3c35906d3b0ef97bbcec
bef9034dcbce33c8f42e52a95e444947ea26fcfb
F20101124_AACJUQ burke_g_Page_149.tif
0412fdcb2f5c54cb8dfa79d2a554198f
f6c15addb9ae9d5afd75cba4b819de09d8314715
F20101124_AACLEB burke_g_Page_131.QC.jpg
e891ce2513c7038d6ba0425edb933755
72da1b9df050080674b761a5928abb9ed9a8b6f7
21028 F20101124_AACLDN burke_g_Page_121.QC.jpg
6bc59bc0373f80d80b2ae13e3adfbeee
e2ec057cb1c43cca1251043c9434b0dcf09b0733
18544 F20101124_AACLCZ burke_g_Page_111.QC.jpg
d2e5fc2344a0d2f9fc92c58fb5cf79f3
5e8619e7c8f0e9517b40a70f5083bf83accf634d
4818 F20101124_AACKYH burke_g_Page_011thm.jpg
fa4e735f0099c28f0b18a85e09c3d065
cd11b51354b11e007e340a7d05a12356a753743c
3212 F20101124_AACKXT burke_g_Page_002.QC.jpg
34159e729a2a617bf28c5e18a007f22e
87d0a8f34b3d09fdd84c13e5eabdd8845ccf96f3
F20101124_AACKAL burke_g_Page_035.tif
fa30e0f2cd0f56662b253cf5b7f7dbcd
a8d84e39234842c3817a54cbc6f5593da5b9c0f6
F20101124_AACJVF burke_g_Page_004.txt
cee62ffb269e161a194f34e8cf1f8a89
6b4dc2f046f3718320c6ce2f0fba03db83f9de1e
434 F20101124_AACJUR burke_g_Page_132.txt
6e79612cd4fd19ba97ebafc5a6ee1fb3
40da5d84aa18f7eb5697cc996e0c01717f784bc9
5410 F20101124_AACLEC burke_g_Page_131thm.jpg
42ac8baafc219f562c036eb1057d89b6
c7eb015332ea8c748d94aa5b642cc12a853650fa
4972 F20101124_AACLDO burke_g_Page_121thm.jpg
bdc33f0a4afd5f93f2c2bba95807a9aa
ceea541abedf1d75e55fe0d864aab5bab89ae005
20243 F20101124_AACKYI burke_g_Page_013.QC.jpg
2a8ec735235e049285363adda561a701
22435a9873bb9749de4a673b8e2da153faaadd5b
3690 F20101124_AACKXU burke_g_Page_003.QC.jpg
3db434afc448f582e58c949b5ca3ba49
4ea8c4611596468fadfd26449d31099e5e24dc44
23191 F20101124_AACKAM burke_g_Page_031.QC.jpg
379230c8286648704d18f7e664a9c8d4
dd63863c3e477d01a5d45689bacb82cae07424c2
84337 F20101124_AACJVG burke_g_Page_070.jpg
91ff7a353cc524976343f2e80fc36666
7928b0ccf7d1b35e53bba23141cecc7fc2b396e2
47407 F20101124_AACJUS burke_g_Page_080.pro
a3ac4ae271f4c8fcf5e54d8c8c6cbb1c
00f9b384ffae4c8814f14502bc520d8cdd4cf7f9
5404 F20101124_AACKBA burke_g_Page_082thm.jpg
fa08ec8ecd54011bc38d2293f64b42ca
d67373877e18ad7bbfd175d3bf84ccb3e3842516
8594 F20101124_AACLED burke_g_Page_132.QC.jpg
169fb974c6ede2a9c5c73c213219e01f
21460bfa01879891a9fb520fb996e4f95ab71bd4
20324 F20101124_AACLDP burke_g_Page_122.QC.jpg
227d8e77b63dbc1aba0cbb2b88fbdfed
d70c885587c8785af7f45408b45a1ea1f4b753bf
22300 F20101124_AACKYJ burke_g_Page_014.QC.jpg
7db51142e875366df92d2e3d1174cf04
2ce8a8538295db387f05ca311906a67d5529f706
93395 F20101124_AACKAN burke_g_Page_151.jp2
8df4b46f727223e169c28b0bb2cf9692
4d62d0e6b5ae9283a94e5608afd05958528899a7
5569 F20101124_AACJVH burke_g_Page_057thm.jpg
d679495633985ce6f3233a7bf4b337be
2801178f2c466f5284dbf05b750a01d07b6483e2
38306 F20101124_AACKBB burke_g_Page_156.pro
854ce4a0743a9ae57ac0f36df2b1446d
ebca2a38bd9a36d6d29cedddad2c91d502a531c4
5815 F20101124_AACLEE burke_g_Page_133thm.jpg
cdbb0a81232cbf92c61f83960c40d63f
826d81f02727d41541f25c2d96a29a1b242675d3
3701 F20101124_AACLDQ burke_g_Page_123thm.jpg
aee12b213065afbbd902fe1b5d5cbf58
66666edd720728d503a88236c34f181a89516f66
5783 F20101124_AACKYK burke_g_Page_014thm.jpg
1854c4328b3945db17c866abe852c8e1
c1810c675c351cbc201a39e4c58a8d7f89994579
1480 F20101124_AACKXV burke_g_Page_003thm.jpg
a55f97a5eb909d6f28efd355f9be3dff
01dd90ed7dd1bbfe193475e56a6c5f7404aa1c81
18164 F20101124_AACKAO burke_g_Page_008.QC.jpg
bf676466f59b78c840ca6eecaf624462
caf59e7201cba9dca317f14cdd686956d42b191f
26146 F20101124_AACJVI burke_g_Page_147.jp2
497c876a58ee121a9583fbbc412b491b
32b041dbdf3acb100346c1627b025569f3066a8f
2697 F20101124_AACJUT burke_g_Page_132thm.jpg
c0ecff493012cdbd7d28d0135880044f
d181a702cc745bcdeb38fc53a13366282264d080
22461 F20101124_AACKBC burke_g_Page_109.QC.jpg
0d863934855c2c682b85707968a7f571
d63d1edab4384ad29234f07081a1cec966eb9aa5
23234 F20101124_AACLEF burke_g_Page_134.QC.jpg
6d45c3de57e53ce7b53a3bc300a7120b
c759fdc98939e63f0dbbaa4110acda11546ed6bf
23865 F20101124_AACLDR burke_g_Page_124.QC.jpg
1088b248da6512d5891f461f0216f532
c98ab030d3677b21222a2da2711d24616ed4b14a
5754 F20101124_AACKYL burke_g_Page_015thm.jpg
6abe611bec62ef1fd52b8c338aeb07c6
fbbf42435ccc94d774d165c14fa0e7a3ca039407
20265 F20101124_AACKXW burke_g_Page_004.QC.jpg
328977c0f348c96bbc287e040b1f3a97
bca625f5d8532c452fa2c947cbd4555dd5124f0f
5792 F20101124_AACKAP burke_g_Page_048thm.jpg
1a0fae6463586434a6b73e812744fa2a
947e8fff604c2a2b459307cbf2b0f93a2b46c650
2040 F20101124_AACJVJ burke_g_Page_049.txt
29e952bfbe922182685d4afe3a93a1f4
a61a15c803263e21f9ca82e6101fa4ea91a9a03b
48116 F20101124_AACJUU burke_g_Page_013.pro
84c1c7e297babc8ac50882393e5ae917
6d12472b3ebeffb19faa7fdf5384ed24d76f5914
F20101124_AACKBD burke_g_Page_033.tif
022cc14846a2a78e6145c76a603787f1
2cb92ecc814d89b5625ce1e0b7ba313208bf8ad1
5809 F20101124_AACLEG burke_g_Page_134thm.jpg
261d65055080c03704e7e810d5b8c4ae
0e12d6a2cdb0c10c775f156574607fa83bac8903
5961 F20101124_AACLDS burke_g_Page_124thm.jpg
751e80280101c37f24444b70890f5fa0
1733178ef205c22f5a27f79b8e96265617f49624
5543 F20101124_AACKZA burke_g_Page_024thm.jpg
631dc5283ff1c74c205dae0ed9a0eedc
c7435c5a590a6e67c8f68317f5fa7bd0f131853b
5596 F20101124_AACKYM burke_g_Page_016thm.jpg
e7351609fc547f8d05d7ba59226025de
0275db87ce3c3017bbe05395212110e2342ebf3f
6775 F20101124_AACKXX burke_g_Page_005.QC.jpg
5a6893a90dcfe3d0a357f9a345d06540
a1c2c3544a6bcb585626d610cf892854f27e52e3
1673 F20101124_AACKAQ burke_g_Page_093.txt
9788338fdac47e7a698aebd844d742db
4ab4fcf45b7402e98ce7c7bc39d35fd6a672d1a4
2260 F20101124_AACJVK burke_g_Page_163.txt
75f32878ef36d3732a29c851b5fb2e9a
f9de6e230035ae6a931ed32968642d93d297cc7a
6007 F20101124_AACJUV burke_g_Page_164thm.jpg
7ffc783354db07247f35274e8baaa3f8
0a3d9a466037e68a9407cb5ba2b1f70a798fe392
24893 F20101124_AACKBE burke_g_Page_125.QC.jpg
e8c2b2629a06e6fb44e126aa93263559
2dce5177be9fbeabe6fa3db17214732d53fb0c29
2961 F20101124_AACLEH burke_g_Page_135thm.jpg
df563569f3305af8df5e44db158cd77f
860b8115d34ba2ba8b73687d0bc25d642d9c6189
6095 F20101124_AACLDT burke_g_Page_125thm.jpg
c44addf92afc118381230084a6baa5c0
c409777b9564c95a345def34f7568662af7c80ce
3927 F20101124_AACKZB burke_g_Page_027thm.jpg
23daec5c658881653bcc3fa7488c7485
1628957460d38e6cf021cff6b873570c34dba8c6
5666 F20101124_AACKYN burke_g_Page_017thm.jpg
f30862a388cc3ed57d4c29322d7310d4
abfa418265008477db94c0c6b464b64512e242fb
2257 F20101124_AACKXY burke_g_Page_005thm.jpg
2d8d529603a3150f953a99be0407f601
92767dcbc2a5803f72ccfe9b89ed9dac22fd0f4f
5966 F20101124_AACKAR burke_g_Page_034thm.jpg
a990a7f16c8dc2bf4ec6307812fd4089
2ece5e962880a96bce6ce0020cf5e41decb9ec02
22139 F20101124_AACJVL burke_g_Page_040.QC.jpg
06deba2b1f87ab8f2d65a5d5f03d53c0
6a57b794d354337799116a20c856b7e61090ec88
1962 F20101124_AACJUW burke_g_Page_039.txt
e0590384718d103e56dff284bf192f6d
cad49d02398c5cebdbe5e070b002bdc997ce4cac
5892 F20101124_AACKBF burke_g_Page_142thm.jpg
5e8f89e133a6682f1d8eb8f950ecec5c
9e3379bc72631187ef3ff67fde5e0156bfe9d86b
23362 F20101124_AACLEI burke_g_Page_136.QC.jpg
e1933db103319c610844565df1d93a07
cc3fcadce3073d5c673415c5b173f6e5d7ad5ee5
8929 F20101124_AACLDU burke_g_Page_126.QC.jpg
3ff28864451e201667faf8077be81f3d
566f71330ad0c8e403349d4dc970ace16b4be0ef
19213 F20101124_AACKZC burke_g_Page_028.QC.jpg
c3e446d0686631a3cf527389497c845f
ea3e3ac7cc310704a48c249606fa15e23dd03e39
22525 F20101124_AACKYO burke_g_Page_018.QC.jpg
38c3727c593e3db9b0dc69b665f88b7d
4e23960722fb2a204c5ff2ec187eb6bd0f29623a
17975 F20101124_AACKXZ burke_g_Page_006.QC.jpg
d8fd9df2f260093c15125a6054f180e5
76c7ce469e864be8fc961a87045e6e77152bd74a
5694 F20101124_AACJWA burke_g_Page_077thm.jpg
5cb47e4568177c52b0cbee84b2e93f94
09ed4931c1ebb089afe90681298e6f03dbe78eb4
F20101124_AACKAS burke_g_Page_025.txt
890bb6344d719aab8831d0bb8ebc006d
808f4f997b9259a665d9b622c59a5edb096f9900
49653 F20101124_AACJVM burke_g_Page_086.pro
84e43980fed5e11c52e64814d4524745
fc25203eb072eb7d00734a496e0a39f032759bb5
934000 F20101124_AACJUX burke_g_Page_092.jp2
048a7099346065a2daf9900fd5c2b590
edd62bba6ddbdca830620d5c99a9760e213043c7
23455 F20101124_AACKBG burke_g_Page_022.QC.jpg
ffac8e7a21389db06cad4e727d672011
6d98d326c0401b2f18d2704d7c745b855185292f
24000 F20101124_AACLEJ burke_g_Page_139.QC.jpg
9c7075aa5fb6aabf22495e39c68a521f
1b5d25616fdf147606f2670e15affcbeddefdf23
2729 F20101124_AACLDV burke_g_Page_126thm.jpg
8497e7d307d6534bbea6b772c81a17d9
94e8c6aaf10e3ef605a329986d2bd99b978b7817
5130 F20101124_AACKZD burke_g_Page_028thm.jpg
2aba820f9dae814cc8a8cae4a220a7a8
b30893042e12907a20def8196f34d8735c8cc4ad
5671 F20101124_AACKYP burke_g_Page_018thm.jpg
7794d83dec07dfd79719cb3b6b81348c
a3671b55971c6a3979f3102dc5d4e5631f6097c7
69780 F20101124_AACJWB burke_g_Page_131.jpg
31ebdd4055184dd8cef8573c6210a929
d363d87af90627ca4f4afed9f6042f4240c467e9
2753 F20101124_AACKAT burke_g_Page_073.txt
39ba475b341f0ce4f4403cd97f87d2a4
6e2ebaca82110f1f278ce8eb0db72d9c0e7efe59
24496 F20101124_AACJVN burke_g_Page_033.QC.jpg
003b91190500f084b3c6ac8ca1c375f2
acd7b5188819c8de4f4482d52e4cea0c3dc947ab
23627 F20101124_AACJUY burke_g_Page_108.QC.jpg
122c038a76c7b37a1c13a2060a030507
51efab0511fb0de1551430279c2f0faa2f3127b5
637186 F20101124_AACKBH burke_g_Page_088.jp2
5fd08c7f12380b84887e754e4a537fbd
e3ada5485a42981eba93d260ff8c77bc6abeb2e2
22245 F20101124_AACLEK burke_g_Page_140.QC.jpg
2eea2c420eb9df996d2eaaae9a1f11ac
097d3f8b7f6343033eaf42428dc454fca779d14c
24059 F20101124_AACLDW burke_g_Page_127.QC.jpg
b33c075ff0b8176673b1de6c501ba826
03800ddf9845007be8afebbb0f583ea92aad6b15
22315 F20101124_AACKZE burke_g_Page_030.QC.jpg
2b671dc7ca571f81bb0c9accbbc59a7c
4b1f06ff3d89a600af5a6f3d3e052c5a0c884228
22635 F20101124_AACKYQ burke_g_Page_019.QC.jpg
b9380dcf6355eef7baef8ce0b6968409
21c6908d0788badb62d2bdec929359d15f9811da
10072 F20101124_AACJWC burke_g_Page_147.pro
8c29bc4636339acf4a14a12a2c24cde0
ef08b2150ff4cc082d1ff9708667f22cb058a96b
67779 F20101124_AACKAU burke_g_Page_077.jpg
183fa62a517ac7903d281f058df8cba1
8aa25eb19445eeedfb2819d1a98cb2910c55925a
5548 F20101124_AACJVO burke_g_Page_045thm.jpg
fd854b038c99b2b27fbc46b7c8abe4ab
6c550842723eda73d8d9e6dc777520cdc3d70185
F20101124_AACJUZ burke_g_Page_090thm.jpg
6ffef00b9367101411625e7ba756c764
3d79e25e3a3ffa9a26ae68083c96404405585d20
22457 F20101124_AACKBI burke_g_Page_150.jpg
1572a8d546161296d3c87e76d3408b1d
9be67edd1b250b6c234957ded291e1de04bee7cd
3288 F20101124_AACLFA burke_g_Page_153thm.jpg
cd692fbfbd690eb18d5b4f6277581440
c5eed1cf77333c6ca9f003a92b99cf3256c34699
12013 F20101124_AACLEL burke_g_Page_141.QC.jpg
e6fc6d75b59831354ac629930e38c6b9
e824f40309ecd9cb84958b885a0d2670085b08ff
5923 F20101124_AACLDX burke_g_Page_127thm.jpg
54a1dbd5ebc621d980b5c5ba457ca1a1
7ccf8884445a281d2b6c395d51e0bd68ed318399
F20101124_AACKZF burke_g_Page_030thm.jpg
d5fa962bfe849171e774396bcf95af56
d3260a4bff97d8e59e223f13797ce5087b0a0817
5738 F20101124_AACKYR burke_g_Page_019thm.jpg
6921979f94a8118eaba4d426bea01c0b
82b46bfd29ba8ec5e9a93d012cbaa05f9809bca2
453 F20101124_AACJWD burke_g_Page_005.txt
3c581c9619c5672ecab87e6ea63fca0f
ffd8bbcdc6ad0017bdb9eb34d6ee8f25fef8731e
75219 F20101124_AACKAV burke_g_Page_140.jpg
80cc6dd53bc472cb09c04563ce4cb9d7
19ea0e28d0ca0f2aac2f52f27bcf38c04b3e7cff
71544 F20101124_AACJVP burke_g_Page_094.jpg
be6d63d26dece072c770ecc7f907b081
f2590fc39ccc22a96129843888c38aa2aa961ad3
22752 F20101124_AACKBJ burke_g_Page_035.QC.jpg
479c19864fd0151d993fdf36480f40c9
46b54a29d61d29c105c5b6d257bfc1f071624af4
5324 F20101124_AACLFB burke_g_Page_155thm.jpg
37fe89c3fd11a1629ded9448b0f34b05
141e61cd414b7088304cfbfaa7e8b6bd99875c9b
3424 F20101124_AACLEM burke_g_Page_141thm.jpg
c46010902dc720793c776cceadef5a09
fd35a4dcdbab9a606750dc4e364bfde740a0a45f
2797 F20101124_AACLDY burke_g_Page_129thm.jpg
81f467302bd3829db22ad8d4045144df
065ed2ee55e1b985c9707d757dc1a34428847ec3
22096 F20101124_AACKZG burke_g_Page_032.QC.jpg
e9db992e032719bca98c303ac2b0ca17
cb6f10ae29c9a0d6b25b5ccc2a26539e9799165e
22656 F20101124_AACKYS burke_g_Page_020.QC.jpg
e7db116bb81fbd64aa3d2de5f892e3a4
b11d32cf0da8739c8e13103e8b86a56d1ecc661a
9206 F20101124_AACJWE burke_g_Page_138.QC.jpg
b0cb6b41164aafbd352f9e0505c1a437
9b683d2d1aa0b8e27d022afd6b35dd972ea0bb0f
23565 F20101124_AACKAW burke_g_Page_054.QC.jpg
e34a7b481354ee4c350c974fbfc9004f
0fcbc45ac91723f42063a4c498b4ffe11d835910
40165 F20101124_AACJVQ burke_g_Page_142.pro
c579ed28dd8ef281a4024759920bd33a
0657647deccb33fb72f86616cfe7470b056adf06
86190 F20101124_AACKBK burke_g_Page_069.jp2
95d0b80eeb1c545881c75ec350da4505
69cfe74c07ec1e77b5ba5385fa2c3ed28a704bc4
24151 F20101124_AACLEN burke_g_Page_142.QC.jpg
e272641258e62b646a224174545e58ab
272a98da1fef9d481281c2749d908790c1279116
23619 F20101124_AACLDZ burke_g_Page_130.QC.jpg
d8243e8ef2e32820011c988388ed33ff
b75e8a5107f19effb8859b47be8a0b8a5867bd47
5457 F20101124_AACKZH burke_g_Page_032thm.jpg
576848197a4b0ad2180fba47fa9afa34
e71ffa8f939968f33cc884c7341e489c0c77ad61
5814 F20101124_AACKYT burke_g_Page_020thm.jpg
a674e36ed3df68cf6438649421422925
275c111d0322721778da769fd7d4533a0a73d364
5074 F20101124_AACJWF burke_g_Page_122thm.jpg
3754e65b920e3241d9a1e109f25a106b
0a18d1677820a85d2b202439ff0b06abe2a2da9c
104833 F20101124_AACKAX burke_g_Page_160.jp2
5c6e1c77c66e18126fb5ad991b04c0fb
dda1ff2e4d0c613e96e1edafdc36312217a0dbe0
70803 F20101124_AACJVR burke_g_Page_057.jpg
cc1d2cef87466993a1cbf5fddf6d61a5
9f9436512fdbd979c4257afc29d02eb4c3c00b59
844431 F20101124_AACKBL burke_g_Page_115.jp2
c197de7d527ddc84078db64e15139b98
64151340b21a0b4c0c92ab119c581ba507775dd7
4578 F20101124_AACLFC burke_g_Page_156thm.jpg
3781076864827b2b2d75fa5cd924940a
b3d7942620a030155de9fc71caebf975d7617327
F20101124_AACLEO burke_g_Page_143thm.jpg
6e7ba7cd642576afcbb9a02ebe7fa858
11ff181ea7db60716574e6bf1b40ac302b79a01d
6106 F20101124_AACKZI burke_g_Page_033thm.jpg
d3ccc9ccad2489416d6fc2620b04ad82
ad6fec8341e92473e7ac24cbc73851fda391fecd
22606 F20101124_AACKYU burke_g_Page_021.QC.jpg
85238f251dfb9bebb6dcae5687b2c775
c6a3a410c8ca9b3a83b1a973fb2cf768d1892617
2290 F20101124_AACJWG burke_g_Page_101.txt
56d6f84c3639cfb5ad0948e6e181bc51
451258a47fe4d527c5a24cb961c0c35011164f1f
64000 F20101124_AACKAY burke_g_Page_008.jpg
082663686fe92c710e6ffc01bc1eddbb
cd73278f4e76a7f5f7468109a2133b62017b9061
2156 F20101124_AACJVS burke_g_Page_083.txt
6d3264cf2af7cadfad9e4768cc070644
8eab9e6ca319a6f781c8d617755957b4c0267258
113593 F20101124_AACKCA burke_g_Page_026.jp2
c94d2a2639373e067810cf9efef7b3a3
2f7a2a99a0c5134ca4cf3469491aa1d562397bed
5693 F20101124_AACKBM burke_g_Page_098thm.jpg
c2d987f03abf8e0ecfd8c161ec98515c
bf456635826de360b6b55b0f58b6418433087414
17276 F20101124_AACLFD burke_g_Page_157.QC.jpg
7753a1513ad5b4e5eeba7795842e3c89
1b02b8f80fb1987706ee520a550f3cb47cf6d1a4
2621 F20101124_AACLEP burke_g_Page_144thm.jpg
dc9cbd5b35af476baedb2defea777ae8
2a1b11c56d8f500bc7294280d10356d65854de43
23261 F20101124_AACKZJ burke_g_Page_034.QC.jpg
53c276080c9bbda44f06ba685fb88220
a0edaebf1c77c255cf3d2ccff9c97710a42087c1
5736 F20101124_AACKYV burke_g_Page_021thm.jpg
945b99dd2002959ef447e1ad82660f11
61c87260c9ab42c6c47caa2a50809cf2f8d9117f
5938 F20101124_AACJWH burke_g_Page_104thm.jpg
1d156f10aab6ffe9565755a63e5aef2a
ce590f21547bb35b78ddb2951cbea64399897b8b
71655 F20101124_AACKAZ burke_g_Page_099.jpg
e8acf8c99d1d82a65c121bb3b15b7525
a6693163aea6ecc84a735bedd19cb6e740d5688c
96252 F20101124_AACJVT burke_g_Page_028.jp2
305f1d8f94ac2730a03b2036f95b5b9c
36f788a2b185911962c7fc84ea9b921abf4b0717
4651 F20101124_AACKCB burke_g_Page_065thm.jpg
9fc8c8a5e04cbef03ba17271b8daa0a7
d2bd2c7ac44615f0ea3a2e0b6fe7d10b53083432
6864 F20101124_AACKBN burke_g_Page_150.pro
b1af0aa16a4212f9ed084014795deb49
465f0c5d0c4c0b67f846bbf1ef8c1b1e6996365b
4300 F20101124_AACLFE burke_g_Page_158thm.jpg
0f6edb053f4827c6a8009cb935a0dbdc
87c98602f7b306e85b195b56e31aa1fc3e930bb4
24013 F20101124_AACLEQ burke_g_Page_145.QC.jpg
04a1b8a719a9e9a29327f361a965ad5e
582e9631278f7376e320a2cd4ddf8c33f345e65c
23790 F20101124_AACKZK burke_g_Page_036.QC.jpg
966625b9d6f8c940ca2d3ec86c61ed5a
4378f205be5ec03b42dba54b99b9154fb929d4ef
71038 F20101124_AACJWI burke_g_Page_017.jpg
7cbea22d65c8c6ad4b55a9a448ba88cb
2dcf892330d17b6eed7851b1b9d79bd71069beec
569871 F20101124_AACKCC burke_g_Page_010.jp2
045799aac9cc93fbee6c063a0406a781
400b0dda009bcd5d7bcf7a64e30109ff1b5e8fa5
2600 F20101124_AACKBO burke_g_Page_103.txt
0462b474130a220f8dfc41a532c37284
41fafe5409fa42d44b27e2d71867fbb6ee980d9d
6384 F20101124_AACLFF burke_g_Page_159.QC.jpg
03034542692b7eb9fa28ed15165a3a1b
f778afe81322383904ddc9a477e9c6a1a468e2c6
5642 F20101124_AACLER burke_g_Page_146thm.jpg
231da6ff723edb4d8db8ecf8023d7284
836a01bb4260e8b39c9fa57f49d2bb21e65d05a3
6034 F20101124_AACKZL burke_g_Page_036thm.jpg
c0d03f4a67ac0a7d00f8f66151bb76d9
dd6f44ebde64995dc2cf088cb516e55d372df85a
5824 F20101124_AACKYW burke_g_Page_022thm.jpg
2559ead9188f78f31e665feb8f993ff9
2812e5ab50c642996a13675d07dc77547c77c486
22223 F20101124_AACJWJ burke_g_Page_057.QC.jpg
32ba4a813995c7c9e90f80bac39e1c14
031743c414b4518d9c570a17cf541fb13d80228c
F20101124_AACJVU burke_g_Page_044.tif
4790aaf2c5ab451bdf2dcb8bf923fe4f
a0d4b17b20f4db21442e6cacef38fb324790bc3c
F20101124_AACKCD burke_g_Page_140.tif
3a6c999307b04491a31bd83ed2106857
976492fcc364df06216fc8936765a4da4daf9b5d
53095 F20101124_AACKBP burke_g_Page_051.pro
1c36cf730a0062936b8161f616043e02
d8a62b71767ff09599b409aea33a0188c25ec1d2
19418 F20101124_AACLFG burke_g_Page_160.QC.jpg
6c55f909745170995f12e7ddf6253f0e
5c4575c8f9b472cf912bd20a6e728ad48d1cb13b
2777 F20101124_AACLES burke_g_Page_147thm.jpg
f07bd276d85613d804e6d497f2bc5e3b
05e9112b1b6f10aaf35d35f809cf5fe56df803ac
22939 F20101124_AACKZM burke_g_Page_037.QC.jpg
e84169b7ff3d1e1106756e632d68e1f9
10fcaa4a22ac234b073a3e30eade490d5409a740
23209 F20101124_AACKYX burke_g_Page_023.QC.jpg
93e1682752b0b71dd8f1130ea04f22a4
6cd54934b2abf8d80fb60974964b92140686b60e
29043 F20101124_AACJWK burke_g_Page_012.pro
1a90e6b7fbb69e4ca53cae0dad2d2be9
c3623583ca4b22b10a2dcfabc5457fa8783e5be4
40576 F20101124_AACJVV burke_g_Page_128.pro
c1badde52756c4d46054801c9272c19b
624f3204f93e6300a977f6da946db14dc9ed020e
F20101124_AACKCE burke_g_Page_116.tif
26caa36ed54f4a2ad3c69680023904eb
98351ed12258276b8bc1094ffd7f6942cd0be723
F20101124_AACKBQ burke_g_Page_151.tif
e14a0dccd7d0876f7ccb46d4a2d24602
b9afb632c1e17c1a09106c8fff93400a7d857c72
5379 F20101124_AACLFH burke_g_Page_160thm.jpg
7b90f3ea0b6c131adfe84e1de04044bd
fab93247b050de12f55ddf0cbf285e2527aae584
5846 F20101124_AACLET burke_g_Page_148thm.jpg
7fc2a2e0f62188edd499bd286439c86d
0b7a97d225f1e258a58ca153b01531b5cc81130c
5860 F20101124_AACKZN burke_g_Page_037thm.jpg
aff57a34ef3b4f82fdf638134cb868f2
2cfd3d7532f68ebf57c7d17d55f2282a68b34454
5964 F20101124_AACKYY burke_g_Page_023thm.jpg
27fbdfd8b7ad81c162f86f4866380323
1e7b3b9580a93fb4e976993cc9cb8c485b0a0f34
2262 F20101124_AACJWL burke_g_Page_107.txt
3f7200fb93ee254a4d4c593b09ed5449
a545fbc7124eb9f4b3a5bd23b13e5da4ab972e2b
90759 F20101124_AACJVW burke_g_Page_011.jp2
b1043673ed4c1679f99c76ad590e752c
549987cae63b3ecd033a89e4850d9df8c2f1b1f5
116554 F20101124_AACKCF burke_g_Page_108.jp2
04400bfd98a9ab89fd1481ca8aeb7bc3
06f387db9d6abdfb446c91ce7308b28f8ca447f2
1932 F20101124_AACKBR burke_g_Page_048.txt
eb477c14f495a1d8f8bfb67e3c51e2c6
97b6929314e64b3dc753409042d445c0f5ad7560
22592 F20101124_AACLFI burke_g_Page_161.QC.jpg
cf64edda64419aac05e426c1d2741e02
2d417aea91b390e12c0247c4a95e5baa55aa5802
6049 F20101124_AACLEU burke_g_Page_149thm.jpg
fb30e49965b55c16800a95b908f4a586
2be93ba8352d285f7edb4755d6e41196a9588c9f
4593 F20101124_AACKZO burke_g_Page_038.QC.jpg
1017dd449c598698fee39c958fde88d0
1974887dcf40b6247a4e307a4572fdcf4b449e14
22038 F20101124_AACKYZ burke_g_Page_024.QC.jpg
332387af33702c5d8f48091230105c13
8938879158ac17b90d20c8bb3dd7cc870700d3f4
5151 F20101124_AACJWM burke_g_Page_072thm.jpg
1358db63f0abcb4cb87cceabf6afc69e
a428f46afe09ea395c6af3bdcd4e9c91185369e0
15058 F20101124_AACJVX burke_g_Page_027.QC.jpg
3dacc39b9751d9cef6f95d60d46de27b
0afad4d97cb8191fb067925a20d5a730f4bc0023
F20101124_AACKCG burke_g_Page_076.tif
6047228959576e173194545e0a7fc65e
1e5911e3e4a9441071d334ba51297679fbe09f59
F20101124_AACJXA burke_g_Page_091.txt
9afcc4f9f8b187f3e3caec5346bcad5e
1e44ed6f277e8c3ca3d4b948dbc6f73ad6c1feba
13709 F20101124_AACKBS burke_g_Page_038.jpg
9df34875d0545fa1f1f983994cde326a
7445a2a5de42f5643d20c8a754f410373995ec01
21874 F20101124_AACLFJ burke_g_Page_162.QC.jpg
9061f8111952982ec29e6cfc28168bf0
2095a88afaa89dce6b3d7138e05b521267dd86a1
6940 F20101124_AACLEV burke_g_Page_150.QC.jpg
1b0700fe885fa9fab2da52fbb8759fff
d56c8bdbc8bd5d94408e311c44cc0ab62be995b3
1716 F20101124_AACKZP burke_g_Page_038thm.jpg
cef63ba2c91fd24e3c0e77f172fcc101
a2315f7de1e7655b572fee40ebf9797833b48b61
2187 F20101124_AACJWN burke_g_Page_110.txt
3b4106c59b6496691a8b44308d1df395
cf1e79cbbeb554d73e43c1c1e4d96ce6dbc8632e
88705 F20101124_AACJVY burke_g_Page_148.jp2
0433323c7466ed3c5de370b17e5187e6
1392ab75d2e1968efc29bac821f6e26a8acbd9b6
34709 F20101124_AACKCH burke_g_Page_081.jpg
0d9f053734adf98440da1d3e825eaec1
17b9cdc1c6cbf03cd265a9643ee21713afdb58ec
4876 F20101124_AACJXB burke_g_Page_157thm.jpg
490d7451041c48361099981a35e6c818
2cfd7d0a2bf0363f859fe783a39bf21f578032c6
5741 F20101124_AACKBT burke_g_Page_031thm.jpg
5b89d21e32635d0c25edb1c80060ac04
7a3c35418359c3c666e47e6a07c8c24e93cd3458
6047 F20101124_AACLFK burke_g_Page_162thm.jpg
96d3a4136d6a9798f5a90419d11bfff3
08f4248a90d4a8d967d97cadf24d300bd632e573
2210 F20101124_AACLEW burke_g_Page_150thm.jpg
8a6c0b4ab8f060adeaaa3c0bfea0aaba
7d61de8e0bd048d49100d8f8c0890b6b5cd82e47
19530 F20101124_AACKZQ burke_g_Page_039.QC.jpg
4a21408b84289f3ee5737d27ad340f23
58661d4462767eeda9dcf58b35b87c603fed7900
56280 F20101124_AACJWO burke_g_Page_116.jpg
bc1970eaa9455b7d2f3630770b798eb9
c3fe17518fe33949a7c7de7b28382fd6d3614829
5317 F20101124_AACJVZ burke_g_Page_013thm.jpg
d3295466b16793df6c0b184e838d65b5
20180dccfb15681211f4aa0396a05a0d66ad79d5
2126 F20101124_AACKCI burke_g_Page_051.txt
72f8382c039d902cefe8db2da8683a48
1ec49905a33e30707547e5007d384423ec12c3d3
56115 F20101124_AACJXC burke_g_Page_023.pro
73031dc5ac3060b6c53a4ff1593cb603
f11756e7f6a0e41bc06cecacf0f1e2184ccb4b96
1582 F20101124_AACKBU burke_g_Page_166.txt
98926becd5a4a7a2fb542ad131e602e7
4bfceba31b679a495ad2e553862e53934bc7eb77
5977 F20101124_AACLFL burke_g_Page_163thm.jpg
080216432f65778c4c8ba28a39209b0d
521a57a02f3717d8f8d2e86c1f7c612e15fdc5f7
5913 F20101124_AACLEX burke_g_Page_151thm.jpg
76fce47d92b75589d476b0b3e579f2c3
ac54a1efb3982ff172b46309a5d78a5e92871cea
5385 F20101124_AACKZR burke_g_Page_039thm.jpg
ff7cacd35a4dcb4bc3903d20557ba094
d820d269fb70017e6480b11368c8fb836e95774a
2850 F20101124_AACJWP burke_g_Page_122.txt
53c72ca17dfd8e77e2684963e84698b1
691a49f8db20138cd9ce2d80c34ca1c28fbcf0c3
81695 F20101124_AACKCJ burke_g_Page_143.jpg
f9fe207b75f43267712c8b5db1826586
9c87b4cc63daf98e427a36c28e7f7e2212435502
F20101124_AACJXD burke_g_Page_068.jp2
c6bd40424f8a88ab4358917adf25e6a3
c7195cc081cc69441643ab03a8ee6909136a3520
891350 F20101124_AACKBV burke_g_Page_043.jp2
fe382d3738bef83701a7e31bd1942308
21135bcaa56ba330692d58938c064d665e86e259
22306 F20101124_AACLFM burke_g_Page_164.QC.jpg
94f5785d2833809050ff68b87f2d080a
aa18db9b26df40f1710c3d36b4f2dd7daa7a9f67
5838 F20101124_AACLEY burke_g_Page_152thm.jpg
0c0aba4ad838de5d15bdd0c58cb201e4
8ac961dd034dc28436eb0b6d7656115a02201340
5554 F20101124_AACKZS burke_g_Page_040thm.jpg
ad7e151ec15f33a0c59c99dbabc6b258
e5212c689c44b1daf8b5351439be72cdecc3ac19
2086 F20101124_AACJWQ burke_g_Page_024.txt
114f3f4c40d64b2451adea6de6a4a2b9
96dec831d308f89c277abc9e2fefff437df57a70
4869 F20101124_AACKCK burke_g_Page_111thm.jpg
424f44ac387fbb7ec6c89c18d72676dd
a6ba5c2634074c0a799b995f3968c8e252f043b6
54709 F20101124_AACJXE burke_g_Page_095.pro
686bd37997efb1f800878bbf97f9ddb0
1e6797b6d1f13d8435fafd5cc65bca4a482e0df6
1814 F20101124_AACKBW burke_g_Page_133.txt
f5bccda842368e0f9cc8b68c0541b4f6
fcfa114fc55e7dc74ea0fa9748250c00d72c39b2
4107 F20101124_AACLFN burke_g_Page_165.QC.jpg
102f1292ddd4fcc7ead80e90b1be203d
f9b7f89e4d1ab201ac413f0b83e19f7838bc99bd
11580 F20101124_AACLEZ burke_g_Page_153.QC.jpg
4ed7997cdf4695beaec536e02ff1da7e
414bbbb8ac8a9e10fffc1706dfda9ad4164f4ff2
17307 F20101124_AACKZT burke_g_Page_042.QC.jpg
a9186d17e4796f21751cb19fa11a1d18
73804868957d838f1348aa65009ee161dcd779cc
27168 F20101124_AACJWR burke_g_Page_059.pro
ee9ca617b6993f69284650202e0fb1df
361ef3c060539e738a8755f64c20a26af170d061
866 F20101124_AACKDA burke_g_Page_081.txt
f0f0c9ee922ba68713444f48d69788af
6420b436cdcfcb475f6ab92bcd5419274cd834e5
5504 F20101124_AACKCL burke_g_Page_079thm.jpg
94bdcbc5c9d6b1c3ad180fcea3822760
a0ebe1d9aa88253fd91b50703f88610e51e4f4eb
22828 F20101124_AACJXF burke_g_Page_015.QC.jpg
204177b1ec80953abdec5f7f75de79a6
b0b994bd6ee9814908f800fe6eca3406e22132a0
66850 F20101124_AACKBX burke_g_Page_075.jpg
874677968610826a333015ecd881f707
c258195967661e88dddcbc11d368c7ce7970591c
4610 F20101124_AACKZU burke_g_Page_042thm.jpg
e186aef28f32cd0625a77dc47b389679
ce91659b87a26e5bcd6699b69ce8a818e7bf2bb9
513 F20101124_AACJWS burke_g_Page_144.txt
b260f096773883cf8755b251585f7ab1
da02671aaee50b29ede065df033ceaea00c1280b
44594 F20101124_AACKCM burke_g_Page_078.pro
ca2abd8abffbd7a958e250f6827ec22e
a3a7e4c142a3179fa4d72f2e785fdb3fffedb827
71623 F20101124_AACJXG burke_g_Page_096.jpg
2f988cf85cb46f6b1e63a10471faef37
6e50391541f001d4c702a4dcffdf96671bc136f0
46031 F20101124_AACKBY burke_g_Page_074.pro
0916b8e592e2d7e66379109848874568
853ad00c69b7b9054af11036e4861a7b5b73f992
19840 F20101124_AACKZV burke_g_Page_043.QC.jpg
262c03f8381d64a6fcab7e3fad72f8c9
1466774ec92ce14e522ccae358952deb34426c02
87452 F20101124_AACJWT burke_g_Page_133.jp2
cdc5839c4f6026d05b5a2a4d67bcf10d
dbdd7b63ca2bca60a7a4981eae4856eda7ba3a63
1918 F20101124_AACKDB burke_g_Page_067.txt
ecce815c02c63624ad7a933f3056a75d
3b3c76d099ac4e2b1cf3c93cbdfa1d139513b1eb
1849 F20101124_AACKCN burke_g_Page_116.txt
604dc734317a992b99ca5120edbb6293
8d21a0f7afbbf7e3a64d2bb03cb95f98ba5485d4
83728 F20101124_AACJXH burke_g_Page_146.jp2
e1a47ddd9b5f8d60f09a70052dee170e
4f0c796e15d6a484e0201f956e6420d0b8b23455
10409 F20101124_AACKBZ burke_g_Page_159.pro
a47c999aa79c8e374046f7384a571036
fa814d05abcabe4a223aed5978c60d5a62d55b1d
5269 F20101124_AACKZW burke_g_Page_043thm.jpg
31f526c3eba52cb3e63176e32ea1f0d2
f035c09dc1d2cd3726b35db524ceaeeeb73a45b1
54694 F20101124_AACJWU burke_g_Page_026.pro
24c1717fc6f96b362480bab4abcf26bc
7793af08e8ced3848e2413186b43362aaf280757
F20101124_AACKDC burke_g_Page_146.tif
ab4d778892d8c383de0cf2668f1da4e5
966b7ad282d262a9b67f041bf451eb9cedd408b1
1675 F20101124_AACKCO burke_g_Page_044.txt
715f693a82d933ca120e38153204960f
1a46de8a12d133a717c1a19f4dbc4db0f8f92a76
23146 F20101124_AACJXI burke_g_Page_104.QC.jpg
b61310fbb861cb729457a853486ea748
d2686f9e817711fdbdabbaef45204e104d40ddab
4641 F20101124_AACKDD burke_g_Page_166thm.jpg
37922f9d165f7ada923962ae0cd5d973
948023659ecdf32f0aca42774c2a98f3c8d761cc
21927 F20101124_AACKCP burke_g_Page_077.QC.jpg
429c6fb87cf8d0b3e6a22b30c7c818f6
c04bba83e499f234d7f042ca97286586f3ad42e8
F20101124_AACJXJ burke_g_Page_025.jpg
c8a6b5b8b983ecc351e8b3fefa5f913c
826d904bacb2607f9b88cb674ed7a26323cf484f
18546 F20101124_AACKZX burke_g_Page_044.QC.jpg
ab135e788a576ee90226a28dc231190f
d867bf3aaeb40403ce6c3f5de19e457a3b60b4aa
F20101124_AACJWV burke_g_Page_085.tif
5bffb1c9b815cd877b8d25bb0bfe39f8
0b7ae0c325af4018defe99b7b399d6532b74c0e2
5263 F20101124_AACKDE burke_g_Page_038.pro
81dba634c1ccf9db599db12e01fcad03
992ad481900d19db3d0b6d625af9db60d04d29c0
20585 F20101124_AACKCQ burke_g_Page_001.jpg
15ac40c345df83058d18e1287be91795
5d1d10ffd7593b1462c815db31ea4b07ef9ec0b1
78090 F20101124_AACJXK burke_g_Page_161.jpg
b49611722a05f6c6f6c6d707b991bdd4
6b6ea08877558d1cbb948e23815dab7a648d999c
5024 F20101124_AACKZY burke_g_Page_044thm.jpg
02fbb4be6934030c955a57d58678d3ec
98972323ed8a583be5f1bca262d9237de08c133c
5275 F20101124_AACJWW burke_g_Page_115thm.jpg
ed5b53e184de6fe4b6f1207e60a3ca4e
c84f553fb1b64d3360c00de2bbaf27b254f91f35
190202 F20101124_AACKDF UFE0011378_00001.mets
939850bc278ce3bfe7e63736808711b0
dff29a156273cef69df45ab393a4491f3ee2b726
114483 F20101124_AACKCR burke_g_Page_058.jp2
185fb8834c4071a352ffa23c39cb5bdc
434e5fbac898c78d6235082b160484807f4a0a17
5994 F20101124_AACJXL burke_g_Page_035thm.jpg
0925aaf32932135322248a60033abcbd
4cd88035e236c82546a376296b4adba5cb903e45
20602 F20101124_AACKZZ burke_g_Page_045.QC.jpg
c4301bb130b313aaf68ebe4941e100bf
7eeb9d8babf262ce17fe4641743c2a3649417f56
59235 F20101124_AACJWX burke_g_Page_059.jp2
f578503443d338a8989ae946fa560ed3
a179542e3ba757dc161454d7314c96106daa015b
47003 F20101124_AACJYA burke_g_Page_039.pro
d1b6689e67fdc5bcdb36c60aed617c0f
e0419ecd7e396ff6da321624a5b3aca2ec121f09
76912 F20101124_AACKCS burke_g_Page_085.jpg
84eeb86325359cae9dcfbe59fef7a60b
fcc6e398c0bd497ae75cb61eba7b1e32982c34a0
62732 F20101124_AACJXM burke_g_Page_004.jpg
61f04956321b0244e728d8d7fa37c64a
412799547790b7fa8f3b5ca34aa5d895779e2085
1641 F20101124_AACJWY burke_g_Page_042.txt
753cc69b48abf05da112297d7cd43a0c
4f807285b04d38ecddc30abe764fe3239eab4aad
1954 F20101124_AACJYB burke_g_Page_071.txt
42b313e1279f8e6e3a9ddd0d1b5d5411
cbb4aedc2d8fe7d33fd267c1676e3d8f7a528d0f
55107 F20101124_AACKCT burke_g_Page_096.pro
fa719d379a23edabf122f19af281620b
f4b26763bf615bfbf3380fc76a854e99b70380a1
63460 F20101124_AACJXN burke_g_Page_012.jp2
cbf243df606acc234fb2bb6f60e053ae
8063aa8473290632d6abccde791e5750640ec8ec
35443 F20101124_AACJWZ burke_g_Page_140.pro
24f8357a4f9ec7d62769b87c07512bcd
9bdaa28e3adb9d0d9c280909e8e39e39200eb314
11343 F20101124_AACKDI burke_g_Page_003.jpg
ce05c8b78e0c125494c9a87d1383dc41
362ec4b116e290ced2342fae4556479a6b1bf5f6
66518 F20101124_AACJYC burke_g_Page_082.jpg
cd2f5b8fa2ae4591424d05acceac3418
77240cbb665dab57357e0fd390d23f884cbe7d27
24788 F20101124_AACKCU burke_g_Page_128.QC.jpg
e672122cc24b126858d5c19a3c8a08f7
c18ddb04edd8e40e960bf8c6bbf08566a9b43d4a
90628 F20101124_AACJXO burke_g_Page_097.jp2
748d962eef88676116efecf131aacb49
4d4ce8b40bf9c107c354166cad7ecbf784a8b6a0
21029 F20101124_AACKDJ burke_g_Page_005.jpg
7c3dac3ec9b134bfad59c63aaf69d0c1
a332a78f8c2b09f9a304d2997476905fb190e6ee
85569 F20101124_AACJYD burke_g_Page_052.jpg
3c2f643cadef8c633066d4e7efd098e3
bfd190793234398dc29d24fb365515378f54bc01
F20101124_AACKCV burke_g_Page_051.tif
f67c55376365179d57b2de4ac00fb72c
9a1194011d276355673ae1183b084e24c3aefb48
1407 F20101124_AACJXP burke_g_Page_002thm.jpg
53f82097cb6c4de87d02852cd4e3de4c
b0e66bc89fa1a04469c5bef1c0ea90107b0a391d
27140 F20101124_AACKDK burke_g_Page_010.jpg
8add95f6772115dcaf3861b409069376
70fd6d3f38ac5a00cca62d3986456853f58f0de7
2029 F20101124_AACJYE burke_g_Page_032.txt
f712a0a0af38a807938243d8d4e97f24
7594eca747470780e2af7301a52c9bd075ca2547
21358 F20101124_AACKCW burke_g_Page_048.QC.jpg
3aa872cbb02a0ed1a09f3daa426ee7d7
397467fa0c25d043a52bdb674d3f6e55b1594a9d
22346 F20101124_AACJXQ burke_g_Page_017.QC.jpg
4bba6cd34a5b75ebabea0bd7afa3921c
99b6857372e73b68cc2596baf693426283225ae2
42480 F20101124_AACKDL burke_g_Page_012.jpg
aa73f0c9d5d18a43aebc19a9ab8915e1
67302ff5ceaa4e5287e949f5872cfa892599179f
F20101124_AACJYF burke_g_Page_101.tif
b92c880975d62d1fa24be2f333ebbb73
a4c6f797f5c97d30ef22b8932b55569566fcdc73
6257 F20101124_AACKCX burke_g_Page_001.QC.jpg
f9c338f5aa1a038e15bd3f26b257f8fa
87d9c45ac61a525bad3e8b2f40da178cd6951da1
1821 F20101124_AACJXR burke_g_Page_130.txt
a7960d22383beb54551fbc5ebef4b9bb
a1c255c56a23df1ce54fdadbd0aee56b4c1a5e60
69401 F20101124_AACKEA burke_g_Page_032.jpg
329d978d59ec29bd5445fff080febe20
baa9d06ed9cbf5e80a351232a86820817883d034
64231 F20101124_AACKDM burke_g_Page_013.jpg
632cbb47c418d4c6dc3f0160ef4bc6c5
5637a0f4d6dbc15ac2eb9e3d0e9645dc77f53c16
F20101124_AACJYG burke_g_Page_056.tif
503304dd82babe1516d51dd8d9cfe6da
0f8aaa94d632484c44b2922beea36480550769d1
66930 F20101124_AACKCY burke_g_Page_052.pro
56fe8d04a073151f5f7ec4a9e54aba6a
f29684aa22feea93f38ae2baf1a3ea38afbb3693
F20101124_AACJXS burke_g_Page_030.txt
7b4b95ab96081bfd024ce231cf291bfd
7b86e39b08efaa60763a0491f835aa0524974d80
73271 F20101124_AACKEB burke_g_Page_034.jpg
44ef39e4cfe115f43b4470a293e395b3
144e12631c70026d49ee002690e3e03c478671da
71095 F20101124_AACKDN burke_g_Page_014.jpg
b14ffa4d061c5f3f76596bd3135d67a0
12c316f88b5a7aacd3245edb57b9e72c4318c4ef
9391 F20101124_AACJYH burke_g_Page_144.QC.jpg
11679bf145b339a559946371d60cf460
79ece6b8a388862c421e0b12c8c2d4e73d379f80
2239 F20101124_AACKCZ burke_g_Page_058.txt
de41cc9cc991f0af19fe4193acc3a6a5
43ac21cd4567c342aae10e73a09f10c5b1e87753
115825 F20101124_AACJXT burke_g_Page_031.jp2
6ccff9269c711bc1d79b6e4f54dbfb39
de919411373d6932228626b2d147504c388947d1
71006 F20101124_AACKDO burke_g_Page_015.jpg
194f9ef61cbce6ba9e12ca52f007c775
0bff86129264f297ac64f5fd54b172ca7e8481e0
6015 F20101124_AACJYI burke_g_Page_137thm.jpg
1de7f6c7abf5694b5da898a73b22c807
cb73ae5b94089d6faf0c4b9251eb246d262c08b2
1788 F20101124_AACJXU burke_g_Page_069.txt
3ab3185377e85941582ab369f21fff7d
b2b52f511a557fc861af6071ebd0882c1001c94c
72222 F20101124_AACKEC burke_g_Page_035.jpg
17c29bb1889143b78c431da636c82592
e8df39a5f2d0e699c09d8f74010976417a1a202b
71462 F20101124_AACKDP burke_g_Page_018.jpg
37509876986d82cddbb4e0ac492ae74a
c828d79a7eef75b09491fe82abd8de0967e5f9e0
F20101124_AACJYJ burke_g_Page_042.tif
9bd5cde3329b3f11f16cd1f8c7bf91be
9672d6b1bf1c95b4d25eb27c8a9e001fa7989773
5785 F20101124_AACJXV burke_g_Page_051thm.jpg
fbc64c871f51aa08e2fbfab9786f39ae
732389d4b94669e0ab72fb12c05f41427bdbac13
72585 F20101124_AACKED burke_g_Page_037.jpg
ee109ba8b8db87be529ebd3ca055b941
8c47e09fbc12b4a721c2bbe2676b3b8552f47d2a
72261 F20101124_AACKDQ burke_g_Page_019.jpg
264e2b5ad18f5c46ac3a8e0452b15329
713144bb9247a425f1649b6d4ecccd39c96d4a78
61060 F20101124_AACJYK burke_g_Page_028.jpg
552cf11f5f47df17bb3a6f95cf1bdf34
46b60a4400bf6450fdcc82cf116a546f18e5d6e7
62227 F20101124_AACKEE burke_g_Page_039.jpg
a86d23916fa46344b3dd023d82a8e85a
671473e4c44f11993211bd546e635cadd972b251
70663 F20101124_AACKDR burke_g_Page_020.jpg
5d2a1711afd8a4e5445379940b8bf875
2469a298a977ec96129f8a9a5ca7471f1942f3f8
23802 F20101124_AACJYL burke_g_Page_133.QC.jpg
7778bb43d1b306e17c819e0ec8ef8f90
cf84d1d85d13631a991799c706f471d98016b959
720 F20101124_AACJXW burke_g_Page_120.txt
54a9355c695f996dc6747844af81f2e7
ec1626d7eec931109bdc644e407ba3e7f53a77a6
69253 F20101124_AACKEF burke_g_Page_040.jpg
01e1b8c6a508789116e183a77190ee05
506e5d310d69f1e8a0dc466721b5154530e05d4e
69843 F20101124_AACKDS burke_g_Page_021.jpg
6ce554c0c9c21bef680b93a935779ab0
5867bc7f11899f2487a05ac2e08c3244888cd8f8
2147 F20101124_AACJYM burke_g_Page_115.txt
515c1d11f6587d84a0dccd3c24089540
75f61a23ff4c7883b6348b572c2ac3041176f2a5
24254 F20101124_AACJXX burke_g_Page_143.QC.jpg
1855b3942e9e51f7c2ebbfd3514c1ba7
091e29601263c4b889b9194ea19e453364fcefa0
72527 F20101124_AACKEG burke_g_Page_041.jpg
bd486a15dbe035b23464c881131c9ca7
514a59611f8ac6002aeeca19a55f80dd7220fac8
73212 F20101124_AACJZA burke_g_Page_108.jpg
3b0e3a0c058a51674e86ea496ac97a65
ac004c1a8d284e0e58b72d11f359a68c7b95f0ff
73949 F20101124_AACKDT burke_g_Page_022.jpg
00f39c298cb189ea9acda77558c829b2
c6d198cac345a17b83cc6b2091e9cc7ae73168f2
19194 F20101124_AACJYN burke_g_Page_074.QC.jpg
86b2b2f5ddd323e02be61a182ed55b14
7e303eb0262847fe4273e5808ae6d11c83f0146b
F20101124_AACJXY burke_g_Page_098.tif
896f597d103878f7cb4a778d5937bee9
f021a2c8e6f568b2734c2808ce25ce2a14f188d8
65391 F20101124_AACKEH burke_g_Page_043.jpg
31e548a5485092c8e1c1cbd1606b36ea
7c4a4079801b486827e7c60167f370ff004df19a
24539 F20101124_AACJZB burke_g_Page_149.QC.jpg
b85d89cf4ac47721d8466b245e5624c7
3e2cfc65442d68c93931aa5d3d2c3882c8b93c0c
72694 F20101124_AACKDU burke_g_Page_023.jpg
45c878bc94141206f71eb7114be1a627
597035b0a727821fa17a97d88cc90ec1e6553f98
F20101124_AACJYO burke_g_Page_163.QC.jpg
87a86356ae523c627c3ddf0636c71e0b
473bfbba40075d6d01c6cbb47c41e2518cc5eb83
56317 F20101124_AACJXZ burke_g_Page_061.pro
3a07295adfeb5f0d0b8e510110ae5f76
7f62d72e59181b004f666ac8301c02740eac9153
66672 F20101124_AACKEI burke_g_Page_045.jpg
d9669129fe355f9ba1c3dc38b0b74370
c7d19838446191c3b536e56e3f159a561c39a4a7
F20101124_AACJZC burke_g_Page_047.tif
c6fb7414080fafe49d056f072537be0e
22fd0a29debd9a15e0b8a4773f9a8d0e75f70369
69835 F20101124_AACKDV burke_g_Page_024.jpg
f8331c10b66d7b0596c4f6b285835583
961bc417945ec71c0a8a2ca29b7943fc5e91562f
5750 F20101124_AACJYP burke_g_Page_026thm.jpg
24e1aa9e8f34f6bf4b6e68c094a1f22c
2e27bbdd779cbb8c682e203f3c13999d24ce1bdb
81568 F20101124_AACKEJ burke_g_Page_046.jpg
0bbfae6e8dabb5ddf517b3089fd961e8
b7159e6c5d53ffd49b3395e489f8b22c4c4d1578
17448 F20101124_AACJZD burke_g_Page_155.QC.jpg
97270e0a2349d67e0831876783c3c330
6c46197f0e59b917c98d3836214505dcd0d52712
73101 F20101124_AACKDW burke_g_Page_026.jpg
1b122cb4821dd6d5c532243e91ada3f7
5fce617dca22514be4c72bdc741cff27907bab4b
9131 F20101124_AACJYQ burke_g_Page_147.QC.jpg
2e22e2b4415690b381b96c1b49659cbb
8e0a5added571de63efc8591ce0586894cd2d3d8
60433 F20101124_AACKEK burke_g_Page_047.jpg
5a89a983b28b5aab54676695703fa4b1
4f05396191b8962b59f8e669d273d09bb57adfb9
79021 F20101124_AACJZE burke_g_Page_007.jpg
469a1a484229f4f9bab0c20235cf7b74
01f4773647ea0da44b5928c89c2ead4ee6b20682
F20101124_AACKDX burke_g_Page_027.jpg
28d163501ffd5d27d8dd83073da1350e
11a20047c37b928d7c241696814b86e6d1b3117a
2305 F20101124_AACJYR burke_g_Page_161.txt
3c8ab9a6d1a80967005ecdcc3fe0b9df
ad0a1ddeef4d15d0faea6424d684ff48e595f7ba
64896 F20101124_AACKFA burke_g_Page_067.jpg
de1cbb89f237f337e82b0491325c8e8f
573d5e2aff56cf98a42360782fe6e7c32168eb9e
60139 F20101124_AACKEL burke_g_Page_049.jpg
e1e3fe3be65c48776747a8a4f0be1a30
33342078628e5050bf95c4440b9f23c95024485b
12373 F20101124_AACJZF burke_g_Page_112.QC.jpg
e320f677cb3ccf43e2cc48f076b22d02
a17fa95fc803b0555b6ac779d5d12e5f83f94b68
71041 F20101124_AACKDY burke_g_Page_029.jpg
66ef2b6757de91b0cac827992b657110
0b0995873cdb4e887f9a365f4230ceba62f7cd81
2039 F20101124_AACJYS burke_g_Page_086.txt
d3ccb862e3da44da9bf2afa16882058d
45d77cd591493d17702f25fb62db66dcfba63a1f
57206 F20101124_AACKFB burke_g_Page_069.jpg
1d0b8c66177aaf4500a0543fcff31b9e
654dff3f9c8dc6f02b26e5ae226759b90b354b43
60137 F20101124_AACKEM burke_g_Page_050.jpg
0c22aa6ac779e21dcc89f161e0c9e301
4b3e082d8f7e9eea87a6b047903205db9757c5a9
21009 F20101124_AACJZG burke_g_Page_041.QC.jpg
783c7267535959830dfc3241d51bff5b
f369909ff56c951b0e8fbab58381bf81229aa7b3
92139 F20101124_AACJYT burke_g_Page_075.jp2
561790dabbe4ef1487400431887cf456
39ebd72554bf2362536bc29f5d7a546588218c2b
52291 F20101124_AACKFC burke_g_Page_071.jpg
cbe52982373741d32a73e774651131e8
9d77d7d06b41f4cc60099b3ffcd35d886c57ceaa
69838 F20101124_AACKEN burke_g_Page_051.jpg
d5c6b1807f829d0ba552c0968a3ca800
faea149198d20c989136ec06c1834b4cd5ee0fe9
2692 F20101124_AACJZH burke_g_Page_003.pro
d4f2ef1d2e54da2780e3e0dbdac7aea9
1bb0915a15f7a1f2d4ed0181b7832af599c7755a
69604 F20101124_AACKDZ burke_g_Page_030.jpg
1fc7f6dc020e96161395149de42ccb2b
d97cd46eb2d0d3d431e074a57d521cfaf53d22f9
55155 F20101124_AACJYU burke_g_Page_046.pro
f92336d7f8b873eb37adfcb299a53366
26d4d323109430e1cd5c8f001868e52025bb1d04
79426 F20101124_AACKEO burke_g_Page_053.jpg
d81bd3c6b3fe7b06e35da6b926b8fb31
67fcaaa612909a00ac636ee8c69e6b3595cb96f9
F20101124_AACJZI burke_g_Page_060.tif
f3fb380bb7e28852ddf6df64abc924b3
b8937b5e29d743b325305bba1aa788031ed247ad
112426 F20101124_AACJYV burke_g_Page_014.jp2
eed386f00057006d3f9608e65565b32a
6a0adfcf3e071126e11e27c389f38a7b68dff23e
57865 F20101124_AACKFD burke_g_Page_072.jpg
a43e79dbd1497a16dda27373f006643a
8d3de59c74012f497f1c81fb296ee4a2a31dc7d4
74468 F20101124_AACKEP burke_g_Page_054.jpg
d149def6ee652bf95ca54e12f0367450
83a709f64cc90cd63b98fa3b3a396c5526fa1f05
22637 F20101124_AACJZJ burke_g_Page_029.QC.jpg
e5859dc6a047b77ba29c37d4f2d388ac
f6891fa004b3613cc28cb5fa277f5742e4ec4788
82286 F20101124_AACJYW burke_g_Page_148.jpg
c30812c2aaa1e108c90279808b851bb4
f09776f80c8be81cc11a29f3892e9818ea4e528d
77235 F20101124_AACKFE burke_g_Page_073.jpg
83e0672636ff0658b378a23ed3b3ddf4
20c858803af9d1b320049d13230004cf8d5e717f
72879 F20101124_AACKEQ burke_g_Page_055.jpg
58122573df6b3db1eae410614ec35034
5bf12a4a1da317c376f726c322717114a3c47cb6
59889 F20101124_AACJZK burke_g_Page_044.jpg
ee18b3974b42a064960dc97ebb5b10ca
71da8d6d68a7ef6b690f6f0905091007f74a444f
62290 F20101124_AACKFF burke_g_Page_074.jpg
f23b536db504617d96ad1ca7371017f9
5cc9637eba3fd940eb4884ffb4f038eff5a3bde0
69085 F20101124_AACKER burke_g_Page_056.jpg
843c69441e9a14f4559b9210e9bfbbd0
28649493f8369a818f1c3d734508180c4abac505
84376 F20101124_AACJZL burke_g_Page_089.jp2
e080f3d50c86e79d378d5fa1d7aef3a5
ba54244edc6c5920af076fbe85ead8afefbc2a62
22718 F20101124_AACJYX burke_g_Page_026.QC.jpg
4e0ca4ead28cc2c252096bf40c23505e
c4d0e406bff452ddb034a7637350858d675b3ae8
78192 F20101124_AACKFG burke_g_Page_076.jpg
cb448cd7d458239bd10e591a15e73f63
899fae9fe35a5add6a09792d2c6be8d251ecbc20
72964 F20101124_AACKES burke_g_Page_058.jpg
f6199b90c05efca3561223b55fcd28ea
a59c5af0b59197f4e90786309d5274eff6afd706
41460 F20101124_AACJZM burke_g_Page_105.jpg
a5db0448136b17ae3d89fa35a7e04e1d
7d732f1d9df8395a0d8a426f4fd2b412bd8906d4
45090 F20101124_AACJYY burke_g_Page_060.pro
18ca8348790d16e053b7aa7474f5b5f7
d662dc49e44a821956b1daf27d4e2d9d35f8cce8
F20101124_AACKFH burke_g_Page_079.jpg
5b78e276da2ced57c0ed9963428046f8
03697ae9a9e2a9f9aade5997f629a2d68e5f8a50
40416 F20101124_AACKET burke_g_Page_059.jpg
7d5963ff24a30fef68ae827adbfd76a6
182dfc7f8232a3684aa18f83d60bef92263686ee
2047 F20101124_AACJZN burke_g_Page_113.txt
b726ace9b1a228535fe79084d8af44c2
e4bb13a5a19867ee515eef9ec58db7af8d1a4b57
112328 F20101124_AACJYZ burke_g_Page_019.jp2
b2a3aab76a6849f19cdfc80c04ff0d34
1efc2ee162042eac29f6b803eb1f78d64b28808b
67788 F20101124_AACKFI burke_g_Page_080.jpg
b3ea8944e7b548761da8e15749f3e0d1
e67f49c8a510b14d42dd1c0e40ad560be52c2c3d
60406 F20101124_AACKEU burke_g_Page_060.jpg
da4b007c912147c65f05def4c0852f7d
772654126435026ecc2aee9c6857e677ac5f0f29
25674 F20101124_AACJZO burke_g_Page_084.QC.jpg
775f59ef3002b61e3980f47227f9c6cb
03d5565899fc7c4095ae28b42fe6abcb44408878
70963 F20101124_AACKFJ burke_g_Page_083.jpg
d8124285fac07017de14a389f55d25f4
70418572dc8e3d926f642341d245f67cc8429bdd
74050 F20101124_AACKEV burke_g_Page_061.jpg
fb85319ae7b067074d8d7db10daa9514
fef103b1a65f7dc032b5d4228836e042f69df360
3858 F20101124_AACJZP burke_g_Page_112thm.jpg
63be1c1634622f0528129ed6ee87dc5e
55e45d08bff0dd06e08fee1afe30331cb1a844ac
89293 F20101124_AACKFK burke_g_Page_084.jpg
e356e97388c5519ce464ca4e63e538e9
13989cc398fcce0c5d92e1c7047d0c0951b71ac4
62794 F20101124_AACKEW burke_g_Page_063.jpg
fe1469dea28e7242c604b37cfc8fde8c
f6d967aca8fae694fc8b43d0ec66ad3e301b8b01
23716 F20101124_AACJZQ burke_g_Page_101.QC.jpg
e6e92cf176a57da8262d9af883241b8e
3d5791b93dd8bace5906c796d5564e9bce89ef49
64995 F20101124_AACKFL burke_g_Page_087.jpg
d4dbd5be866912b91ce6a202f860765d
8b52c48b533fba11934a2dce2f3f7c51b81cbbef
51571 F20101124_AACKEX burke_g_Page_064.jpg
0d650a77fbbce71301c6006151f4e39a
f628d199289ec148979b726de33e9ccb2abee36b
1742 F20101124_AACJZR burke_g_Page_111.txt
bd45be2192a339548a8dfcecf6ca6b94
0241bcfd6d1664eb0bc8c8b6cca61b6b1f20a2fd
38818 F20101124_AACKGA burke_g_Page_112.jpg
aa6e03eb6c21921c4cb6566bd66fe96c
5863f849f600da78c52fd9941336e67967d74c3b
46914 F20101124_AACKFM burke_g_Page_088.jpg
23f4506570895f464a9ab54f81568de2
85e543695ec434beff071b12f0c9fdb669fb1baa
46301 F20101124_AACKEY burke_g_Page_065.jpg
5ae7c6b36a21da54865956494fe8d3c1
fcc506ad1028022b1564b3d5b53e6c13647eefcb
110769 F20101124_AACJZS burke_g_Page_018.jp2
3befed23e81430dc31b825a718a78d2e
cd9b042e9ce815de93ed30178e3973b655402d2d
14326 F20101124_AACKGB burke_g_Page_117.jpg
c5ec7acfbc74d803cfe6f3715cb9b3b8
9bdaf2f3cbcb478505329ad22cd7c5e57541aa57
52617 F20101124_AACKFN burke_g_Page_091.jpg
b4716ecd415380c46ca1a3d5528571ab
02afe8bd2839bdc3d6378feb9d809dccce6f6fd2
63287 F20101124_AACKEZ burke_g_Page_066.jpg
01a85fc8891b2bad47bb701260f971b3
60cc4ae48c7b6dd11aa6e4b346dc51896ccb1908
9168 F20101124_AACJZT burke_g_Page_132.pro
2ba68f06780893ff8abec363964cf8b8
93fd98cc8df7654459b1264a2375bcc03e9542ea
42010 F20101124_AACKGC burke_g_Page_118.jpg
b613c49e301a073ad1f3355d800de25c
bce68c820dca4dcffc90af633ee66f026770d9a6
66337 F20101124_AACKFO burke_g_Page_092.jpg
b4411f8ce0ef2a92a3a20b37d038fa19
712de5924cee203fd16a742029bcff32458b3970
74491 F20101124_AACJZU burke_g_Page_107.jpg
929ca3783bd585fcc35945e9af123b09
7a6b7d1a57ccfa2f9f98f7ef87bd583d8247e8ec
33375 F20101124_AACKGD burke_g_Page_119.jpg
5062d34badeb1ab04f3ff193dd0ef15b
39ba9127aade772d8cbf06261f85c9f94e1d435a
42609 F20101124_AACKFP burke_g_Page_093.jpg
2811f6141112b05dd610e40b83eb2598
df86daae65690fa7197713fe90366538c73a4615
110223 F20101124_AACJZV burke_g_Page_016.jp2
ebac01a150f462be84da071672c113de
ae4e37cace3a13d73e139c5b81751cf3d9a12799
60430 F20101124_AACKFQ burke_g_Page_097.jpg
03150986b5ac43024e66f0115efa65a8
3a47346e251867b10de1634b770230b8054da4af
12858 F20101124_AACJZW burke_g_Page_059.QC.jpg
ed712e397090dfd0a3abaa19649620e7
fe643f2fc35ab17240a336e56a7d62c21fdb61dc
76481 F20101124_AACKGE burke_g_Page_121.jpg
c6e6068fdfcb5d17b1d5b471f1a998c4
24ed8a3f5fa9156da885ed7828513ba8c1a7b915
68288 F20101124_AACKFR burke_g_Page_098.jpg
4cfb9e021e10f800382c46fa5a921133
fb8da6c1915d7efadfeb9a13ea68a8379cf1a2fd
114253 F20101124_AACJZX burke_g_Page_055.jp2
01f7bc35d2ec046d64e2b5aeb327f907
78b3057867d182a08e566c1bcd52cb478fc64009
74985 F20101124_AACKGF burke_g_Page_122.jpg
cc3b12d2194729cdb82472da2e93faa6
175a987677f20e12cba5a8653cfe47bee00b37cc
75469 F20101124_AACKFS burke_g_Page_101.jpg
da75566698b6bf62bbfe9f1289119518
b827617b4833dbb1d4006a61f0625938529b8c61
83934 F20101124_AACKGG burke_g_Page_125.jpg
3c6c069eac67b97a5859c7af2fd20f0c
d8c290dafb155225a48372fe8e93a6ae593481b0
68614 F20101124_AACKFT burke_g_Page_102.jpg
a4388c90593881d1d988ad863aee2021
a9cf75bbd1307366792e7305b4c1f29b1aa228ef
F20101124_AACJZY burke_g_Page_161thm.jpg
0c2bb086c705b96e1753e1df0b8cb2d8
6ecfc21f9ad27328b2ca8176c908f1ea5c8d0a53
28326 F20101124_AACKGH burke_g_Page_126.jpg
15d8409dd265a7042d6327333b7547d9
fe49ce0d0041d13c603d34893080cf9939b11ee9
76507 F20101124_AACKFU burke_g_Page_103.jpg
180e32359b668c6d1d7e8f9efa0e16a4
77c47a72ddaa21ce2a68bebb1cc5fd39b77ee628
573 F20101124_AACJZZ burke_g_Page_135.txt
75c73af8afeb05ad3202b185cc75a3ed
2fbf5b75a443955dd546ba239af97d1833d95b81
83419 F20101124_AACKGI burke_g_Page_128.jpg
5235e05644c01e45801bbb0d6a9efa36
0d27e9e01b32073189c07ba0ada930e8295493e0
72103 F20101124_AACKFV burke_g_Page_104.jpg
11355bf2c0877275778d4e643c973443
1919f593cccc49680e8969d4278fd84b22f8c59c
30536 F20101124_AACKGJ burke_g_Page_129.jpg
6e72a64299fb29f6df260c2ab04bb908
a25327f84f8419004d30de05ae19a12f0a784b95
67389 F20101124_AACKFW burke_g_Page_106.jpg
5f4d2d19c9fd2c10a40c7902946cb5a4
e4f358588f2e6e9ca0cba3e92767ec947d28e4d7
81206 F20101124_AACKGK burke_g_Page_130.jpg
71dc7a70e2caa7017a0cafbab76441be
9f79abffba3aa72b3a929d9846b428580fa71315
70287 F20101124_AACKFX burke_g_Page_109.jpg
61520e830ada1c16c78179270e7d1750
fa5f000978ea6eae15986560b9bc5508982fae22
49932 F20101124_AACKHA burke_g_Page_156.jpg
c453f231b2a2d4d9fbc794c992c60f83
8016ec4348f3b87fa072f733b3196f775bc551b9
81981 F20101124_AACKGL burke_g_Page_133.jpg
26268469932626eddc9dd7b997ac2bde
2b8049df6d5f30b87a384981e7078b016d494e09
71652 F20101124_AACKFY burke_g_Page_110.jpg
85b23f2b62f3934971416bc7ef491597
7149ccfd331ac3fd1d6107caeeec4b8037f53b7a
53241 F20101124_AACKHB burke_g_Page_157.jpg
371eaac8c61f4cbd7207ae86cda07109
ed3390e181dd33dcb181692127ada26d6d5e19c9
77621 F20101124_AACKGM burke_g_Page_134.jpg
6c70aaccab125b0e5f9105b43c044e97
add0c145f0cda2294f98f009241807ad682c4478
58676 F20101124_AACKFZ burke_g_Page_111.jpg
752020e32a6e89c54bd140ff2c92e08e
aaf2c911cc0a1899789c0b48ecf9eb6648f31843
44918 F20101124_AACKHC burke_g_Page_158.jpg
c3d13fa1df69a98699e98e396ecc16ab
e01650506e43b3351d35c7bdd28683c9f2cee90f
33558 F20101124_AACKGN burke_g_Page_135.jpg
f9c79adf61e220101f2ae6719666be3b
d558e0d24a3a1b4bb624563317c7e301f1c3643b
20688 F20101124_AACKHD burke_g_Page_159.jpg
bc2c43432e8f7dc5e305797a96eab36a
a5b4f9f11d8fd4019c03255e5ba73e86a0bbc25b
77814 F20101124_AACKGO burke_g_Page_136.jpg
57f401be39c3000fa85a9687bb9b4def
808f7c7ed076b29563cc749aed0b1a4e2536d18a
73172 F20101124_AACKHE burke_g_Page_162.jpg
5f9b645f305f92c37c3c3b2fe4d26e97
808a7d8555ea68cde88286a979332729711ba197
82385 F20101124_AACKGP burke_g_Page_137.jpg
827e8df1a9bdd44f136f151064947389
9cc10ac7e9c9c01f75a1c56dc6b8aaa9b924d61b
28965 F20101124_AACKGQ burke_g_Page_138.jpg
8d11fc883d9408efda44e187d106a31d
d45a08d0982a513e8aaf62a7ff096106dfb9d31d
77520 F20101124_AACKHF burke_g_Page_163.jpg
315d3396cb806010439af594bc37ba06
19114f337a9bbc7556c4d4fc4b7b5b78b8288d05
82643 F20101124_AACKGR burke_g_Page_139.jpg
f161dfe9cbc342b90763f15b05f29189
6bf836c08f2e33ac12bf04a7b0cd97921ae8b913
75742 F20101124_AACKHG burke_g_Page_164.jpg
d717d3bdae3c49fd30762765e1b236d9
9ca138c86a687eb1c6a7784fe17cbcd1a67118d1
38938 F20101124_AACKGS burke_g_Page_141.jpg
5beaf7b23ccc2d0ac243352d46746577
aa77ff27f230e93cad8156de30e36462788fea9d


Permanent Link: http://ufdc.ufl.edu/UFE0011378/00001

Material Information

Title: Sourcing Strategies in a Supply Chain
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0011378:00001

Permanent Link: http://ufdc.ufl.edu/UFE0011378/00001

Material Information

Title: Sourcing Strategies in a Supply Chain
Physical Description: Mixed Material
Copyright Date: 2008

Record Information

Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
System ID: UFE0011378:00001


This item has the following downloads:


Full Text











SOURCING STRATEGIES
IN A SUPPLY CHAIN















By

GERARD JOSEPH BURKE JR.


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


2005

































Copyright 2005

by

Gerard Joseph Burke Jr.
















This work is dedicated to my family, and especially my wife, Amy, for your

love and support.















ACKNOWLEDGMENTS

The hard work and dedication required by me to create this dissertation

were made possible by the personal and practical support of my family, friends,

committee members and other members of the Decision and Information Sciences

Department. I wish to express my gratitude by specifically acknowledging each

group of supporters.

I wish to thank God; my parents, Jerry and Carlyn; my wife, Amy; my chil-

dren, Maddie, Marley, Ella and James; my grandfather, Frank Burke (whose

support is priceless); and my mother-in-law, Deb Crenshaw for their love, inspira-

tion, and support. I also thank my Hoosier family, Hank and Betty Tallman, Jim

and Joni Ping, Jeannie Meenach, and Drew and Pam Kissel for their unwavering

encouragement, love and much needed breaks from my studious endeavors.

Although this next group for gratitude is legally not considered family,

they are closer to me than friendship describes. I wish to thank Lou and Sandy

Paganini, and Ed and Atwood Brewton for their support, faith, and perspective.

Also, many thanks go to the Kempers and NX 1- -; for their friendship.

My doctoral studies were greatly enriched by the comradery of my fellow doc-

toral students Mark Cecchini, Selcuk Colak, Enes Eryarsoy, Ling He, Jason Dean,

Yuwen C'!I C i ( '!ii- Z!i i1;: Fidan Boylu, and Michelle Hanna. Additionally, I

would like to thank Pat Brawner, Shawn Lee, and Cindy Nantz for assisting me

throughout this tribulation.

My capability to complete this dissertation was developed in large part by the

seminars and coursework taught by members of my dissertation committee. I wish

to thank Janice Carrillo, Selcuk Erenguc, Anand Paul, and Joe Geunes for their









instruction and service on my dissertation committee during my doctoral studies.

I also wish to thank Janice Carrillo for candidly sharing her experiences in the

academic profession. Finally, I wish to thank Professor Asoo Vakharia, my advisor

and committee chair, for his expert guidance, timely responses to the drafts of each

chapter of my dissertation, and genuine interest in my personal and professional

well-being.















TABLE OF CONTENTS
page

ACKNOWLEDGMENTS ................... ...... iv

LIST OF TABLES ...................... ......... ix

ABSTRACT ...................... ............. xi

CHAPTER

1 INTRODUCTION ........................... 1

1.1 Supply C('!i ii Management-An Overview ...... ..... ... 1
1.2 Strategic Issues in SC \ I ....................... 3
1.2.1 Product Strategy .................. ..... 4
1.2.2 Network Design .................. ..... 5
1.3 Operational Issues in SC \I .............. .. 7
1.3.1 Transportation .................. .... 7
1.3.2 Transformation .................. ..... 8
1.3.3 Information Sharing .................. 9
1.4 Focus of this Research: Strategic Sourcing . . ..... 10
1.5 Organization of this Dissertation .................. .. 14
1.6 Statement of Contribution .................. .... 14

2 LITERATURE REVIEW .................. ....... .. 16

2.1 Overview .. .. .. .. .. .. .. .. .. ...... 16
2.2 Strategic Evolution of Sourcing .................. .. 16
2.3 Buyer-Supplier Relationships ................ .. .. 17
2.4 Strategic Sourcing .................. ...... .. .. 18
2.4.1 Qualification Criteria ................ .. .. 18
2.4.2 Selection and Allocation .................. .. 19

3 STRATEGIC SOURCING DECISIONS WITH STOCHASTIC SUP-
PLIER RELIABILITY .................. ........ .. 27

3.1 Introduction .................. ........... .. 27
3.2 Sourcing M odel .................. ......... .. 28
3.3 Model Development .................. ....... .. 30
3.4 Analysis .................. ............. .. 30
3.4.1 Heterogeneous Suppliers .................. .. 31
3.4.2 Heterogeneous Cost Suppliers ............... .. 35









3.4.3 Heterogeneous Reliability Suppliers .
3.4.4 Homogeneous Suppliers .. .......
3.5 Numerical Analysis .. ............
3.5.1 Experimental Design .. .......
3.5.2 R results . . . . .
3.6 Conclusions . . . . .

4 IMPACT OF SUPPLIER PRICING SCHEMES AND
SOURCING STRATEGIES .. ...........

4.1 Introduction . . . . .
4.2 Sourcing M odel .. ..............
4.2.1 Prelim inaries .. ............
4.2.2 Supplier Pricing Schemes and Capacity
4.3 Analysis and Insights .. ...........
4.3.1 Constant Price .. ............
4.3.2 Linear Discount Price .. ........
4.3.3 Incremental Units Discount Price .
4.3.4 All Units Discount Price .. ......
4.3.5 Summary of Insights from Analysis .
4.4 Application . . . . .
4.5 Optimal Algorithm for Incremental Quantity D
4.5.1 Algorithm Description .. ........
4.5.2 Computation and Validity of LBP and U
4.5.3 The Branching Process .. .......
4.5.4 Formal Statement of the Algorithm .
4.6 Conclusions . . . . .


5 STRATEGIC SOURCING WITH DIVERSIFICATION
T IO N . . . . . . .


CAPACITY ON


iscounted

Bf ...


Sourcing


CONSIDERA-


5.1 Introduction .. ............
5.2 Integrated Selection/Allocation Model
5.2.1 Preliminaries .. .........
5.2.2 Model Development .......
5.3 A analysis . . . . .
5.3.1 No Diversification Benefit .
5.3.2 Diversification Benefit ......
5.4 Model Extensions .. ..........
5.5 Numerical Analysis .. .........
5.5.1 Experimental Design ......
5.5.2 Results . . . .
5.6 Conclusions and Implications ......









6 SUMMARY ..............


Key Results and Directions for
Key Results and Directions for
Key Results and Directions for


Future Research from
Future Research from
Future Research from


A PPE N D IX . . . . . . . . ..

A PROOFS FOR CHAPTER 3 .. ....................

A.1 Proof of Corollary 3.1 .......................
A.2 Proof of Theorem 3.1 .. .....................
A.3 Proof of Corollary 3.2 .........................
A.4 Proof for Corollary 3.3 .. ....................
A.5 Proof of Corollary 3.4 .........................
A.6 Proof of Corollary 3.5 ......................
A.7 Proof of Corollary 3.6 .........................
A.8 Proof of Theorem 3.2 .. .....................
A.9 Proof of Theorem 3.3 .. .....................
A.10 Proof of Theorem 3.4 .........................
A.11 Proof of Corollary 3.7 .........................

B PROOFS FOR CHAPTER 4 .. ....................


Proof of Theorem 4.1 .
Proof of Result 4.1
Tiinpi r Di~nrmnt Pricinr


Test Problems Data


B.4 Incremental and All-Unit Discount Pricing Test Problems Data.


C PROOFS FOR CHAPTER 5 .


Proof of Theorem 5.1 .
Proof of Theorem 5.2 .
Proof of Theorem 5.5 .


REFERENCES .......... ...

BIOGRAPHICAL SKETCH ......


Ilpter 3
Ilpter 4
Ilpter 5















LIST OF TABLES
Table page

3-1 Model Notation .................. ........... .. 29

3-2 Description of Numerical Examples ................ 40

3-3 Results for Numerical Examples .................. .. 41

4-1 Linear Discount Heuristic Performance ................. .. 56

4-2 Incremental Units Discount Heuristic Performance .......... ..58

4-3 All-Units Discount Heuristic Performance ............... ..61

4-4 Product A Bid Information Data .................. .. 62

4-5 Product B Bid Information Data .................. .. 63

4-6 CPO Product A Solutions Comparison. ............... 64

4-7 CPO Product B Solutions Comparison ................. .. 64

4-8 Number of Subproblems Solved for each Test Problem . ... 68

5-1 Sensitivity Analysis of the Key Parameters . . ..... 88

5-2 Impact of Minimum Order Quantity on the Sourcing Strategy . 89

5-3 Interactions between Minimum Order Quantities and Reliabilities 90

5-4 Parameter Values for Ranked Supplier C'I ,i o,:teristics . ... 91

5-5 Selected Results for Capacity Adjusted Model A ........... .91

B-1 Linear Discount Pricing Test Problems Data (1-12) . . ... 109

B-2 Linear Discount Pricing Test Problems Data (13-24) . ... 110

B-3 Linear Discount Pricing Test Problems Data (25-30) . ... 111

B-4 Incremental and All-Units Pricing Test Problems Data (1-3) ..... ..112

B-5 Incremental and All-Units Pricing Test Problems Data (4-6) ..... ..115

B-6 Incremental and All-Units Pricing Test Problems Data (7-9) ..... ..118

B-7 Incremental and All-Units Pricing Test Problems Data (10-12) 121










8 Incremental and

9 Incremental and

10 Incremental and

11 Incremental and

12 Incremental and

13 Incremental and


All-Units

All-Units

All-Units

All-Units

All-Units

All-Units


Pricing

Pricing

Pricing

Pricing

Pricing

Pricing


Test Problems

Test Problems

Test Problems

Test Problems

Test Problems

Test Problems


Data

Data

Data

Data

Data

Data


(13-15)

(16-18)

(19-21)

(22-24)

(25-27)

(28-30)


. .. 124

. .. 127

. 130

. 133

. 136

. 139















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 Phil. .. 1hi

SOURCING STRATEGIES
IN A SUPPLY CHAIN

By

Gerard Joseph Burke Jr.

August 2005

C'! In': Asoo J. Vakharia
Major Department: Decision and Information Sciences

The focus of this dissertation is on supply chain management (SC\ ), and

more specifically on the upstream connection between a firm and its suppliers.

My research examines single versus multiple supplier sourcing strategies under

three specific scenarios. In general, my dissertation seeks to characterize when a

buying firm should single source its requirements instead of employing a diversified

purchasing policy under various commonly encountered operating scenarios.

First, the effects of upstream and downstream uncertainty on a firms sourcing

strategy are examined. Our results show that order-splitting (i.e., choosing a

multiple supplier strategy) is an optimal choice for the firm even when suppliers are

completely heterogeneous in terms of their reliability and costs. Additionally, the

choice of single versus multiple sourcing depends to some extent on supplier prices.

This latter result motivates investigating our second scenario to gain insight into

how alternate supplier pricing schemes may impact sourcing decisions.

The second scenario we examine is motivated not only through the results

obtained under the stochastic supply setting described earlier but also through an

understanding of the supplier selection and quantity allocation decisions made by









a n i "r office products retailer located in Florida. The pricing schemes quoted by

its suppliers tend to be either a constant price, a linearly discounted price, or a

staged quantity discount (i.e., all-units and/or incremental discounted price) price.

For each type of pricing scheme, we develop a unique optimization model where

the objective is to minimize the sum of concave cost functions while satisfying the

firms total requirements. We adapt existing branch and bound algorithms in order

to identify the optimal number of suppliers who should receive an order.

Finally, we incorporate explicit diversification benefits (due to channel power

leverage and price competition) into a newsvendor framework to analyze a firms

sourcing decisions when suppliers are unreliable. Analysis reveals that a trade-off

between the marginal benefit of diversification and the marginal cost of shifting

allocated order quantities away from lower cost suppliers need to be assessed.

Managerially, this model stresses the importance of consistency between a firms

sourcing strategy and its corporate strategy.















CHAPTER 1
INTRODUCTION

1.1 Supply C(' ii, Management-An Overview

A supply chain can be visualized as a network of firms servicing and being

serviced by several other firms. However, it is conceptually easier to imagine a

chain as a river, originating from a source, moving downstream and terminating at

a sink. The supply chain extends upstream to the sourcing of raw materials and

downstream to the afterlife activities of the product, such as disposal, recycling and

remanufacturing. Regardless of magnitude, every supply chain can be visualized as

consisting of sourcing stages, manufacturing stages and distribution stages.

Each of these stages p .1i both a primary (usually physical transformation or

service creation) and a dual (market mediator) role. The approach taken to execute

activities in support of both roles depends on the strategy of the supply chain,

which in turn, is a function of the serviced products' demand pattern (Fisher,

1997). Depending upon the structure of the chain (in terms of products and

processes employ, 1), channel power can reside with the sourcing (e.g., monopolist

supplier of key commodities such as oil), manufacturing (e.g., dominant producer

of a unique product such as semiconductors), or distribution (e.g., key distributor

of consumer items) stages in the supply chain. Relative power in the supply chain

influences strategic positioning of each link in the chain. Thus, managing supply

chains is a negotiation between the objectives of constituent's benefit at each

stage and the impact of each constituent's objective to the overall objective of

maximizing the benefit of the entire chain.

The contribution captured at each stage depends on the nature of the dealings

between the buyer and supplier. The traditional model is characterized by an









adversarial relationship where a buyer awards contracts to one or more competing

suppliers based on price and other pertinent criteria. This paradigm has been

widely criticized as short sighted by proponents of partnered buyer-supplier

relationships. The partnered approach favors a smaller or even single supplier base

for more supplier management initiatives to minimize inventory investments and

encourage collaboration in, among other things, new product development.

Supply chain management (SC'\ ) is the art and science of creating and

accentuating synergistic relationships among the trading members that constitute

supply and distribution channels. Supply chain managers strive to deliver desired

goods or services on time to the appropriate place in the ordered quantity in the

most effective and efficient manner. Usually this is achieved by negotiating a

balance between conflicting objectives of customer satisfaction and cost efficiencies.

Each link in each supply chain represents an intersection where supply meets

demand, and directing the product and information flows at these crossroads is at

the core of SC'\ I. The integral value proposition of an integrated supply chain is

as follows. Total performance of the entire chain is enhanced when all links in the

chain are simultaneously optimized as compared to the resulting total performance

when each individual link is separately optimized. Supply chain performance as

a whole hinges on achieving fit between the nature of the products it supplies,

the competitive strategies of the interacting firms, and the overall supply chain

strategy.

Coordination of the individual links in the chain is essential to achieve this

objective. The ability of trading partners to jointly communicate in real time and

the transactional ease of digital dealings allow web-connected firms to virtually

integrate. The Internet and information technology in general facilitate the

integration of multitudes of channel enterprises. On-line collaboration enables









better informed economic decision making, reduces the costs of order placement,

tracking and receipt, and enhances customer satisfaction.

Information technologies are a key driver of modern operational efficiency,

and efficient operational execution is a driver of effective SC'\ \. Selection of trading

partners, location of facilities, manufacturing schedules, transportation routes and

modes, and inventory levels and location are the fundamental operations decisions

that run supply chains. These operational dimensions are the tributaries that pilot

the chain downstream through its channel to end demand. Accurate and timely

integrated information navigates the chain from source to sink.

A supply chain is a collection of multiple suppliers', manufacturers' and

distributors' processes. Each process employs a distinct focus and a related

dimension of excellence. Key issues in managing an entire supply chain relate to

tactical and strategic analysis of coordinated decisions in logistics, manufacturing,

distribution, and after sales activities of service and disposal or recycling; analyzing

product strategies; and network design decisions.

The motivation for this research is derived from the debate as to the best num-

ber of suppliers to employ for satisfying a buyer's requirements. Further, the buyer

considered is an intermediary in the supply chain and therefore must incorporate

downstream demand into its sourcing decision. Essentially, the decisions analyzed

address the question of whether a single sourcing strategy is optimal or not. To

understand the relevance of strategic sourcing decisions, it must be understood how

a firm's supply chain strategy is anchored to its sourcing strategy.

1.2 Strategic Issues in SC' I

A supply chain is only as strong as its weakest link. How the chain defines

strength is at the core of a supply chain's strategy, and therefore design. Is strength

anchored in efficiency or responsiveness? Regardless of which strategic position is

chosen, a firm's ability to maintain a competitive advantage will depend on how









well it reinforces its firm level value proposition with functional and departmental

strategic decision-making. By analyzing product demand characteristics and the

supply chain's capabilities, and crafting a fit between them, an individual supply

chain manager can be assured that the specific product and process strategy

emplolv ,1 does not create dissonance within his firm and further throughout the

entire supply chain.

1.2.1 Product Strategy

Achieving a tight fit between the competitive strategies of supply chain

members and the supply chain itself is gained by evaluating the characteristics

of the products serviced by the chain. "The root cause of the problems plaguing

many supply chains is a mismatch between the type of product and the type of

supply chain" (Fisher, 1997, p.106). Critical product attributes are (a) the demand

pattern; (b) the life-cycle; (c) variety of offerings; and (d) the product delivery

strategy. Fisher (1997) categorizes a product as being either functional (basic,

predictable, long-lived, low profit margin) or innovative (differentiated, volatile,

short-lived, high profit margin). Further, using the product life cycle argument,

innovative products (if successful) will eventually evolve to become functional

products. The types of supply chains needed to effectively service these two

categories of products are quite distinct. An efficient or low cost supply chain is

more appropriate for a functional product while a responsive or customer attuned

supply chain better services an innovative product. Obviously, a spectrum of

chain varieties exists between the end points of responsiveness and efficiency, and

hence, most supply chains are hybrids which target responsiveness requirements

for each product serviced while exploiting commonalities in servicing all products

to gain economies of scope. Thus, the strategic position of a supply chain balances

customer satisfaction demands and the firm's need for cost minimization.









Information technologies enable both efficient and responsive supply chains

since they have the potential to provide immediate and accurate demand and order

status information. Efficiency gains via information technologies are gleaned from

decreased transactional costs resulting from order automation and easier access

to information needed by chain members. Likewise, responsiveness gains can be

obtained by a quicker response to customer orders. Hence, in practice, it seems

to have become standard practice for all supply chains to utilize some form of

information technology to enable not only a more efficient physical flow of their

products but also to simultaneously improve their market mediation capability.

However, the efficiency of physical flow primarily depends on a supply chain's

infrastructure.

1.2.2 Network Design

In general, network design determines the supply chain's structure. The

significant capital investments required in building such a structure indicate the

relative long run or strategic importance of network decisions. Network decisions

in a supply chain involve facility focus, facility location, capacity 1p1 '::i:lr' and

sourcing/distribution channels (Chopra and Meindl, 2001). Each network design

decision impacts the firm's ability to provide value. Therefore, these decisions must

incorporate their strategic influence into the analysis.

Facility focus relates to how network investments facilitate the supply chain

strategy. If the facility in question is a manufacturing plant and the plant is set up

to produce only a specific product type, the chain will be more efficient but less

flexible than it would be if the plant produced multiple product types.

Facility location decisions are essential to a firm's strategy. The cost ramifi-

cations of a sub-optimal location decision could be substantial. Further, shutting

down or moving a facility is significant not only in terms of financial resources, but

also in terms of the impact on employees and communities. Other factors which









should be considered are the available infrastructure for physical and informa-

tion transportation, flexibility of production technologies employ, ,1 external or

macroeconomic influences, political stability, location of competitors, availability of

required labor and materials, and the logistics costs contingent on site selection.

Depending on the expected level of output for a facility, capacity allocations

should be made so that idle time is minimal. Under-utilization results in lower

return on investment and is sure to get the attention of company executives.

On the other hand, under allocating capacity (or large utilizations) will create a

bottleneck or constricted link in the supply chain. This will result in unsatisfied

demand and lost sales or increased costs as a result of satisfying demand from a

non-optimal location. The capacity allocation decision is a relatively long-term

commitment, which becomes more significant as sophistication and price of the

production technology increase.

The most basic question of an enterprise is: Who will serve our needs and

whose needs will we serve? This is a recurring question. Decisions regarding the

suppliers to a facility and the demand to be satisfied by a facility determine the

costs of material inputs, inventory, and delivery. Therefore, as forces driving supply

and/or demand change, this decision must be reconsidered. The objective here is

typically to match suppliers and markets to facilities in order to minimize not only

the system-wide costs but also the customer responsiveness of the supply chain.

Each of these network design decisions is not made in isolation since there is a

need to prioritize and coordinate their combined impact on the firm and its supply

chain. In general, network configuration is the structure of the supply chain and it

is within this structure that operations strategies and tactics are implemented to

reinforce the overall strategy of the entire chain. Of particular relevance for this

research is a firm's sourcing strategy.









1.3 Operational Issues in SC'\

SC'\I has evolved from process reengineering efforts to coordinate and integrate

production planning at the factory level in order to expand the scope of strategic

fit (Chopra and Meindl, 2001). Positive results from these intra-functional efforts

have extended the SC'\ philosophy throughout the enterprise. Further, process

improvements at the firm level highlighted the need for suppliers and customers

of supply chain managed firms to adopt an integrated SC'\ philosophy. Making

a supply chain's linkages as frictionless as possible is the tactical goal of such

an integrated philosophy. Key tactical coordination decisions for SC'\ relate to

transportation, transformation, and information transmission.

1.3.1 Transportation

Transportation decisions impact product flow not only between supply chain

members but also to the market place. In many supply networks, transportation

costs account for a significant portion of total supply chain cost. In determining the

mode(s) and routes) to employ through the supply chain, transportation decisions

seek to strike a balance between efficiency and responsiveness so as to reinforce

the strategic position of the supply chain. For example, an innovative product's

typically short life-cycle may warrant expensive air freight speed for a portion or

all of its movement through the chain, while a commodity is generally transported

by slow but relatively economical water or rail freight. Shipping via truck is also

used frequently. Trucking is more responsive and more expensive than rail, and less

responsive and less expensive than air. Most supply chains employ an intermodal

strategy (e.g., raw materials are transported by rail or ship, components by truck,

and finished goods by air).

A supply chain's transportation network decisions are inextricably linked to

strategic network design decisions. Transportation network design choices drive

routing decisions in the supply network. The major decisions are whether to ship









directly to buyers or to a distribution center, and whether a routing scheme is

needed. As consumers' expectations regarding merchandise availability and delivery

become more instantaneous, the role of a supply chain's transportation network is

more critical.

1.3.2 Transformation

"A transformation network links production facilities conducting work-in

process inventories through the supply chain" (Erenguc, Simpson, and Vakharia,

1999, p.224). Suppliers linked to manufacturers linked to distribution systems can

be viewed as a transformation network hinging on the manufacturer. Transforming

supplies begins at the receiving stations of manufacturers. The configuration of

manufacturing facilities and locations of transformation processes are determined

by plant level design decisions. The manufacturing process strategy employ -l at a

specific plant largely drives the decisions. While an assemble-to-order (ATO) plant

may have very little investment in production, it requires larger investment in sub-

assembly inventories. On the other hand, a make-to-stock (AI I S) facility may have

little or no investment in process inventories, it typically requires larger investments

in raw materials and finished goods inventories. A make-to-order (\ITO) facility

may have significant investment in components and production facilities, with few

raw materials and finished goods inventories. A product's final form can also take

shape closer to the end consumer. To keep finished goods inventory costs as low as

possible, and better match end demand, a supply chain may employ postponement

to delay customizing end products.

Major design decisions such as facility configuration and transformation

processes are considered longer term decisions. These decisions constrain the short

to mid-term decisions addressed in a plant's ..-:-regate plan. An ..--.regate plan is a

general production plan that encompasses a specific planning horizon. Information

required to develop an effective .,-.-.regate plan include accurate demand forecasts,









reliable supply delivery schedules, and the cost trade-offs between production and

inventory. Each supply chain member develops an..-.- regate plan to guide medium-

term tactical decisions. To ensure that these individual plans support each other,

the planning process must be coordinated. The degree and scope of coordination

will depend on the economics of collaborative planning versus the costs of under-

supply and over-supply. In general, a manufacturer should definitely involve 1n! i 1r

suppliers and buyers in .-- regate planning. Whether this planning information

trickles to other supply chain members (a key for the success of integrated supply

chain management) will depend on the coordination capabilities of successive l .-ri

of members emanating from a collaborative planning center, which is often the

i_ ii r manufacturer.

The execution of the .-'-regate plan is a function of the information inputs

into the ..-.- regate plan. It is vital that these inputs be as accurate as possible

throughout the entire supply chain. Integrated planning in a supply chain requires

its members to share information. The initiator of integrated planning is typi-

cally an intermediary. To understand why, we must understand the dynamics of

distribution.

1.3.3 Information Sharing

A distribution channel is typically composed of a manufacturer, a wholesaler,

a distributor, and a retailer. The "bull-whip effect" is a classic illustration of

dysfunction in such a channel due to the lack of information sharing. This effect

is characterized by increasing variability in orders as the orders are transferred

from the retailer upstream to the distributor, then to the wholesaler, and finally

to the manufacturer. Distorted demand information induces amplifications in

order variance as orders flow upstream. Therefore, the manufacturer bears the

greatest degree of order variability. This is a i1i i i r reason manufacturers initiate

collaborative efforts with downstream channel members. To anticipate the quantity









of product to produce and when, a manufacturer must compile demand forecasts

from downstream supply chain members. Forecasting accuracy is paramount

because it is the basis for effective and efficient management of supply chains.

A major challenge of SC'\ is to minimize costs and maintain flexibility

in the face of uncertain demand. This is accomplished through capacity and

inventory management. Similarly, marketers attempt to maximize revenues

through demand management practices of pricing and promotion. Therefore,

it is vital that marketing and operations departments collaborate on forecasts

and share harmonious incentive structures. The degree of coordination between

order acquisition, supply acquisition, and production process directly affects how

smoothly a firm operates. Likewise, buyers', suppliers', and producers' coordination

levels directly affect how smoothly the supply chain operates. Summarily, accurate

information flows between channel members are essential for effective SC'\!.

1.4 Focus of this Research: Strategic Sourcing

The strategic importance of sourcing activities is inherent in purchasing's

supply chain position. Purchasing activities link the firm to the upstream value sys-

tem and allow a buying firm to obtain appropriate inputs from external suppliers.

Procurement activities in large part support a firm's inbound logistics and are vital

to value creation (Porter, 1985). A firm's sourcing strategy is therefore a key driver

of an effective supply chain (value system).

Innovations in technology and increased global competition provide opportu-

nities and challenges that drive firms to continuously evaluate and modify their

sourcing strategies. Moreover, recent studies reveal that the long-term implications

of poor supply chain management are far reaching, ultimately impacting both firm

performance and market value. Since a typical manufacturing firm spends 55' of

earned revenue on purchased materials (Leenders and Fearon, 1998), disruptions

due to supply inadequacies could have a major impact on profitability. Hendricks









and Singhal (2001) reinforce this by showing that in the 90 di- prior and sub-

sequent to a reported supply chain problem stemming from supplier glitches, the

buying firm's average shareholder return typically decreases by 1"' Clearly, a

manufacturer's operations strategy and financial livelihood rely on its chosen sup-

plier pool and thus decisions with regard to suppliers are fundamental to successful

enterprise.

A firm's sourcing strategy is characterized by three key interrelated decisions

(Burke and Vakharia, 2002): (a) criteria for establishing a supplier base; (b) criteria

for selecting suppliers (a subset of the base) who will receive an order from the

firm; and (c) the quantity of goods to order from each supplier selected. To start

with, criteria for developing a supplier base are typically based upon the firm's

perception of the supplier's ability to fulfill the objectives of quality, quantity,

delivery and price. While the supplier's price may be the most important criteria

for profit maximization, the other dimensions can also affect the overall profitability

of the firm. Scoring models are generally used to evaluate suppliers for inclusion in

the base. In general, this approach ranks each supplier in terms of objectives and

then, based on a relative weighting of each of the objectives, a total score for each

potential supplier is derived. Next, by specifying a threshold score, all suppliers

who achieve this threshold are included in the base. Objectives used in developing

rankings vary across firms. For example, Sun Microsystems ranks its suppliers

with a "scorecard" based on quality, delivery, technology, and supplier support

(Holloway et al., 1996).

From the approved supply base, the specific subset of suppliers which will

actually receive an order must be determined. Since all suppliers in the base meet

the quality, delivery, and other objectives of the firm, dominant industry practice

appears to base this decision primarily on cost considerations. While the supplier's

price may be the most important criterion for profit maximization, some buying









firms impose alternate criteria related to robust delivery reliability capabilities.

Once the selected set of suppliers (a subset of the base) is determined, the firm

must allocate products) requirements among them. For the allocation decision,

supplier yields (in terms of percentage of ;ood" units), order quantity policies,

lead times, and transportation costs are salient.

A firm may choose either a specialized (i.e., single supplier) or a generalized

(i.e., multiple suppliers) strategic sourcing position, and allocation of requirements

will seek to optimize the value to the firm from this decision-making process.

Since it is the collective suppliers' capabilities that can enable or limit supply

chain performance at its inception, a firm's sourcing strategy is vital to successful

enterprise.

Single-sourcing strategies strive for partnerships between buyers and suppliers

to foster cooperation and achieve shared benefits. The tighter coordination between

buyer and suppliers) required for successful just-in-time (JIT) inventory initiatives

encourages supplier alliances to streamline the supply network and tends to shift

supply relations toward single sourcing.

Managing more than one source is obviously more cumbersome than dealing

with a single source. However, web-based SC'\ applications enable closer man-

agement of diverse suppliers, streamline supply chain processes and drive down

procurement costs. For example, GM utilized Internet tools to purchase more of

its total purchasing budget on-line, which resulted in a streamlined procurement

process and decreased vehicle delivery times (Veverka, 2001). Other documented

benefits of effectively utilizing web-based procurement tools include a reduction in

price of materials, administrative costs, inventory costs, and purchase and fulfill-

ment cycles. Consequently, firms that prefer single-sourcing for ease of management

can embrace multiple-sourcing via information technology-based SC'\ applications









as a more viable strategy to capture risk-pooling benefits. A shortcoming of In-

ternet procurement tools is their limited capability to provide complex decision

support for strategic sourcing decisions (Aberdeen Group, 1999). Nonetheless, the

differential tactical or low-cost competitive advantages supported by single-sourcing

over multiple-sourcing are diminished by the proliferation of Internet procurement

capable firms.

Moreover, single-source dependency exposes the buying firm to a greater

risk of supply interruption. Toyota's brake valve crisis in 1997 provides a recent

example of realized supply risk resulting from a single sourcing strategy in a JIT

inventory system. In 1997, Toyota's assembly plants were forced to shut down after

a fire at Aisin's main plant. This single-supplier's particular facility provided 9,n'

of all brake parts components and practically all brake valves for Toyota, before it

was destroyed (Nishiguchi and Beaudet, 1998). It is estimated that the impact on

Toyota's net income from this single event at Aisin was a decrease of $300 million.

Thereafter, Toyota sought at least two suppliers for each part (Treece, 1997).

Operationally, multiple-sourcing provides greater assurance of timely delivery,

and greater upside volume flexibility due to the diversification of the firm's total

requirements (Ramasesh, et. al., 1991).

Single-sourcing all of a firm's requirements also exposes the buying firm to

hold-up risk. Land Rover's contractual problem with its only chassis supplier is

an example of the operational difficulties this situation creates (Lester, 2002).

Strategically, supplier power exerted on the buyer is weakened when the firm splits

its total requirements among multiple sources. Multiple sourcing hedges the risks of

creating a monopolistic (sole source) supply base and supplier forward integration

(N. v ii oi 1989). In sum, the choice of multiple-sourcing versus single-sourcing

depends on the trade-off between the benefits of multiple-sourcing versus those of

single-sourcing.









1.5 Organization of this Dissertation

To position this research in the broader area of supply chain management

research, this chapter has provided an overview of supply chain management

and strategic sourcing. The remainder of this proposal is organized as follows.

('!C lpter two provides a comprehensive literature review of research in supply

chain management that relates to supplier selection and requirements allocation

decisions. In ('!i lpters 3-5 we seek to characterize conditions under which a firm

should choose either a specialized (i.e., single supplier) or a generalized (i.e.,

multiple suppliers) sourcing strategy under different scenarios. More specifically,

chapter 3 incorporates the impact of variability in both supply and demand in

addressing sourcing decisions; chapter 4 examines how the sourcing strategy is

moderated when suppliers offer a variety of common quantity discount schemes;

and chapter 5 integrates the impact of explicit diversification benefits on the

sourcing decision. C'!i lpter 6 provides a summary of key results from this research,

and discussion of opportunities for future research.

1.6 Statement of Contribution

Research in the area of supply chain management continues to be active. This

research focuses on the upstream connections of a buying firm to an approved

supplier base under various commonly encountered industrial operating condi-

tions. We add to the existing body of knowledge in supply chain management by

surveying closely related existing research, and developing and analyzing realistic

models for strategic sourcing decision making. Our focus on optimal decision mak-

ing in regard to the number of suppliers to requisition highlights the importance

of risk mitigation through supplier diversification. In chapter 3 we illustrate this

value to the firm by modeling the firm's purchase decision with both upstream and

downstream uncertainty, we are able to analytically solve this sourcing problem

for optimal solutions and show that in some scenarios cost dominates reliability.









Further, our analysis is uncommonly general as compared to existing work for

this particular problem. In chapter 5 we explicitly consider strategic benefits of

diversification with unreliable supply and downstream product demand uncertainty.

This approach allows firm level flexibility in deriving the most strategically desired

size of the selected supplier pool, and gauges the marginal cost trade off of selecting

higher unit cost supply against the imputed benefits of supply diversification.

C'! lpter 4 considers strategic behavior of capacitated suppliers via quantity dis-

count pricing quotes. We develop constant, linear discount, incremental discount

and all-units discount pricing schemes sourcing models. For the discount pricing

models we find that they are concave minimization problems and combinatorial in

nature. We develop an optimal branch and bound algorithm and well-performing

heuristics to aid in strategic sourcing decision making in this operating envi-

ronment. In all this research provides guidance for practitioners and academics

concerned with supplier management in regards to the frequent decision making

tasks of choosing the appropriate number of suppliers to source from and their

respective order quantity allocations.















CHAPTER 2
LITERATURE REVIEW

2.1 Overview

The first section in this chapter describes the evolution of the role of sourcing

from a purely transactional pursuit to one of strategic influence. Next, a review

of literature examining the buyer-supplier relationship reveals the divide between

proponents of single sourcing and multiple sourcing. The literature reviewed in

the final section focuses on supplier qualification, supplier selection, and quantity

allocation criteria.

2.2 Strategic Evolution of Sourcing

Ellram and Carr (1994) provide a history and review of literature related to

sourcing's strategic importance. They note that even as the oil crisis of 1973-1974

highlighted the perils of raw material supply shortages, research on industrial

buying behavior largely viewed the purchasing function as administrative. It

was not until the 1980s when Porter's Five Forces model gained popularity

that the strategic role of the interface between suppliers and buyers was better

understood. This has led to more contemporary research investigating the strategic

impact of sourcing as an integrative link between the firm and its suppliers.

As the interface between suppliers and the firm, purchasing's influence on firm

performance increases as supplier contribution to the firm increases. Further, the

inclusion of purchasing (sourcing) strategy in corporate strategy is more vital

as global competition and the pace of technological change quickens. A firm's

supplier management orientation is reflected in its contracting policies for external

purchases. Cohen and Agrawal (1999) model the trade-offs between short term and

long term contractual relationships. Short term contracts provide more flexibility









and avoid fixed investments, but also forgo improvement and price certainty

benefits afforded from long term contracts. Their analysis reveals that short term

contracting is optimal under a wide range of conditions. In a survey of supply

managers they find that management intend to develop long term relationships,

but often engage in short term contracting. This conflict over the optimal buyer-

supplier relationship is not exclusive to the ranks of supply managers.

2.3 Buyer-Supplier Relationships

Research on the number of sources for fulfillment of product requirements

is somewhat controversial. At one extreme, we have empirical evidence of many

firms shrinking their supplier base per item and ordering the in i i i ly of total units

required from a single source (e.g., Spekman (1988) and Pilling and Zhang (1992)).

Further, the documented benefits of single-sourcing such as quantity discounts

from order consolidation, reduced order lead times, and logistical cost reductions

as a result of a scaled down supplier base. Hahn et al., (1986), and Bozarth et al.,

(1998) reinforce this evidence. In fact, Mohr and Spekman (1994) contend that

single-sourcing performance benefits often outweigh the benefits of a price centric

multiple-sourcing strategy. These benefits are also enhanced by incorporating

performance improvement criteria in managing supplier relationships (Fawcett and

Birou, 1992).

In contrast, Bhote (1987) observed that relationship management costs, in

terms of time and capital, may outweigh the benefits of single-sourcing. The

primary rationale driving this argument is that single-sourcing requires the

firm and the supplier to develop a partnership based on trust. In line with this

i -.... ii.- McCutcheon and Stewart (2000) assert that the parties must achieve

goodwill trust to have a successful partnership. Further, they conjecture that this

level of trust is rarely attained. Adversarial history between channel members and

competition for larger shares of a product's limited total margin drive distrust. A









significant amount of research exists on procurement competitions where supply

base size and order allocation is of concern. Elmaghraby (2000) and Minner

(2003) both offer excellent overviews of the supply chain literature relevant to

requirements allocation decisions.

2.4 Strategic Sourcing

As previously stated, a firm's sourcing strategy can be characterized by three

key interrelated decisions (Burke and Vakharia, 2002): (a) criteria for establishing

a supplier base; (b) criteria for selecting suppliers (a subset of the base) who will

receive an order from the firm; and (c) the quantity of goods to order from each

supplier selected. Literature related to these key decisions is now reviewed.

2.4.1 Qualification Criteria

Quality and delivery are strategically important supplier qualification crite-

ria. In the context of these criteria, as a firm's supply management orientation

increases, supplier and buyer performance increases (Shin, et al., 2000). This rein-

forces Dickson's (1966) survey of vendor selection criteria, which ranks quality and

delivery, respectively, as the first and second most important supplier attributes.

Similarly, in a survey of companies at different levels of the U.S auto industry,

Choi and Hartley (1996) found that price is one of the least important factors for

supplier selection across the supply chain, while consistency (defined as the mar-

riage of quality and delivery) was the most important criterion. A review of prior

research rates quality as the single extremely important criterion, while delivery is

of considerable importance (Weber et al., 1991).

Measurements of quality are difficult to generalize across industries. Garvin

(1987) outlines several dimensions of quality, including performance, features,

reliability, conformance, durability, serviceability, aesthetics and perceived quality.

In practice, a firm may use quality measures to screen out all suppliers who fail

to meet some pre-specified minimal quality level. For example, Hillview Hospital









utilized such a policy in eliminating potential suppliers for hand soap based upon

the soap's ability to eradicate infection-causing germs (Maurer, 1997). Through a

qualification process, an approved supply base can be assembled to ensure minimal

capabilities on key dimensions.

2.4.2 Selection and Allocation

As noted earlier, supplier selection is concerned with identifying the subset

of qualified suppliers who will be considered for order placement, and allocation

focuses on splitting the required quantity between the selected suppliers. Obviously

these decisions are interdependent and are also driven by the total delivered

costs to the firm of an order quantity from each supplier. Pan (1989) proposes a

linear programming model to optimally identify the number of suppliers and their

respective quantity allocations to meet pre-specified product requirements. Other

constraints incorporated are related to .,i regate incoming quality, lead times, and

service level. The overall objective is to minimize the price per unit as a weighted

average of selected suppliers' prices. It is assumed that product requirements are

deterministic and supply is reliable and unlimited. In reality, however, it is common

for suppliers to quote alternative pricing schemes and uncertainty exists in both

supply and demand markets. Since the focus of this dissertation is on incorporating

these considerations in making supplier selection and quantity allocation decisions,

prior research relating to this area are reviewed next.

Analytical studies on supplier selection and quantity allocation decisions

show that in certain cases, multiple-sour' in: order-splitting, or diversification

is preferable to single-sourcing. Horowitz (1986) provides an economic analysis

of dual sourcing a single input at differing costs. It is shown that uncertainty in

supply price and risk-aversion of the buyer motivate a firm to place positive orders

from the high cost seller. Kelle and Silver (1990) investigate a continuous review

inventory policy replenished by suppliers with stochastic delivery lead-times, and









find that order-splitting among multiple sources reduces safety stock without

increasing stockout probability. Ramasesh et al. (1991) also analyze a reorder point

inventory model with stochastic supply lead-time, and find that in the presence

of low ordering costs and highly variable lead-times, dual sourcing can be cost

preferable.

Gerchak and Parlar (1990) examine second-sourcing in an EOQ context to

reduce the effective yield randomness of a buying firm's purchase quantity. The

benefits of diversification are traded-off against the costs of managing a larger

supply base to determine whether second-sourcing is worthwhile. They also analyze

the optimal number of identical sources. Rosenthal et al. (1995) introduce a

mixed integer programming model for solving a supplier selection problem with

bundling. The suppliers are capacitated, offer different prices, differing quality

levels, and discount bundles. Agrawal and N i1iini -; (1997) examine a single period

supplier selection and allocation problem with normally distributed supply and

deterministic demand for a single product with fixed ordering costs. They are able

to show that for two non-identical suppliers, the expected profit function is concave

in the number of suppliers.

Parlar and Wang (1993) compare the costs of single versus dual-sourcing for a

firm assuming that the overall objective is to minimize purchasing and inventory

related costs. In their approach, they assume that actual incoming quantities are

a function of a random variable representing the yield. Separately using an EOQ

and newsboy based ordering policy, they are able to show that in certain cases

dual-sourcing dominates single-sourcing. Both of these studies ignore the supplier

capacity issue in making supplier selection and quantity allocation decisions.

Further, Parlar and Wang (1993) note that supplier yields and demand uncertainty

pihl a critical role in the analysis.









Other analytical studies similar to this research examine supplier selection

and order allocation decisions with stochastic demand for the product purchased.

Gallego and Moon (1993) employ Scarf's ordering rule for a distribution free

optimal newsboy order quantity. They maximize profit against the worst possible

distribution of demand with known mean and variance. Separate extensions

incorporate a second purchasing opportunity, fixed ordering costs, random yields,

and multiple items into the analysis. In particular, the case of random supplier

yields assumes that each unit supplied has the same probability of being good, and

the buyer p ', for all units. Bassok and Akella (1991) introduce the Combined

Component Ordering and Production Problem (CCOPP). The problem is one of

selecting ordering and production levels of a component and a finished good for a

single period with supply and demand uncertainty. In their model the distribution

of supply depends on the order quantity given to a single source.

Anupindi and Akella (1993) consider a two supplier, single product procure-

ment problem with stochastic supply and demand. They '-.-. -I that minimum

order quantity policies of suppliers may affect their findings. Gurnani et al. (2000)

simultaneously determine ordering and production decisions for a two component

assembly system facing random finished product demand and random yield from

two suppliers, each providing a distinct component. They also consider a joint sup-

plier option and determine the value to the assembler of reliable component supply.

Diversification occurs when positive orders are placed with the joint supplier and

individual component supplier. They show that if there are not mismatched initial

inventories of components, each component will be single sourced. Kim, et al.

(2002) model a capacitated manufacturer's supply base configuration problem

for multiple parts used to produce multiple products with independent stochastic

demands. They develop an iterative algorithm to determine the manufacturer's

end product mix and specify the suppliers who will be used to satisfy the parts









requirements for each end product. In their model suppliers are capacitated and

reliable. Yano and Lee (1995) also offer a review of the normative literature which

addresses the random yield problem.

Order quantity or lot sizing decisions can be largely influenced by alternative

supplier pricing schemes. Quantity discounts, especially all-units and incremental,

are common pricing practices. In a quantity discount schedule, the range of

potential purchase quantities is segmented by quantity break points. Order

quantities that fall between two break points qualify for a specific unit price.

Typically the larger the quantity ordered, the lower is the unit price. The key

difference between an all-unit discount and incremental schedule is that in the

former, all units ordered are supplied at the unit price, while in the latter; only the

number of units in a specific break point are supplied at the unit price. Suppliers

offer these types of schedules to encourage buyers to procure larger quantities and

to reap the operating advantages associated with these larger quantities (such

as economies of scale). Both the buyer and supplier can realize higher overall

profits by making decisions jointly. SC' I coordination research in part resides in

the research efforts for joint optimal supplier pricing schemes and buyer lot size

decisions that consider quantity discounts.

Operations literature takes a coordinated cost minimization approach for

analyzing the impact of pricing schemes on supply chains. Munson and Rosenblatt

(2001) analyze a three-level distribution channel for a single product where a

supplier provides an order quantity discount schedule to a manufacturer. The

manufacturer then proposes an order quantity discount schedule downstream to a

retailer. The manufacturer solicits and proposes order quantity discounts upstream

and downstream respectively, and is termed the channel captain for his dictatorial

role. As such, the manufacturer retains all channel savings from coordinated

quantity discounts. Numerical experiments provide a ,I' increase in manufacturer









savings from two way order quantity discounts over retailer only quantity discounts.

Further, optimizing over both supplier and retailer simultaneously results in savings

greater than the sum of optimizing over each individually.

Weng (1995) develops an integrated model of quantity discounts and channel

coordination. In his model, operating costs are a function of purchase quantity and

demand is a function of selling price. In this scenario, profit increases result from

cost reductions and demand enhancement. However, a quantity discount scheme

alone is not enough to achieve joint profit maximization. A fixed p ,iment or

franchise fee in concert with a quantity discount schedule is necessary to motivate

system optimal decision making.

Abad (1988a, 1988b) also incorporates price dependent demand into lot-sizing

models with alternate supply acquisition schemes. Optimal lot size and selling

price are simultaneously solved under linear and constant price-elastic demand. A

supplier offers an all-unit quantity discount in Abad (1988a) and an incremental

quantity discount in (1988b). An iterative procedure is developed to handle the

lot-size and selling price interdependency. Burwell et al. (1997) extend the work of

Abad by developing an optimal lot sizing and selling price algorithm for a single

item given supplier offered quantity and freight discounts and price dependent

downstream demand. Numerical results indicate that the maximized profit with

all-units dominates the maximized profits under incremental and mixed quantity

and freight discounts.

It is clear from the above studies that order quantity discounts can influence

decision making at each supply chain link regardless of analytical perspective.

Benton and Park (1996) provide a thorough review of the lot-sizing literature

with quantity discount considerations for time phased as well as non-time phased

demands. Their i::;.n, ,- further classifies research based on the type of discount

(all-units or incremental) and the perspective of the modeling effort (buyer or









supplier-buyer). The scope of this dissertation is one of non-time phased demand

from a buyer perspective. As such, the remainder of this section reviews literature

with similar scope.

Benton (1991) uses Lagrangian relaxation to evaluate a purchasing manager's

resource constrained order quantity decisions given alternative pricing schedules

from multiple suppliers. The example decision maker has a limited budget and

storage space for ten items offered by three vendors, each quoting three all-units

discount levels for each item. The objective is to minimize total acquisition and

inventory costs. The manager must choose a single supplier for all items. However,

if multiple sourcing is allowed, the optimal objective is S' lower than the single

source optimum. Rubin and Benton (1993) use Lagrangian relaxation to formulate

a separable dual problem. A branch and bound algorithm with a best bound

branching rule is developed to close any duality gap. To allow multiple sourcing

among the multiple items, a merged discount schedule is constructed which quotes

the lowest price among all suppliers for each quantity of each item. A shortcoming

of this model is that there are no order cost savings from consolidated delivery or

order placement. As an extension to this work, Rubin and Benton (2003) analyze

the same purchasing scenario, except that suppliers quote incremental quantity

discount schedules instead of all-units. A similar solution methodology is employ, -1

They note the best feasible solution obtained from their relaxation algorithm

should be acceptable for use unless a large duality gap exists. Further, they observe

that in numerical studies the primal cost minimization objective is rather flat. This

implies that good solutions are rather easy to find, but fathoming is slow. Thus, an

optimal solution search is probably more costly than beneficial.

Another Lagrangian relaxation based heuristic is developed by Guder et al.

(1994) to solve a buyer's multiple item material cost minimization problem with

incremental discounts offered by a single supplier. The buyer has a single resource









constraint and demands for items are independent. The complexity of the problem

lies in evaluating all feasible price level sequences. For large problems with n items

and m price breaks, the optimal price level sequence can be obtained in O(nm).

For 100 items with 8 price breaks each, numerical experiments result in precise

solutions quickly. Their heuristic is adapted from Pirkul and Aras (1985) which

analyzed the all-units version of this problem.

The i i i i ily of research on purchase decisions involving alternate pricing

schemes assumes that demand for the purchased product is known. Modeling

efforts that consider stochastic downstream demand and upstream quantity

discounts are Jucker and Rosenblatt (1985), Khouja (1996a,b), and Lin and

Kroll (1997). All of these analyze the implications of quantity discounts on the

newsboy problem. Jucker and Rosenblatt (1985) develop a marginal analysis

based algorithm to minimize total costs of purchasing and transportation with

separate but simultaneous purchasing and transportation quantity discounts.

Khouja (1996a,b) implements a bisection method to solve for a retailer's optimal

order quantity with supplier offered all-units discounts and progressive multiple

retail discounts. Lin and Kroll (1997) develop a solution technique for solving a

risk constrained newsboy problem with all-units or incremental supplier pricing

schemes.

As evidenced by the proliferation of research on integrating supply chains and

adoption of SC' I practices in industry, the importance of supplier management on

a firm's competitive strategy is now widely accepted. A firm's supplier management

approach is characterized by its sourcing strategy. This research provides manage-

rial insight by examining the influence of alternate pricing schemes from suppliers,

uncertainty in product supply and demand, and explicit treatment of diversifi-

cation benefits on a firm's strategic sourcing decisions. Of particular interest is

characterizing when it is optimal for a firm to single-source versus multiple-source.







26

Our modeling and analytical approaches to these problems of supplier selection

and order quantity allocation position this research among the existing literature

reviewed in this chapter.















CHAPTER 3
STRATEGIC SOURCING DECISIONS WITH STOCHASTIC SUPPLIER
RELIABILITY

3.1 Introduction

Supplier sourcing strategies are a crucial factor driving supply chain success.

While many firms utilize a single supplier for a particular item, others diversify

their supply risk by sourcing from multiple suppliers. In particular, a firm's

allocation decision determining an appropriate supplier set and order allotment

impacts on all competitive dimensions for the delivery of finished goods to its

customers, including cost, quality, reliability and flexibility. Benefits of a single

supplier strategy have been touted in the popular press, such as JIT replenishment

and increased quality levels. More recently, the benefits of supplier diversification

as a possible defense against supply disruption has gained attention. In this

chapter, we investigate the implications of uncertain supplier reliability on a firm's

sourcing decisions in an environment with uncertain demand. In particular, we

characterize circumstances under which a firm should diversify its orders amongst

several suppliers to increase its total profit, rather than utilizing a single supplier

sourcing strategy.

The modeling framework developed in this chapter is similar to Anupindi and

Akella (1993). They analyze both single and multi-period models with uncertain

supplier reliability and stochastic demand. In particular, in a two supplier setting,

they characterize scenarios under which it would be optimal to source from one

vs. two suppliers based on initial inventory levels. While their general results

are derived for the two supplier case with no specific distributional assumptions,

they also an i1 -- specific scenarios utilizing an exponential demand distribution,









and normal and gamma distributions to reflect supplier reliability. Parlar and

Wang (1993) also consider a two supplier newsboy model with random yields.

They obtain first order optimality conditions, and establish the concavity of the

objective. Furthermore, they utilize a linear approximation to obtain optimal order

quantities for each supplier.

Another closely related work is Agrawal and N i1ii (1997). They analyze

a variation of the single period newsvendor model where demand is actually

deterministic, but supplier reliabilities are normally distributed. In addition,

they examine the optimal number of suppliers and corresponding order quantities

assuming that a fixed order cost is incurred for each supplier with a positive order.

Optimal policies for N suppliers are derived for the case of homogenous suppliers

(i.e., all suppliers have similar reliability distributions and costs). For the case with

heterogeneous supplier reliability and similar supplier costs, they show optimality

conditions for two suppliers and conjecture that these also hold for cases with more

than two suppliers.

3.2 Sourcing Model

In this section, we introduce the model notation and variables. First, we

introduce the decision variables. Let N reflect the number of available suppliers.

We assume that this set has been pre-qualified such that they all meet minimum

sourcing standards set by the firm. The key decision variable in the model is to

determine the number of units to purchase from supplier i, where (i = 1,... ,N).

Although we do not include an explicit variable reflecting the optimal number of

suppliers, it can be determined implicitly by identifying the number of suppliers

with a non-zero order quantity.

Next, we introduce the supplier specific parameters. We assume that each

supplier quotes a unit cost of ci. In addition, the firm has some knowledge of

the historical reliability in terms of the number of good units delivered for each










supplier. We treat this quality or yield reliability for each supplier as a random

variable, ri. Let gi(ri) denote the continuous probability density function associated

with yield for each supplier i. We also assume that the density function is twice

differentiable with ri and ai representing the mean and standard deviation.

Finally, several firm specific parameters typically associated with a single-

period newsboy framework are relevant here. We assume that the total demand is

unknown but represented by a stochastic parameter x with a probability density

function f(x) and a cumulative distribution function denoted by F(x). Also, the

unit price (p), salvage value (s) and underage costs (u) are also assumed to follow

standard assumptions associated with a newsboy model, including p > ci > s. A

complete list of the model variables and parameters is included in Table 3-1 below.

Table 3-1: Model Notation

Variable Description
N Number of suppliers available
qi Number of units to purchase from supplier i (decision variable)
x Demand (random variable)
f(x) Probability density function associated with demand
F(x) Cumulative distribution function associated with demand
p Mean demand
a Min value parameter for uniform demand distribution
b Max value parameter for uniform demand distribution
ci Unit cost for supplier i
ri Yield for supplier i (random variable)
gi (r) Probability density function associated with yield for supplier i
Gi(ri) Cumulative distribution function associated with yield for supplier i
ri Mean yield factor for supplier i
ai Standard deviation of the yield factor for supplier i
Vi Second moment of the yield factor for supplier i
p Unit price
s Unit salvage value
u Unit underage cost
ci Unit cost for supplier i
Q Total number of good units received
Ki Critical ratio derived from cost for supplier i
bi Uniform demand adjusted by critical ration for supplier i
b' Uniform demand adjusted critical ration for homogeneous suppliers









3.3 Model Development

The objective of the firm is to determine the appropriate order quantities for

each supplier such that the expected profit associated with satisfying demand is

maximized. Utilizing the framework from the traditional newsboy problem (Silver,

Pyke and Peterson 1998), the objective function in Equation (3.1) maximizes the

single period (season) expected profits for the firm. Note that we assume that

the buying firm only p i,,- suppliers for "good"' units delivered. In addition, a

non-negativity constraint (Equation 3.2) is also included in our formulation.

Max E() =


gi (1) 92(r2)... gNrN) (px ciriqi + s(Q x))f(x)dx drN...dr2dr1
0 0 0 0i 1


91 () J 92 J 9N (N) j Q Cqi ux Q))f (x)dx drN...dr2drl
0 0 0 Q i=1
(3.1)

subject to

S> 0 Vi (3.2)
N
where Q = riq
i=i

3.4 Analysis

In this section, we describe the optimal supplier sourcing strategy under

certain conditions. To make our analysis more tractable, we assume that demand is

uniformly distributed with parameters [a, b].1 To support subsequent analysis, we

first ensure that a key result holds.



1 Given that price per unit p is assumed to be fixed, it is reasonable to assume
that demand around this fixed price is uniformly distributed.









Corollary 3.1: The expected profit function shown in Equation (3.1) is concave

in the order quantities qi for N suppliers when demand is uniformly distributed

between [a, b].

Proof: See APPENDIX A.

It is interesting to note that Agrawal and N ,liini (1997) were unable to

prove concavity of the profit function for N heterogeneous suppliers with normally

distributed reliability functions and deterministic demand. However, by assuming

that demand is uniformly distributed, we can derive a closed form expression for

optimal supplier order quantities making no distributional assumptions for supplier

reliability.

We start our analysis for the general case of heterogeneous suppliers. Then we

analyze the case where suppliers are homogeneous with respect to the reliability

distributions but have differing cost structures. Next, we consider the case where

the reverse is true (i.e., all suppliers have homogeneous cost structures but different

reliability distributions). Finally, we analyze the case where all suppliers are

completely homogeneous.

3.4.1 Heterogeneous Suppliers

When all suppliers have heterogeneous cost and reliability functions, a key

result is that there is no one dominant sourcing strategy. That is, under certain

circumstances, it will be optimal to place a single order with the lowest cost sup-

plier; under other circumstances, an order is placed with a subset of the suppliers.

However, it is possible to analytically determine the exact order quantities for each

supplier as shown in Theorem 3.1 below.









Theorem 3.1: When suppliers are heterogeneous with respect to costs and reliabil-

ity parameters, then the optimal sourcing quantity for each supplier i is:

N _2
E (bi bj) j + bi

q i (3.3)

j =1

where b, Ki(b a) + a, and K, p+-

Proof: See APPENDIX A.

Of course, the result of Theorem 3.1 needs to be moderated to account for a

negative order quantity (i.e.,, qj* as computed above in Equation (3.3) is less than

zero for some supplier j). Later, we analyze the case including the non-negativity

constraints. We now focus on further characterizing the situation where a single

supplier sourcing strategy is appropriate. Although the results of Theorem 3.1

do not provide guidelines into a generalizable optimal sourcing strategy, there are

certain specific insights for the heterogeneous supplier case which can be obtained

as discussed below.

Corollary 3.2: The firm will ahv-,- order a positive quantity from the lowest cost

supplier.

Proof: See APPENDIX A.

The key insight stemming from this result is that supplier cost structures

dominate the reliability distributions for selection. Based on this, we can derive

additional insights into optimal sourcing strategies for the firm. We first start by

indexing suppliers (i = 1,.., N) in increasing order of costs (i.e., cl < c2 < c3 <

.. CN). Corollary 3.3 below specifies when a single supplier sourcing strategy is

optimal for the firm.









Corollary 3.3: The firm will source all its requirements from the lowest cost

supplier (i.e., use a single sourcing strategy) if and only if:

(a7)2< bb- (c c)(b -a)_ for j = 2,..N (3.4)
-r, bj (p + u cj)b + (cj s)a

Proof: See APPENDIX A.

The expression on the left hand side (LHS) of Equation (3.4) reflects the

coefficient of variation for the first supplier, while the expression on the right hand

side (RHS) of Equation (3.4) reflects the cost differential between the first and

any of the jth suppliers. Based on this result, a sensitivity analysis of the key

parameters in the LHS and RHS of Equation (3.4) leads us to conclude that the

single sourcing strategy is an optimal choice when:

1. The mean reliability of the low cost supplier is relatively high, rl;

2. The standard deviation of reliability for the low cost supplier is relatively low,

oi;

3. The coefficient of variation for the reliability of the low cost supplier is

relatively low, ';

4. The difference in costs between the low cost supplier (i.e., supplier 1) and the

next highest cost supplier (i.e., supplier 2) is relatively large;

5. The minimum demand parameter is relatively high, a;

6. The maximum demand is relatively low, b;

7. The mean demand is relatively low, a-;

8. The spread in demand is relatively high, (b a);

9. The unit price is relatively low, p;

10. The unit underage cost is relatively low, u;

11. The unit salvage value cost is relatively high, s.

First, consider the impact of supplier reliability on the supplier sourcing decision.

When the coefficient of variation of the first supplier is relatively small, then the









first supplier will optimally receive the complete order. Note that the reliability

distribution is such that most values are less than 10i' Consequently, this

necessitates that a < r < 1. For example, Agrawal and N liini -; (1997) assume

that 3a < r. Therefore, in many cases, the coefficient of variation in the left hand

side of Equation (3.4) is fairly small such that the equation holds and a single

supplier strategy is appropriate.

A surprising feature of the relationship shown in Equation (3.4) is that

the expression is independent of the reliability distribution of suppliers 2,..,N.

Specifically, the only parameter associated with every other supplier that impacts

on the single vs. dual sourcing decision is the unit cost. Therefore, if a supplier has

relatively higher costs than other suppliers, then he/she will not receive an order

from the firm.

Another key parameter driving the single supplier decision is the demand

distribution, as shown in parts (e)-(h) above. In particular, higher levels of mean

demand lead a firm to diversify its total order and source from multiple suppliers.

However, a counter-intuitive result concerns the impact of the variance or spread in

the demand on the optimal sourcing policy. Note that the (b a) expression in the

right hand side of Equation (3.4) is reflective of the variability of demand. From

part (h), higher levels of variability in demand are associated with a single supplier

strategy. Therefore, it appears that there are some interesting interactions between

the uncertainty in demand and uncertainty in supply. When the uncertainty

in demand is low, then it is optimal for a firm to hedge against the uncertainty

in supply by diversifying its total orders amongst several suppliers. Conversely,

when the uncertainty in demand is high, then the firm limits its financial risk and

optimally relies only on the single lowest cost supplier.

Now, consider the case where Equation (3.4) does not hold, and a multiple

supplier sourcing strategy is optimal. From Corollary 3.2, we know that the lowest









cost supplier will aliv-, receive a positive order. Corollary 3.4 discusses the order

quantity for the lowest cost supplier relative to the other suppliers by simply

stating that, although the lowest cost supplier will receive an order, the actual

order size may be lower than other higher cost suppliers. To illustrate, consider the

case where a dual supplier strategy is optimal. In this situation, if the standard

deviation of the first lowest cost supplier is fairly high, then the second higher cost

supplier may receive a higher order.

Corollary 3.4: When a multiple supplier sourcing strategy is optimal, the lowest

cost supplier will not necessarily receive the highest order quantity.

Proof: See APPENDIX A.

The logic used in the two supplier example follows through in determining

an optimal subset of suppliers which receive a positive order quantity. Corollaries

3.5 and 3.6 summarize the importance of supplier cost in determining this optimal

subset of suppliers.

Corollary 3.5: A higher cost supplier will never receive a positive order when a

lower cost supplier's order quantity is equal to zero.

Proof: See APPENDIX A.

Corollary 3.6: The optimal subset of suppliers (n*) receiving a positive order

quantity is the lowest cost subset of suppliers such that the following relationships

hold:
n*-l 2 n* 2
j 1 j \ j/i


j=1 j=1
b > and b,,+l < -- (3.5)



Proof: See APPENDIX A.

3.4.2 Heterogeneous Cost Suppliers

To extend the basic results derived from the analysis of heterogeneous suppli-

ers, we now consider the situation where supplier reliability is homogeneous, but

the costs are not. In particular, let each supplier have a cost of ci and reliability









parameters ri = r, gi = g, Gi = G, and ai = a, Vi. Theorem 3.2 characterizes the

optimal solution under these conditions.

Theorem 3.2: When suppliers (a) have heterogeneous cost structures and (b) have

identical reliability distributions, then there is no one dominant supplier sourcing

strategy. The optimal order quantity for each supplier i is:


r r2 (b b) + bta2
2 + 4 Vi (3.6)
SN Fr2 2 + ,74

Proof: See APPENDIX A.

While the firm optimally places an order with the lowest cost supplier, the

orders for the remaining suppliers are determined via cost differentials and variance

reduction. The results of Theorem 3.2 are similar to those of Theorem 3.1, in that

the lowest cost supplier will receive a positive order, while the others may not.

Anupindi and Akella (1993) analyze the two supplier case with stochastic

demand and supplier reliability where one supplier has a relative cost advantage.

They show that when the initial on-hand inventory falls below some minimal level,

it is optimal for the firm to source from both suppliers. In contrast, Theorem 3.2

and Equation (3.4) for a N supplier setting with uniform demand shows that, in

the absence of capacity constraints, there are circumstances under which it is never

optimal to order from more than one supplier when reliability distributions of

suppliers are identical.

3.4.3 Heterogeneous Reliability Suppliers

In this case, we focus on a situation where supplier costs are roughly equiv-

alent (i.e., c = cVi), but each supplier has a unique historical reliability function

associated with the number of goods actually delivered. Theorem 3.3 characterizes

the first order conditions of optimal sourcing strategy for this case.









Theorem 3.3: When suppliers (a) have identical cost structures and (b) have unique

reliability distributions, then it is optimal to order from all suppliers. The optimal

order quantity for each supplier i is:
ri bl
q- Vi (3.7)

j= 1

where b' K(b a) + a, and K p+u-c
p+-I- s"
Proof: See APPENDIX A.

From the expressions for the optimal order quantity, the firm optimally

orders different non-zero order quantities from all suppliers. Furthermore, it

seems that the supplier reliabilities directly impact on the order quantities such

that the buying firm realizes the diversification benefits. The order quantity

for a particular supplier depends not only upon its unique reliability function,

but also on the reliability of the other available suppliers. This optimal order

quantity for an individual supplier increases in response to the following: (a) an

increase in the mean reliability of that supplier, (b) a decrease in the standard

deviation of reliability of that supplier, (c) a decrease in the mean reliability of

other suppliers, and (d) an increase in the standard deviation of reliability of other

suppliers. Consequently, the order quantity for the first supplier is adjusted for the

uncertainty in the reliability of other suppliers. Further, a firm optimally sources

from the entire supplier pool for this case.

3.4.4 Homogeneous Suppliers

Finally, consider the scenario where the suppliers are roughly equivalent in

costs and reliability expectations. Let c = c, gi = g, Gi = G, ri = r, and oi = a, Vi.

Theorem 3.4 characterizes the optimal solution under these conditions.

Theorem 3.4: When suppliers (a) have identical cost structures and (b) have

identical reliability distributions, then it is optimal to order the same amount from









all suppliers. The optimal order quantity for each supplier is:

rb'
=* r (3.8)
q 2 + Nr2 (3.

Proof: See APPENDIX A.

This result indicates that it is optimal for the firm to diversify its supply base

and place an equal order from every qualified supplier. Furthermore, it is obvious

that the total quantity sourced by the firm increases in the number of available

suppliers (N). This result concurs with those derived by Agrawal and N ili1i i

(1997) for a setting with normally distributed reliabilities and constant demand.

Corollary 3.7: Suppose m of the N suppliers (with m < N) have identical cost and

reliability distributions. Then, the same order quantity should be placed with all m

suppliers.

Proof: See APPENDIX A.

Comparing the results of Theorems 3.1-3.4, these advocate different sourcing

strategies. From Theorems 3.1 and 3.2, if suppliers have unique unit costs, then a

single supplier solution may be optimal. Conversely, from Theorems 3.3 and 3.4, a

multiple supplier solution is ahv--, optimal such that the firm reaps the benefits of

diversification. Hence, it appears that single sourcing strategies could be optimal

only when there are differences in costs across suppliers. On the other hand, if

costs across suppliers are identical then multiple sourcing strategies are an optimal

choice regardless of the suppliers' reliability distributions.

In the next section, we conduct a numerical analysis to further explore some

of the interactions between supplier reliability and firm demand and also to

investigate the impact of minimum order quantities on optimal sourcing strategies.









3.5 Numerical Analysis

3.5.1 Experimental Design

We utilize a set of parameter values which satisfy the assumptions of the

newsvendor model. For the buying firm, we let p = 19, s = 2, and u = 6. Demand

is uniformly distributed over [300,700] and thus the corresponding mean demand

(i.e., (a + b)/2) is 500 while the spread of demand (i.e., (b a)) is 400. For these

examples, we consider a possible supplier set (each meeting minimal qualification

criteria) that consists of three suppliers. In addition, we assume that the supplier

reliabilities are also uniformly distributed over the range 1 [ Li /2, i + Li

/2] with a mean of 1 and a spread of Li. Furthermore, we assume that all three

suppliers have similar reliability distributions with a mean of 0.7 and a spread of

0.1. For the base case examples, we assume that the suppliers have not specified

any minimal order quantities. Later, we investigate the impact of such constraints

on the optimal solutions.

Three sets of examples are included which investigate the impact of changes

in various parameters given that the supplier set has certain cost structures. The

first two sets of examples illustrate the case where the suppliers have heterogeneous

costs, while the third set of examples illustrates the case where the suppliers have

homogeneous costs. Note that the heterogeneous suppliers are alv--, rank ordered

such that the first supplier has the lowest cost and the last supplier has the highest

cost. For the first set of examples, the supplier costs are set at cl = 6.75, C2 = 7.00,

and c3 =7.25. For the second set of examples, the supplier costs are set at Cl =

6.95, C2 = 7.00, and c3 = 7.05. For the third set of examples, the supplier costs

are equal with cl = c2 = C3 = 7.00. The experimental design for the numerical

examples is contained in Table 3-2.

We refer to the first example for each set (i.e., Examples 1, 2 and 3) as the

base case example which reflects the parameter values discussed here. Within









Table 3-2: Description of Numerical Examples

Example Cost Type c1,c2,c3 Parameter Changed Parameter Value
1 Heterogeneous [6.75,7,7.25] Base Case 1
1A Heterogeneous [6.75,7,7.25] Mean Demand (a+b)/2= 5200
1B Heterogeneous [6.75,7,7.25] Spread Demand (b-a) = 800
1C Heterogeneous [6.75,7,7.25] Mean Reliability Supplier 1 M1 .5
1D Heterogeneous [6.75,7,7.25] Spread Reliability Supplier 1 L1 .5
1E Heterogeneous [6.75,7,7.25] Min Order Quantity Supplier 1 minqi =1000
2 Heterogeneous [6.95,7,7.05] Base Case 2
2A Heterogeneous [6.95,7,7.05] Mean Demand (a+b)/2 5200
2B Heterogeneous [6.95,7,7.05] Spread Demand (b-a) 800
2C Heterogeneous [6.95,7,7.05] Mean Reliability Supplier 1 M = .5
2D Heterogeneous [6.95,7,7.05] Spread Reliability Supplier 1 L1 .5
2E Heterogeneous [6.95,7,7.05] Min Order Quantity Supplier 1 minqi 1000
3 Homogeneous [7,7,7] Base Case 3
3A Homogeneous [7,7,7] Mean Demand (a+b)/2= 5200
3B Homogeneous [7,7,7] Spread Demand (b-a) = 800
3C Homogeneous [7,7,7] Mean Reliability Supplier 1 M1 .5
3D Homogeneous [7,7,7] Spread Reliability Supplier 1 L1 .5
3E Homogeneous [7,7,7] Min Order Quantity Supplier 1 minqi = 300
3F Homogeneous [7,7,7] Min Order Quantity All minq 300


each set of examples, we vary one specific parameter relative to those used in the

base case. The first column in Table 3-2 contains the example reference with the

corresponding number (i.e., 1-3) denoting the particular set the example belongs

to, and the corresponding letter (i.e., A-F) denoting a parameter variation relative

to the base case example for that set. The second column specifies the type of

cost structure for the suppliers (i.e., heterogeneous vs. homogeneous), while the

third column identifies the particular unit costs used for each example. Finally, the

fourth and fifth columns describe the parameters and their corresponding values

that are changing relative to the base case example. To illustrate, in Example 1A

the three suppliers have heterogeneous costs (ci = 6.75, C2 = 7, and c3 = 7.25) and

expected demand (i.e., (a + b)/2 = 5200) much higher than that of the base case.

3.5.2 Results

Table 3-3 contains a summary of the results of the numerical examples. The

particular performance metrics included are the optimal order quantity for each

supplier, the total units ordered, the number of suppliers which receive a positive










order, and the firm profit. In the remainder of this section, we discuss highlights

from these numerical examples.

Table 3-3: Results for Numerical Examples

Example Parameter Changed ql* q2* q3* Total n* Profit
1 Base Case 1 880 0 0 880 1 $5,353
1A Mean Demand 5619 1968 0 7587 2 $61,751
1B Spread Demand 1048 0 0 1048 1 $4,604
1C Mean Reliability Supplier 1 1231 0 0 1231 1 $5,335
1D Spread Reliability Supplier 1 174 700 0 875 2 $5,218
1E Min Order Quantity Supplier 1 0 874 0 874 1 $5,199
2 Base Case 2 803 73 0 876 2 $5,230
2A Mean Demand 3259 2529 1798 7586 3 $61,183
2B Spread Demand 1038 0 0 1038 1 $4,458
2C Mean Reliability Supplier 1 759 333 0 1092 2 $5,220
2D Spread Reliability Supplier 1 60 772 42 874 3 $5,202
2E Min Order Quantity Supplier 1 0 802 72 874 2 $5,199
3 Base Case 3 292 292 292 876 3 $5,211
3A Mean Demand 2529 2529 2529 7586 3 $61,158
3B Spread Demand 346 346 346 1037 3 $4,430
3C Mean Reliability Supplier 1 249 349 349 946 3 $5,210
3D Spread Reliability Supplier 1 17 429 429 875 3 $5,208
3E Min Order Quantity Supplier 1 300 288 288 875 3 $5,211
3F Min Order Quantity All Suppliers 438 0 438 875 2 $5,208


First, consider the impact of the supplier cost structures as shown in the base

case examples. Examples 1 and 2 confirm Theorem 3.2 for the case with hetero-

geneous costs and homogeneous reliabilities. In Example 1, the cost differentials

are significant enough such that the firm optimally places a single order with the

lowest cost supplier. In contrast, Example 2 shows the situation where the cost

differentials are small enough such that the firm optimally places an order with

the two lowest cost suppliers. Example 3 essentially illustrates the results shown in

Theorem 3.4 for homogeneous suppliers. Specifically, the total order is equally split

among all suppliers. A comparison of the optimal profit levels for these examples

shows that the cost benefits associated with having the lowest cost supplier (cl

6.75) available outweigh the diversification benefits. Specifically, the profit for

Example 1 is higher than that of Examples 2 or 3.









Second, the impact of changes in demand on the optimal supplier sourcing

strategies confirms the analytic results shown in Corollary 3.3. In Examples 1A and

2A, the optimal number of suppliers with a positive order increases in response to

higher levels of mean demand. Therefore, if a firm anticipates a significant increase

in demand, it should consider enlarging its supplier base. In Example 1B, the firm

still sources from a single supplier, but the total order quantity is increased to

buffer against demand uncertainty. In Example 2B, the firm decreases the number

of suppliers which receive a positive order to only 1, but it increases its total order

size to that single supplier. For the case with homogeneous suppliers, the optimal

number of suppliers remains the same but the total order quantity increases in

response to either an increase in the mean demand or an increase in demand

variability. To summarize, the firm optimally decreases the number of suppliers

receiving a positive order and increases the total order quantity in response to

higher levels of demand uncertainty.

Third, consider the impact of supplier reliability on the optimal sourcing

strategy. In general, it appears that the mean supplier reliability impacts the

corresponding order quantity, but does not affect the optimal number of suppliers.

The spread in supplier reliability, however, impacts on both the corresponding

order quantities and the optimal number of suppliers when the supplier costs

are heterogeneous. In Examples 1C, 2C, and 3C, when the mean reliability of

the first supplier is reduced, the firm optimally lowers its order quantity to the

first supplier but sources from the same number of suppliers. For the case with

heterogeneous costs (i.e., Examples 1D and 2D), in response to an increase in

the spread associated with the reliability for supplier 1, the firm optimally lowers

its order quantity to the first supplier and increases the number of suppliers

that receive positive orders. For the case with homogeneous costs (i.e., Example

3D), the order quantities are adjusted downward for the supplier with increased









reliability uncertainty while the order quantities for other suppliers are adjusted

upward. Recall that for the case with homogeneous supplier costs, it is al--iv

optimal to source from the full pool of suppliers. Therefore, an increase in supplier

reliability uncertainty does not impact the optimal number of suppliers which

receive positive orders.

Next, consider the impact of supplier dictated minimum order quantities

on the firm's optimal sourcing strategy. Note that a minimum order quantity is

essentially analogous to a fixed order cost associated with a particular supplier.

Therefore, the minimum order quantity can be considered a proxy for costs

associated with maintaining the buyer/supplier relationship. Examples 1E and 2E

illustrate the situation where the lowest cost supplier also has a fairly high minimal

order quantity (i.e., min ql=1000). In this situation, the lowest cost supplier no

longer receives an order from the firm. Instead, the firm optimally sources from

an alternative set of suppliers which have slightly higher costs, thus forgoing the

stringent minimum order quantity restrictions associated with the low cost supplier.

Interestingly, the optimal number of suppliers remains the same as the base case for

these examples. In Example 1E, the firm optimally sources the entire order from

supplier 2 instead of supplier 1. In Example 2E, the firm optimally sources from

suppliers 2 and 3 instead of suppliers 1 and 2. The total profit for both of these

examples decreases relative to the corresponding base cases because the firm no

longer sources from the lowest cost supplier.

The effect of the minimum order quantity is slightly different for the case with

homogeneous suppliers. When one of the suppliers has a significant minimum order

quantity, the firm still sources from all three suppliers for a total order quantity

and profit which are essentially the same as the base case, (see Example 3E).

However, the distribution of orders is changed such that the order to the restricted

supplier is increased and the orders to the remaining suppliers are decreased. When









all suppliers have a similar minimum order quantity which is significant, then the

optimal number of suppliers receiving a positive order is reduced, and the results

from Theorem 3.3 no longer hold. In Example 3F, the optimal number of suppliers

is decreased such that the optimal order quantity for each supplier slightly exceeds

the minimal restriction.

Finally, consider the performance of the homogeneous set of suppliers as a

result of changes in supplier related factors. In Examples 3C-3F, the individual

order quantities for each supplier changes in response to changes in the mean

supplier reliability, the spread in supplier reliability, and supplier dictated minimum

order quantities. However, the resulting profit remains very close to that of the

base case shown in Example 3. For Examples 3D-3F, the total order quantity

is also similar to the base case. Therefore, it appears that the scenario with

homogeneous suppliers is fairly robust to changes in an individual supplier's

reliability or minimum order quantities. This result may have some bearing on

risk averse decision maker operating in an environment where consistency in

performance and output is desirable.

3.6 Conclusions

This chapter provides structural and numerical results for determining an

appropriate supplier sourcing strategy in the presence of upstream and downstream

uncertainty. A firm's sourcing strategy is characterized by three interrelated

decisions: (1) the criteria for qualifying as an approved vendor; (2) the suppliers)

selection from the approved base for order placement; and (3) the order quantities

to place with each selected supplier. Our analytical results directly address the

second and third decisions, while our numerical results offer managerial guidelines

for making the first decision.

In the context of the supplier selection decision, our results are in line with

observed practice. For example, Verma and Pullman (1996) find that while supply









managers recognize the importance of quality, cost primarily drives their supplier

selection decisions. In our model, a supplier's cost (and not its reliability) is the

key factor which comes into pl1 i when a firm is deciding whether or not to place

an order with that supplier. Consequently, the lowest cost supplier in the pre-

qualified pool will ah--iv receive a positive order. An exception to this rule shown

through numerical examples is when the lowest cost supplier has a restrictively high

minimum order quantity. It follows that if all pre-qualified suppliers have similar

costs, then it's optimal to place an order with all suppliers in the pool.

When addressing the second sourcing decision, a firm may decide to take an

extreme approach and limit its supplier selection to only a single supplier. In the

chapter, we derive a simple ratio to analytically determine whether or not a single

supplier strategy is appropriate. This ratio reflects a trade-off between the first

supplier's reliability and its cost advantage relative to other suppliers. Essentially,

if the lowest cost supplier has a reliability distribution with a high mean and a low

standard deviation, and has a large cost advantage, then a single supplier strategy

is warranted.

Another key factor influencing the single supplier sourcing decision is the

firm's anticipated demand. Both analytic and numerical results confirm that a

single supplier strategy is favorable when the mean demand is low. However, if a

firm anticipates a significant increase in demand, it should consider enlarging its

supplier base even when the low cost supplier could provide the full order quantity.

Surprisingly, an increase in the variability in demand also favors a single sourcing

strategy. In this case, the firm limits its financial risk by sourcing only with the

single lowest cost supplier when the firm anticipates great uncertainty in demand.

Other factors contributing to a single supplier sourcing strategy are also discussed

in the paper.









The third sourcing decision concerns how much to order from each selected

supplier. Analytic expressions are developed which determine the optimal order

quantity for each selected supplier under a v ,ii I. of circumstances. The most

general case is shown in Theorem 3.1 which addresses the situation where all

suppliers have different costs and reliability functions. In this case, each supplier

will receive an order amount based on its unit cost, mean reliability, and variance

in reliability. Note that while the lowest cost supplier is guaranteed to receive a

positive order, he/she won't necessarily receive the largest order. In contrast, we

also analyze the situation where the suppliers are homogeneous in their costs and

reliability functions in Theorem 3.4. In this situation, all pre-qualified suppliers

receive an equivalent order quantity.

Finally, the first sourcing decision addresses what criteria a firm should use

to pre-qualify a set of suppliers. While there are many factors impacting this

decision that are not addressed here, we can extrapolate some insights concerning

the appropriate number of suppliers that should be pre-qualified. The fixed

costs of qualifying a supplier to ensure that it meets a minimal set of criteria

based on quality, costs, and delivery can be exorbitant. It follows that a firm

should judiciously target a set of suppliers which supports its anticipated sourcing

strategy. For the set of numerical examples that we considered, sourcing only from

a single supplier with very low costs resulted in a higher profit than other multi-

supplier sourcing strategies. Therefore, if a firm anticipates that a single supplier

has much lower costs, then it might be better off to focus on qualifying that single

supplier to ensure that the supplier's reliability, quality and delivery are sufficient

to meet demand. Essentially, additional efforts can be made to ensure consistent

reliability of the lowest cost supplier.

However, if several suppliers are very close in cost, the company should con-

sider qualifying them all such that they meet specific reliability criteria. Indeed, the









numerical examples show that our results for the case with homogeneous suppliers

are fairly robust with respect to changes in demand and supplier reliability. Specif-

ically, when all suppliers in the pre-qualified pool have similar costs and reliability,

then the total quantity ordered and the total firm profit is fairly constant even

when demand and supplier reliability factors change.

There are several future areas of research related to this model which warrant

further investigation. First, a more detailed model could be developed which

addresses the first supplier sourcing decision, or appropriate criteria for qualifying

suppliers. Second, we assume that demand is uniformly distributed to facilitate the

development of simplified expressions. Other types of demand distributions could

be explored to enhance the generalizability of the results derived here. Lastly, the

focus of this model is on decision making at the buying firm; future research could

incorporate the supplier's decision making process as well.















CHAPTER 4
IMPACT OF SUPPLIER PRICING SCHEMES AND CAPACITY ON
SOURCING STRATEGIES

4.1 Introduction

Consider the following procurement process followed by a in, i i' office products

retailer with headquarters in Florida, USA. The centralized purchasing organization

(CPO) for the retailer is responsible for the procurement of all commodity type

products which are sold through the retail outlets. Orders for these types of

products are placed by the CPO on a periodic basis and in general, a four step

process characterizes the procurement process. First, for each time period, the

CPO ..-i'-regates the total estimated requirements for each product based on input

from the retail outlets. Second, using a web-based interface (which is set-up to

allow access only to pre-qualified suppliers), the CPO posts timing and quantity

requirements for the commodity productss. Third, suppliers respond by quoting

prices and quantity limitations, if any, for delivering the products to the central

warehouse maintained by the CPO. Finally, a CPO ,i lv-i analyzes the supplier

submitted information and allocates requirements to suppliers.

The final step of this process for supplier selection and order placement

was biased towards selecting a single supplier who could supply the entire set

of requirements rather than an explicit focus on costs quoted by the suppliers.

The primary motivation for such a strategy (i.e., single sourcing) was that it was

easier to manage order receipts from one supplier at the central warehouse. In a

more recent analysis of supplier responses to the CPO posting of requirements

and timing information, it was found that suppliers were starting to offer quote









pricing schemes and thus, the supplier selection and order placement decision was

becoming increasingly difficult.

This scenario motivates this chapter's analysis of supplier selection and order

placement decisions in the presence of alternative supplier pricing schemes. It

is well known that sourcing decisions in this setting are extremely complex and

also require a frequent reassessment. For example, suppliers often offer discount

schedules to induce larger purchases by offering progressively lower unit prices for

progessively larger purchase quantities. Even for a single product's purchasing

decision, if it is available from many vendors, each with various qualifying order

sizes, identifying the optimal selected supplier set and corresponding quantity

allocations is a difficult decision. Further complicating matters is that decisions

must often be made quickly and with limited information due to time pressures

(Rubin and Benton, 1993). This fast paced decision environment is not conducive

to optimally solving combinatorial problems by complete enumeration. Therefore,

heuristic procedures that produce optimal or near optimal feasible solutions are of

significant value for decision makers.

More specifically, this chapter examines supplier selection and quantity

allocation decisions for acquisition of a single product's total requirements from a

pool of suppliers offering quantity discount schedules. We examine these decisions

for environments where suppliers offer constant, linear discount, incremental

discount, and all-units discount pricing schemes. The remainder of this paper is

organized as follows. In the next section, we review the relevant prior literature.

Section 3 develops our general sourcing model for each of the pricing schemes

and due to the complexities associated with this model, we propose heuristics for

obtaining feasible solutions very quickly. In Section 4, we analyze the performance

of these heuristics through an extensive numerical ,in 1, -i and this is followed by

an application of our approach to data obtained from the office products retailer in









Section 5. A branch and bound algorithm for finding a minimum cost solution to

the supplier selection and quantity allocation problem with incremental discounts

is developed and stated in section 6. Finally, Section 7 discusses implications and

conclusions of our research.

4.2 Sourcing Model

4.2.1 Preliminaries

Our analysis of the supplier sourcing decisions focuses on a single product,

single period analysis of a system consisting of N suppliers (i = 1,..., N) and

a single buying firm. In general, we assume the following three-stage sequential

decision framework for our analysis. At the first stage, the firm F communicates

the total quantity of the single product (Q) which it will procure from the sup-

pliers. Following this, in the next stage, each supplier i discloses a pricing scheme

(f/(.)) and a related maximum quantity which it can provide to the firm (yi). After
receiving this information, the firm makes the supplier sourcing decision (qi) for

each supplier in the third stage.

We model the problem for a buying firm which is either a channel intermediary

with a fixed quantity contract from a set of downstream firms or a manufacturer

making procurement quantity decisions using automated materials planning

systems (such as MRP). In both cases it is reasonable to assume that the buying

firm can declare with reasonable certainty the total quantity Q to be procured from

the suppliers. As noted earlier, zero fixed ordering costs for the buying firm are

being assumed in line with our motivating example of the office products retailer.

There are three additional factors that need to be clarified in the context of

our analysis. First, each supplier is an independent operator and hence, there are

no opportunities for supplier collusion/collaboration in our setting. Second, the

pricing scheme disclosed by each supplier (i.e., fi(.)) is all inclusive and includes

the logistics/transportation cost. Finally, supply lead times are assumed to be









relatively constant and thus, are not incorporated in our analysis. In the next

section, we describe the alternative supplier pricing schemes (and in some cases, the

related supplier capacity) parameters which are investigated in this paper.

4.2.2 Supplier Pricing Schemes and Capacity

There are several types of supplier pricing schemes which are used in the mar-

ketplace. These schemes are primary drivers for analyzing the sourcing decisions

since they have a direct impact on firm-level profits. Further, supplier capacity also

can drive some of these pricing schemes. On reviewing prior literature in the field,

the four distinct supplier pricing schemes which we incorporate along with capacity

consideration are:

1. Constant Price

Under this scheme each supplier i discloses a constant price per unit (i.e.,

fi(qi) = ci) it can provide the firm within the capacity range [0, y].

2. Linear Discount Price

Under this scheme each supplier i discloses a linearly declining price in the

quantity qi purchased from supplier i. Thus, fi(qi) = a biq, (at, bi > 0)

within the capacity range [0, yi]. We also assume that the linearly discounted

pricing scheme disclosed is such that ai biyi > 0 for each supplier i.

3. Incremental Units Discounted Price

Under this scheme each supplier i discloses the traditional incremental

units discounting scheme (\ !111n 1-1 2001) which is dependent on the

quantity qi purchased from supplier i. To specify such a scheme, we first

define k = 1,..., Ki as the index for discount classes offered by sup-

plier i. Corresponding to each discount class k for a supplier i, define

[ik, uik] as the minimum and maximum quantities for the class such that

UiK = i. Based on these definitions, the pricing scheme can be specified as









fi(qii) (i ui) "_'' *. if lik < qi < Uik for k 1,..., Ki. It is
assumed that i, > ir,_. > ... > i,,, for each supplier i.

4. All Units Discount Price

Under this scheme each supplier i discloses the traditional all-units discount-

ing scheme (\ i!ln 1- 2001) which is dependent on the quantity qi purchased

from supplier i. As with the incremental units scheme, let k = 1,..., Ki

as the index for discount classes offered by supplier i, and corresponding to

each discount class k for a supplier i, define [lik, ik] as the minimum and

maximum quantities for the class, respectively, such that uiK yi. Based

on these definitions, the pricing scheme can be specified as fi(qi) = ,./ if

lik < i q< ik for k = 1,..., K. It is assumed that > _. > ... > ; for

each supplier i.

Next, we formulate the sourcing model for each pricing scenario and provide

analytical/experimental insights into the sourcing strategy under each case.

4.3 Analysis and Insights

4.3.1 Constant Price

Given that fi(qi) = c for each supplier i, our sourcing model for this case is as

follows:

Minimize Zc = ciqi (4.1)
i= 1
subject to:


q Q (4.2)
i=1
0 < q< Vi (4.3)


Analysis of this model leads us to the following theorem which characterizes the

optimal sourcing strategy for this case.

Theorem 4.1:Under constant supplier prices, the optimal sourcing policy for the

firm is:









Index suppliers in non-decreasing order of prices (i.e., cl < c2 < ... < Cn).

If yl > Q, then source the complete requirement Q from supplier, i.e., q* = Q

and qj = 0 V j 2,..., n.

If yl < Q, then the following algorithm determines the optimal sourcing

strategy for the firm:

1. Set i-1.

2. Order qi = min{Q, yi} units from supplier i. If qi 0, then set qj = 0

Vj = + 1,..., n and Stop, else goto 3.

3. Set Q = Q qi, i = i + 1 and repeat 2.

Proof: See APPENDIX B.

The results of this theorem show that under this scenario, the optimal sourcing

strategy will ahv--l- choose to source from the lowest cost supplier. Further,

multiple sourcing is optimal only when the lowest cost supplier does not have

adequate capacity.

4.3.2 Linear Discount Price

Given that fi = a biqi for each supplier i, our sourcing model for this pricing

scheme is:

Minimize ZL (ai biqi)qi (4.4)
i=i
subject to:


q = Q (4.5)
i=1
0< qi< Vi (4.6)


For the case where each supplier has adequate capacity (i.e., yi > Q Vi)

to meet the .,.-:-regate requirement Q, it is obvious that the firm will choose to

source the complete requirement Q from supplier j such that ajQ bjQ2

min{aiQ biQ211 < i < n}. Thus, in this case, the single sourcing strategy is an

optimal choice.









On the other hand, if there is at least one supplier i such that yi < Q, there

is no guarantee that this type of single sourcing strategy is optimal. However, by

observing that the objective function is strictly concave in qi, a general result which

can characterize an optimal solution for our sourcing model under this pricing

scheme is as follows.

Result 4.1: There exists at least one optimal solution to our sourcing model

such that qi = 0 or qi = yi for all i suppliers except that there may be at most one

supplier j for which 0 < qj < yj.

Proof: See APPENDIX B.

Given that the optimal solution is difficult to obtain explicitly, we propose

a heuristic that builds upon this result. The algorithm first starts by ranking

suppliers in non-decreasing order of the total costs or average costs assuming

that each supplier is given an order for the maximum he/she can supply or firm's

requirements. Next, the first phase of our procedure considers all suppliers in

this list in a sequential manner. If the supplier capacity is less than the firm

requirements, we place an order for the maximum quantity the supplier can deliver.

Then we update the requirements for the firm based on this allocation, and again

rank the remaining suppliers in non-decreasing order of total or average costs and

the process repeats itself until the remaining requirements are zero. In the final

phase of the heuristic, we consider switching the partial order quantity among all

suppliers who have been allocated a positive order quantity. Details of our heuristic

are as follows.

1. Define the active supplier set, Q, as consisting of all suppliers.

2. For each supplier i included in f, compute ri = ai bi(min{Q, yi}).

3. Rank suppliers in increasing order of r, and index suppliers in this ranked list

[1] [2],...,[N ].
4. Set j = 1.









5. Set q[j] = min{Q, y[j]}. Remove supplier j from f.

6. Set Q =Q q[]. If Q > 0 go to 2, else go to 7.

7. For all suppliers k, with qk > 0, explore all possible improvements in the

solution by switching the partial order quantity within this selected supplier

set.

8. Store the best solution as Solution A

9. Repeat the above process, except at step 2, calculate r = min{Q, yi} (ai -

bi(min{Q, yi})), and in step 6, store the best solution as Solution B.

10. C'! ....-- solution A or Solution B based on the better objective function value.

To evaluate the solution quality of this heuristic, we carried out numerical ex-

periments by randomly generating 30 test problems (APPENDIX B). For every

problem, the supply base size, N, was set to 10, and the total requirements, Q,

equals 2000 units. Further, for each supplier i, the pricing and capacity parameters

were randomly generated from a uniform distribution as follows: base price, ai

U[0,200]; price elasticity, bi U[0,1]; and capacity, y, U[0,1000].

The results of this evaluation are presented in Table 4-1. The gap reported

in this table is the percentage deviation of the heuristic solution from the optimal

solution obtained using LINGO. As can be seen, in 20 of the 30 problems, the

heuristic solution was optimal. For the remaining 10 problems, the worst case

heuristic solution gap was 5.7-'. while the average gap was 1.81 Further all

the heuristic solutions were obtained is less than 1 second using a PC (with 256

KB RAM, a Pentium III Processor, and 550 Mhz clock speed) while to obtain the

optimal solutions using LINGO on a similar machines, the run times were between

several minutes and a over 2 hours. This leads us to conclude that our heuristic

provides good quality solutions in reasonable computation times.










Table 4-1: Linear Discount Heuristic Performance


Instance Heuristic Optimal Gap % Instance Heuristic Optimal Gap %
1 88282.77 88282.77 1 II' 16 168636.10 1 I. .'. 10 II' .
2 103315.00 103315.00 1 II 17 12I,. I7.20 120457.20 1 II .
3 128455.30 128455.30 II "' 18 98735.40 98583.63 0.15%
4 89650.40 88850.70 i i '. 19 94898.40 94898.40 0.00 %
5 129274.70 127915.70 1 .' 20 174835.10 174835.10 II I .
6 130886.20 128455.30 1 "'. 21 39921.43 39921.43 i II .
7 58198.44 58198.44 I II I' 22 45553.20 43856.09 7 .
8 79593.48 79593.48 i i' 23 89040.30 88585.22 0.51%
9 79593.48 79593.48 II I'. 24 111166.30 111166.30 II I .
10 119205.40 119205.40 "iiI'. 25 66280.70 66051.12 i ..
11 79593.48 79593.48 ii' 26 86099.20 81393.94 5.7'.
12 42926.90 41538.80 27 53897.25 53897.25 II I' .
13 110474.80 110474.80 II I. 28 119360.00 119360.00 II I .
14 69594.40 69444.00 0._2' 29 55034.56 55034.56 iII .
15 174675.70 174675.70 II I. 30 195287.90 195287.90 II I .
Mean Gap Ii I.' Worst Gap 5.7'


4.3.3 Incremental Units Discount Price

In this case, fik(qik) = i. (Uij


Uij-1) + "', (qik


lik) if lik < qik < Uik for


k = 1,..., Ki. The sourcing model for this case is:


Minimize ZIU


n Ki
S fik (ik)1ik
i=1 k=1


subject to:


n K,

i 1 k 1
Ki
ik <_ 1 Vi
k-1
likyik < qik < Uikyik Vi, k


Yik [0, 1]


For the case where each supplier discloses a pricing scheme such that uiK, > Q

Vi, to meet the .i,.:.-regate requirement Q, it is trivial to show that the firm will

choose to source the complete requirement Q from supplier j such that ',, =


(4.7)


(4.8)


(4.9)

(4.10)

(4.11)









minl
single sourcing strategy is an optimal choice for this situation.

As with the linear discount pricing scheme, the dominance of the single

sourcing strategy is questionable when every supplier does not have adequate

capacity to meet the .,-- -regate requirements Q. In this case, again, we note that

result 4.1 stated above for the linear discount pricing scheme still holds since our

objective function is piecewise linear concave in qi. Based on this, we propose the

following heuristic algorithm for obtaining solutions to the sourcing model under

this pricing scenario.

1. Define the active supplier set, f, as consisting of all suppliers.

2. For each supplier i included in f,

If UiK, < Q, compute r fiK (tiKi)/iK, and set qi = iK,,

otherwise,

DO k 2, ..., Ki

If lik < Q < uik, ri fik(Q)/Q and q = Q.

END

3. Rank suppliers in increasing order of ri and index suppliers in this ranked list

[1],[2],...,[N], and set j = i.

4. Set Q = Q q[j]. Remove supplier j from f. If Q > 0 go to 2, else go to 5.

5. For all suppliers k, with qk > 0, explore all possible improvements in the

solution by switching the partial order quantity within this selected supplier

set.

6. Store the best solution as Solution A

7. Repeat the above process, except at step 2, calculate in the first if statement

ri = fi(uiK,), and in the DO loop ri = fik(Q); and in Step 6, store the best

solution as Solution B.

8. C'! ....- solution A or Solution B based on the better objective function value.










To evaluate the solution quality of this heuristic, we carried out numerical exper-

iments by randomly generating 30 test problems (APPENDIX B). As with the

linear discount pricing case, for every problem, the supply base size, N, was set to

10, and the total requirements, Q, equals 2000 units. Further, for each supplier i,

the number of price breaks, upper and lower bounds of quantities for each price

break, and unit prices for each price break were randomly generated from a uni-

form distribution as follows: base price, K, U[0,10]; lil = 0, and lik ik + 1 for

k 2,... K; ul U [0,100], and 4ik lik+ U[0,100] for k 2,... K; and i ,, U[0,1]

and it. = t'A-_1- 0.05 for k= 2,..., Ki.

Table 4-2: Incremental Units Discount Heuristic Performance
Instance Heuristic Optimal Gap % Instance Heuristic Optimal Gap %
1 2944.16 2944.16 I I' 16 2435.61 2435.61 i I .
2 2620.90 2613.21 ( 17 3046.59 3046.59 I I I' .
3 2937.55 2937.55 I I 18 3202.77 3202.77 II I' .
4 2680.84 2680.84 I I' 19 2993.12 2993.12 II I' .
5 2477.86 2477.86 II I' 20 2465.22 2465.22 II I' .
6 2592.35 2592.35 I I' 21 2888.12 2888.12 II I' .
7 3306.59 3306.59 II I' 22 2841.58 2841.58 II I' .
8 2964.60 2962.86 III.'. 23 3178.13 3178.13 II I' .
9 2696.40 2685.76 0.I I' 24 3178.13 3178.13 II I' .
10 2574.28 2574.28 II I' 25 2853.53 2853.53 II I' .
11 2670.19 2670.19 I I' 26 2992.17 2992.17 II I' .
12 2365.78 2365.78 I I 27 2753.51 2753.51 II I' .
13 3017.81 3017.81 II I' 28 2850.40 2850.40 II I' .
14 2964.46 2964.46 I I' 29 2699.23 2699.23 II I' .
15 2546.91 2546.91 II "' 30 2704.60 2704.60 II I' .
Mean Gap 0.1i '. Worst Gap 0. II'.


The results of this evaluation are presented in Table 4-2. The gap reported

in this table is the percentage deviation of the heuristic solution from the optimal

solution obtained using LINGO. As can be seen, in 27 of the 30 problems, the

heuristic solution was optimal. For the remaining 3 problems, the worst case

heuristic solution gap was 0.. 11'. while the average gap was 0.2!'. As with the

linear discount case, all the heuristic solutions were obtained is less than 1 second

using a PC (with 256 KB RAM, a Pentium III Processor, and 550 Mhz clock

speed) while to obtain the optimal solutions using LINGO on a similar machine,









the run times were between several minutes and a over 2 hours. This leads us to

conclude that our heuristic provides very good quality solutions fairly quickly.

4.3.4 All Units Discount Price

In this case, fik(qik) if li k < qik < Uik for k = 1,..., Ki for each supplier

i. The sourcing model is now formulated as follows.

n Ki
Minimize ZAU -' i ',/1 (4.12)
i= k= 1
subject to:

nK
Y Y q-Q (4.13)

Ki

k=1
likyik < qik < Uikyik Vi, k (4.15)

Yik e [0, 1] (4.16)


For the case where each supplier has adequate capacity (i.e., uiK, > Q

Vi) to meet the .,:--regate requirement Q, it is trivial to show that the firm will

choose to source the complete requirement Q from a single supplier j such that

Vj minli<
quantity Q.

Again, the dominance of the single sourcing strategy is questionable when

every supplier does not have adequate capacity to meet the .,.:: -regate requirements

Q. Further, in this case, our objective is discontinuous and thus, the result stated

for the prior two cases (for a continuous concave objective function) do not

necessarily apply. On the other hand, the property stated in the result is quite easy

to incorporate in a heuristic and hence, the following heuristic was proposed to

generate feasible solutions to our problem.

1. Define the active supplier set, f, as consisting of all suppliers.









2. For each supplier i included in Q,

If UiKi < Q, compute ri = and set qi = uiK,

otherwise,

DOk 1, 2, ..., Ki

If lik < Q < Uik, ri = and qi = Q.

END

compute ri = max{i ,l .ij < Q < uij}.

3. Rank suppliers in increasing order of ri and index suppliers in this ranked list

[1],[2],...,[N], and set j i.

4. Set Q = Q q[j]. Remove supplier j from Q. If Q > 0 go to 2, else go to 5.

5. For all suppliers k, with qk > 0, explore all possible improvements in the

solution by switching the partial order quantity within this selected supplier

set.

6. Store the best solution as Solution A item Repeat the above process, except

at step 2, calculate in the first if statement ri = I iK,, and in the DO loop

ri = Q; and in Step 6, store the best solution as Solution B.

7. C'! ..... solution A or Solution B based on the better objective function value.

To evaluate the solution quality of this heuristic, we carried out numerical exper-

iments by randomly generating 30 test problems (APPENDIX B). The parameter

setting for each problem are identical to those generated for the incremental units

discount price case except that for each supplier i, the unit price = ,., for

k = 1,... ,Ki.

The results of this evaluation are presented in Table 4-3. The gap reported

in this table is the percentage deviation of the heuristic solution from the optimal

solution obtained using LINGO. As can be seen, in 17 of 30 problems, the heuristic

solution was optimal. For the remaining 13 problems, the worst case heuristic

solution gap was 1. ".'-. while the average gap was 0.-''. All the heuristic










solutions were obtained is less than 1 second using a PC (with 256 KB RAM,

a Pentium III Processor, and 550 Mhz clock speed) while to obtain the optimal

solutions using LINGO on a similar machine, the run times were between several

minutes and a over 2 hours. This leads us to conclude that our heuristic provides

fairly good quality solutions fairly quickly.

Table 4-3: All-Units Discount Heuristic Performance

Instance Heuristic Optimal Gap % Instance Heuristic Optimal Gap %
1 2675.01 2675.01 ii I' 16 2116.36 2116.36 II I' .
2 2313.10 2308.95 0.1.' 17 2753.34 2753.34 I I' .
3 2746.50 2736.80 ii '.. 18 3023.82 3023.82 II I' .
4 2465.49 2465.49 I i 19 2756.77 2756.77 iII I' .
5 2185.60 2181.90 0.17% 20 2245.90 2240.93 0..'".
6 2303.10 2267.30 1 .. 21 2529.90 2527.17 0.11%
7 3117.94 3117.94 II I'. 22 2497.80 2494.66 0.1 :
8 2694.41 2694.41 I I' 23 2931.00 2920.13 0. :' .
9 2381.40 2380.86 0.1I" 24 2161.84 2161.84 II I' .
10 2319.18 2319.18 I I' 25 2578.03 2578.03 II I' .
11 2433.14 2433.14 I I' 26 2650.97 2650.97 II I' .
12 2101.40 2099.66 ii i' 27 2530.91 2530.91 II I' .
13 2777.31 2777.31 I I' 28 2523.10 2523.10 II I' .
14 2727.40 2721.61 0.2 !'. 29 2493.43 2493.43 II I' .
15 2262.40 2259.67 0.1.;'. 30 2461.70 2449.75 0. '".
Mean Gap 0.1 Worst Gap 1 ".


4.3.5 Summary of Insights from Analysis

Based on our analysis, we can offer the following general conclusions. In

regard to the firm's sourcing strategy, for any pricing scheme, the only time a

single supplier sourcing strategy is preferred is when the lowest cost supplier has

adequate capacity to meet the entire demand for the firm. In all other cases, a

multiple sourcing strategy is the general choice. In regard to supplier quantity

allocations, for the case where a multiple supplier sourcing strategy is preferred, we

can determine the optimal supplier quantity allocations only when each supplier

quotes a constant price. When suppliers quote quantity discount pricing schemes,

the optimal quantity allocations are difficult to determine. However, we have

developed efficient heuristics which can generate supplier quantity allocations which










are close to optimal. We now turn to an application of our heuristic approaches for

data obtained from an office products manufacturer.

4.4 Application

The Central Purchasing Organization (CPO) of a n, i i" office products retailer

is frequently faced with complex sourcing decisions due to the presence of quantity

discount schemes embedded within various suppliers' bids. Given the CPO's total

quantity commitments and supplier bid data for two distinct commodity products

we apply our modeling approach.

The data shared by the CPO is for two commodity products among hundreds

of commodity stock keeping units (SKU) that must be frequently replenished for

retail sale. Total quantity requirements for each product are ..-.-- egated from retail

locations by the CPO. The CPO then posts a request for quotation (RFQ) for

each product. Two distinct sets of suppliers quote on product A and product B

respectively. The portfolio of bids for product A consists of six suppliers, four with

constant price quotes and two with discount pricing. Product B's bid portfolio

consists of eight suppliers, four with constant price quotes and four with discount

pricing. All suppliers in both portfolios also provide limitations on their respective

unit capacities. Tables 4-4 and 4-5 contain bid information for products A and B

respectively.

Table 4-4: Product A Bid Information Data
TQC 9855 Price Break Min Price Break Max Unit Price
Supplier Al 0 1000 623
1001 2100 534
2101 3200 465
Supplier A2 0 2100 452
Supplier A3 0 2650 457
Supplier A4 0 1000 449
Supplier A5 0 700 654
701 1920 494
Supplier A6 0 2200 453










From the bid information we generate a data set for each product's suppliers.

Each data set is implemented twice. First it is assumed that quantity discount

quotes are of the incremental variety. Secondly, an all-units quantity discount

structure is assumed. Further, constant or fixed pricing can be modeled as an

incremental or an all-units discount with only one discount level. Therefore, our

heuristic can handle portfolios of supplier bids that consist of constant price quotes

as well as more sophisticated bids with discount pricing.

Table 4-5: Product B Bid Information Data

TQC=7680 Price Break Min Price Break Max Unit Price
Supplier B1 0 1200 634
Suppler B2 0 600 875
601 1300 800
1301 2500 725
2501 2900 634
Supplier B3 0 700 790
701 1600 710
1601 3000 620
Supplier B4 0 1460 621
Supplier B5 0 1275 625
Supplier B6 0 2600 632
Supplier B7 0 400 922
401 1400 822
1401 2000 722
2001 2600 622
Supplier B8 0 800 821
801 1750 700
1751 2400 610


Using our heuristics we determine supplier selection and quantity allocation for

this industry problem. Tables 4-6 and 4-7 provide a summary of heuristic solution

performance versus LINGO's Global Solver optimal solutions for products A and B

respectively.

For the incremental instances, the heuristic solution is optimal and the CPO

should source from five suppliers for each product. In the all-units instances the

heuristic arrives at near optimal solutions that vary from the optimal solutions

in both the pool of selected suppliers and allocations among them. For example,

it is optimal to source from five suppliers for product A and the heuristic selects









Table 4-6: CPO Product A Solutions Comparison

Incremental qA1 qA2 qA3 qA4 qA5 qA6 Cost Gap
Optimal 0 2100 2650 1000 1905 2200 4658920
Heuristic 0 2100 2650 1000 1905 2200 4658920 I'.
All Units qA1 qA2 qA3 qA4 qA5 qA6 Cost Gap
Optimal 2101 2100 2454 1000 0 2200 4493243
Heuristic 2905 2100 2650 0 0 2200 4507675 :'.

Table 4-7: CPO Product B Solutions Comparison

Incremental qBl qB2 qB3 qB4 qB5 qB6 qB7 qB8 Cost Gap
Optimal 1200 0 1145 1460 1275 2600 0 0 4976485
Heuristic 1200 0 1145 1460 1275 2600 0 0 4976485 I' .
All Units qBl qB2 qB3 qB4 qB5 qB6 qB7 qB8 Cost Gap
Optimal 0 0 3000 279 0 0 2001 2400 4741881
Heuristic 0 0 3000 1460 820 0 0 2400 4743160 0.1:'.


only four. Optimally, supplier A4 is selected and allocated 1000 units by decreasing

allocation to suppliers Al and A3. For product B the heuristic and optimal

solution consists of four selected suppliers. However, the optimal solution selects

supplier B7 and not supplier B5 and our heuristic selects supplier B5 instead of B7.

Also, the optimal allocation to supplier B4 is 279 units and our heuristic allocates

1460 units. Therefore, while the optimality gap is slight for the all-units instances,

the sourcing strategy decisions of supplier selection and quantity allocation are

quite different.

4.5 Optimal Algorithm for Incremental Quantity Discounted Sourcing

In this section we present a branch and bound algorithm for optimally solving

our sourcing problem when all suppliers quote prices with incremental quantity

discounts. The sourcing problem is exactly as described in section 4.3.3. As such

we need to solve for the minimum of the sum of piecewise linear concave functions.

Again, by leveraging Result 4.1, we search for a solution such that at most one

supplier's allocation is not at a boundary of zero or his capacity. Since our sourcing

problem with incremental quantity discounts is a special case of the separable

piecewise linear concave cost allocation problem, we develop and outline an









optimal algorithm for solving the separable piecewise linear concave cost allocation

problem. The mathematical formulation of the problem is as follows:

(SPLCCAP): Minimize f(x)

subject to x ED n C

where x = (x,... ,x,) E R. Given are f(x) = 1I fi(xi) and for each

i, fi concave and bounded on [li, ui]; C = 1 Ci, Ci = [li, ui], and li, ui E R+;

D x: 1,i i = Q, Q E R-; and D C assumed to be bounded.

4.5.1 Algorithm Description

In this algorithm we branch if the solution at the current node, p allocates

any xi such that l' < xi < ui. Branching corresponds to partitioning the subset

of solutions in a hyper-rectangle CP E C into two subsets of solutions in the

two hyper-rectangles CV and C' where Cq U C = CP. Bounding corresponds to

determining a lower bound on the optimal value of (SPLCCAP), f*, in CP.

As the branch and bound tree is generated, the nodes are numbered accord-

ingly. Let NP, for p = 0,1,2,..., denote node p, where No is the root node and

Co = C. At stage k of the algorithm nodes 2k and 2k 1 are generated. A hyper-

rectangle CP = xi : lf < xi < uf, i = 1,2,..., n is associated with each node, NP.

The branching process to be discussed later determines the appropriate values for li

and uf for all i at each node, p. Let LBP be a lower bound on the optimal value of

f over GC CP. The calculation and validity of LBP is described shortly.

Each solution, xP, to a linear knapsack problem at each node, NP, is a feasible

solution to (SPLCCAP). Therefore the upper bound, UBf, on f* is the minimum

value of f over all encountered feasible solutions, xP. The incumbent solution, XUB,

is the feasible solution, x' such that UBf = f( x).

In the algorithm, a candidate list is maintained which includes nodes from

which to branch. As the branch and bound tree grows, for each node, NP that is

created, if LBP < UBf, node N1 is added to the candidate list and a comparison









is executed to determine whether UBf can be updated. If LBP > UBf, node N1

is pruned. Further, if LBP < UBf, a separation variable is determined and from

this separation variable node NP is partitioned into two new nodes. For the new

nodes, lower bounds are calculated and if necessary, pruning is performed and UBf

is updated. The process of node selection, partitioning and obtaining bounds on

f* is repeated until the candidate list becomes empty. When the candidate list is

empty the algorithm terminates with an optimal solution, XUB and f* f(xuB).

4.5.2 Computation and Validity of LBP and UBf

The algorithm determines bounds on each concave subproblem by solving

a linear programming relaxation. For each concave term, fi(xi) a linear under-

estimator is constructed, gi(xi) such that gf(lI) fi(lf) and gp( ) = fi(uf). In

fact, gi(xi) is the convex envelope of f (xi) over [lf, u ]. Also, from Falk and Soland

(1969), it is well known that the convex envelope of f(x) = Y 1 fi(xi) over a
rectangular set CP is gP(x) = i gf (xi) such that x E [1i, ul]. Based on this, we

know:

g (x) < f(x if x4 E [i, uf].
Therefore, the lower bound LBP, on f* over D Q C is given by the optimal

solution to the linear knapsack problem (LKPP).

(LKP ) minimize gP(x)

subject to x E D C.
Let LBP = gP(x*), where x* is an optimizer of (LKPP). Thus from the

discussion above and the fact that (D Cp) C (D C), is a valid lower bound on

f*. Thus, at node NP of the branch and bound tree, we solve the lower bounding

subproblem (LKPP).

From subproblems (LKPP),p = 0, 1,..., a solution x is produced by the

algorithm. Since for each p = 0, 1,..., x E D n C, then each x' is a feasible









solution to (SPLCCAP). The validity of the upper bound on f*, UBf, follows from

the algorithm setting UBf= min f(xP) : p = 0,1,...

A lower bound on f* over D Q C, LBf is now expressed. Let NBk denote the

set of nodes not branched from at stage k of the algorithm. A node NP E NBk due

to fathoming or current inclusion in the candidate list. It can be shown that:

LBf minNPENBkLBP.

Therefore at each stage k, k = 0,1,... of the algorithm LBf < f* < UBf.

4.5.3 The Branching Process

At stage k of the algorithm, a node NP is selected from the candidate list.

Branching is invoked and node N1 is partitioned into two new nodes N2 -1 and

N2k. This is accomplished by first choosing a separation variable, xP which satisfies

fs(x) gf(x) > 0 (i.e., 1 4 < x4 < uw). Note If fs consists of t, linear intervals
or price break points, BPy, (j 1,... ,ts), then BPb-1 < xb < BPb for some

b E 2,..., t,. Then the interval [1$, ui] is partitioned in the following manner to

generate nodes N2k-1 and N2k and their respective sets C2k-1 and C2k
[/2k-1 2k-l1 p a [12k 2k 2 P p
[k1 ak \11= ^s[1P, BPb] and [2k, ul B, uP,

Recall that each fi is piecewise linear concave over [li, ui]. Therefore each

fi consists of ti intervals (price break points) over [li, uj]. Furthermore, since the

branching rule guarantees that lP and uP for all p = 0, 1, 2,..., correspond to price

break points, there a finite number of hyper-rectangles that can be generated. Also,

each invocation of the branching process will create two new hyper-rectangles. Due

to these constructs the algorithm is finite.

4.5.4 Formal Statement of the Algorithm

1. Set k = 0 and solve (LKPk). Set UBf = f(xo) and XUB = x. If LBo < UBf,

add No to the candidate list, set k = 1 and go to step 2. Otherwise, go to

step 4.

2. If the candidate list is empty, go to step 4. Otherwise, go to step 3.









3. Remove any node N1 from the candidate list. If LBP > UBf go to step 2.

Otherwise, generate nodes N2k-1 and N2k and find LB2k-1 and LB2k. If

necessary, update UBf and x. If LBM < UBf for each j = 2k 1, 2k, add

node NM to the candidate list. k = k + 1. Go to step 2.

4. XUB is an optimal solution for (SPLCCAP) with f* = UBf. Terminate.

At each node of the branch and bound tree we generate a new LP, each of

which is a knapsack problem (KP) of size Q. Using a greedy approach based on

each suppliers linearized unit cost, we can optimally solve each KP by inspection.

Further, each KP solution is feasible to our original NLP. By inputting our

test problems, this branch and bound algorithm validates the optimal solutions

obtained using LINGO's Global Solver. Table 4-8 provides a summary of the

number of subproblems solved for each test problem.

Table 4-8: Number of Subproblems Solved for each Test Problem

Test Problem 1 2 3 4 5 6 7 8 9 10
Sub Problems 3 7 5 3 3 11 1 3 7 5

Test Problem 11 12 13 14 15 16 17 18 19 20
Sub Problems 1 7 5 5 3 5 3 3 1 5

Test Problem 21 22 23 24 25 26 27 28 29 30
Sub Problems 3 5 3 3 1 7 1 5 1 5


4.6 Conclusions

The analysis of alternate supplier base pricing schemes in this chapter provides

guidance for a buying firm's optimal sourcing strategy. For the constant cost case,

if the lowest cost supplier has enough capacity to satisfy all of the buying firm's

product requirements, single sourcing is optimal, otherwise multiple sourcing is

best. For sourcing scenarios where supplier bases offer quantity discount pricing

schemes, if all suppliers in the base possess enough capacity to individually provide

Q units, then it is optimal to single source from the least cost supplier evaluated









at Q units. However, if even one supplier in the base is incapable of providing Q

units, then single sourcing may not be the optimal sourcing strategy.

In cases where quantity discount quoting suppliers' capacity is individually

inadequate for a buying firm's product requirements, a firm's sourcing problem

can become extremely complex. Therefore, identifying an optimal solution may

be resource (time) prohibitive for supply chain sourcing professionals. In such

cases, the heuristics developed in this chapter can be expected to efficiently provide

good quality solutions. In sum, the heuristic solutions to randomly generated test

problems arrived at the optimal solution in i.'. of the instances. Furthermore, in

the 3%-'. of non-optimal solution instances, a respectable .6 !'. average optimality

gap exists.















CHAPTER 5
STRATEGIC SOURCING WITH DIVERSIFICATION CONSIDERATION

5.1 Introduction

To be successful in the long term, a firm's sourcing strategy tod i must be

consistent with the vision of the firm for tomorrow. This vision is at the core of

a firm's corporate strategy. Corporate strategy is not only a company's pattern

of purposes and goals that defines the type of company that it is, it is also the

policies that direct goal achievement (Andrews, 1980). The intent of the general

model developed in this chapter is to provide a strategic sourcing framework to

capture subjective elements of a firm's corporate strategy that directly influence

sourcing decisions. The general approach in this chapter for examining supplier

selection and quantity allocation decisions is two fold. First, an integrated model

is developed to address both of these decisions. Given uncertain product demand,

we simultaneously consider supplier costs, supplier reliabilities, supplier capacities,

manufacturer inventory costs, and manufacturer diversification benefits in making

these integrated decisions. Second, using our model, we characterize conditions

under which single-sourcing and multiple-sourcing strategies are optimal. A key

feature of the model facilitating this analysis is an explicit treatment of supply

diversification benefits. These executive level valuations of alternate supply base

sizes are assumed to be consistent with the firm's competitive strategy. This

strategic direction may be presently inconvenient, in that it may indicate that

current profits should be foregone in lieu of potential future gains.

This work is most closely related to those of Pan (1989) and Parlar and Wang

(1993). Pan (1989) proposes a linear programming model to optimally identify the

number of suppliers and their respective quantity allocations to meet pre-specified









product requirements. Other constraints incorporated are related to i.-.- egate

incoming quality, lead times, and service level. The overall objective is to minimize

the price per unit as a weighted average of selected suppliers' prices. It is assumed

that product requirements are deterministic and supply is reliable and unlimited.

Parlar and Wang (1993) compare the costs of single versus dual-sourcing for a

firm assuming that the overall objective is to minimize purchasing and inventory

related costs. In their approach, they assume that actual incoming quantities are

a function of a random variable representing the yield. Separately using an EOQ

and newsboy based ordering policy, they are able to show that in certain cases

dual-sourcing dominates single-sourcing. Both of these studies ignore the supplier

capacity issue in making supplier selection and quantity allocation decisions.

Further, Parlar and Wang (1993) note that supplier yields and demand uncertainty

plh- a critical role in the analysis.

We build upon both of these studies by analyzing the simultaneous supplier

selection and quantity allocation decisions for a single firm facing supply unreliabil-

ity and demand uncertainty. Further, we incorporate an explicit benefit related to

requirements diversification among the supplier base in identifying optimal sourc-

ing strategies. In the next section, the integrated supplier selection and quantity

allocation model which forms the basis of our analysis is described. In Section 5.3,

we characterize the optimal sourcing strategies under various scenarios. In Section

5.4, we proceed to discuss extensions to this modeling effort and this is followed by

an extensive numerical cin ii ; in Section 5.5. Finally, conclusions and managerial

implications are discussed in Section 5.6.

5.2 Integrated Selection/Allocation Model

5.2.1 Preliminaries

Our examination of the supplier selection and quantity allocation decisions

focuses on a single-period analysis of a two stage supply chain consisting of N









suppliers (i = 1,..., N) and one buying firm. All N suppliers are assumed to have

been pre-screened by the firm and are thus, included in the supplier base. The

firm faces an uncertain single-period demand w (with f(w) and F(w) representing

density and distribution functions, respectively) for the product requirements

which it satisfies through procurement from the N suppliers. We assume that this

product is being supplied to the next stage of the supply chain at a unit price p.

Excess inventory of the product is disposed of at the end of the single period by

the firm which receives a price of s per unit while unsatisfied demand "(< -I the

firm u per unit. For each supplier i we assume that the firm has information on (a)

the cost per unit ce; (b) the capacity (in units) yi; and (c) the reliability index ri

representing the historical percentage of ;ood" units (i.e., 0 < ri < 1 Vi) received

from the supplier1 In line with the assumptions in single-period inventory models,

we assume that p > ci > s for all i (Silver, Pyke, and Peterson, 1998). We assume

zero fixed order placement costs for the firm since in current industrial settings

where orders are issued online (through, for example, B2B exchanges), these costs

are negligible (N !iiin .- 2001).

A final component of our model characterizes the diversification benefit func-

tion d(-). The motivation for incorporating this function stems from observed

industry practices. For example, consider HP's sophisticated Procurement Risk

Management program which was initially aimed at better managing the pro-

curement risk of critical memory components. Incorporated in this program is a

portfolio approach to assess and mitigate pricing and availability exposure to insure



1 In contrast to Parlar and Wang (1993), we assume that the firm has exact
knowledge on the reliability of each supplier. This is primarily due to the fact that
the uncertainty in this parameter required Parlar and Wang (1993) to impose some
very restrictive conditions on other parameters in order to show that dual-sourcing
is preferred to single-sourcing. We do, however, analyze the impact of changes in
this parameter on the resulting selection/allocation policies.









margin stability. HP expects the revenue contribution of this program to approach

$1 billion as it expands its use throughout the company. This financial benefit

stems from reduced unit costs and avoiding costs that arise when product delivery

is d. 1 -i, .1 due to sourcing allocation. Additionally, as a result of this program, HP

has learned that by segmenting its expected requirements it can target contracts to

take advantage of a supplier's particular strengthss. This results in more efficient

supply chain practices that create shared savings for HP and its suppliers (Shah,

2002). However, these supply chain savings are sometimes difficult to quantify.

The issue of positive diversification benefits is well documented through anec-

dotal and/or case examples. In addition to HP, Unifine Richardson's decision to

change from a single honey supplier to multiple honey suppliers to meet more strin-

gent regulation requirements; and Wendy's decision to find a second high-volume

supplier of chicken given the increasing demand for its chicken products illustrates

the importance of these diversification benefits (Prahinski, 2002; Lambert and

Ki., ii. i r, 2004). Toyota and Honda also have a policy of sourcing all components

from a minimum of two or three suppliers (Liker and Choi, 2004).

The diversification function d(.) reflects buyer specific supply chain efficiency

savings and strategic positioning benefits. In general, this function captures the net

benefits of choosing to source product requirements from multiple suppliers and is

analogous to the risk-averse expected utility function maximized in the portfolio

selection problem (Gerchak and Parlar, 1990). By choosing multiple suppliers the

firm can reduce the risk associated with selecting a single supplier (Ramasesh et

al., 1991). Consequently, the diversification function is essentially insurance against

supply disruptions attributable to the size of the supply base for a specified part.

We also recognize that the there is a potential decline in diversification

benefits if the number of selected suppliers is too large due to excessive individual

order related costs. We incorporate this by assuming that that d(X) is strictly









concave in X where X represents the number of suppliers selected by the firm.

To support this specific functional form, consider the findings of Agarwal and

N i1,1,1i 1- (1997). These authors show that expected firm profits are concave in

the number of suppliers selected in a setting characterized by stochastic supplier

quantity reliability. Further, the notion of such benefits being piece-wise concave

is documented in prior work (e.g., Gerchak and Parlar, 1990; and Ramasesh et al.,

1991). Thus, our d(X) function could be regarded as a proxy for diversification

benefits associated with uncertainty in the quantities delivered by each supplier.

While the diversification function may be difficult to quantify, a value/utility

function procedure outlined below can be used to gauge diversification valuations of

a firm's decision makerss.

1. To start with assume that diversification benefits are not included. Then

using the result of Theorem 5.1, determine the total quantity ordered by the

firm MQ,, = q[1] and the corresponding expected firm level profits be MZc.

2. Set j 2

3. Pose the following question to the firm level sourcing manager:

Selecting j 1 suppliers, ;,;, will order MQ,, units and expect profits of

MZnc. Assuming that ;. ;' decide to select j suppliers for ordering MQ,,

units, ix,,' expected profits will decline for certain by $ (cj- cj-1) per unit

ordered from supplier j What $ amount would compensate ;/.';, for the

decline in expected profits -- i,,',,.;:,j that ;,. ', decide to: (a) source a., r1i.

MQ,, units from all j suppliers?; and (b) order at least 1 unit from supplier

j.

Record this $ amount as d(j).

4. If j = N stop else set j = j + 1 and repeat Step (3).

The process outlined is based on established work in utility theory and de-

cision making under uncertainty, and is an adaptation of the reference gamble









procedure. Furthermore, it utilizes results from Theorem 5.1 where no diversifica-

tion benefits are realized. The outcome of this process is to determine a v,' i" for

d(i) for i = 2,... ,n) (d(1) = 0 since no diversification benefits are realized by the

firm with a single supplier). However, elicitation of these values from procurement

policy setters or decision makers is not he focus of our analysis.

The key decision variables for our selection/allocation model are both the

number of suppliers and the order quantity for each supplier. We define the binary

decision variable xi to be 1 if we choose to source from supplier i, and 0 otherwise;

and the related allocation quantity qi (in units) procured from supplier i.

5.2.2 Model Development

Let us start by expressing the profit function without diversification benefits

as2 :

NV ,[\N i W < vN i
PW Li 1 iriqi + s[i i if w < N rqi
N _N N
Si=1 riqi i=- cri iw q ] if w > 7_ riqi

Since this profit is uncertain and depends on the exact realization of demand w,

we use the traditional newsboy analysis for this profit function (Silver, Pyke, and

Peterson, 1998), and determine the expected profits as:

N N
E(II) = (p s)p ciqiri + s rjqj (p s + u)ES
i= 1 i 1



2 Similar to Agarwal and N ilini ,1 (1997), this profit realization characterizes a
situation where the firm compensates suppliers for only the sood" units supplied
(i.e., riqj). Later in the chapter, we consider the impact of the firm compensating
the supplier for the complete order quantity (i.e., qi).









where


p = mean demand

ES =expected number of units short
N N
J [w- ( qsr)]f(w)dw
i1 qii) i

and depending upon the distribution of demand, we can specify the expected

shortage ES3 .

Based on this, the firm's expected profit (including the diversification benefit)

maximization sourcing model can be defined as follows:

Maximize Zq,, = E(II) + d(X)

N N
(p s)p ciqiri + s (p- s + u)ES + d(X) (5.1)
i= ii=

subject to:


qi < yxi Vi (5.2)
N
X = i (5.3)
i=1
qi > 0 Vi (5.4)

i = {0,1} Vi (5.5)


where constraint set (5.2) integrates capacity limitations when xi = 1 (or supplier i

is selected) or forces the quantity allocation decision qi to be 0, if xi = 0 (or when

supplier i is not selected), constraint (5.3) determines the total number of suppliers

chosen for sourcing total product requirements, and equations (5.4) and (5.5)



3 For example, if demand is normally distributed with mean p and standard de-
viation a, then ES = G(Z) where G(.) is the loss function and Z is the standard
normal deviate (i.e., Z (E, qri -).









are the non-negativity and binary restrictions on the decision variables qi and xi,

respectively. In the next section, we proceed to analyze this model and characterize

the optimal solutions and sourcing strategies for several cases.

5.3 Analysis

The focus of our paper is on investigating sourcing strategies for the supply

chain. Primarily we are interested in identifying when it is optimal for the manu-

facturer to use multiple suppliers versus a single supplier. We start our analysis by

first assuming a zero diversification benefit. These results will serve as a base case

in our in ,-i-

5.3.1 No Diversification Benefit

Assuming that the firm does not obtain any explicit diversification benefit for

choosing to source from more than one supplier, let us examine the structure of the

optimal sourcing policies. When the diversification benefit d(X) = 0, our integrated

sourcing model can be formulated as follows.

N N
MaximizeZ, (p s) ciqir + s qir (p- s + u)ES (5.6)
i= 1 i= 1

subject to:


qi < yi Vi (5.7)

qi > 0 Vi (5.8)


Let us first analyze this problem assuming that supplier capacity is not a

significant issue. Note that this scenario may be relevant to smaller manufacturing

firms with larger suppliers. In this case, it is relatively easy to show that the firm

commits all its requirements to a single supplier and this supplier is, as would be

expected, the one which offers the lowest per unit cost to the firm. The theorem

given below formalizes this result.









Theorem 5.1: 4 When the suppliers are uncapacitated and there are no diversifica-

tion benefits and there is a unique least cost supplier, then it is optimal for the firm

to order its total requirements from the least cost supplier. Under this scenario, the

total usable quantity ordered from the least cost supplier (i.e., q[i]r[i]) is determined

such that:
P C[i]1+ U
F(q[] c[ +
p-s+u

where c[1] is the cost per unit charged by the lowest cost supplier.

Proof: See APPENDIX C.

One surprising result of this single sourcing strategy is that supplier reliabili-

ties do not impact the supplier choice (i.e., the supplier choice is based strictly on

cost considerations regardless of the quantity reliability parameter ri). On further

investigation, we find that this result is solely due to the fact that the manufacturer

only incurs the purchasing cost for ;ood" units (i.e., incurs cost ci per unit for riqi

units). In certain situations, the cost of defective units in a delivery may need to be

absorbed by the manufacturer. To reflect this scenario, the uncapacitated supplier

model without diversification benefits can be reformulated as:



4 While this result is analytically trivial, it is driven by the fact that for all sup-
pliers j = 1,..., N, cjrj is a constant. However, it points to the fact that when
the firm uses historical information of supplier reliabilities and only compensates
a supplier for the ;ood" units received, then only cost considerations pl. i a role
in determining which supplier will be chosen to received the complete order. This
result also forms the basis of how the firm's single supplier selection decision is
moderated should it decide to compensate suppliers for all units ordered (i.e., the
firm absorbs the complete costs of defective units which could occur in situations
where suppliers hold more 1" ,v i in the channel). In this case, we show that ratio
of costs to reliability drives the choice of the single supplier who will receive the
entire order from the firm.









Maximize E(HI)
N N OC N
= (p- s)>- cqi + s qir- (p- s+ u) [w ( qr)] f (w)dw (5.9)
i=1 i=1 ii1 iri) i 1

subject to:

q > 0 Vi (5.10)

It is easy to show that E(H)' is strictly concave in qi, and thus, the FOC are nec-

essary and sufficient to identify a global optimal solution to this model. Essentially,

the manufacturer determines the total ;ood" quantity received from all suppliers

(i.e., z 1 qiri) such that:

p (c i/ri) + U
F( qi) + (5.11)
i= 1 p-s+u

The issue, of course, is which supplier's reliability adjusted unit cost (i.e., ce/ri) is

relevant in determining this total quantity. The optimal policy, which can be easily

verified, is for the manufacturer to place an order for the total quantity from a

single supplier with the lowest cost/reliability ratio. Hence, if suppliers are indexed

in order of decreasing cost/reliability ratio such that:

C[] C[] [N] (5.12)
[l] r[2] r[N]

then the manufacturer determines the quantity to purchase from supplier [1] such

that:
p- (C[1]/r[1]) + U (5.t3)
p- (c[]/r[i]) + u
F(+q[r[l]) = (5.13)
p-s+u
and orders zero units from all other suppliers. Thus, even in this case, it is optimal

for the manufacturer to adopt a single sourcing strategy except that the choice of

the supplier is based on the lowest cost/reliable unit.

Given that eliminating capacity constraints results in single sourcing, let us

now proceed to examine how this solution changes if supplier capacity constraints









do not necessarily permit the firm to place orders for all requirements with the

least cost supply, i` The theorem below characterizes the optimal supplier

selection and quantity allocation policy with capacitated suppliers.

Theorem 5.2: When suppliers are capacitated and there are no diversification

benefits, then the optimal number of suppliers selected and the corresponding

quantity allocated to each supplier can be determined as follows.

Step 1: Index all suppliers in increasing order of cost per unit (i.e., c[1] < c[] <

C[3] < C[N]).
Step 2: For each supplier [i] (i 1,..., N), determine Q[j] such that:

F(Q[i) P-c[i]+

and based on this determine:

For i 1, t[] =- Q[]. Otherwise, let t[] = Q[] 1 y[jlr[j]

Step 3: The optimal number of suppliers selected (k) is max{1 < k < Njt[k] >

0}.

Step 4: The quantities allocated to supplier j = 1,..., k 1 are determined

such that q[j] = Y[j, and the quantity allocated to supplier k is q[k] = min{t[k], [k]}.

The total quantity ordered by the firm from all suppliers can be determined as

min{Q[k], 1 ]}

Proof: See APPENDIX C.

An interesting observation based on these results is that the firm's optimal

total order quantity when suppliers are capacity constrained is alv--, lower than

the optimal total order quantity when the lowest cost supplier's capacity is not

binding. This leads to the general result that expected profits for the firm dealing



5 All the analysis in the remainder of this chapter reflects the original assump-
tion that suppliers are compensated by the firm for only the good" units received
by the firm rather than all the units supplied.










with capacitated suppliers are never higher than the profits realized by a firm

dealing with uncapacitated suppliers. This serves as a rationale for the observed

industry practice of firms expending resources to encourage lower cost suppliers to

increase capacity.

5.3.2 Diversification Benefit

In this section, we analyze our complete model which includes explicit benefits

derived from the size of the selected supplier pool. To start with assume that

suppliers are uncapacitated. In this case, our sourcing model is:

N N
Maximize Zqi,x (p s)p ciqir + s qi -
i= 1 i 1
-(p- s+ u)ES +d(X) (5.14)


subject to:

N
X = xi (5.15)
i=1
q, > 0 Vi (5.16)

x { = {0,1} Vi (5.17)


The optimal solution to this problem is characterized in Theorem 5.3 below.

Theorem 5.3: If the diversification benefits are positive and suppliers are uncapaci-

tated, then the optimal number of suppliers to source from is v*, where 1 < v* < n,

and is determined such that v* maximizes the diversification benefit function d(X).

The proof of this theorem is a direct extension of Theorem 5.1. Note that

based on Theorem 5.1, we choose to source the entire quantity from









the lowest cost supplier. If the diversification benefit function is maximized

when we choose v* suppliers where 1 < v* < n,6 then we order the total

requirements from the lowest cost supplier and simply include the others (i.e.,

suppliers 2,..., v*; indexed in order of increasing unit costs) in the supplier

selection set.

In this setting the diversification benefits drive the choice of the optimal num-

ber of suppliers. In particular, if it is relatively costless to source from additional

suppliers at a negligible level, then the firm can reap the v* selected supply pool

diversification benefits without incurring additional costs. When the marginal

benefits of sourcing from an additional supplier are positive, we can alv--x find an

order quantity that is sufficiently small enough to warrant selecting that supplier.

Qualitatively, the supplementary suppliers (2,..., v*) provide the firm with those

suppliers to consider for emergency supply.

For the capacitated supplier case, the structural insights into the optimal

sourcing strategy are characterized in the following theorem. The proof follows

directly from Theorem 5.3 and is omitted.

Theorem 5.4: If the suppliers are capacitated and diversification benefits are

positive, identify the number of suppliers v* where 1 < v* < n, such that v*

maximizes the diversification benefit function d(X). Using the results of Theorem

5.2, identify the optimal number of suppliers k* and optimal order quantities for

the capacitated suppliers problem without diversification benefits. If v* > k* > 1,

then allocate a minimal quantity c > 0 to suppliers k* + 1,... ,v* (assuming that

suppliers are indexed in increasing order of unit costs).



6 Obviously if v* = 1, then we would simply source the entire quantity from a
single lowest cost supplier









The implications of Theorem 5.3 and 5.4 for decision making concerning

an appropriate supplier base are clear. Recall from the introduction that there

are three interrelated decisions with regards to a firms sourcing strategy (Burke

and Vakharia, 2002) (a) criteria for establishing a supplier base; (b) criteria for

selecting suppliers (a subset of the base) who will receive an order from the firm;

and (c) the quantity of goods to order from each supplier selected. Theorem 5.3

offers direct managerial guidance concerning the second decision, or the number of

suppliers who will receive an order from the firm. Furthermore, as a consequence

of the positive diversification benefits, the total order quantity will slightly exceed

that of the original solution for the capacitated suppliers. Finally, there may

also be benefits beyond those directly captured by the model for ordering at

a negligible level from some of the suppliers. Depending upon the particular

contracts negotiated with these suppliers concerning upside order flexibility, the

manufacturing firm could potentially place larger orders with these suppliers in the

event that lower cost suppliers cannot deliver good units.

Note that if 1 < v* < k*, then the optimal v* that maximizes the diversi-

fication benefit function d(X) is actually smaller than the number of suppliers

needed to satisfy total requirements due to capacity limitations. In this case the

firm foregoes profit from product sales for a greater strategic benefit of a smaller

selected supplier pool. Analytically, the optimal strategy would be to source from

any number of suppliers m* such that v* < m* < k*. We now turn to describing an

extension to our modeling effort.

5.4 Model Extensions

In this section, we extend our model to examine the impact of including a

minimal order quantity when sourcing from a supplier. In certain situations, firms

may have to commit to ordering some minimal order quantity from each supplier

in order to reap the benefits of diversification. Suppose that we add an additional









constraint which reflects the minimum order quantity for each supplier as shown

below.

i > Zixi Vi (5.18)

While the complete solution algorithm for this model is fairly complicated, we

can easily obtain boundaries for the optimal number of suppliers. As compared

to Theorems 5.3 and 5.4 in the previous section, we now have non-negligible costs

associated with including additional suppliers in our supplier base. In particular,

the costs are associated with shifting enough units from a lower cost supplier to a

higher cost supplier to meet that higher cost supplier's minimum order quantity.

The intuition developed in Theorems 5.3 and 5.4 as to the appropriate number of

suppliers is still relevant. In particular, we would want to consider all candidate

solutions for the number of suppliers where the marginal diversification benefits are

non-negative. Therefore, its likely that the optimal number of suppliers will be less

than or equal to the number determined in Theorem 5.3.

While these diversification benefits may be difficult to quantify precisely, the

firm can analyze the marginal costs associated with a re-allocation strategy to

evaluate the diversification benefits. To illustrate, if the order quantity for the

n + Ith supplier is fairly low, and the cost per unit of the n + Ith supplier is only

slightly higher than for the nth supplier (i.e., when regret is more likely), then it

may be worthwhile to source from n + 1 suppliers. Recall the Toyota dilemma

discussed in the introduction where Toyota was forced to shut down an assembly

plant because of a problem with its sole supplier (Nishiguchi and Beaudet, 1998).

In this case, a secondary supplier hedges against the potential costs incurred

from problems associated with a single supplier strategy. The marginal costs of

including a secondary supplier at a minimum required level can be utilized as

a proxy for the "insui ,in premium necessary to reap the benefits of a larger

pool of suppliers. Likewise, the implicit risk premium for single sourcing can be









determined by comparing procurement costs for single versus multiple supplier

allocation strategies.

If we assume that the firm has already determined the subset of suppliers

that will receive orders, then Theorem 5.5 below specifies the structure of a simple

algorithm which can be used to determine optimal order quantities.

Theorem 5.5: When each supplier has both maximum and minimum limitations

placed on the size of the order, then the optimal quantity allocated to each supplier

can be determined as follows.

Step 1: Index all chosen suppliers in increasing order of cost per unit (i.e.,

C[1] < C[2] < C[3] ... < C[X]).
Step 2: For each supplier [i] (i 1,... ,X), determine Q[j] such that:

F(Qli) P-cl8+1
and based on this determine:

t[i] Q[i] E i 1 y[jr[j Ej i+l z[jr[j]
Step 3: The quantity allocated to supplier i is q[] = min{max{ t[i], :i }, 'l ['I }

and the total quantity ordered by the firm from all suppliers can be determined as
Yx
i= i q[ij"
Proof: See APPENDIX C.

From Theorem 5.5, we know that at most one of the chosen suppliers will

be unconstrained. Suppose that supplier [i] is unconstrained (i.e., q[i] = t[]).

Then, the optimal order quantity for the lower cost suppliers (j = 1,..., i 1)

is determined by the capacity constraint for each supplier. Similarly, the optimal

order quantity for the higher cost suppliers (j = i + 1,... X) is determined

by the minimum order quantity dictated by each supplier. Interestingly, the

total order quantity in this situation (ie. sum of all orders placed to the subset of

suppliers) is determined by the cost of the unconstrained supplier [i] and is such

that F(Q[) = p-c+
8-S+1









However, because the generalized structure of this problem has a wide range of

supplier options corresponding to alternate cost levels with differing combinations

of minimum and maximum order quantities, a simple solution algorithm cannot be

easily derived. This leads us to the problem of determining the optimal subset of

potential suppliers that will receive an order from the firm, in addition to allocating

appropriate order quantities. Branch and bound methodologies can be utilized

based on Theorem 5.5 to enumerate all possible subsets of suppliers for an optimal

solution. In addition to the optimal mathematical solution, firms may want to

consider different qualitative evaluation measures in determining an appropriate

subset of suppliers to source from. For example, in an international sourcing

context, firms may wish to pick a subset of suppliers in a variety of countries,

thereby hedging against country specific risks such as changing political climate

and/or exchange rates.

We now turn to an extensive numerical analysis in order to illustrate some of

our results and explore the sensitivity of these results for key parameters in our

analysis.

5.5 Numerical Analysis

Analytic results have been presented offering insights concerning the optimal

choice of suppliers and appropriate order quantities for a manufacturer. In this

section, we present the results of a numerical study to illustrate several key

cases for which the analytical insights cannot be obtained. We also examine the

sensitivity of our results based on changes in the key input parameters. Our

intention is to show an overview of these examples which offer insights concerning

the relative impact of these factors on a firm's sourcing strategy.

5.5.1 Experimental Design

The parameters and functions were chosen to capture the underlying assump-

tions outlined in Section 3. The explicit numerical parameters selected for the base









case example reflect those shown in Jucker and Rosenblatt (1985). For the manu-

facturer: (a) price/unit (p) = $ 19; (b) salvage value (s) = $ 2/unit; (c) lost sales

cost (u) = $ 6/unit; and (d) demand is assumed to be uniformly distributed with

parameters [300,700]. Our supplier base consists of 5 suppliers (i.e., i = 1,..., 5)

with identical reliabilities (ri = 0.9), minimum order quantities (z = 200 Vi), and

capacities (yi = 300 Vi). Suppliers are assumed to be heterogeneous with respect

to costs (i.e, ci) and these parameter settings are ci = 6.5; c2 = 7; c3 = 8; C4 = 9;

and s5 = 10. While the diversification benefit function is discrete, we assume that

it is roughly quadratic in the number of suppliers (X) and when X = 1,..., 5,

this function is defined as d(X) = d d2(d3 X)2 and when X = 0, d(0) = 0.

This functional form of d(X) was chosen since the single parameter d3 represents

the optimal number of suppliers which maximizes this function. All the results

discussed next were obtained using LINGO optimization software.

5.5.2 Results

Table 5.1 summarize the results of a set of numerical examples which show

sensitivity of the optimal supplier strategy to changes

in parameter values. Model A represents the case where both diversification

benefits and supplier minimum order quantities are included. For the remaining

examples in Table 5.1, the parameter changes are specified in the variable range

column. Not surprisingly, models B and C in Table 1 show that an increase in the

price or salvage value of the items increases the total quantity ordered, the total

number of suppliers sourced from and the total profit. Similarly, an increase in the

underage costs increases the total quantity ordered, the total number of suppliers,

but decreases the total profit earned.

Model E is intended to illustrate the impact of changes in the diversification

benefit function on the optimal sourcing strategy. In this model, the peak in the

magnitude of diversification benefits earned (i.e., di) is varied between --"0 and









$2000 (it is set to $1000 for the base case example). In response to an increase

in the peak value of the diversification function, the total order quantity and the

optimal number of suppliers remains the same, while the profit increases. This

would indicate that the optimal sourcing policy is fairly robust in that it is not

sensitive to large increases in the peak diversification value for this example.

Models F and G illustrate how the optimal sourcing policy changes with

alterations in the first supplier's cost and reliability. While small increases in

the first supplier's cost does not change the optimal number of suppliers, it does

decrease the total quantity ordered and the total profit. Similarly, an increase

in the first supplier's reliability decreases the total number of units ordered and

increases profit. In general, the firm simply compensates for small changes in

reliability by ordering proportionately more items since it does not p i for the bad

units.
Table 5-1: Sensitivity Analysis of the Key Parameters

Model Parameter Firm Profit n* Q* Nonzero Allocations
Range ($) qi q2 q3
A NA 6103.59 3 691 291 200 200
B p [10,25] [1737,9065] [2,3] [600,701] [300,300] [300,201] [0,200]
C s=[-6,6] [5375,6703] [2,3] [591,704] [300,300] [291,254] [0,200]
D u [0,12] [61'.I., .i1.,.] [3,3] [660,701] [260,300] [200,201] [200,200]
E di [250, 2000] [5354, 7103] [3,3] [691,691] [291,291] [200,200] [200,200]
F c = [6.25,6.75] [6171, .11 '.' ] [3,3] [667,658] [29(., -i.] [200,200] [200,200]
G ri [0.5,1] [6045,6104] [3,3] "II ','.] [300,262] [200,200] [200,200]


Next we show the impact of the minimum order quantity constraints on the

optimal sourcing strategy (see model extensions described in Section 5.5). In this

case, we choose three scenarios to illustrate our results. In all three scenarios,

the results are generated assuming capacitated suppliers. Table 5.2 contains the

results for these three scenarios in the following manner: Model A is the same as in

Table 5.1; model H represents the case where diversification benefits are included

without supplier minimum order quantities; and model I represents the case where

supplier minimum order quantities are incorporated in the absence of diversification