SixSigma

ThecommonSixSigmasymbol

Partofaseriesofarticleson

Industry

Manufacturingmethods
BatchproductionJobproduction
Flowproduction
Improvementmethods
RCMTPMVDM
QRMTOC
LeanSixSigmaTQMZD
Informationandcommunication
ISA88ISA95ERP
SAPIEC62264B2MML
Processcontrol
PLCDCS
SixSigmaisasetoftechniquesandtoolsforprocessimprovement.ItwasintroducedbyengineerBill
SmithwhileworkingatMotorolain1986.[1][2]JackWelchmadeitcentraltohisbusinessstrategyatGeneral
Electricin1995.[3]Today,itisusedinmanyindustrialsectors.[4]
SixSigmaseekstoimprovethequalityoftheoutputofaprocessbyidentifyingandremovingthecausesof
defects(errors)andminimizingvariabilityinmanufacturingandbusinessprocesses.Itusesasetofquality
managementmethods,mainlyempirical,statisticalmethods,andcreatesaspecialinfrastructureofpeople

withintheorganization,whoareexpertsinthesemethods.EachSixSigmaprojectcarriedoutwithinan
organizationfollowsadefinedsequenceofstepsandhasspecificvaluetargets,forexample:reduce
processcycletime,reducepollution,reducecosts,increasecustomersatisfaction,andincreaseprofits.
ThetermSixSigmaoriginatedfromterminologyassociatedwithmanufacturing,specificallyterms
associatedwithstatisticalmodelingofmanufacturingprocesses.Thematurityofamanufacturingprocess
canbedescribedbyasigmaratingindicatingitsyieldorthepercentageofdefectfreeproductsitcreates.
Asixsigmaprocessisoneinwhich99.99966%ofallopportunitiestoproducesomefeatureofapartare
statisticallyexpectedtobefreeofdefects(3.4defectivefeatures/millionopportunities),although,as
discussedbelow,thisdefectlevelcorrespondstoonlya4.5sigmalevel.Motorolasetagoalof"sixsigma"
forallofitsmanufacturingoperations,andthisgoalbecameabywordforthemanagementand
engineeringpracticesusedtoachieveit.
Contents

1Doctrine
2Methodologies
o 2.1DMAIC
o 2.2DMADVorDFSS
o 2.3Qualitymanagementtoolsandmethods
3Implementationroles
o 3.1Certification
4Etymologyof"sixsigmaprocess"
5Roleofthe1.5sigmashift
o 5.1Sigmalevels
6Software
7Application
8Criticism
o 8.1Lackoforiginality
o 8.2Roleofconsultants
o 8.3Potentialnegativeeffects
8.3.1Overrelianceonstatisticaltools
8.3.2Stiflingcreativityinresearchenvironments
o 8.4Lackofsystematicdocumentation
o 8.51.5sigmashift
9Seealso
10References
11Furtherreading

Doctrine
SixSigmadoctrineasserts:

Continuouseffortstoachievestableandpredictableprocessresults(e.g.,byreducingprocess
variation)areofvitalimportancetobusinesssuccess.
Manufacturingandbusinessprocesseshavecharacteristicsthatcanbemeasured,analyzed,
controlledandimproved.
Achievingsustainedqualityimprovementrequirescommitmentfromtheentireorganization,
particularlyfromtoplevelmanagement.

FeaturesthatsetSixSigmaapartfrompreviousqualityimprovementinitiativesinclude:

AclearfocusonachievingmeasurableandquantifiablefinancialreturnsfromanySixSigma
project.
Anincreasedemphasisonstrongandpassionatemanagementleadershipandsupport.
Aclearcommitmenttomakingdecisionsonthebasisofverifiabledataandstatisticalmethods,
ratherthanassumptionsandguesswork.

Theterm"sixsigma"comesfromstatisticsandisusedinstatisticalqualitycontrol,whichevaluatesprocess
capability.Originally,itreferredtotheabilityofmanufacturingprocessestoproduceaveryhigh
proportionofoutputwithinspecification.Processesthatoperatewith"sixsigmaquality"overtheshort
termareassumedtoproducelongtermdefectlevelsbelow3.4defectspermillionopportunities
(DPMO).[5][6]SixSigma'simplicitgoalistoimproveallprocesses,butnottothe3.4DPMOlevelnecessarily.
Organizationsneedtodetermineanappropriatesigmalevelforeachoftheirmostimportantprocesses
andstrivetoachievethese.Asaresultofthisgoal,itisincumbentonmanagementoftheorganizationto
prioritizeareasofimprovement.
"SixSigma"wasregisteredJune11,1991asU.S.ServiceMark74,026,418.In2005Motorolaattributed
overUS$17billioninsavingstoSixSigma.[7]
OtherearlyadoptersofSixSigmaincludeHoneywell(previouslyknownasAlliedSignal)andGeneral
Electric,whereJackWelchintroducedthemethod.[8]Bythelate1990s,abouttwothirdsoftheFortune
500organizationshadbegunSixSigmainitiativeswiththeaimofreducingcostsandimprovingquality.[9]
Inrecentyears,somepractitionershavecombinedSixSigmaideaswithleanmanufacturingtocreatea
methodologynamedLeanSixSigma.[10]TheLeanSixSigmamethodologyviewsleanmanufacturing,which
addressesprocessflowandwasteissues,andSixSigma,withitsfocusonvariationanddesign,as
complementarydisciplinesaimedatpromoting"businessandoperationalexcellence".[10]Companiessuch
asGE,[11]Verizon,GENPACT,andIBMuseLeanSixSigmatofocustransformationeffortsnotjuston
efficiencybutalsoongrowth.Itservesasafoundationforinnovationthroughouttheorganization,from
manufacturingandsoftwaredevelopmenttosalesandservicedeliveryfunctions.
TheInternationalOrganizationforStandardization(ISO)haspublishedin2011thefirststandard"ISO
13053:2011"definingaSixSigmaprocess.[12]Other"standards"arecreatedmostlybyuniversitiesor
companiesthathavesocalledfirstpartycertificationprogramsforSixSigma.
Methodologies
SixSigmaprojectsfollowtwoprojectmethodologiesinspiredbyDeming'sPlanDoCheckActCycle.These
methodologies,composedoffivephaseseach,beartheacronymsDMAICandDMADV.[9]

DMAICisusedforprojectsaimedatimprovinganexistingbusinessprocess.[9]DMAICis
pronouncedas"duhmayick"(<dmek>).
DMADVisusedforprojectsaimedatcreatingnewproductorprocessdesigns.[9]DMADVis
pronouncedas"duhmadvee"(<dmdvi>).

DMAIC

ThefivestepsofDMAIC

Mainarticle:DMAIC
TheDMAICprojectmethodologyhasfivephases:

Definethesystem,thevoiceofthecustomerandtheirrequirements,andtheprojectgoals,
specifically.
Measurekeyaspectsofthecurrentprocessandcollectrelevantdata;calculatethe'asis'Process
Capability.
Analyzethedatatoinvestigateandverifycauseandeffectrelationships.Determinewhatthe
relationshipsare,andattempttoensurethatallfactorshavebeenconsidered.Seekoutrootcause
ofthedefectunderinvestigation.
Improveoroptimizethecurrentprocessbasedupondataanalysisusingtechniquessuchasdesign
ofexperiments,pokayokeormistakeproofing,andstandardworktocreateanew,futurestate
process.Setuppilotrunstoestablishprocesscapability.
Controlthefuturestateprocesstoensurethatanydeviationsfromthetargetarecorrectedbefore
theyresultindefects.Implementcontrolsystemssuchasstatisticalprocesscontrol,production
boards,visualworkplaces,andcontinuouslymonitortheprocess.

SomeorganizationsaddaRecognizestepatthebeginning,whichistorecognizetherightproblemtowork
on,thusyieldinganRDMAICmethodology.[13]
DMADVorDFSS

ThefivestepsofDMADV
Mainarticle:DFSS

TheDMADVprojectmethodology,knownasDFSS("DesignForSixSigma"),[9]featuresfivephases:

Definedesigngoalsthatareconsistentwithcustomerdemandsandtheenterprisestrategy.
MeasureandidentifyCTQs(characteristicsthatareCriticalToQuality),measureproduct
capabilities,productionprocesscapability,andmeasurerisks.
Analyzetodevelopanddesignalternatives
Designanimprovedalternative,bestsuitedperanalysisinthepreviousstep
Verifythedesign,setuppilotruns,implementtheproductionprocessandhanditovertothe
processowner(s).

Qualitymanagementtoolsandmethods
WithintheindividualphasesofaDMAICorDMADVproject,SixSigmautilizesmanyestablishedquality
managementtoolsthatarealsousedoutsideSixSigma.Thefollowingtableshowsanoverviewofthemain
methodsused.

5Whys
Statisticalandfittingtools
o Analysisofvariance
o Generallinearmodel
o ANOVAGaugeR&R

Regressionanalysis
Correlation
Scatterdiagram
Chisquaredtest
Axiomaticdesign
BusinessProcessMapping/Checksheet
Cause&effectsdiagram(alsoknownasfishboneorIshikawadiagram)
Controlchart/Controlplan(alsoknownasaswimlanemap)/Runcharts
Costbenefitanalysis
CTQtree
Designofexperiments/Stratification
Histograms/Paretoanalysis/Paretochart
Pickchart/Processcapability/Rolledthroughputyield
QualityFunctionDeployment(QFD)
QuantitativemarketingresearchthroughuseofEnterpriseFeedbackManagement(EFM)systems
Rootcauseanalysis
SIPOCanalysis(Suppliers,Inputs,Process,Outputs,Customers)
COPISanalysis(Customercentricversion/perspectiveofSIPOC)
Taguchimethods/TaguchiLossFunction
Valuestreammapping
o
o
o
o

Implementationroles
OnekeyinnovationofSixSigmainvolvestheabsolute"professionalizing"ofqualitymanagementfunctions.
PriortoSixSigma,qualitymanagementinpracticewaslargelyrelegatedtotheproductionfloorandto
statisticiansinaseparatequalitydepartment.FormalSixSigmaprogramsadoptakindofeliteranking
terminology(similartosomemartialartssystems,likeKungFuandJudo)todefineahierarchy(andspecial
careerpath)thatincludesallbusinessfunctionsandlevels.
SixSigmaidentifiesseveralkeyrolesforitssuccessfulimplementation.[14]

ExecutiveLeadershipincludestheCEOandothermembersoftopmanagement.Theyare
responsibleforsettingupavisionforSixSigmaimplementation.Theyalsoempowertheotherrole
holderswiththefreedomandresourcestoexplorenewideasforbreakthroughimprovementsby
transcendingdepartmentalbarriersandovercominginherentresistancetochange.[15]
ChampionstakeresponsibilityforSixSigmaimplementationacrosstheorganizationinan
integratedmanner.TheExecutiveLeadershipdrawsthemfromuppermanagement.Champions
alsoactasmentorstoBlackBelts.
MasterBlackBelts,identifiedbyChampions,actasinhousecoachesonSixSigma.Theydevote
100%oftheirtimetoSixSigma.TheyassistChampionsandguideBlackBeltsandGreenBelts.
Apartfromstatisticaltasks,theyspendtheirtimeonensuringconsistentapplicationofSixSigma
acrossvariousfunctionsanddepartments.
BlackBeltsoperateunderMasterBlackBeltstoapplySixSigmamethodologytospecificprojects.
Theydevote100%oftheirvaluedtimetoSixSigma.TheyprimarilyfocusonSixSigmaproject
executionandspecialleadershipwithspecialtasks,whereasChampionsandMasterBlackBelts
focusonidentifyingprojects/functionsforSixSigma.
GreenBeltsaretheemployeeswhotakeupSixSigmaimplementationalongwiththeirotherjob
responsibilities,operatingundertheguidanceofBlackBelts.

Specialtrainingisneeded[16]forallofthesepractitionerstoensurethattheyfollowthemethodologyand
usethedatadrivenapproachcorrectly.Thistrainingisveryimportant[citationneeded].

Someorganizationsuseadditionalbeltcolours,suchasYellowBelts,foremployeesthathavebasictraining
inSixSigmatoolsandgenerallyparticipateinprojectsand"Whitebelts"forthoselocallytrainedinthe
conceptsbutdonotparticipateintheprojectteam."Orangebelts"arealsomentionedtobeusedfor
specialcases.[17]
Certification
Mainarticle:ListofSixSigmacertificationorganizations
GeneralElectricandMotoroladevelopedcertificationprogramsaspartoftheirSixSigmaimplementation,
verifyingindividuals'commandoftheSixSigmamethodsattherelevantskilllevel(GreenBelt,BlackBelt
etc.).Followingthisapproach,manyorganizationsinthe1990sstartedofferingSixSigmacertificationsto
theiremployees.[9][18]CriteriaforGreenBeltandBlackBeltcertificationvary;somecompaniessimply
requireparticipationinacourseandaSixSigmaproject.[18]Thereisnostandardcertificationbody,and
differentcertificationservicesareofferedbyvariousqualityassociationsandotherprovidersagainsta
fee.[19][20]TheAmericanSocietyforQualityforexamplerequiresBlackBeltapplicantstopassawritten
examandtoprovideasignedaffidavitstatingthattheyhavecompletedtwoprojectsoroneproject
combinedwiththreeyears'practicalexperienceinthebodyofknowledge.[18][21]
Etymologyof"sixsigmaprocess"
Theterm"sixsigmaprocess"comesfromthenotionthatifonehassixstandarddeviationsbetweenthe
processmeanandthenearestspecificationlimit,asshowninthegraph,practicallyno[notincitationgiven]items
willfailtomeetspecifications.[5]Thisisbasedonthecalculationmethodemployedinprocesscapability
studies.
Capabilitystudiesmeasurethenumberofstandarddeviationsbetweentheprocessmeanandthenearest
specificationlimitinsigmaunits,representedbytheGreekletter(sigma).Asprocessstandarddeviation
goesup,orthemeanoftheprocessmovesawayfromthecenterofthetolerance,fewerstandard
deviationswillfitbetweenthemeanandthenearestspecificationlimit,decreasingthesigmanumberand
increasingthelikelihoodofitemsoutsidespecification.OneshouldalsonotethatcalculationofSigma
levelsforaprocessdataisindependentofthedatabeingnormallydistributed.Inoneofthecriticismsto
SixSigma,practitionersusingthisapproachspendalotoftimetransformingdatafromnonnormalto
normalusingtransformationtechniques.ItmustbesaidthatSigmalevelscanbedeterminedforprocess
datathathasevidenceofnonnormality.[5]

Graphofthenormaldistribution,whichunderliesthestatisticalassumptionsoftheSixSigmamodel.The
Greekletter(sigma)marksthedistanceonthehorizontalaxisbetweenthemean,,andthecurve's
inflectionpoint.Thegreaterthisdistance,thegreateristhespreadofvaluesencountered.Forthegreen
curveshownabove,=0and=1.Theupperandlowerspecificationlimits(USLandLSL,respectively)are
atadistanceof6fromthemean.Becauseofthepropertiesofthenormaldistribution,valueslyingthat
farawayfromthemeanareextremelyunlikely.Evenifthemeanweretomoverightorleftby1.5at
somepointinthefuture(1.5sigmashift,colouredredandblue),thereisstillagoodsafetycushion.Thisis

whySixSigmaaimstohaveprocesseswherethemeanisatleast6awayfromthenearestspecification
limit.
Roleofthe1.5sigmashift
Experiencehasshownthatprocessesusuallydonotperformaswellinthelongtermastheydointhe
shortterm.[5]Asaresult,thenumberofsigmasthatwillfitbetweentheprocessmeanandthenearest
specificationlimitmaywelldropovertime,comparedtoaninitialshorttermstudy.[5]Toaccountforthis
reallifeincreaseinprocessvariationovertime,anempiricallybased1.5sigmashiftisintroducedintothe
calculation.[5][22]Accordingtothisidea,aprocessthatfits6sigmabetweentheprocessmeanandthe
nearestspecificationlimitinashorttermstudywillinthelongtermfitonly4.5sigmaeitherbecausethe
processmeanwillmoveovertime,orbecausethelongtermstandarddeviationoftheprocesswillbe
greaterthanthatobservedintheshortterm,orboth.[5]
Hencethewidelyaccepteddefinitionofasixsigmaprocessisaprocessthatproduces3.4defectiveparts
permillionopportunities(DPMO).Thisisbasedonthefactthataprocessthatisnormallydistributedwill
have3.4partspermillionoutsidethelimits,whenthelimitsaresixsigmafromthe"original"meanofzero
andtheprocessmeanisthenshiftedby1.5sigma(andtherefore,thesixsigmalimitsarenolonger
symmetricalaboutthemean).[5]Theformersixsigmadistribution,whenundertheeffectofthe1.5sigma
shift,iscommonlyreferredtoasa4.5sigmaprocess.However,itshouldbenotedthatthefailurerateofa
sixsigmadistributionwiththemeanshifted1.5sigmaisnotequivalenttothefailurerateofa4.5sigma
processwiththemeancenteredonzero.[5]Thisallowsforthefactthatspecialcausesmayresultina
deteriorationinprocessperformanceovertimeandisdesignedtopreventunderestimationofthedefect
levelslikelytobeencounteredinreallifeoperation.[5]
Theroleofthesigmashiftismainlyacademic.Thepurposeofsixsigmaistogenerateorganizational
performanceimprovement.Itisuptotheorganizationtodetermine,basedoncustomerexpectations,
whattheappropriatesigmalevelofaprocessis.Thepurposeofthesigmavalueisasacomparativefigure
todeterminewhetheraprocessisimproving,deteriorating,stagnantornoncompetitivewithothersinthe
samebusiness.Sixsigma(3.4DPMO)isnotthegoalofallprocesses.
Sigmalevels

Acontrolchartdepictingaprocessthatexperienceda1.5sigmadriftintheprocessmeantowardthe
upperspecificationlimitstartingatmidnight.Controlchartsareusedtomaintain6sigmaqualityby
signalingwhenqualityprofessionalsshouldinvestigateaprocesstofindandeliminatespecialcause
variation.
Seealso:Threesigmarule
Thetable[23][24]belowgiveslongtermDPMOvaluescorrespondingtovariousshorttermsigmalevels.
Itmustbeunderstoodthatthesefiguresassumethattheprocessmeanwillshiftby1.5sigmatowardthe
sidewiththecriticalspecificationlimit.Inotherwords,theyassumethataftertheinitialstudydetermining
theshorttermsigmalevel,thelongtermCpkvaluewillturnouttobe0.5lessthantheshorttermCpk
value.So,forexample,theDPMOfiguregivenfor1sigmaassumesthatthelongtermprocessmeanwillbe
0.5sigmabeyondthespecificationlimit(Cpk=0.17),ratherthan1sigmawithinit,asitwasintheshort
termstudy(Cpk=0.33).Notethatthedefectpercentagesindicateonlydefectsexceedingthespecification
limittowhichtheprocessmeanisnearest.Defectsbeyondthefarspecificationlimitarenotincludedin
thepercentages.
Sigma
level
1
2
3
4
5
6
7

Sigma(with1.5
shift)
0.5
0.5
1.5
2.5
3.5
4.5
5.5

Percent
defective
691,462 69%
308,538 31%
66,807 6.7%
6,210 0.62%
233
0.023%
3.4
0.00034%
0.019 0.0000019%
DPMO

Percentage
yield
31%
69%
93.3%
99.38%
99.977%
99.99966%
99.9999981%

Shortterm
Cpk
0.33
0.67
1.00
1.33
1.67
2.00
2.33

Longterm
Cpk
0.17
0.17
0.5
0.83
1.17
1.5
1.83

Software
Mainarticle:ListofSixSigmasoftwarepackages
Application
Mainarticle:ListofSixSigmacompanies
SixSigmamostlyfindsapplicationinlargeorganizations.[25]AnimportantfactorinthespreadofSixSigma
wasGE's1998announcementof$350millioninsavingsthankstoSixSigma,afigurethatlatergrewto
morethan$1billion.[25]AccordingtoindustryconsultantslikeThomasPyzdekandJohnKullmann,
companieswithfewerthan500employeesarelesssuitedtoSixSigmaimplementationorneedtoadapt
thestandardapproachtomakeitworkforthem.[25]SixSigmahowevercontainsalargenumberoftools
andtechniquesthatworkwellinsmalltomidsizeorganizations.Thefactthatanorganizationisnotbig
enoughtobeabletoaffordBlackBeltsdoesnotdiminishitsabilitiestomakeimprovementsusingthisset
oftoolsandtechniques.TheinfrastructuredescribedasnecessarytosupportSixSigmaisaresultofthe
sizeoftheorganizationratherthanarequirementofSixSigmaitself.[25]
Criticism
Lackoforiginality
NotedqualityexpertJosephM.JuranhasdescribedSixSigmaas"abasicversionofqualityimprovement",
statingthat"thereisnothingnewthere.Itincludeswhatweusedtocallfacilitators.They'veadoptedmore

flamboyantterms,likebeltswithdifferentcolors.Ithinkthatconcepthasmerittosetapart,tocreate
specialistswhocanbeveryhelpful.Again,that'snotanewidea.TheAmericanSocietyforQualitylongago
establishedcertificates,suchasforreliabilityengineers."[26]
Roleofconsultants
Theuseof"BlackBelts"asitinerantchangeagentshas(controversially)fosteredanindustryoftrainingand
certification.CriticsarguethereisoversellingofSixSigmabytoogreatanumberofconsultingfirms,many
ofwhichclaimexpertiseinSixSigmawhentheyhaveonlyarudimentaryunderstandingofthetoolsand
techniquesinvolvedorthemarketsorindustriesinwhichtheyareacting.[27]
Potentialnegativeeffects
AFortunearticlestatedthat"of58largecompaniesthathaveannouncedSixSigmaprograms,91percent
havetrailedtheS&P500since".Thestatementwasattributedto"ananalysisbyCharlesHollandof
consultingfirmQualpro(whichespousesacompetingqualityimprovementprocess)".[28]Thesummaryof
thearticleisthatSixSigmaiseffectiveatwhatitisintendedtodo,butthatitis"narrowlydesignedtofix
anexistingprocess"anddoesnothelpin"comingupwithnewproductsordisruptivetechnologies."[29][30]
Overrelianceonstatisticaltools
Amoredirectcriticismisthe"rigid"natureofSixSigmawithitsoverrelianceonmethodsandtools.In
mostcases,moreattentionispaidtoreducingvariationandsearchingforanysignificantfactorsandless
attentionispaidtodevelopingrobustnessinthefirstplace(whichcanaltogethereliminatetheneedfor
reducingvariation).[31]Theextensiverelianceonsignificancetestinganduseofmultipleregression
techniquesincreasestheriskofmakingcommonlyunknowntypesofstatisticalerrorsormistakes.A
possibleconsequenceofSixSigma'sarrayofPvaluemisconceptionsisthefalsebeliefthattheprobability
ofaconclusionbeinginerrorcanbecalculatedfromthedatainasingleexperimentwithoutreferenceto
externalevidenceortheplausibilityoftheunderlyingmechanism.[32]Oneofthemostseriousbutalltoo
commonmisusesofinferentialstatisticsistotakeamodelthatwasdevelopedthroughexploratorymodel
buildingandsubjectittothesamesortsofstatisticalteststhatareusedtovalidateamodelthatwas
specifiedinadvance.[33]
AnothercommentreferstotheoftenmentionedTransferFunction,whichseemstobeaflawedtheoryif
lookedatindetail.[34]Sincesignificancetestswerefirstpopularizedmanyobjectionshavebeenvoicedby
prominentandrespectedstatisticians.Thevolumeofcriticismandrebuttalhasfilledbookswithlanguage
seldomusedinthescholarlydebateofadrysubject.[35][36][37][38]Muchofthefirstcriticismwasalready
publishedmorethan40yearsago.Referto:Statisticalhypothesistesting#Criticismfordetails.
ArticlesfeaturingcriticshaveappearedintheNovemberDecember2006issueofUSAArmyLogistician
regardingSixSigma:"Thedangersofasingleparadigmaticorientation(inthiscase,thatoftechnical
rationality)canblindustovaluesassociatedwithdoublelooplearningandthelearningorganization,
organizationadaptability,workforcecreativityanddevelopment,humanizingtheworkplace,cultural
awareness,andstrategymaking."[39]
NassimNicholasTalebconsiderriskmanagerslittlemorethan"blindusers"ofstatisticaltoolsand
methods.[40]Hestatesthatstatisticsisfundamentallyincompleteasafieldasitcannotpredicttheriskof
rareeventssomethingSixSigmaisspeciallyconcernedwith.Furthermore,errorsinpredictionarelikely
tooccurasaresultofignoranceforordistinctionbetweenepistemicandotheruncertainties.Theseerrors
arethebiggestintimevariant(reliability)relatedfailures.[41]
Stiflingcreativityinresearchenvironments

ABusinessWeekarticlesaysthatJamesMcNerney'sintroductionofSixSigmaat3Mhadtheeffectof
stiflingcreativityandreportsitsremovalfromtheresearchfunction.ItcitestwoWhartonSchool
professorswhosaythatSixSigmaleadstoincrementalinnovationattheexpenseofblueskiesresearch.[42]
ThisphenomenonisfurtherexploredinthebookGoingLean,whichdescribesarelatedapproachknownas
leandynamicsandprovidesdatatoshowthatFord's"6Sigma"programdidlittletochangeitsfortunes.[43]
AccordingtoanarticlebyJohnDodge,editorinchiefofDesignNews,useofSixSigmaisinappropriateina
researchenvironment.Dodgestates[44]"excessivemetrics,steps,measurementsandSixSigma'sintense
focusonreducingvariabilitywaterdownthediscoveryprocess.UnderSixSigma,thefreewheelingnature
ofbrainstormingandtheserendipitoussideofdiscoveryisstifled."Heconcludes"there'sgeneral
agreementthatfreedominbasicorpureresearchispreferablewhileSixSigmaworksbestinincremental
innovationwhenthere'sanexpressedcommercialgoal."
Lackofsystematicdocumentation
OnecriticismvoicedbyYasarJarrarandAndyNeelyfromtheCranfieldSchoolofManagement'sCentrefor
BusinessPerformanceisthatwhileSixSigmaisapowerfulapproach,itcanalsoundulydominatean
organization'sculture;andtheyaddthatmuchoftheSixSigmaliteratureinaremarkableway(sixsigma
claimstobeevidence,scientificallybased)lacksacademicrigor:
Onefinalcriticism,probablymoretotheSixSigmaliteraturethanconcepts,relatestotheevidenceforSix
Sigmassuccess.Sofar,documentedcasestudiesusingtheSixSigmamethodsarepresentedasthe
strongestevidenceforitssuccess.However,lookingatthesedocumentedcases,andapartfromafewthat
aredetailedfromtheexperienceofleadingorganizationslikeGEandMotorola,mostcasesarenot
documentedinasystemicoracademicmanner.Infact,themajorityarecasestudiesillustratedon
websites,andare,atbest,sketchy.TheyprovidenomentionofanyspecificSixSigmamethodsthatwere
usedtoresolvetheproblems.IthasbeenarguedthatbyrelyingontheSixSigmacriteria,managementis
lulledintotheideathatsomethingisbeingdoneaboutquality,whereasanyresultingimprovementis
accidental(Latzko1995).Thus,whenlookingattheevidenceputforwardforSixSigmasuccess,mostlyby
consultantsandpeoplewithvestedinterests,thequestionthatbegstobeaskedis:arewemakingatrue
improvementwithSixSigmamethodsorjustgettingskilledattellingstories?Everyoneseemstobelieve
thatwearemakingtrueimprovements,butthereissomewaytogotodocumenttheseempiricallyand
clarifythecausalrelations.
[31]
1.5sigmashift
ThestatisticianDonaldJ.Wheelerhasdismissedthe1.5sigmashiftas"goofy"becauseofitsarbitrary
nature.[45]Itsuniversalapplicabilityisseenasdoubtful.
The1.5sigmashifthasalsobecomecontentiousbecauseitresultsinstated"sigmalevels"thatreflect
shorttermratherthanlongtermperformance:aprocessthathaslongtermdefectlevelscorrespondingto
4.5sigmaperformanceis,bySixSigmaconvention,describedasa"sixsigmaprocess."[5][46]Theaccepted
SixSigmascoringsystemthuscannotbeequatedtoactualnormaldistributionprobabilitiesforthestated
numberofstandarddeviations,andthishasbeenakeyboneofcontentionoverhowSixSigmameasures
aredefined.[46]Thefactthatitisrarelyexplainedthata"6sigma"processwillhavelongtermdefectrates
correspondingto4.5sigmaperformanceratherthanactual6sigmaperformancehasledseveral
commentatorstoexpresstheopinionthatSixSigmaisaconfidencetrick.[5]