The Mexico‐China Sourcing Game: Teaching Global Dual Sourcing · 2010-03-03 · planning horizon....

Preview:

Citation preview

Page | 1

TheMexico‐ChinaSourcingGame:TeachingGlobalDualSourcing1

GadAllon&JanA.VanMieghemKelloggSchoolofManagement

NorthwesternUniversity2001SheridanRoad,Evanston,IL,60208

g‐allon@kellogg.northwestern.edu,vanmieghem@kellogg.northwestern.edu

Jul20,2009;RevisedNov9,2009;Jan18,2010;Jan252010;Feb10,2010Wedescribeathree‐hourclassonglobaldualsourcingbuiltaroundagamethatdemonstratesthechallengesinmakingoperationaldecisionsandtransfersrecentacademicinsightstotheclassroom.Studentteamsmanageafirmwithaccesstoaresponsivebutexpensivesupplysource(Mexico)andacheapbutremotesource(China).Eachteammustdetermineasourcingstrategytosatisfyrandomdemandthatisrevealedthroughoutthegame.Ineachperiod,teamsplaceorderstobothsourcesandmanagetwoassets:inventoryandtheirbankaccount.Thegoalistomaximizeeachteam’svalue(finalbankbalance).Duringthedebriefings,weanalyzethepoliciesusedbydifferentteamsalongbothfinancialandoperationalmetrics,presenttheoptimalstrategy,andsummarizetheexperientiallearningpoints.Keywords:dualsourcing,strategicsourcing,experientiallearning,inventorymanagement,totallandedcost,simulationgame.

1.IntroductionInmanyretailsettings,andbusinessingeneral,firmscansourcegoodsfrommultiplesuppliers.Weinvestigateatypicalchoicebetweentwosuppliers:Oneislowcostbuthaslongleadtimeswhiletheotherprovidesquickresponsebutathighercost.Thisclassicalproblemisfacedbymanycompaniesandmoststudentsunderstanditsrelevance.However,effectivemanagementofdualsourcingissurprisinglychallengingandconveyingitscomplexitythroughtraditionalmeanssuchasacasestudyoralectureisdifficult:Solvingadualsourcingcaserequiresanalytictoolsthatarenotreadilyavailabletostudentsorinstructors.Alectureondualsourcingrunstheriskthatstudentsmaynotappreciatetheaddedlevelofcomplexityinday‐to‐day,aswellasstrategic,dualsourcingrelativetosinglesourcing.Wehavedevelopedasourcinggameasanexperientiallearningtoolthataddressestheaboveshortcomings.Inthisgame,studentsplaytheroleofsourcingmanagerswhomustmakestrategicallocationdecisionsaswellasplaceday‐to‐dayorderstotwosuppliers.Duringthisprocess,theyexperiencetheoperational,financialandservicerelatedconsequencesoftheirdecisions.

1 Submitted to INFORMS Transactions on Education’s special issue on Teaching Service and Retail Operations Management

Page | 2

Thesourcinggameisdesignedwithfourpedagogicalobjectivesinmind:1. Todevelopintuitiononpossiblesourcingstrategiesandtoappreciate

simple,robustpoliciestoguidesourcingdecisions.2. Tohighlighttheroleofworkingcapitalestimationintheconceptoftotal

landedcost.3. Toappreciatetheaddedcomplexityofmanagingasupplyportfolioover

singlesourcing.4. Tounderstandthefutilityofguessingandover‐reactingtodemand.

ThegamehasbeenusedwithMBAandexecutiveeducationstudentsandalthoughsomeknowledgeofbasicinventorymodelsisuseful,itisnotnecessary.Intheremainder,wewilldescribesequentiallythethree‐hourclasswehavetaughtinthreeparts:PartAintroducesglobalsourcingandrelatedtheory;PartBdescribesthegame;andPartCsummarizesthedebriefingphaseaswellasadiscussionofrecentacademicresearch.Weconcludethispaperbysummarizingstudentreactionsandpossibleextensions.

2.ClassPartA:IntroductiontoGlobalSourcing

Thissectiondescribesthefirstpartoftheclassthatintroducesglobalsourcingandrelatedtheory(about30to45minutesofclasstime).

2.1.TheMotivatingCaseAspreparationfortheclass,thestudentsareaskedtoreadtheMexico‐Chinamini‐caseinVanMieghem(2008,pp.230‐232).Thecasedescribesa$10billionhigh‐techU.S.manufacturerofwirelesstransmissioncomponents.Thecompanywasatacrossroadsregardingitsglobalnetwork;thecasefocusesonitstwoassemblyplants,oneinChinaandtheotherinMexico.WhiletheChinesefacilityenjoyedlowercosts,oceantransportationmadeitsorderlead‐times5to10timeslongerthanthosefromMexico.Withhighlyuncertainproductdemand‐‐coefficientsofvariationsofmonthlydemandforsomeproductswereashighas1.25‐‐solesourcingwasunattractive:MexicowastooexpensiveandChinatoounresponsive.Thefirmhadtodecidehowitcouldbestutilizethesetwosources.Atthestrategiclevel,thisamountstoproperlyallocatingproductdemandtoeachsource.Strategicallocationreferstotheexpectedcumulativedemandallocatedtoeachsourceovertheplanninghorizon.Atthetacticallevel,thefirmhadtochooseadynamicorderingpolicythatimplementsthatstrategicallocationatlowestcost.Inpractice,specifyingstrategicallocationsandorderingpoliciesarekeytasksofanysourcingstrategy‐‐beitglobalordomestic‐‐becauseitaffectscostsandsuppliermanagement.

Page | 3

2.2.Discussiononpractice:Toconnectthegametorealsourcingpractices,andifstudentshavepriorexperiencewithsimilarsettings,westartwithanopendiscussioncoveringtwoquestions:1. Whatpolicieshaveyouusedinpractice?Thisdiscussiontypicallyrevealssomeofthe

followingpractices:a. Allocateproductstoplantsbasedsolelyuponhistoricallocations(e.g.,new

productswithinafamilyareallocatedtoplantsintheexactsamefashionasexistingproductsinthatfamilyregardlessofsupplyanddemandcharacteristics).

b. Allocateproductstoplantsbaseduponinternalpolitics(e.g.,“pet”productionfacilitiesgethighervolume,lowercomplexityproductstokeeputilizationhighandcoststructurelow).

c. Allocateproductstoplantsbaseduponbasicunderstandingofcosts(e.g.,China’scostbaseislower–thussourcefullrequirementsoffshore).

d. Productallocationsfollowasimpleprimary‐secondaryallocation(e.g.,ifprimarylocationcannotfulfilldemand–requirementscascadetothesecondarylocation).

e. Product‐to‐plantallocationdecisionsarerarelyre‐visitedoradjusted(e.g.,allocationdecisionisa“lifesentence”).

2. Whatdata/metricsareusedtomakesourcingdecisions?Thisdiscussionfollowsthepracticesaboveandhighlightsthetypicalcost,leadtimes,andserviceriskmetrics.Theinstructorthancanaskthenaturalquestion:“Howcanwecombinethesevariouscomponentsintoasinglemetric?”Thisbringsustotheconceptoftotallandedcost.

2.3.TheConceptofTotalLandedCost(TLC)Toaddressthemotivatingcase,thenaturalfirststepistocomparethetwoextremesolutions:single‐sourcingfromMexicoversussingle‐sourcingfromChina.Thisrequiresaccountingforthedifferenceincost‐of‐good‐sold,shippingcosts,duties,andworking‐capitalrequirements2.Forthatpurposewereviewtheconceptoftotallandedcost(TLC),whichrepresentstheend‐to‐endcosttotransforminputsatthesourcetooutputsatdestination.TheTLCcapturesnotonlythetraditionalcostofgoodssold,butalsoaccountsforsupplychaincostssuchastransportation,customs,duties,taxes,aswellasrequiredworkingcapitalcarryingcoststoachievedesiredservicelevelsandprotectagainstsupplyanddemandrisks;seeFig1.

2 It should be noted that working capital includes the pipeline inventory as well as the safety stock. The pipeline inventory depends on the transformation process, which takes place between the moment of ordering and the moment of receipt of the order. If part of the lead-time is caused by inflexibility in production due to set-up times, then there is no per-se investment in material during the lead-time period. Thus, in order to accurately assess the working capital requirements one needs to know the precise timing of material inflow to build a sequence of cash flows over time.

Page | 4

Wewillrefertoallbutworkingcapitalcostcomponentsasthe“sourcingcost.”Computingthesourcingcostistediousyetstraightforward.Incontrast,workingcapitalisdrivenbypipelineandsafetyinventory,whichdependonlead‐times,volatilityandservicelevels.

Figure1:TotalLandedCost(TLC)isthetotalcostincurredfromsourcetodestination.

2.4.ComputingTLCforsinglesourcingWorkingcapitalandinventoryrequirementsareeasilyestimatedforsinglesourcingusingreadilyavailablestandardinventoryformulae.Beforemovingtodualsourcingandtothegame,weremindthestudentsoftheconceptofsafetystockandcomputethetotallandedcostforeachofthelocationsseparately.ThiscomputationrevealsthatsinglesourcingfromChinaappearstobethefavoritesolution.Askingthestudents’reactiontothisanalysisgeneratesadiscussionofthemeritsofsourcingfromMexicoandoftheexpectedbenefitsofdualsourcinginaquesttocombinethestrengthsofbothsources.Thisalsoleadstoadiscussionontheriskofusingasinglesource.Weusuallydiscusstwotypesofsupplyrisks:stochasticleadtimesandsupplydisruptions(suchasnaturaldisastersstrikes.)Thisnaturallyleadstothequestionofhowtodesignaneffectivedualsourcingpolicy,whichisthetransitionpointtothegame.

3.ClassPartB:TheMexicoChinaSourcingGameThissectioncoversthemainpartoftheclassbydescribingthedifferentstagesofthegame(about1to1½hourofclasstime).

3.1.Gamesetup(mostlydoneinadvanceofclass)

Theideabehindthegameisforstudentstoactassourcingmanagersofafirmmakingperiodicorderingdecisionsforanewproduct.Demandishighlyuncertainandaprobabilisticforecastisprovided.

Page | 5

Setting:Aclasswithasfewas5orasmanyas60studentsdividedinto5to10groups.Props:Eachstudentgroupneeds(atleast)onelaptopcomputer.Eachgroupreceivesa

log‐incodefortheweb‐sitehostingthegame3.TheinstructorneedsalaptopthatisconnectedtotheInternetandtoaclassroomscreenprojector.

Setup:SimilartotheMexico‐Chinamini‐case,eachteamrepresentsanidentical

companythatisintroducinganewproductthatcanbesourcedfromChinaorMexico.Eachteammanagestwoassets:cashandinventory.

Thedemanddistributionfortheproductisknownandshowntoallgroupsinadvance.(ThespecificdistributionthatwehaveusedisGammawithmean10andstandarddeviation15,reflectingahighlyvolatiledemand.)Theactualdemandrealizationisidenticalforallgroupsandprojectedontotheclassroomscreendynamicallyovertime;seeFig2.

EachteamcanplaceanordertoMexicoandanordertoChinaineachperiod;sourcesdifferincostandinleadtimes.Specifically,ordersplacedtoMexicohaveasourcingcostof$8,000perunitandarriveatthebeginningofthesubsequentperiod.OrdersplacedtoChinahaveasourcingcostof$7,000perunitbutarriveonlyafter4timeperiods.Theproductislaunchedinthe4thperiod,allowingthegroupstobuildpipelineinventoryfrombothsourcesbeforesalesstart.Theproductsellsfor$10,000perunit.Thestartingversionofthegameassumesthatanyunitdemandedbutnotavailableon‐handisbackloggedforacostof$20thousandperunitandeachteamplaysinaseparateyetidenticalmarket.

Allteamsstartwithzeromoneyinthebankandcanborrowtofinanceinventory.Anybankbalance(bothnegativeandpositive)incursa1%interestrateperperiod.Allteambankbalancesareprojectedontheclassroomscreencontinuously(whichallowsteamstocomparetheirpositiontootherteamsandcreatesasenseofcompetitionandexcitement).Anyleftoverinventoryattheendofthegameisliquidatedatzerovalue(althoughtheinstructorcaneasilychangethesalvagevalue).

Theobjectivefortheteamsistomaximizetheirfirmvalue,whichistheirbankbalance,attheendofthegame.Theterminationtimeofthegameisdeterminedbytheinstructorbutisnottoldtostudentsinadvancetopreventend‐of‐horizoneffects).

3 For access to the game, please contact the authors.

Page | 6

Figure2:Duringthegame,thetimeaswellascorrespondingdemandandbankbalancesareprojectedonthe

classroomscreen.

3.2.Kick‐offandBrainstorming(between15to30minofclasstime).

Thegoalofthisstageistoalloweachgrouptodevelopasourcingstrategy.EachgroupisgivenanExcelspreadsheetthatallowsstudentstosimulateandtesttheirstrategies.Specifically,thespreadsheetshowshowthepipelineinventory,sales,andbankbalanceevolvegivenuser‐enteredordersanddemand.Groupsareaskedtothinkthroughhowtheywouldreact(intermsofplacingorders)givenanewdemandrealizationandacertaininventorystatus.Tosteertheirthinking,eachgroupisalsoaskedtoestimatethefractionoforderstheywillplacetoChinaandtoMexicoandhandthisintotheinstructor.

3.3.Playingthedualsourcinggame(between45to60minofclasstime)

Thegamestartswithafourperiod“pre‐launchstage”duringwhichstudentscanfilltheirpipelinebeforesalesstart.Demandandsalesstartfollowingthepre‐launchstage.Duringeachperiod,allgroupsareinformedsimultaneouslyofthesameperioddemandasshowninFig.2.(Showing all students the same demand streamcreatesapositiveclassatmosphere,manifestedbystudentsshouting“YES!”or“OhJeez”etc.,whereallstudentstogetherexperiencethegame.)Ineachperiod,thetypicalinventoryactionstakeplace:(i) Teamsobservetheperioddemand(whichiszeroduringpre‐launch);(ii) Salesaredeterminedautomaticallyastheminimumofperioddemandandon‐

handinventory(seeFig.3forascreenshotoftheteam’swebbrowserinterface);(iii) Previousordersarereceivedautomatically;(iv) Teamsplaceorderstothetwosources;(v) Eachteam’spipelineinventorystatusanditsbankbalanceareupdated

Page | 7

automaticallyforthenextperiod.Theinstructorterminatesthegameatatimeofhisorherchoosing.(Wehavetypicallyplayed30to35periods,forabout50‐60minutes.)Itiscriticalnottoinformstudentsoftheterminationperiodinadvancetopreventend‐of‐horizoneffects.

Figure3:Screenshotofateam'swebbrowserinterfacewherestudentscanplaceordersandtracktheir

performance.

4.ClassPartC:DebriefandConnectiontoRecentResearchThissectiondescribesthelastpartoftheclassduringwhichwedebriefthegameinsights,discussrecentacademicresearch,andsummarizetake‐aways(about30minofclasstime).

4.1.Debriefingthegame

Afterendingthegame,theinstructorasksdifferentgroups(includingthe“winning”team,the“runner‐up,”andalowperformer)fortheirsourcingstrategy.Teamsareencouragedtosharetheirthoughtsandexperiencesonthebenefitsoftheirstrategyandwhattheywoulddodifferentlynexttime.Duringthatdiscussion,everyargumentisvalidatedagainstmetricscompiledduringthegame.Theinstructorshowsa“dashboard”(Fig.4)thatsummarizestheperformanceofeachteamalongfinancialandoperationalmetrics:(i) Financialmetricsincludethevalueofeachteam’sfirm(bankvaluesattheendof

thegame),costs(totalnumberofunitssourced),andrevenues(thecumulativenumberofunitssold),and

Page | 8

(ii) Operationalmetricsincludeeachteam’sactualstrategicallocation(fractionoftotalunitssourcedfromeachlocation),averageon‐handinventory,andservicelevelmeasuredbythefillrate(totalnumberofunitssolddividedbythetotaldemand).

Wearealsoabletoshowtheactualordersplacedovertimefromeachsourceforeachgroup.

Figure4:Adashboardthatsummarizesfinancial(toprow)andoperational(bottomrow)metricsofallteamsis

projectedontheclassroomscreenduringthedebriefingofthegame.

Thediscussionleadstotheobservationthatdifferentteamsmayperformequallywellfinanciallywhileemployingdifferentstrategies.Thegroupsusuallydifferwithrespecttoaverageon‐handinventoryaswellasthefractionofunitssourcedtoChinavs.Mexico.Thecommonthreadforall“successful”groupsistheabilitytoexecutethestrategyandnotchangeitinthefaceoflower(orhigher)thanexpecteddemandorinventories.Thisnaturallyleadstothefollowingquestions:Whatarethefeaturesofagooddualsourcingpolicy?Whatguidelinescanacademictheoryprovide?Answeringthesequestionsbringsustothenexttopic.

4.2.RecentAcademicResearch:TBSanddualindexpolicies

Theobjectivesofthegameincludeansweringthefollowingtwoquestions:(a)whatfeaturesdoeffectivedualsourcingpoliciesshare?And(b)howcanwedeterminenear‐optimalpoliciesandaveragesourcingallocation?Effectivedualsourcingpoliciestakeintoaccountnotonlytheon‐handinventoryandthecurrentdemand,butalsotheentirepipelinestatusaswellastheentiredemandforecast.Researchhasshownthatthestructureoftheoptimalpolicyisverycomplex.Therefore,thedualsourcingliteraturehastraditionallyfocusedondetermining

Page | 9

sophisticateddynamicpoliciesthatapproachoptimalperformance.Typically,thesepoliciesarecharacterizedbyoneortwotargetinventorylevels(base‐stocklevels)andkeeptrackofoneortwoinventorypositions(indices).Forexample,VeeraraghavanandScheller‐Wolf(2006)showthatadual‐indexpolicywithtwotargetlevelsperformsclosetooptimally,whichrepresentsstate‐of‐the‐artdualsourcingresearch.Unfortunately,determiningthetargetlevelsofsophisticatedpoliciessuchasthedualindexpolicyrequiressophisticatedcomputationalwork(optimizationviasimulation).Inaddition,underthesepolicies,theassociatedstrategicsourcingallocationcanbedeterminedonlythroughsimulation.Forastrategicsourcingmanager,itwouldbedesirabletohaveasimplepolicywithsimpleguidelinesthatdeterminethestrategicallocationduringtheplanningphasewithoutsignificantlycompromisingperformance.Recently,AllonandVanMieghem(2009)havepresentedsuchapolicythatisusedinpractice:thetailoredbase‐surge(TBS)policy.Underthispolicy,aconstant(“standing”)orderisplacedtothelowcostsupplierineachperiod.Theresponsivesourceisusedonlytobringthetotalinventory(on‐hand+pipeline)toasingletargetlevel.TheTBSpolicythusechoesafundamentaltenetinstrategy:italignstheorderingpatternswiththecorecompetenciesofthesuppliers(Fig.5).TheconstantbaseallocationallowsChinatofocusoncostefficiencywhileMexico'squickresponseisutilizedonlydynamicallytoguaranteehighservice.

Figure5:ATailoredBase‐SurgePolicyordersaconstantfractionfromthelowcost

sourceandordersfromtheresponsivesourcetorespondtosurges.

Besidesitssimplicity,theTBSpolicydirectlydeterminesthestrategicallocation,whichequalsthestandingorderdividedbytheaveragedemand.JanssenandDeKok(1999)studyasimilarpolicytotheTBSandshowusingnumericalstudythatafirmshouldallocatecloseto90%ofitsdemandtothecheapersource.AllonandVanMieghem(2009)optimizedtheTBSpolicyanalyticallyandpresentasimplesquare‐rootformulatospecifythenear‐optimalstrategicallocation:

Page | 10

,

wherehistheunitholdingcost,Δcistheunitsourcingcostdifferential,λisthedemandrate,andσisthesupply‐demandvolatility.4Simpleformulaealsoexistforthecorrespondingtargetinventorylevelandcost.Whenfocusingonlyondemandvolatility,theformulasimplifiesfurtherto:

whereCOVDisthecoefficientofvariationofthesingle‐perioddemanddistribution.Theformulaalsoprovidesinsightbyidentifyingthekeydriversofdualsourcingandquantifyingtheirinteraction.Specifically,theallocationtothelowcostsupplierishighwhen:1. Thekeymonetarytrade‐offΔc/hislarge,whichimpliesalargecostadvantageora

smallcostofcapital(smallcommercialrisk)2. Theexpecteddemandrateλishigh3. Thesupply‐demandvolatilityσissmall.Thisariseswithstableorhigh‐volume

products(inthematurityphaseoftheproductlifecycle)andrequiresstable(level)productioninChina.Thisfactoralsoshowshowthesourcingallocationshouldchangeasaproductmovesthroughitsproductlifecycle.

Studentsareaskedtocomputethestrategicallocationusingthesquare‐rootformulafortheparametersofthegame,whichiseasy:

• h=about$7,000*1%/period=$70/period• Δc=$1,000• COVD=1.5

sothatthenear‐optimalbaseallocationisapproximately1–√(.07/2)*1.5=1‐.28=.72(Asahome‐workassignment,theinstructorcanaskforasimilarcomputationforeachofthedifferentSKUsinthemini‐case.).Intheclasswealsocomparetheperformanceofdifferentclassicalpoliciesforafinitehorizonproblem,similartothegame.Table1comparestheprofitsof4strategies:singlesourcingfromMexico,singlesourcingfromChina,thetailoredbase‐surgepolicy,andadualbase‐stockpolicy.Foreachpolicy,wehaveoptimizedviasimulationoverthepolicyparameters.Forsinglesourcingpolicieswecomputedtheoptimalbase‐stockpolicy.FortheTailoredBase‐Surgepolicyweusedtheformulagiveninclass,andforthedualbase‐stockpolicyweused,again,simulation‐basedoptimizationtocomputetheoptimalthresholdlevels.Foreachpolicywepresentboththevalueafirmwouldhaveobtainedusingthatspecificpolicyandthespecificdemandsamplepathplayedinthegame,aswellastheexpectedvaluecorrespondingtothedemanddistribution.Wealsocomparethesepolicieswiththecaseinwhichthefirmhasperfectdemandinformation.Withperfectdemandforesight,theoptimalpolicyissimple:orderthe

4 Specifically, σ2 is the sum of the squared coefficients of variation of the interdemand times and the China supply times.

Page | 11

exactdemandquantitiesfourperiodsearlyfromChina.Thecomparisonwithperfectinformationhighlightsthedramaticcostofuncertainty.Table 1

Strategy OptimalTargetInventoryLevels

ActualValue|gamedemand

ExpectedValue

SingleSourceMexico

23 $557 $422

SingleSourceChina 48 $423 $569

TailoredBaseSurge 18,standingordertoChina=5

$616 $547

DualBaseStock 10,45 $514 $586

Perfectinformation+SinglesourceChina

OrderexactdemandfourperiodsearlierfromChina

$2,321 $1,878

Asonemayexpect,theDualBaseStockpolicyoutperformstheotherpolicies,inexpectation.However,onthespecificsamplepathplayedinthegame,theTBSgeneratesahigherprofit.Itisimportanttonotethat,inexpectation,SingleSourcingfromChinainthisgamewouldresultinhigherprofitsthantheTBSpolicy.However,ouranalyticmodeloftheTBSdoesnotcapturedynamiccontrolfromChina,sinceitessentiallyassumestheleadtimesaretoolongtoallowforclosed‐loopcontrol.Therefore,asinglesourcingpolicythatdoesallowfordynamiccontrolmaydobetterthantheTBSpolicy.(Alsonotethatthesignificantdifferencesbetweentheexpectedvalueofapolicyandtheprofitunderthespecificrealizationofdemandareduetothehighvolatilityofdemand.)

5.SummaryandExtensions

Thissectiondescribeshowweconcludetheclassandsummarizesthecontributionsofthegameandpossibleextensions,aswellasstudentreactions.

Page | 12

5.1.Summarizingkeytake‐away’softhegameWeendtheclassbysummarizingthekeyinsightsfromthegameas:1. Dualsourcingisasurprisinglycomplexmanagementproblemwithpotentially

highrewards.Thefirststepindetermininganeffectivedualsourcingpolicyistoconsideritsentiresource‐to‐destinationcost.Whileitisstraightforward(yettedious)tocomputemostcomponentsofthistotallandedcost,theimpactoninventoryandthusworkingcapitalisimportantyetdifficulttoassessunderdualsourcing.

2. Effectivedualsourcingusessupplierswithverydifferentstrengths.Inthisgame,wecombinedalowcost(butslow)supplierwithafast(butexpensive)source.Effectivedualsourcingpoliciesarealsotailoredtothestrengthsofeachsource.Thetailoredbase‐surgepolicyhasastandingorderwiththelowcostsupplier(allowingthatsuppliertoleveltheworkloadandreducecostsfurther)andordersfromthefastsourceonlytoreacttodemandsurges.Suchapolicyallowsthefirmtoreducetotalcostbyplacingthemajorityoforderstothelowcostsupplierwhileguaranteeinghighservicelevelsbyplacingoccasional,smallordersfromtheresponsive,yetexpensivesupplier.

3. Whiledeterminingoptimaldualsourcingpoliciesiscomputationallycomplex(andrequiressimulation),itisvaluable—andoftensufficientinpractice‐‐tohavesomeguidingformulaetosupportdualsourcingdecisions.Thesquare‐rootformulaprovidesasimplestartingpointwhendeterminingstrategicsourcingallocations.

4. Successfuldualsourcingoftenrequireshavingastrategyandstickingtoit.Havingasuboptimalpolicyistypicallybetterthancontinuouslychangingthedecisionprocess.Forexample,astrategyshouldallowthemanagertoreacttohighdemandsbyincreasingorders;however,reducingthetargetinventorylevel(base‐stock)whenobservingafewlowdemandperiodsusuallyleadstolowperformance.

Thegamealsoprovidesabackgroundtoillustratethebenefitsthatcompanieshaveexperiencedaftertransitioningfromsoletodualsourcing,including:

• Significantairfreightreduction–bymovingtodualsourcingsomecompanieshaveseen70%+reductionsinairfreight.

• Fasterresponsetochangesindemand,asopposedtomissingdemand.Undersinglesourcingfromalowcostsource,bythetimeademandchangewasrecognized,itwastoolateintheseasontorespond.

• Improvedabilitytomanagenewproductintroductions.Whenproductsweresolesourcedfromanoffshorelowcostsupplier,6to9monthsofproductwereinthepipelinebeforethecompanyrecognizedpoorsell‐throughwhichleftthemwithsignificantobsoleteinventory.

Page | 13

5.2.PossibleCustomizationsandExtensionsoftheGameThegamecaneasilybeextendedalongseveraldimensions,inorderofsuggestedimportanceandeaseofimplementation:

• Customization:thegamesoftwarecurrentlyallowstheinstructortochangethenamesandlocationsofthesources.Forexample,thefastsourcecanbedomesticandtwolocationscouldbeFranceandPoland.Theinstructorcanalsomodifythedemandrealizationandanyotherparameters.

• Operationalextensions:Changingthebackloggingassumptioninthegameandallowingforlostsalesisastraightforwardmodificationtothegame.Soisaddingaphysicalholdingcosttothecurrentfinancialopportunitycostorafixedorderingcost.Inaddition,itiseasytomodifythesoftwareandallowforstochasticleadtimes(butwefoundthatthismakesthegamemuchhardertounderstandforstudentswithoutofferinganynewinsights).

• Financialextensions:Itisstraightforwardtohaveahigherinterestrateforborrowingthansaving.Onecanalsointroduceadefaultthreatbyaddingaconstraintonmaximalborrowing.

• Non‐stationaryparameters:Forexample,insteadofassumingi.i.d.demand,thedemanddistributioncouldreflectthedifferentstagesoftheproductlifecycleofintroduction,growth,maturity,anddecline.Thesamecanbedonewithprices,costs,exchangerates,andinterestrates.

• Competitivemarket:Forexample,anydemandunmetbyoneteamspillsovertotheotherteams.Thisaddsanotherlayerofcomplexity,evenforasimplesingle‐period,single‐sourcemodelwhereonecanchoosedifferentexcessdemandsplittingrules;seeLippmanandMcCardle(1997).

• Design‐marketing‐salesdecisions:Forexample,teamscanboostdemandbyspendingmoneyonmarketingcampaignsandsalesincentives.Thiswouldnecessarilyimplythatteamsnolongerobserveacommondemand.Anotherextensioncouldbetoincludeproductandsupplynetworkdesign,similartotheGlobalSupplyChainManagementSimulationofEnspireLearning.5Whilethatsimulationfocusesonthedesignoftheproductsandthesupplynetwork,itallowstheplayerstomodifythesourcingallocationinalimitedmanner.Incontrast,ourgameissimplerandfocusesonmanaginganexistingsupplynetworkinadynamicsetting,

• Multipleitemsormultiplestages:whilewedonotsuggestthis,onecouldextendtomultipleproductsortoamulti‐stagesupplychainalongthelinesofthefamousBeerGame.Inessence,thiswouldleadtoaBeerGamewithdualsourcing.

5 http://www.enspire.com/simulations/gscms. Also distributed by Harvard Business Publishing under Prod. #: 6107-HTM-ENG.

Page | 14

Thegamecanalsobeusedtoconductexperimentstotestresearchhypotheses.Forexample,teamsmaystickclosertotheiroriginalplanandpolicyiftheyexperiencethegameasynchronouslyorarenotinformedofotherteams’bankbalances.Whilethisshowshowthegamecanbeextended,westronglyadvisetoplaythegameinitscurrentsimpleformatbecauseitallowsstudentstomaptheirdecisionstosubsequentresultsandthusenhancesthelearningfromthegame.

5.3.InstructorExperiencesofarSofar,theauthorshaveplayedthegameinabout10businessschoolclasssettings:withfull‐timeMBAstudentsinanoperationsstrategyelective;withfacultymembersataresearchconference;andinseveralexecutiveeducationcourses.Nextwewillplayitwithourdoctoralstudents.Webelievethegamewouldalsoworkwellwithgraduate,aswellasadvancedundergraduate,studentsinengineering.Theresponsehasbeenuniformlypositive.Executivesappreciatetherealismofthegameandthecomplexityinherentindualsourcing.Afterplayingthegameusingonlyintuition,theclassenhancestheirperceivedvalueoftheacademicresearchandsimpleguidelineformula.Studentsoftensuggestthatthegameshouldbeplayedwithacrossfunctionalteam,representingtheirmarketing,sales,financial,andoperationalgroups.Suchexperiencewouldconveytheimportanceofinter‐functionalcoordinationandcollaborativeforecasting.Fromaninstructor’sperspective,thegameiseasytoexplain,hasminimalrequirements(itneedsonlyafewlaptopsconnectedtotheInternet),andissimpletosetup.Asingleinstructorcaneasilyrunthegamewithasfewas5studentsorasmanyas60studentswithoutneedingadditionalassistantsordiminishingtheeffectivenessofthegame.Thegamenotonlysuccessfullyachievesthepedagogicalobjectives;italsohighlightsthevalueofacademicresearchforarealisticandimportantbusinessproblem.

AcknowledgmentsWearegratefultoCortJacobiandRuchirNandaofDeloitteConsultingforongoingcollaborationandinsightfuldiscussionsondualsourcing.References

Allon,G.andJ.A.VanMieghem(2009).“GlobalDualSourcing:TailoredBaseSurgeAllocationtoNearandOffshoreProduction.”ToappearinManagementScience.Lippman,S.A.andK.F.McCardle(1997)TheCompetitiveNewsboy.OperationsResearch45(1)54‐65.

Page | 15

VanMieghem,J.A.(2008).OperationsStrategy:PrinciplesandPractice.DynamicIdeas,Belmont,MA.Veeraraghavan,S.andScheller‐Wolf,A.A.(2006)NoworLater:ASimplePolicyforEffectiveDualSourcinginCapacitatedSystemsOperationsResearch56,;850‐864.Janssen,F.anddeKok,T.(1999)“Atwo‐supplierinventorymodel”,InternationalJournalofProductionEconomics,59(1‐3);395‐‐403.

Recommended