KanbanSchedulingSystemsTheChallengeofSimplicityWhatIsKanbanSchedulingKanbanschedulingsystemsareamongthemostsimple,effectiveandinexpensivemeansformanufacturingproductionandinventorycontrol.Theconceptisproven.FromNagoyatoWichitaFalls;fromWindsortoGeelong;frommicroelectronicstoheavysteel–Kanbanschedulingsystemsreduceinventory,eliminatestockouts,displacemassivecomputersandslashoverhead.Theyimprovebothserviceandquality.Sowhydoesn'teverymanufactureremploythismiracle?Inmanysituations,itisinappropriate––othermethodsworkbetter.EvenWhenKanbanisanexcellentchoice,firmsmayignoreit.Kanbanschedulingoftenevokesstrongemotionalresponsesfrombothproponentsanddetractorsandsetsavarietyoforganizationalphenomenaatworkagainstit.Kanbanschedulingsystemsoperatelikesupermarkets.Asmallstockofeveryitemsitsinadedicatedlocationwithafixedspaceallocation.Customerscometothestore.Theyvisuallyselectandpurchasetheiritems.Anelectronicsignalgoestothesupermarket'sregionalwarehousewhichdetailswhichitemshavesold.Thewarehousepreparesa(usually)dailyreplenishmentdeliveryoftheexactitemssold.Inlarge,modernsupermarketsKanbansignalscomefromcheckoutscanners.Theytravelelectronically(usuallyonceeachday)tothewarehouse.Somestoresstilluseavisualsystem.Here,aclerkwalkstheaislesdaily.Fromemptyspaceshededuceswhatsold.Theclerkthenordersreplacements.Thissignalmightgobytelephone,FAXorcourier.Oftenittravelsonthereturningdeliverytruck.Anothervariationisthebreadtruck.Heredriversfollowafixedroutefromstoretostore.Theyhaveasupplyofbakeryitemsintheirtruck.Ateachstop,theyexaminethestockandreplenishwhathasbeensold.AManufacturingExampleKanbanschedulinginmanufacturingworksinthesameway.Theessentialelementsofasystemare: Stockpoint(s) WithdrawalSignal ImmediateFeedback FrequentReplenishmentInthemanufacturingkanbansystemshownbelow,amachineshopsuppliescomponentpartstofinalassembly.Assemblyisamanualoperationwithlittlesetup.Assemblyproducesinlotsizesofoneaccordingtocustomerrequirements.Machiningismoreautomatedthanassemblyandhassignificantsetupcosts.Machiningmustproduceinbatchestoamortizethesetupandsequencepartstominimizetoolchanges. Asmallquantityofeverypartismaintainedatthemachineshop.Byobservingtheremainingquantities,themachinistsknowwhatproductsneedtobemadenext.KanbanandOtherMethodsProductionControlcoordinatesmulti-stepprocesses,oftenwithmultipleproducts.Kanbanisjustoneofseveralwaystoachievethiscoordination.PhysicalLinkingisanotherwaytocoordinate.Here,eachpartintheprocessmovesinsynchronizationandeachstepstartssimultaneously.Processesmusthavethesamelotsizeandco-location.InBroadcast,afinalassemblyoperationbuildsdirectlytoschedule.Thescheduleissimultaneously"Broadcast"toupstreamsubassemblyandsupplyoperations.Theybuildtheneededpartsin"Line-SetOrder"withasmalltimeoffsetfordelivery.Thissystemdoesnotrequireco-location.Itdoesrequireidenticallotsizes(usuallyone)forallprocesses.Kanbanschedulingsystemsareusefulwhenlotsizesdifferbetweenprocesssteps,processesareunbalancedorwhendistanceintroducestimelagorvariability.Thesesystemsslightlyde-coupletheprocesses.MaterialsRequirementsPlanning(MRP)worksfromBillsOfMaterial(BOM),routings,inventoryrecordsandforecasts.Itplanseachprocessstepforeachproduct,subassemblyanditem.Thesystemaccumulatesdemandforeachworkcenterandeachtimeperiod.MRPallowseffectiveschedulingunderthemostdifficultconditions.Thisabilitytoconnectadisjointedproductioncomeswithaprice.MRPpermitssomeformsofsloppyengineering.Theadministrativecostsarehigh.Throughputtimesarelongandinventoryturnslow.Errorsininventory,BOM'sorleadtimesdisruptthesystem.About50%-80%ofinstalledMRPsystemsdonotmeettheiruser'sneeds.Re-OrderPoint(ROP)systemsstoreeachitemandissuetodownstreamworkcentersonrequest.ROPsignalsaresupplywhentheinventoryisjustsufficienttocovertheresupplytime.ROPsystemsaresimple.Theyrequiresteadyandpredictablewithdrawalratesandpredictablereplenishmenttimes.But,theseconditionsarerare.Typicalsystemshaveveryhighinventoriesandexperiencefrequentstockouts.KanbanCartAHierarchyOfMethodologyThefigurebelowshowshowthemethodsformahierarchyofsimplicityandflexibility.Thebestsystemisthesimplest.Wheretheprocessallows,PhysicalLinkisthesystemofchoice.Broadcast,Kanban,MRPandROPfollowindesirability.Asystemdesignerwouldexamineeachprocessandeachproductgroupinturn.He/sheattemptstoapplyphysicallinkage.Ifthenecessaryconditionsdonotexistandprocesschangeisimpractical,PhysicalLinkageisrejectedandBroadcastisconsidered.Thisprocessfollowsdowntotheleastdesirablesystem,ROP.Inpractice,Kanbanschedulingsystemsareoftenagoodchoice.TheycanbeatransitionbetweenMRPandROPapproachesandPhysicalLinkage.DesigningTheKanbanSystemPreparingforaKanbanschedulingsystemcanbeformalwithelaborateanalysesandsimulations.Itcanalsobeveryinformalwithfine-tuningdoneontheproductionfloor.Wesuggestthefollowingsteps:1.AnalyzeProduct-VolumeForUpstreamWorkCenter2.AnalyzeDownstreamOrderPatterns3.IdentifyKanbanProducts4.IdentifyAppropriateLotSizes5.IdentifyContainers6.IdentifySignalMechanism7.SpecifyStockpoint(s)8.SpecifyInitialKanbanQuantities9.DevelopUpstreamSchedulingAlgorithm10.OperateFinetuneDesigningeachconnectionandproductusingtheabovemethodscouldbeveryelaborateandtimeconsumingiffollowedrigorously.Withexperience,however,mostdesignersordesignteamslearntoperformmoststepsmentallyandinformally.Inouronlineseminars,weinitiallystresstheformalapproachandinsistoncompletedocumentationofeachstep.Onceparticipantsunderstandtheprocess,theycanperformthestepsinformallyandveryquickly.DailyOperationsInthebestsystems,operatorsorteamsscheduletheirownwork.Theyhavecurrentandaccurateinformationofdownstreamproductionneeds.Theschedulingbucketisrarelylargerthanaday.Insomesystems,itmaybeonlyminutes.Operatorsexaminereturningkanbancardsorsignalstodeterminewherestockisloworhigh.Aidssuchasboardswithred,yellowandgreenzonescanassist.Knowingthemostfavorablesequenceforchangeovers,theoperatorfirstschedulesitemsintheredzones.Productsintheyellowandgreenzonesthenfollow.Operatorsmightalsohavealistofincomingorders.Withthis,theyidentifyanyunusuallylargeorder(s)thatwilloverwhelmthestock.HowManyKanban?Thenumberofcardsorcontainersisakeyissue.Excesscardsandinventoryencouragesloppyschedulingandalaissez-faireattitude.Insufficientcardsadverselyaffectscustomers.Severalmethodologiescandetermineanoptimumlevel: BoundaryAnalysis PredeterminedFormula FactorAnalysis Computersimulation Trial&ErrorInBoundaryAnalysis,keypeoplementallystepthroughsystemoperationusingProduct-Volumeandotherdata.Withsimplifiedmethodsandriskestimates,theyputupperandlowerlimitsonthestock. Predeterminedformulaeareavailable.Suchformulaeusuallycontain"guess"factorsandunavailabledata.Theyoftenapplytospecificsituationsandhavelimitedusefulness.FactorAnalysisidentifiesthevariousfactorsthataffectstocks.Examplesaresetupcosts,ordervolatility,qualityissues,stockoutsconsequencesandgrossvolume.ComputerSimulationusesspecializedsoftwaretobuildamodelofthesystem.Theanalystcanvaryparametersandarriveatdesiredlevels.Simulationisexpensivetodoproperly.Oversimplifiedsimulationsrarelyshowtrueoperation,theroleofhumanintelligenceoreffectsofcontinuousimprovement. WithTrialandError,yousetupthesystemandgo.Ifitoperatespoorly,addmorestock.Ifitoperatestoosmoothly,takesomeaway.AttheidealKanbanlevelthesystemoperates"neartheedge"andsometimesdemandsextraordinarymeasures. Inpractice,mostdesignersuseacombinationofBoundaryAnalysis,FactorAnalysisandTrial&Error.Sincemostsystemsareeasytomodify,thisworksquitewell.AKanbanSystemCaseStudyAkanbansystemschedulestheproductionofsixpeopleassemblingindustrialaircleanersfromsheetmetalandpurchasedparts.Theybuild15basicunitsandmanyvariations.Tenassemblycellshavefixtures,toolsandpartsreadyatalltimes.Eachcellproducesoneortwobasicmodels.Onetothreepeoplecanstaffanycell.Anadjacentwarehouseholdsasmallfinishedstockofeachstandardmodel.Chartsonthefollowingpagedisplaythedatathatwasthebasisforsystemdesign.TheProduct-VolumeAnalysisshowseachbasicmodelandtheaveragenumberofunitssoldeachday.Thesearelong-termaveragesanddonotreflectthedailyorderflow.Threemodelsrepresentthebulkofdemand.Fiveadditionalmodelshavemediumdemandlevels.Sevenmodelshaveextremelylowdemand.Thedesignersdecidedtohaveastockpointinthewarehousewhereorderscouldbequicklypulledandshipped.Onlythehighandmediumdemanditemswouldhaveakanbanstock.Whenaunitwaswithdrawnfromthekanbanstock,acardwouldgobacktotheproductionareatosignalthatwithdrawal.Ifalow-volumeitemappearedonanorder,aspecialone-timekanbancardwouldimmediatelygotoproductionwheretheitemwouldbebuiltbeforeday’send.TheDailyOrderProfileshowsthetotalnumberofunitsorderedeachdayforatwo-monthperiod.Itappearsthatonmostdays,thedemandisfairlyconstantatabout25units.However,veryheavydaysseemtopunctuatethissteadypatternatabout3-weekintervals.TheOrderProfileisanotherwaytopresentthissamedata.Itshowsonmostdaysordervolumeclustersaround20-25unitswithafewdaysofveryheavyvolume.Ifthekanbansystemwasdesignedtosatisfytheheavyorderdays,inventorieswouldbehuge.Therewassomedoubtwithinthedesignteamaboutthefeasibilityofkanban.Weirdorderpatternslikethisshouldand,inthiscasedid,raisequestions: Whythesuddenincreases? Whyarethesomewhatregular? Wheredoesthisvolumeoriginate?Uponinvestigation,thedesignersfoundthatmostorderscamefromdomesticdistributors.ThesedistributorsorderedinverysmallquantitiesbutthereweremanydistributorsaroundtheU.S.Thecumulativedailyvolumefromthesemanysmallorderstendedtobesteadyatabout20-25units.Thelargespikesresultedfromasingle,overseasdistributorthatorderedlargequantitiesforoceanshipment.Thiswasadifferentmarketanddifferentdistributionchannel.Fastdeliveryoftheselargeorderswasnecessaryforthedomesticmarketbutnotfortheoverseasdistributor.Itwasdecidedtodesignthekanbansystemforthedomesticordersandmakespecialarrangementsfortheoverseasdistibutor.Whenanoverseasorderarrived,itwouldbemixedwiththedomesticproductionoveraperiodofseveralweeks.Thisapproachsmoothedproductionandreducedtheneedforkanbanstock.KanbanStockpointDailyOperationsTheTeamLeaderscansincomingorders.Hepreparesone-timecardsforlargeordersandcustomizeditems.TheLeadersortscardscomingfromthewarehouse.Allcardsthengoonaboardarrangedbyassemblycell.Cellswithcardsintheredzonehavepriority.Ifnecessary,additionalpeopleworkanoverloadedcell.Thewarehousepicksstandardordersfromstockandsendscardstoproduction.Theycombinestandarditemswithanycustomizeditemsarrivingfromproductionandshiptheorders.Inasecondphaseofthisproject,sheetmetalandweldingoperationsmoveddirectlyadjacenttotheassemblycells.Theyhavededicatedpeopleandequipment.TheweldingTeamLeaderexamineseachassemblycellforstocksofweldedcabinets.HealsocheckstheBoard.Thisdailycheckingconstitutesthesignalforreplacement.Operatorsweldthenecessaryreplacementcabinetsandplacethemonapaintline.Thisreplenishmentisnormally24hours.Itmaybeaslittleasfourhours.Theweldingdepartmentstockssheetmetalcomponentsinlargewiretainers.Eachwiretainerhasspecialshelvesandbrackets.Itholdsafixednumberofeachitemonaparticularcabinet.AminimumquantitysignalsoperatorsortheTeamLeadertosendthebaskettoSheetMetalforreplenishment.ThesheetmetalTeamsetsupandbuildscomponentstoreplenishthebasketandreturnsittotheWelding.Thisnormallyoccurswithin24hours.Highervolumecabinetsmayhaveseveralidenticalbasketstomaintainweldingproductionduringreplenishment.ThiscompletesystemusesKanban,DirectLinkandRe-OrderPoint.ABroadcastsystemoverlaystheothersystemssinceallteamleadershaveaccesstothefinalassemblyKanbanboard.Thiskanbansystemeliminated96%offinishedgoodsinventory,simplifiedschedulingandeliminatedlossesfromobsolescentproduct.Formoredetailedinformationontheleanmanufacturing,pleasevisitathttp:www.mqip.com