航空公司收益管理模型pros公司首席科學(xué)家報告（英文版）課件

RevenueManagement

andDynamicPricing:

PartIE.AndrewBoydChiefScientistandSeniorVP,ScienceandResearchPROSRevenueManagementaboyd@RevenueManagement

andDynamicOutlineConceptExampleComponentsReal-TimeTransactionProcessingExtracting,Transforming,andLoadingDataForecastingOptimizationDecisionSupportNon-TraditionalApplicationsFurtherReadingandSpecialInterestGroupsOutlineConceptRevenueManagement

andDynamicPricingRevenueManagementinConceptRevenueManagement

andDynamicWhatisRevenueManagement?BeganintheairlineindustrySeatsonanaircraftdividedintodifferentproductsbasedondifferentrestrictions$1000Yclassproduct:canbepurchasedatanytime,norestrictions,fullyrefundable$200Qclassproduct:Requires3weekadvancedpurchase,Saturdaynightstay,penaltiesforchangingticketafterpurchaseQuestion:Howmuchinventorytomakeavailableineachclassateachpointinthesalescycle?WhatisRevenueManagement?BegWhatisRevenueManagement?RevenueManagement:ThescienceofmaximizingprofitsthroughmarketdemandforecastingandthemathematicaloptimizationofpricingandinventoryRelatednames:YieldManagement(original)RevenueOptimizationDemandManagementDemandChainManagementWhatisRevenueManagement?RevRudimentsStrategic/Tactical:MarketingMarketsegmentationProductdefinitionPricingframeworkDistributionstrategyOperational:RevenueManagementForecastingdemandbywillingness-to-payDynamicchangestopriceandavailableinventoryRudimentsStrategic/Tactical:IndustryPopularityWasbornofabusinessproblemandspeakstoabusinessproblemAddressestherevenuesideoftheequation,notthecostside2–10%revenueimprovementscommonIndustryPopularityWasbornofIndustryAccolades“Nowwecanbealotsmarter.Revenuemanagementisallofourprofit,andmore.” BillBrunger,VicePresidentContinentalAirlines“PROSproductshavebeenakeyfactorinSouthwest'sprofitperformance.”KeithTaylor,VicePresidentSouthwestAirlinesIndustryAccolades“NowwecanAnalystAccolades“RevenuePricingOptimizationrepresentthenextwaveofsoftwareascompaniesseektoleveragetheirERPandCRMsolutions.”–ScottPhillips,MerrillLynch“Oneofthemostexcitinginevitabilitiesaheadis‘yieldmanagement.’”–BobAustrian,BancofAmericaSecurities“RevenueOptimizationwillbecomeacompetitivestrategyinnearlyallindustries.”–AMRResearchAnalystAccolades“RevenuePricAcademicAccolades“Anareaofparticularinteresttooperationsresearchexpertstoday,accordingtoTrick,isrevenuemanagement.”InformationWeek,July12,2002.Dr.TrickisaProfessoratCMU

andPresidentofINFORMS.AcademicAccolades“AnareaofAcademicAccoladesAswemoveintoanewmillennium,dynamicpricinghasbecometherule.“Yieldmanagement,”saysMr.Varian,“iswhereit’sat.”“ToHalVarianthePriceisAlwaysRight,”strategy+business,Q12000.Dr.VarianisDeanoftheSchoolofInformationManagementandSystemsatUCBerkeley,andwasrecentlynamedoneofthe25mostinfluentialpeopleineBusinessbyBusinessWeek(May14,2001)AcademicAccoladesAswemoveiApplicationAreasTraditionalAirlineHotelExtendedStayHotelCarRentalRailTourOperatorsCargoCruiseNon-TraditionalEnergyBroadcastHealthcareManufacturingApparelRestaurantsGolfMore…ApplicationAreasTraditionalNoDynamicPricingThedistinctionbetweenrevenuemanagementanddynamicpricingisnotaltogetherclearArefareclassesdifferentproducts,ordifferentpricesforthesameproduct?RevenuemanagementtendstofocusoninventoryavailabilityratherthanpriceRealityisthatrevenuemanagementanddynamicpricingareinextricablylinkedDynamicPricingThedistinctionTraditionalRevenueManagementNon-traditionalrevenuemanagementanddynamicpricingapplicationareashavenotevolvedtothepointofstandardindustrypracticesTraditionalrevenuemanagementhas,andwefocusprimarilyontraditionalapplicationsinthispresentationTraditionalRevenueManagementRevenueManagement

andDynamicPricingManagingAirlineInventoryRevenueManagement

andDynamicAirlineInventoryAmid-sizecarriermighthave1000dailydepartureswithanaverageof200seatsperflightlegEWRSEALAXIAHATLORDAirlineInventoryAmid-sizecaAirlineInventory200seatsperflightleg200x1000=200,000seatspernetworkday365networkdaysmaintainedininventory365x200,000=73millionseatsininventoryatanygiventimeThemechanicsofmanagingfinalinventoryrepresentsachallengesimplyduetovolumeAirlineInventory200seatsperAirlineInventoryRevenuemanagementprovidesanalyticalcapabilitiesthatdriverevenuemaximizingdecisionsonwhatinventoryshouldbesoldandatwhatpriceForecastingtodeterminedemandanditswillingness-to-payEstablishinganoptimalmixoffareproductsAirlineInventoryRevenuemanagFareProductMixShoulda$1200SEA-IAH-ATLMclassitinerarybeavailable?A$2000Yclassitinerary?EWRSEALAXIAHATLORDFareProductMixShoulda$1200FareProductMixShoulda$600IAH-ATL-EWRBclassitinerarybeavailable?An$800Mclassitinerary?EWRSEALAXIAHATLORDFareProductMixShoulda$600FareProductMixOptimizationputsinplaceinventorycontrolsthatallowthehighestpayingcollectionofcustomerstobechosenWhenitmakeseconomicsense,fareclasseswillbeclosedsoastosaveroomforhigherpayingcustomersthatareyettocomeFareProductMixOptimizationpRevenueManagement

andDynamicPricingComponentsRevenueManagement

andDynamicTheReal-TimeTransactionProcessorRealTimeTransactionProcessor(RESSystem)RequestsforInventoryTheReal-TimeTransactionProcTheRevenueManagementSystemRevenueManagementSystemForecastingOptimizationExtract,Transform,andLoadTransactionDataRealTimeTransactionProcessor(RESSystem)RequestsforInventoryTheRevenueManagementSystemRAnalystsRevenueManagementSystemForecastingOptimizationExtract,Transform,andLoadTransactionDataRealTimeTransactionProcessor(RESSystem)RequestsforInventoryAnalystDecisionSupportAnalystsRevenueManagementSysTheRevenueManagementProcessRevenueManagementSystemForecastingOptimizationExtract,Transform,andLoadTransactionDataRealTimeTransactionProcessor(RESSystem)RequestsforInventoryAnalystDecisionSupportTheRevenueManagementProcessReal-TimeTransactionProcessorTheoptimizationparametersrequiredbythereal-timetransactionprocessorandsuppliedbytherevenuemanagementsystemconstitutetheinventory

control

mechanismReal-TimeTransactionProcessoReal-TimeTransactionProcessorDFWEWRYAvailMAvailBAvailQAvail11060200DFW-EWR:$1000Y$650M$450B$300QReal-TimeTransactionProcessoReal-TimeTransactionProcessorNestedleg/classavailabilityisthepredominantinventorycontrolmechanismintheairlineindustryDFWEWRYAvailMAvailBAvailQAvail11060200DFW-EWR:$1000Y$650M$450B$300QMClassBooking10959Real-TimeTransactionProcessoReal-TimeTransactionProcessorAfareclassmustbeopenonbothflightlegsifthefareclassistobeopenonthetwo-legitinerarySATDFWEWRYClassMClassBClassQClass501000YClassMClassBClassQClass11060200Real-TimeTransactionProcessoExtract,Transform,andLoadTransactionDataComplicationsVolumePerformancerequirementsNewproductsModifiedproductsPurchasemodificationsExtract,Transform,andLoadTExtract,Transform,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,Transform,andLoadTDemandModelsandForecastingHowshoulddemandbemodeledandforecast?Smallnumbers/levelofdetailUnobserveddemandandunconstrainingElementsofdemand:purchases,cancellations,noshows,goshowsDemandmodel…theprocessbywhichconsumersmakeproductdecisionsDemandcorrelationanddistributionalassumptionsSeasonalityDemandModelsandForecastingHDemandModelsandForecastingHolidaysandrecurringeventsSpecialeventsPromotionsandmajorpriceinitiativesCompetitiveactionsDemandModelsandForecastingHOptimizationOptimizationissuesConvertibleinventoryMovableinventory/capacitymodificationsOverbooking/oversaleofphysicalinventoryUpgrade/upwardsubstitutableinventoryProductmix/competitionforresources/networkeffectsOptimizationOptimizationissueDecisionSupportDecisionSupportRevenueManagement

andDynamicPricingNon-TraditionalApplicationsRevenueManagement

andDynamicTwoNon-TraditionalApplicationsBroadcastBusinessprocessessurroundingthepurchaseandfulfillmentofadvertisingtimerequiremodificationoftraditionalrevenuemanagementmodelsHealthcareBusinessprocessessurroundingpatientadmissionsrequirere-conceptualizationoftherevenuemanagementprocessTwoNon-TraditionalApplicatioNewAreasContractsandlongtermcommitmentsofinventoryCustomerlevelrevenuemanagementIntegratingsalesandinventorymanagementAlliancesandcooperativeagreementsNewAreasContractsandlongteRevenueManagement

andDynamicPricingFurtherReadingandSpecialInterestGroupsRevenueManagement

andDynamicFurtherReadingForanentrypointintotraditionalrevenuemanagementJefferyMcGillandGarrettvanRyzin,“RevenueManagement:ResearchOverviewandProspects,”TransportationScience,33(2),1999E.AndrewBoydandIoanaBilegan,“RevenueManagementande-Commerce,”underreview,2002FurtherReadingForanentrypoSpecialInterestGroupsSpecialInterestGroupsRevenueManagement

andDynamicPricing:

PartIIE.AndrewBoydChiefScientistandSeniorVP,ScienceandResearchPROSRevenueManagementaboyd@RevenueManagement

andDynamicOutlineSingleFlightLegLeg/ClassControlBidPriceControlNetwork(O&D)ControlControlMechanismsModelsOutlineSingleFlightLegRevenueManagement

andDynamicPricingSingleFlightLegRevenueManagement

andDynamicLeg/ClassControlDFWEWRYAvailMAvailBAvailQAvail11060200DFW-EWR:$1000Y$650M$450B$300QAtafixedpointintime,whataretheoptimalnestedinventoryavailabilitylimits?Leg/ClassControlDFWEWRYAvailAMathematicalModelGiven:FareforeachfareclassDistributionoftotaldemand-to-comebyclassDemandassumedindependentDetermine:OptimalnestedbookinglimitsNote:Cancellationstypicallytreatedthroughseparateoptimizationmodeltodetermineoverbooking

levelsAMathematicalModelGiven:AMathematicalModelWheninventoryispartitionedratherthannested,thesolutionissimplePartitioninventorysothattheexpectedmarginalrevenuegeneratedofthelastseatassignedtoeachfareclassisequal(forsufficientlyprofitablefareclasses)AMathematicalModelWheninvenAMathematicalModelNestedinventorymakestheproblemsignificantlymoredifficultduetothefactthatdemandforonefareclassimpactstheavailabilityforotherfareclassesTheproblemisill-posedwithoutmakingexplicitassumptionsaboutarrivalorderEarlymodelsassumedlow-before-highfareclassarrivalsAMathematicalModelNestedinvAMathematicalModelThereexistsasubstantialbodyofliteratureonmethodsforgeneratingoptimalnestedbookingclasslimitsMathematicsbasicallyconsistsofworkingthroughthedetailsofconditioningonthenumberofarrivalsinthelowervaluefareclassesAnheuristicknownasEMSRbthatmimicstheoptimalmethodshascometodominateinpracticeAMathematicalModelThereexisAnAlternativeModelThelow-before-higharrivalassumptionwasaddressedbyassumingdemandarrivesbyfareclassaccordingtoindependentstochasticprocesses(typicallynon-homogeneousPoisson)Sincemanypractitionersconceptualizedemandas

totaldemand-to-come,modelsbasedonstochasticprocessesfrequentlycauseconfusionAnAlternativeModelThelow-beALegDPFormulationWithPoissonarrivals,anaturalsolutionmethodologyisdynamicprogrammingStagespace:timepriortodepartureStatespacewithineachstage:numberofbookingsStatetransitionscorrespondtoeventssuchasarrivalsandcancellationsALegDPFormulationWithPoiss…TT-1T-2T-310nn+1n+2n+3…SeatsRemainingTimetoDepartureCancellationNoEvent/RejectedArrivalAcceptedArrival…………………TT-1T-2T-310nn+1n+2n+3…SeatsALegDPFormulationV(t,n):Expectedreturninstaget,staten

whenmakingoptimaldecisionsV(t,n)=maxu[p0(0+V(t-1,n)) Noevent

(1-p0)c(0+V(t-1,n-1))+ Cancel

(1-p0)(fi<u)i(0+V(t-1,n))

Arrival/Reject

(1-p0)(fiu)i(fi+V(t-1,n+1))] Arrival/Acceptu(t,n):Optimalpricepointformaking accept/rejectdecisionswheneventin

staget,statenisabookingrequestALegDPFormulationV(t,n):EALegDPFormulationDPhastheinterestingcharacteristicthatitcalculatesV(t,n)forall(t,n)pairsProvidesvaluableinformationfordecisionmakingPresentscomputationalchallengesThisnaturallysuggestsanalternativecontrolmechanismtonestedfareclassavailabilityBidpricecontrolALegDPFormulationDPhasthe882591639492847884768473200……8823916194908820915891878817200………nn+1n+2n+3SeatsRemainingTT-1T-2T-310TimetoDeparture………………8480V(t,n)=

ExpectedRevenue882591639492847884768473200……8882591639492…nn+1n+2n+3SeatsRemainingT…8480V(t,n)=

ExpectedRevenueV(t,n+1)–V(t,n)=

MarginalExpectedRevenue345338330…T…352882591639492…nn+1n+2n+3SeatsRnn+1n+2n+3SeatsRemainingBidPriceControl:Withn+1seatsremaining,acceptonlyarrivalswithfaresinexcessof345345338330…T…352nn+1n+2n+3SeatsRemainingBidPBidPriceControlLikenestedbookinglimits,thereexistsasubstantialliteratureondynamicprogrammingmethodsforbidpricecontrolWhilebidpricecontrolissimpleandmathematicallyoptimal(foritsmodelingassumptions),ithasnotyetbeenbroadlyacceptedintheairlineindustrySubstantialchangestotheunderlyingbusinessprocessesBidPriceControlLikenestedbBidPriceControlSolutionsfromdynamicprogrammingcanalsobeconvertedtonestedbookinglimits,butthistechniquehasnotbeenbroadlyadoptedinpracticeBidpricecontrolcanbeimplementedwithroughlythesamenumberofcontrolparameters(bidprices)asnestedfareclassavailabilityBidPriceControlSolutionsfroRevenueManagement

andDynamicPricingNetwork(O&D)ControlControlMechanismsRevenueManagement

andDynamicNetworkControlNetworkcontrolrecognizesthatpassengersflowonmultipleflightlegsAnissueofglobalversuslocaloptimizationProblemiscomplicatedformanyreasonsForecastsofmanysmallnumbersDataLegacybusinesspracticesNetworkControlNetworkcontrolInventoryControlMechanismTheinventorycontrolmechanismcanhaveasubstantialimpactonRevenueMarketinganddistributionChangestoRESsystemChangestocontractsanddistributionchannelsInventoryControlMechanismTheExample:

LimitationsofLeg/ClassControlSATDFWEWRSupply:1seatontheSAT-DFWleg1seatontheDFW-EWRlegDemand:1$300SAT-DFWYpassenger1$1200SAT-DFW-EWRYpassenger$1200Y$300YExample:

LimitationsofLeg/ClExample:

LimitationsofLeg/ClassControlOptimalleg/classavailabilityistoleaveoneseatavailableinYclassoneachlegSATDFWEWRYClassMClassBClassQClass1000YClassMClassBClassQClass1000Example:

LimitationsofLeg/ClExample:

LimitationsofLeg/ClassControlSATDFWEWR$1200Y$300YWithleg/classcontrol,thereisnowaytoclose

SAT-DFWYwhileleavingSAT-DFW-EWRYopenSupply:1seatontheSAT-DFWleg1seatontheDFW-EWRlegDemand:1$300SAT-DFWYpassenger1$1200SAT-DFW-EWRYpassengerExample:

LimitationsofLeg/ClLimitationsofLeg/ClassControlThelimitationsofleg/classavailabilityasacontrolmechanismlargelyeliminaterevenueimprovementsfromanythingmoresophisticatedthanleg/classoptimizationForthisreason,carriersthatadoptO&DcontrolalsoadoptanewinventorycontrolmechanismRequirestremendouseffortandexpensetoworkaroundthelegacyinventoryenvironmentLimitationsofLeg/ClassContrAlternativeControlMechanismsWhiletherearemanypotentialinventorycontrolmechanismsotherthanleg/classcontrol,twohavecometopredominateO&DrevenuemanagementapplicationsVirtualnestingBidpriceNotethattheconceptofitinerary/fareclass(ODIF)inventorylevelcontrolisimpracticalAlternativeControlMechanismsVirtualNestingAprimalcontrolmechanismsimilarinflavortoleg/classcontrolAsmallsetofvirtualinventorybucketsaredeterminedforeachlegNestedinventorylevelsareestablishedforeachbucketEachleginanODIFismappedtoaleginventorybucketandanODIFisavailableforsaleifinventoryisavailableineachlegbucketVirtualNestingAprimalcontroVirtualNestingSAT-DFW-EWRYmapstovirtualbucket3onlegSAT-DFWandvirtualbucket1onlegDFW-EWRTotalavailabilityof10forSAT-DFW-EWRYSATDFWEWRBucket1Bucket2Bucket3Bucket410060100Bucket1Bucket2Bucket3Bucket440000VirtualNestingSAT-DFW-EWRYmVirtualNestingSAT-DFWYmapstovirtualbucket4onlegSAT-DFWSAT-DFWYisclosedSATDFWEWRBucket1Bucket2Bucket3Bucket410060100Bucket1Bucket2Bucket3Bucket440000VirtualNestingSAT-DFWYmapsBidPriceControlAdualcontrolmechanismAbidpriceisestablishedforeachflightlegAnODIFisopenforsaleifthefareexceedsthesumofthebidpricesonthelegsthatareusedBidPriceControlAdualcontroBidPriceControlSATDFWEWR$1200YBidPrice=$400BidPrice=$600SAT-DFW-EWRYisopenforsalebecause

$1200$400+$600

BidPriceControlSATDFWEWR$120BidPriceControlSATDFWEWRBidPrice=$400BidPrice=$600$300YSAT-DFWYisclosedforsalebecause

$300<$400BidPriceControlSATDFWEWRBidBidPriceControlSATDFWEWRIntermediatecontrolbetweenoptimizationpointsisachievedbyhavingadifferentbidpriceforeach

seatsoldininventory654321$664$647$632$619$610$600SeatBidPrice654321$434$425$417$410$405$400SeatBidPriceBidPriceControlSATDFWEWRInteBidPriceControlSATDFWEWRAfteraseatissoldthebidpriceincreases,reflectingthereducedinventoryavailability654321$664$647$632$619$610$600SeatBidPrice654321$434$425$417$410$405$400SeatBidPriceBidPriceControlSATDFWEWRAfteVirtualNestingAdvantagesVerygoodrevenueperformanceComputationallytractableRelativelysmallnumberofcontrolparametersComprehensibletousersAcceptedindustrypracticeDisadvantagesNotdirectlyapplicabletomulti-dimensionalresourcedomainsProperoperationrequiresconstantremappingofODIFstovirtualbucketsVirtualNestingAdvantagesBidPriceControlAdvantagesExcellentrevenueperformanceComputationallytractableComprehensibletousersBroaderusethanrevenuemanagementapplicationsPlacesamonetaryvalueonunitinventoryDisadvantagesGrowinguseracceptance,buthasnotreached

thesamelevelasprimalmethodsBidPriceControlAdvantagesRevenueManagement

andDynamicPricingNetwork(O&D)ControlModelsRevenueManagement

andDynamicAModelThedemandallocationmodel(alsoknownasthedemand-to-comemodel)hasbeenproposedforuseinrevenuemanagementapplications,butistypicallynotemployedForallofitslimitations,thedemandallocationmodelbringstolightmanyoftheimportantissuesinrevenuemanagementAModelThedemandallocationmDemandAllocationModelMax

iIrixis.t. iI(e)xice eE (e)

xidi iI (i)

xi0

iI

I=setofODIFsE=setofflightlegsce=capacityofflightedi=demandforODIFiri=ODIFirevenueI(e)=ODIFsusingflightexi=demandallocatedtoODIFiDemandAllocationModelMax iLeg/ClassControlMax

iIrixis.t. iI(e)xice eE (e)

xidi iI (i)

xi0

iI

Thevariablesxicanberolleduptogenerateleg/classavailabilityLeg/ClassControlMax iIrVirtualNestingMax

iIrixis.t. iI(e)xice eE (e)

xidi iI (i)

xi0

iI

OnceODIFshavebeenassignedtolegbuckets,thevariablesxicanberolleduptogenerateleg/classavailabilityVirtualNestingMax iIriBidPriceControlMax

iIrixis.t. iI(e)xice eE (e)

xidi iI (i)

xi0

iI

ThedualvariableseassociatedwiththecapacityconstraintscanbeusedasbidpricesBidPriceControlMax iIrNetworkAlgorithms:

Leg/ClassControlNetworkalgorithmsforgeneratingnestedleg/classavailabilityarenottypicallyusedLimitationsofthecontrolmechanismandfarestructureeliminatemuchofthevalueNetworkAlgorithms:

Leg/ClassNetworkAlgorithms:

VirtualNestingControlOptimizationconsistsofdeterminingtheODIFtoleg/bucketmapping,andthencalculatingnestedleg/bucketinventorylevelsBestmappingsprorateODIFfarestolegs,andthengroupsimilarproratedfaresintothesamebucketThebestprorationmethodsdependondemandforecastsandrealizedbookings,andchangedynamicallythroughoutthebookingcycleWithODIFsmappedtobuckets,nestedbucketinventorylevelsarecalculatedusingthenestedleg/bucketalgorithmofchoiceNetworkAlgorithms:

VirtualNeNetworkAlgorithms:

BidPriceControlBidpricesarenormallygenerateddirectlyorindirectlyfromthedualsolutionofanetworkoptimizationmodelNetworkAlgorithms:

BidPriceResourceAllocationModelObservationsA200legnetworkmayhave10,000activeODIFs,leadingtoanetworkoptimizationproblemwith10,000columnsand10,200rowsWith20,000passengers,theaveragenumberofpassengersperODIFis2Typically,20%oftheODIFswillcarry80%ofthetraffic,withalargenumberofODIFscarryingontheorderof.01orfewerpassengersper

networkdayResourceAllocationModelObserResourceAllocationModelMax

iIrixis.t. iI(e)xice eE (e)

xidi iI (i)

xi0

iI

ManysmallnumbersResourceAllocationModelMax LevelofDetailProblemThelevelofdetailproblemremainsapracticalconsiderationwhensettingupanyrevenuemanagementsystemWhatlevelofdetaildotheexistingdatasourcessupport?Whatlevelofdetailprovidesthebestrevenueperformance?Atwhatpointdoesforecastnoiseovercomeimprovementsfrommoresophisticated

optimizationmodels?LevelofDetailProblemThelevLevelofDetailProblemAsarule,evenwiththemanysmallnumbersinvolved,networkoptimizationalgorithmsperformconsistentlybetterthannon-networkalgorithmsDualsolutionsaretypicallymuchmorerobustandofbetterqualitythansolutionsconstructedfromprimalODIFallocationsLevelofDetailProblemAsaruRevenueManagement

andDynamicPricingNetwork(O&D)ControlOptimizationChallengesRevenueManagement

andDynamicANetworkDPFormulationNetworkDPformulationStagespace:timepriortodepartureStatespacewithineachstage:multidimensional,withnumberofbookingsoneachofMflightsStatetransitionscorrespondtoeventssuchasODIFarrivalsandcancellationsANetworkDPFormulationNetworANetworkDPFormulationV(t,n1,…,nM):Expectedreturninstaget,state (n1,…,nM)whenmakingoptimaldecisionsu(t,n1,…,nM,k):Optimalpricepointformaking accept/rejectdecisionswheneventin

staget,state(n1,…,nM)isabookingrequest forODIFkANetworkDPFormulationV(t,n1ANetworkDPFormulationObservationsA200legnetworkwithanaverageof150seatsperflightlegwouldhave150200statesperstageWith10,000activeODIFs,assumingonlysinglepassengerarrivalsandcancellations,eachstatewouldhave~20,000possiblestatetransitionsGivesriseto~20,000“bidprices”perstateANetworkDPFormulationObservAnAlternativeViewofDPConsiderabookingrequestatt