Lean Startup in the Enterprise - Lean Startup Conference 2012
The lean startup
-
date post
17-Sep-2014 -
Category
Business
-
view
366 -
download
0
description
Transcript of The lean startup
AcclaimforTHELEANSTARTUP
AcclaimforTHELEANSTARTUP
“TheLeanStartupisn’tjustabouthowtocreateamoresuccessful
entrepreneurialbusiness;it’saboutwhatwecanlearnfromthose
businessestoimprovevirtualyeverythingwedo.IimagineLean
Startupprinciplesappliedtogovernmentprograms,tohealthcare,
andtosolvingtheworld’sgreatproblems.It’sultimatelyananswer
tothequestionHowcanwelearnmorequicklywhat
worksand
discardwhatdoesn’t?”
—TimO’Reily,CEO,O’ReilyMedia
“EricRiesunravelsthemysteriesofentrepreneurshipandreveals
thatmagicandgeniusarenotthenecessaryingredientsforsuccess
butinsteadproposesascienticprocessthatcanbelearnedand
replicated.Whetheryouareastartupentrepreneurorcorporate
entrepreneur,thereareimportantlessonshereforyouonyour
questtowardthenewandunknown.”
—TimBrown,CEO,IDEO
“Theroadmapforinnovationforthetwenty-firstcentury.Theideas
inTheLeanStartupwilhelpcreatethenextindustrialrevolution.”
—SteveBlank,lecturer,StanfordUniversity,
UCBerkeleyHassBusiness
School
“Everyfoundingteamshouldstopforforty-eighthoursandread
TheLeanStartup.Seriously,stopandreadthisbooknow.”
—ScotCase,CEO,StartupAmericaPartnership
“Thekeylessonofthisbookisthatstartupshappeninthe
present
—thatmessyplacebetweenthepastandthefuturewherenothing
happensaccordingtoPowerPoint.Ries’s‘readandreact’approach
tothissport,hisrelentlessfocusonvalidatedlearning,thenever-
endinganxietyofhoveringbetween‘persevere’and‘pivot,’albear
witnesstohisappreciationforthedynamicsofentrepreneurship.”
witnesstohisappreciationforthedynamicsofentrepreneurship.”
—GeofreyMoore,author,CrossingtheChasm
“Ifyouareanentrepreneur,readthisbook.Ifyouarethinking
aboutbecominganentrepreneur,readthisbook.Ifyouarejust
curiousaboutentrepreneurship,readthisbook.StartingLeanis
today’sbestpracticeforinnovators.Doyourselfa
favorandread
thisbook.”
—RandyKomisar,foundingdirectorofTiVoandauthorofthe
bestselingTheMonkandtheRiddle
“Howdoyouapplythefty-year-oldideasofLeantothefast-
paced,high-uncertaintyworldofstartups?Thisbookprovidesa
briliant,wel-documented,andpracticalanswer.Itissureto
becomeamanagementclassic.”
—DonReinertsen,author,ThePrinciplesofProductDevelopment
Flow
“Whatwouldhappenifbusinesseswerebuiltfromthegroundup
tolearnwhattheircustomersrealywanted?TheLeanStartupis
thefoundationforreimaginingalmosteverythingabouthowwork
works.Don’tletthewordstartupinthetitleconfuseyou.Thisisa
cookbookforentrepreneursinorganizationsofalsizes.”
—RoyBahat,president,IGNEntertainment
“TheLeanStartupisafoundationalmust-readforfounders,
enablingthemtoreduceproductfailuresbybringingstructureand
sciencetowhatisusualyinformalandanart.Itprovides
actionablewaystoavoidproduct-learningmistakes,rigorously
evaluateearlysignalsfromthemarketthroughvalidated
learning,
anddecidewhethertopersevereortopivot,alchalengesthat
heightenthechanceofentrepreneurialfailure.”
—NoamWasserman,professor,HarvardBusinessSchool
“Oneofthebestandmost
insightfulnewbookson
entrepreneurshipandmanagementI’veeverread.Shouldbe
entrepreneurshipandmanagementI’veeverread.Shouldbe
requiredreadingnotonlyfortheentrepreneursthatIworkwith,
butformyfriendsandcoleaguesinvariousindustrieswhohave
inevitablygrappledwithmanyofthechalengesthatTheLean
Startupaddresses.”
—EugeneJ.Huang,partner,TrueNorthVenturePartner
“Inbusiness,a‘lean’
enterpriseissustainableeciencyinaction.
EricRies’srevolutionaryLeanStartupmethodwilhelpbringyour
newbusinessideatoanendresultthatissuccessfulandsustainable.
You’lndinnovativestepsandstrategiesforcreatingand
managingyourownstartupwhilelearningfromthereal-life
successesandcolapsesofothers.Thisbookisamust-readfor
entrepreneurswhoaretrulyreadytostartsomethinggreat!”
—KenBlanchard,coauthorofTheOneMinute
Manager®
andTheOneMinuteEntrepreneur
Copyright©2011byEricRies
Allrightsreserved.
PublishedintheUnitedStatesbyCrownBusiness,animprintoftheCrownPublishing
Group,adivisionofRandomHouse,Inc.,NewYork.www.crownpublishing.com
CROWNBUSINESSisatrademarkandCROWNandtheRisingSuncolophonare
registeredtrademarksofRandomHouse,Inc.
LibraryofCongressCataloging-in-PublicationData
Ries,Eric,1978–
Theleanstartup/EricRies.
—1sted.
p.cm.
1.Newbusinessenterprises.2.Consumers’preferences.3.Organizational
effectiveness.I.Title.
HD62.5.R5452011
658.1′1—dc222011012100
eISBN:978-0-307-88791-7
BookdesignbyLaurenDong
IllustrationsbyFredHaynes
JacketdesignbyMarcusGosling
v3.1
ForTara
Contents
Cover
TitlePage
Copyright
Dedication
Introduction
PartOneVISION
1.Start
2.Define
3.Learn
4.Experiment
PartTwoSTEER
5.Leap
6.Test
7.Measure
8.Pivot(orPersevere)
PartThreeACCELERATE
9.Batch
10.Grow
11.Adapt
12.Innovate
13.Epilogue:WasteNot
14.JointheMovement
Endnotes
Disclosures
Disclosures
Acknowledgments
AbouttheAuthor
Introduction
Stopmeifyou’veheardthisonebefore.Briliantcolegekids
sitinginadormareinventingthefuture.Heedlessofboundaries,
possessedofnewtechnologyandyouthfulenthusiasm,theybuild
anewcompanyfromscratch.Theirearlysuccessalowsthemto
raisemoneyandbringanamazingnewproducttomarket.They
hiretheirfriends,assembleasuperstarteam,anddaretheworldto
stopthem.
Tenyearsandseveralstartupsago,thatwasme,buildingmyrst
company.Iparticularlyrememberamomentfrombackthen:the
momentIrealizedmycompanywasgoingtofail.Mycofounder
andIwereatourwits’end.Thedot-combubblehad
burst,andwe
hadspentalourmoney.Wetrieddesperatelytoraisemore
capital,andwecouldnot.Itwaslikeabreakupscenefroma
Holywoodmovie:itwasraining,andwewerearguinginthe
street.Wecouldn’tevenagreeonwheretowalknext,andsowe
partedinanger,headinginoppositedirections.Asametaphorfor
ourcompany’sfailure,thisimageofthetwoofus,lostintherain
anddriftingapart,isperfect.
Itremainsapainfulmemory.Thecompanylimpedalongfor
monthsafterward,butoursituationwashopeless.Atthetime,it
hadseemedweweredoingeverythingright:wehadagreat
product,abriliantteam,amazingtechnology,andthe
rightideaat
therighttime.Andwerealywereontosomething.Wewere
buildingawayforcolegekidstocreateonlineprolesforthe
purposeofsharing…withemployers.Oops.Butdespitea
promisingidea,wewerenonethelessdoomedfromdayone,
becausewedidnotknowtheprocesswewouldneedtousetoturn
becausewedidnotknowtheprocesswewouldneedtousetoturn
ourproductinsightsintoagreatcompany.
Ifyou’veneverexperiencedafailurelikethis,itishardto
describethefeeling.It’sasiftheworldwerefalingoutfromunder
you.Yourealizeyou’vebeenduped.Thestoriesinthemagazines
arelies:hardworkandperseverancedon’tleadtosuccess.Even
worse,themany,many,manypromisesyou’vemadetoemployees,
friends,andfamilyarenotgoingtocometrue.Everyonewho
thoughtyouwerefoolishforsteppingoutonyourownwilbe
provenright.
Itwasn’tsupposedtoturnoutthatway.Inmagazinesand
newspapers,inblockbustermovies,andoncountlessblogs,wehear
themantraofthesuccessfulentrepreneurs:throughdetermination,
briliance,greattiming,and—aboveal—agreatproduct,youtoo
canachievefameandfortune.
Thereisamythmakingindustryhardatworktoselusthatstory,
butIhavecometobelievethatthestoryisfalse,theproductof
selectionbiasandafter-the-factrationalization.Infact,having
workedwithhundredsofentrepreneurs,Ihaveseenrsthandhow
oftenapromisingstartleadstofailure.Thegrimrealityisthatmost
startupsfail.Mostnewproductsarenotsuccessful.Mostnew
venturesdonotliveuptotheirpotential.
Yetthestoryofperseverance,creativegenius,andhardwork
persists.Whyisitsopopular?Ithinkthereissomethingdeeply
appealingaboutthismodern-dayrags-to-richesstory.Itmakes
successseeminevitableifyoujusthavetherightstu.It
meansthat
themundanedetails,theboringstu,thesmalindividualchoices
don’tmater.Ifwebuildit,theywilcome.Whenwefail,asso
manyofusdo,wehaveaready-madeexcuse:wedidn’thavethe
rightstu.Weweren’tvisionaryenoughorweren’tintheright
placeattherighttime.
Aftermorethantenyearsasanentrepreneur,Icametoreject
thatlineofthinking.Ihavelearnedfrombothmyownsuccesses
andfailuresandthoseofmanyothersthatit’stheboringstuthat
matersthemost.Startupsuccessisnotaconsequenceofgood
genesorbeingintherightplaceattherighttime.Startupsuccess
canbeengineeredbyfolowingtherightprocess,
whichmeansit
canbeengineeredbyfolowingtherightprocess,whichmeansit
canbelearned,whichmeansitcanbetaught.
Entrepreneurshipisakindofmanagement.No,youdidn’tread
thatwrong.Wehavewildly
divergentassociationswiththesetwo
words,entrepreneurshipandmanagement.Lately,itseemsthatone
iscool,innovative,andexcitingandtheotherisdul,serious,and
bland.Itistimetolookpastthesepreconceptions.
Letmetelyouasecondstartupstory.It’s2004,andagroupof
foundershavejuststartedanewcompany.Theirpreviouscompany
hadfailedverypublicly.Theircredibilityisatanal-timelow.They
haveahugevision:tochangethewaypeoplecommunicate
by
usinganewtechnologycaledavatars(remember,thiswasbefore
JamesCameron’sblockbustermovie).Theyarefolowinga
visionarynamedWilHarvey,whopaintsacompelingpicture:
peopleconnectingwiththeirfriends,hangingoutonline,using
avatarstogivethemacombinationofintimateconnectionandsafe
anonymity.Evenbeter,insteadofhavingtobuildaltheclothing,
furniture,andaccessoriestheseavatarswouldneedto
accessorize
theirdigitallives,thecustomerswouldbeenlistedtobuildthose
thingsandselthemtooneanother.
Theengineeringchalengebeforethemisimmense:creating
virtualworlds,user-
generatedcontent,anonlinecommerceengine,
micropayments,and—lastbutnotleast—thethree-dimensional
avatartechnologythatcanrunonanyone’sPC.
I’minthissecondstory,too.I’macofounderandchieftechnology
ocerofthiscompany,whichiscaledIMVU.Atthispointinour
careers,mycofoundersandIaredeterminedtomakenewmistakes.
Wedoeverythingwrong:insteadofspendingyearsperfectingour
technology,webuildaminimumviableproduct,an
earlyproduct
thatisterrible,fulofbugsandcrash-your-computer-yes-realy
stabilityproblems.Thenweshipittocustomerswaybeforeit’s
ready.Andwechargemoneyforit.Aftersecuringinitialcustomers,
wechangetheproductconstantly—muchtoofastbytraditional
standards—shippingnewversionsofourproductdozensoftimes
everysingleday.
Werealydidhavecustomersinthoseearlydays—truevisionary
earlyadopters—andweoftentalkedtothemandaskedfortheir
earlyadopters—andweoftentalkedtothemandaskedfortheir
feedback.Butweemphaticalydidnotdowhattheysaid.We
viewedtheirinputasonlyonesourceofinformation
aboutour
productandoveralvision.Infact,weweremuchmorelikelyto
runexperimentsonourcustomersthanweweretocatertotheir
whims.
Traditionalbusinessthinkingsaysthatthisapproach
shouldn’t
work,butitdoes,andyoudon’thavetotakemywordforit.As
you’lseethroughoutthisbook,theapproachwepioneeredat
IMVUhasbecomethebasisforanewmovementofentrepreneurs
aroundtheworld.Itbuildsonmanypreviousmanagementand
productdevelopmentideas,includingleanmanufacturing,design
thinking,customerdevelopment,andagiledevelopment.It
representsanewapproachtocreatingcontinuous
innovation.It’s
caledtheLeanStartup.
Despitethevolumeswritenonbusinessstrategy,thekey
atributesofbusinessleaders,andwaystoidentifythenextbig
thing,innovatorsstilstruggletobringtheirideastolife.Thiswas
thefrustrationthatledustotryaradicalnewapproachatIMVU,
onecharacterizedbyanextremelyfastcycletime,afocusonwhat
customerswant(withoutaskingthem),andascienticapproachto
makingdecisions.
ORIGINSOFTHELEANSTARTUP
Iamoneofthosepeoplewhogrewupprogrammingcomputers,
andsomyjourneytothinkingaboutentrepreneurshipand
managementhastakenacircuitouspath.Ihavealwaysworkedon
theproductdevelopmentsideofmyindustry;mypartnersand
bossesweremanagersormarketers,andmypeersworkedin
engineeringandoperations.Throughoutmycareer,Ikepthaving
theexperienceofworkingincrediblyhardonproducts
that
ultimatelyfailedinthemarketplace.
Atrst,largelybecauseofmybackground,Iviewedtheseas
technicalproblemsthatrequiredtechnicalsolutions:beter
architecture,abeter
engineeringprocess,beterdiscipline,focus,or
architecture,abeterengineeringprocess,beterdiscipline,focus,or
productvision.Thesesupposedxesledtostilmorefailure.SoI
readeverythingIcouldgetmyhandsonandwasblessedtohave
hadsomeofthetopmindsinSiliconValeyasmymentors.Bythe
timeIbecameacofounderofIMVU,Iwashungryfornewideas
abouthowtobuildacompany.
Iwasfortunatetohavecofounderswhowerewilingto
experimentwithnewapproaches.Theywerefedup—asIwas—by
thefailureoftraditionalthinking.Also,wewereluckytohave
SteveBlankasaninvestorandadviser.Backin2004,Stevehadjust
begunpreachinganewidea:thebusinessandmarketing
functions
ofastartupshouldbeconsideredasimportantasengineeringand
productdevelopmentandthereforedeserveanequalyrigorous
methodologytoguidethem.HecaledthatmethodologyCustomer
Development,anditoeredinsightandguidancetomydailywork
asanentrepreneur.
Meanwhile,IwasbuildingIMVU’sproductdevelopmentteam,
usingsomeoftheunorthodoxmethodsImentionedearlier.
Measuredagainstthe
traditionaltheoriesofproductdevelopmentI
hadbeentrainedoninmycareer,thesemethodsdidnotmake
sense,yetIcouldseersthandthattheywereworking.Istruggled
toexplainthepracticestonewemployees,investors,andthe
foundersofothercompanies.Welackedacommonlanguagefor
describingthemandconcreteprinciplesforunderstandingthem.
Ibegantosearchoutsideentrepreneurshipforideasthatcould
helpmemakesenseofmyexperience.Ibegantostudy
other
industries,especialymanufacturing,fromwhichmostmodern
theoriesofmanagementderive.Istudiedleanmanufacturing,a
processthatoriginatedinJapanwiththeToyotaProduction
System,acompletelynewwayofthinkingaboutthemanufacturing
ofphysicalgoods.Ifoundthatbyapplyingideasfromlean
manufacturingtomyownentrepreneurialchalenges—withafew
tweaksandchanges—Ihadthebeginningsofa
frameworkfor
makingsenseofthem.
ThislineofthoughtevolvedintotheLeanStartup:the
applicationofleanthinkingtotheprocessofinnovation.
IMVUbecameatremendoussuccess.IMVUcustomershave
IMVUbecameatremendoussuccess.IMVUcustomershave
createdmorethan60milionavatars.Itisaprotablecompany
withannualrevenuesofmorethan$50milionin2011,employing
morethanahundredpeopleinourcurrentocesin
Mountain
View,California.IMVU’svirtualgoodscatalog—whichseemedso
riskyyearsago—nowhasmorethan6milionitemsinit;more
than7,000areaddedeveryday,almostalcreatedbycustomers.
AsaresultofIMVU’ssuccess,Ibegantobeaskedforadviceby
otherstartupsandventurecapitalists.WhenIwoulddescribemy
experiencesatIMVU,Iwasoftenmetwithblankstaresorextreme
skepticism.Themostcommonreplywas“That
couldneverwork!”
Myexperiencesoewinthefaceofconventionalthinkingthat
mostpeople,evenintheinnovationhubofSiliconValey,could
notwraptheirmindsaroundit.
ThenIstartedtowrite,rston
ablogcaledStartupLessons
Learned,andspeak—atconferencesandtocompanies,startups,and
venturecapitalists—toanyonewhowouldlisten.Intheprocessof
beingcaledontodefendandexplainmyinsightsandwiththe
colaborationofotherwriters,thinkers,andentrepreneurs,Ihada
chancetoreneanddevelopthetheoryoftheLeanStartupbeyond
itsrudimentarybeginnings.Myhopealalongwastondwaysto
eliminatethetremendouswasteIsawalaroundme:
startupsthat
builtproductsnobodywanted,newproductspuledfromthe
shelves,countlessdreamsunrealized.
Eventualy,theLeanStartupideablossomedintoaglobal
movement.Entrepreneursbeganforminglocalin-
persongroupsto
discussandapplyLeanStartupideas.Therearenoworganized
communitiesofpracticeinmorethanahundredcitiesaroundthe
world.1Mytravelshavetakenmeacrosscountriesandcontinents.
EverywhereIhaveseenthesignsofanewentrepreneurial
renaissance.TheLeanStartupmovementismaking
entrepreneurshipaccessibletoawholenewgenerationoffounders
whoarehungryfornewideasabouthowtobuildsuccessful
companies.
Althoughmybackgroundisinhigh-techsoftware
entrepreneurship,themovementhasgrownwaybeyondthose
entrepreneurship,themovementhasgrownwaybeyondthose
roots.Thousandsof
entrepreneursareputingLeanStartup
principlestoworkineveryconceivableindustry.I’vehadthe
chancetoworkwithentrepreneursincompaniesofalsizes,in
dierentindustries,andeveningovernment.Thisjourneyhastaken
metoplacesIneverimaginedI’dsee,fromtheworld’smostelite
venturecapitalists,toFortune500boardrooms,tothePentagon.
ThemostnervousIhaveeverbeeninameetingwaswhenIwas
atemptingtoexplainLeanStartupprinciplestothechief
informationoceroftheU.S.Army,whoisathree-stargeneral
(fortherecord,hewasextremelyopentonewideas,evenfroma
civilianlikeme).
PretysoonIrealizedthatitwastimetofocusontheLean
Startupmovementfultime.Mymission:toimprovethesuccess
rateofnewinnovativeproductsworldwide.Theresultisthebook
youarereading.
THELEANSTARTUPMETHOD
Thisisabookfor
entrepreneursandthepeoplewhoholdthem
accountable.TheveprinciplesoftheLeanStartup,whichinform
althreepartsofthisbook,areasfolows:
1.Entrepreneursareeverywhere.Youdon’thavetoworkina
garagetobeinastartup.Theconceptofentrepreneurshipincludes
anyonewhoworkswithinmydenitionofastartup:ahuman
institutiondesignedtocreatenewproductsandservicesunder
conditionsofextremeuncertainty.Thatmeans
entrepreneursare
everywhereandtheLeanStartupapproachcanworkinanysize
company,evenaverylargeenterprise,inanysectororindustry.
2.Entrepreneurshipismanagement.Astartupisaninstitution,
notjustaproduct,andsoitrequiresanewkindofmanagement
specicalygearedtoitscontextofextremeuncertainty.Infact,asI
wilarguelater,Ibelieve“entrepreneur”shouldbeconsidereda
wilarguelater,Ibelieve“entrepreneur”shouldbe
considereda
jobtitleinalmoderncompaniesthatdependoninnovationfor
theirfuturegrowth.
3.Validatedlearning.Startupsexistnotjusttomakestu,make
money,orevenservecustomers.Theyexistto
learnhowtobuilda
sustainablebusiness.Thislearningcanbevalidatedscienticalyby
runningfrequentexperimentsthatalowentrepreneurstotesteach
elementoftheirvision.
4.Build-Measure-Learn.Thefundamentalactivityofa
startupis
toturnideasintoproducts,measurehowcustomersrespond,and
thenlearnwhethertopivotorpersevere.Alsuccessfulstartup
processesshouldbegearedtoacceleratethatfeedbackloop.
5.Innovationaccounting.To
improveentrepreneurialoutcomes
andholdinnovatorsaccountable,weneedtofocusontheboring
stu:howtomeasureprogress,howtosetupmilestones,andhow
toprioritizework.Thisrequiresanewkindofaccountingdesigned
forstartups—andthepeoplewhoholdthemaccountable.
WhyStartupsFail
Whyarestartupsfailingsobadlyeverywherewelook?
Therstproblemisthealureofagoodplan,asolidstrategy,
andthoroughmarketresearch.Inearliereras,
thesethingswere
indicatorsoflikelysuccess.Theoverwhelmingtemptationisto
applythemtostartupstoo,butthisdoesn’twork,becausestartups
operatewithtoomuchuncertainty.Startupsdonotyetknowwho
theircustomerisorwhattheirproductshouldbe.Astheworld
becomesmoreuncertain,itgetsharderandhardertopredictthe
future.Theoldmanagementmethodsarenotuptothetask.
Planningandforecastingareonlyaccuratewhenbasedon
along,
stableoperatinghistoryandarelativelystaticenvironment.Startups
stableoperatinghistoryandarelativelystaticenvironment.Startups
haveneither.
Thesecondproblemisthatafterseeingtraditional
management
failtosolvethisproblem,someentrepreneursandinvestorshave
thrownuptheirhandsandadoptedthe“JustDoIt”schoolof
startups.Thisschoolbelievesthatifmanagementistheproblem,
chaosistheanswer.Unfortunately,asIcanatestrsthand,this
doesn’tworkeither.
Itmayseemcounterintuitivetothinkthatsomethingas
disruptive,innovative,andchaoticasastartupcanbemanagedor,
tobeaccurate,mustbe
managed.Mostpeoplethinkofprocessand
managementasboringanddul,whereasstartupsaredynamicand
exciting.Butwhatisactualyexcitingistoseestartupssucceedand
changetheworld.Thepassion,energy,andvisionthatpeoplebring
tothesenewventuresareresourcestooprecioustowaste.Wecan—
andmust—dobeter.Thisbookisabouthow.
andmust—dobeter.Thisbookisabouthow.
HOWTHISBOOKISORGANIZED
Thisbookisdividedinto
threeparts:“Vision,”“Steer,”and
“Accelerate.”
“Vision”makesthecaseforanewdisciplineofentrepreneurial
management.Iidentifywhoisanentrepreneur,deneastartup,
andarticulateanewwayfor
startupstogaugeiftheyaremaking
progress,caledvalidatedlearning.Toachievethatlearning,we’l
seethatstartups—inagarageorinsideanenterprise—canuse
scienticexperimentationtodiscoverhowtobuildasustainable
business.
“Steer”divesintotheLeanStartupmethodindetail,showingone
majorturnthroughthecoreBuild-Measure-Learnfeedbackloop.
Beginningwithleap-of-faithassumptionsthatcryoutforrigorous
testing,you’llearnhowtobuildaminimumviableproducttotest
thoseassumptions,anewaccountingsystemforevaluatingwhether
you’remakingprogress,andamethodfordecidingwhetherto
pivot(changingcoursewithonefootanchoredtothe
ground)or
persevere.
In“Accelerate,”we’lexploretechniquesthatenableLean
StartupstospeedthroughtheBuild-Measure-Learnfeedbackloop
asquicklyaspossible,evenastheyscale.We’lexplorelean
manufacturingconceptsthatareapplicabletostartups,too,suchas
thepowerofsmalbatches.We’lalsodiscussorganizationaldesign,
howproductsgrow,andhowtoapplyLeanStartupprinciples
beyondtheproverbialgarage,eveninsidetheworld’s
largest
companies.
MANAGEMENT’SSECONDCENTURY
Asasociety,wehaveaprovensetoftechniquesformanagingbig
companiesandweknowthebestpracticesforbuildingphysical
products.Butwhenitcomestostartupsandinnovation,wearestil
shootinginthedark.Wearerelyingonvision,chasingthe“great
shootinginthedark.Wearerelyingonvision,chasingthe“great
men”whocanmakemagichappen,ortryingtoanalyze
ournew
productstodeath.Thesearenewproblems,bornofthesuccessof
managementinthetwentiethcentury.
Thisbookatemptstoputentrepreneurshipandinnovationona
rigorousfooting.Weareat
thedawnofmanagement’ssecond
century.Itisourchalengetodosomethinggreatwiththe
opportunitywehavebeengiven.TheLeanStartupmovementseeks
toensurethatthoseofuswholongtobuildthenextbigthingwil
havethetoolsweneedtochangetheworld.
PartOne
PartOne
VISION
1START
ENTREPRENEURIALMANAGEMENT
Buildingastartupisanexerciseininstitutionbuilding;thus,it
necessarilyinvolvesmanagement.Thisoftencomesasasurprise
toaspiringentrepreneurs,becausetheirassociationswiththese
twowordsaresodiametricalyopposed.Entrepreneursarerightly
waryofimplementingtraditionalmanagement
practicesearlyonin
astartup,afraidthattheywilinvitebureaucracyorstiflecreativity.
Entrepreneurshavebeentryingtotthesquarepegoftheir
uniqueproblemsintotheroundholeofgeneralmanagementfor
decades.Asaresult,manyentrepreneurstakea“justdoit”atitude,
avoidingalformsofmanagement,process,anddiscipline.
Unfortunately,thisapproachleadstochaosmoreoftenthanitdoes
tosuccess.Ishouldknow:myrststartupfailureswere
alofthis
kind.
Thetremendoussuccessofgeneralmanagementoverthelast
centuryhasprovidedunprecedentedmaterialabundance,butthose
managementprinciplesareilsuitedtohandlethechaos
and
uncertaintythatstartupsmustface.
Ibelievethatentrepreneurshiprequiresamanagerialdisciplineto
harnesstheentrepreneurialopportunitywehavebeengiven.
Therearemoreentrepreneurs
operatingtodaythanatany
previoustimeinhistory.Thishasbeenmadepossiblebydramatic
previoustimeinhistory.Thishasbeenmadepossiblebydramatic
changesintheglobaleconomy.Tocitebutoneexample,oneoften
hearscommentatorslamentthelossofmanufacturingjobsinthe
UnitedStatesovertheprevioustwodecades,butonerarelyhears
aboutacorrespondinglossofmanufacturingcapability.That’s
becausetotalmanufacturingoutputintheUnitedStatesis
increasing(by15percentinthelastdecade)evenasjobscontinue
tobelost(seethechartsbelow).Ineect,thehugeproductivity
increasesmadepossiblebymodernmanagementandtechnology
havecreatedmoreproductivecapacitythanrmsknowwhat
todo
with.1
Wearelivingthroughanunprecedentedworldwide
entrepreneurialrenaissance,butthisopportunityislacedwithperil.
Becausewelackacoherentmanagementparadigmfornew
innovativeventures,we’rethrowingourexcesscapacityaround
withwildabandon.Despitethislackofrigor,wearendingsome
waystomakemoney,butforeverysuccesstherearefartoomany
failures:productspuledfromshelvesmereweeks
afterbeing
launched,high-prolestartupslaudedinthepressandforgotena
fewmonthslater,andnewproductsthatwindupbeingusedby
nobody.Whatmakesthesefailuresparticularlypainfulisnotjust
theeconomicdamagedonetoindividualemployees,companies,
andinvestors;theyarealsoacolossalwasteofourcivilization’s
mostpreciousresource:thetime,passion,andskilofitspeople.
TheLeanStartupmovementisdedicatedtopreventing
these
failures.
THEROOTSOFTHELEANSTARTUP
TheLeanStartuptakesits
namefromtheleanmanufacturing
revolutionthatTaichiOhnoandShigeoShingoarecreditedwith
developingatToyota.Leanthinkingisradicalyalteringtheway
supplychainsandproductionsystemsarerun.Amongitstenetsare
drawingontheknowledgeandcreativityofindividualworkers,the
shrinkingofbatchsizes,just-in-timeproductionandinventory
control,andanaccelerationofcycletimes.Ittaughttheworldthe
dierencebetweenvalue-creatingactivitiesandwaste
andshowed
howtobuildqualityintoproductsfromtheinsideout.
TheLeanStartupadaptstheseideastothecontextof
entrepreneurship,proposingthatentrepreneursjudgetheirprogress
dierentlyfromthewayotherkindsofventuresdo.
Progressin
dierentlyfromthewayotherkindsofventuresdo.Progressin
manufacturingismeasuredbytheproductionofhigh-quality
physicalgoods.Aswe’lseeinChapter3,theLeanStartupusesa
dierentunitofprogress,caledvalidatedlearning.Withscientific
learningasouryardstick,wecandiscoverandeliminatethesources
ofwastethatareplaguingentrepreneurship.
Acomprehensivetheoryofentrepreneurshipshouldaddressal
thefunctionsofanearly-stageventure:visionandconcept,product
development,marketingandsales,scalingup,partnershipsand
distribution,andstructureandorganizationaldesign.Ithasto
provideamethodformeasuringprogressinthe
contextofextreme
uncertainty.Itcangiveentrepreneursclearguidanceonhowto
makethemanytrade-odecisionstheyface:whetherandwhento
investinprocess;formulating,planning,andcreatinginfrastructure;
whentogoitaloneandwhentopartner;whentorespondto
feedbackandwhentostickwithvision;andhowandwhento
investinscalingthebusiness.Mostofal,itmustalow
entrepreneurstomaketestablepredictions.
Forexample,considertherecommendationthatyoubuildcross-
functionalteamsandholdthemaccountabletowhatwecal
learningmilestonesinsteadoforganizingyourcompanyintostrict
functionaldepartments(marketing,sales,
informationtechnology,
humanresources,etc.)thatholdpeopleaccountableforperforming
welintheirspecializedareas(seeChapter7).Perhapsyouagree
withthisrecommendation,orperhapsyouareskeptical.Either
way,ifyoudecidetoimplementit,Ipredictthatyoupretyquickly
wilgetfeedbackfromyourteamsthatthenewprocessisreducing
theirproductivity.Theywilasktogobacktotheoldwayof
working,inwhichtheyhadtheopportunityto“staye
cient”by
workinginlargerbatchesandpassingworkbetweendepartments.
It’ssafetopredictthisresult,andnotjustbecauseIhaveseenit
manytimesinthecompaniesIworkwith.Itisastraightforward
predictionoftheLeanStartuptheoryitself.Whenpeopleareused
toevaluatingtheirproductivitylocaly,theyfeelthatagooddayis
oneinwhichtheydidtheirjobwelalday.WhenIworkedasa
programmer,thatmeanteightstraighthoursof
programming
withoutinterruption.Thatwasagoodday.Incontrast,ifIwas
withoutinterruption.Thatwasagoodday.Incontrast,ifIwas
interruptedwithquestions,process,or—heavenforbid—meetings,I
feltbad.WhatdidIrealyaccomplishthatday?Codeandproduct
featuresweretangibletome;Icouldseethem,understandthem,
andshowthemof.Learning,bycontrast,isfrustratinglyintangible.
TheLeanStartupaskspeopletostartmeasuringtheir
productivitydierently.Becausestartupsoftenaccidentalybuild
somethingnobodywants,itdoesn’tmatermuchiftheydoiton
timeandonbudget.Thegoalofastartupistogureouttheright
thingtobuild—thethingcustomerswantandwilpay
for—as
quicklyaspossible.Inotherwords,theLeanStartupisanewway
oflookingatthedevelopmentofinnovativenewproductsthat
emphasizesfastiterationandcustomerinsight,ahugevision,and
greatambition,alatthesametime.
HenryFordisoneofthemostsuccessfulandcelebrated
entrepreneursofaltime.Sincetheideaofmanagementhasbeen
boundupwiththehistoryoftheautomobilesinceitsrstdays,I
believeitistingtousetheautomobileasametaphorfora
startup.
Aninternalcombustionautomobileispoweredbytwoimportant
andverydierentfeedbackloops.Therstfeedbackloopisdeep
insidetheengine.BeforeHenryFordwasafamousCEO,hewasan
engineer.Hespenthisdaysandnightstinkeringinhisgaragewith
theprecisemechanicsofgetingtheenginecylinderstomove.Each
tinyexplosionwithinthecylinderprovidesthemotive
forcetoturn
thewheelsbutalsodrivestheignitionofthenextexplosion.Unless
thetimingofthisfeedbackloopismanagedprecisely,theengine
wilsputerandbreakdown.
StartupshaveasimilarenginethatIcaltheengineof
growth.
Themarketsandcustomersforstartupsarediverse:atoycompany,
aconsultingrm,andamanufacturingplantmaynotseemlike
theyhavemuchincommon,but,aswe’lsee,theyoperatewiththe
sameengineofgrowth.
Everynewversionofaproduct,everynewfeature,andevery
Everynewversionofaproduct,everynewfeature,andevery
newmarketingprogramisanatempttoimprovethisengineof
growth.LikeHenryFord’stinkeringinhisgarage,notalofthese
changesturnouttobeimprovements.Newproductdevelopment
happensintsandstarts.Muchofthetimeinastartup’slifeis
spenttuningtheenginebymakingimprovementsin
product,
marketing,oroperations.
Thesecondimportantfeedbackloopinanautomobileis
betweenthedriverandthesteeringwheel.Thisfeedbackisso
immediateandautomaticthatweoftendon’tthink
aboutit,butit
issteeringthatdierentiatesdrivingfrommostotherformsof
transportation.Ifyouhaveadailycommute,youprobablyknow
theroutesowelthatyourhandsseemtosteeryouthereontheir
ownaccord.Wecanpracticalydrivetherouteinoursleep.YetifI
askedyoutocloseyoureyesandwritedownexactlyhowtogetto
youroce—notthestreetdirectionsbuteveryactionyouneedto
take,everypushofhandonwheelandfootonpedals—
you’dnd
itimpossible.Thechoreographyofdrivingisincrediblycomplex
whenoneslowsdowntothinkaboutit.
Bycontrast,arocketshiprequiresjustthiskindofin-advance
calibration.Itmustbe
launchedwiththemostpreciseinstructions
onwhattodo:everythrust,everyringofabooster,andevery
changeindirection.Thetiniesterroratthepointoflaunchcould
yieldcatastrophicresultsthousandsofmileslater.
Unfortunately,toomanystartupbusinessplanslookmorelike
theyareplanningtolauncharocketshipthandriveacar.They
prescribethestepstotakeandtheresultstoexpectinexcruciating
detail,andasinplanningtolauncharocket,theyareset
upinsuch
awaythateventinyerrorsinassumptionscanleadtocatastrophic
outcomes.
OnecompanyIworkedwithhadthemisfortuneofforecasting
signicantcustomeradoption—inthemilions—foroneof
itsnew
products.Poweredbyasplashylaunch,thecompanysuccessfuly
executeditsplan.Unfortunately,customersdidnotocktothe
productingreatnumbers.Evenworse,thecompanyhadinvestedin
massiveinfrastructure,hiring,andsupporttohandletheinuxof
customersitexpected.Whenthecustomersfailedtomaterialize,the
customersitexpected.Whenthecustomersfailedtomaterialize,the
companyhadcommiteditselfsocompletelythatthey
couldnot
adaptintime.Theyhad“achievedfailure”—successfuly,faithfuly,
andrigorouslyexecutingaplanthatturnedouttohavebeenuterly
flawed.
TheLeanStartupmethod,incontrast,isdesignedtoteach
you
howtodriveastartup.Insteadofmakingcomplexplansthatare
basedonalotofassumptions,youcanmakeconstantadjustments
withasteeringwheelcaledtheBuild-Measure-Learnfeedback
loop.Throughthisprocessofsteering,wecanlearnwhenandifit’s
timetomakeasharpturncaledapivotorwhetherweshould
perseverealongourcurrentpath.Oncewehaveanenginethat’s
revvedup,theLeanStartupoersmethodstoscaleand
growthe
businesswithmaximumacceleration.
Throughouttheprocessofdriving,youalwayshaveaclearidea
ofwhereyou’regoing.Ifyou’recommutingtowork,youdon’tgive
upbecausethere’sadetourin
theroadoryoumadeawrongturn.
Youremainthoroughlyfocusedongetingtoyourdestination.
Startupsalsohaveatruenorth,adestinationinmind:creatinga
thrivingandworld-changingbusiness.Icalthatastartup’svision.
Toachievethatvision,startupsemployastrategy,whichincludesa
businessmodel,aproductroadmap,apointofviewaboutpartners
andcompetitors,andideasaboutwhothecustomerwilbe.The
productistheendresultofthisstrategy(seethecharton
thispage).
Productschangeconstantlythroughtheprocessofoptimization,
whatIcaltuningtheengine.
Lessfrequently,thestrategymayhave
tochange(caledapivot).However,theoverarchingvisionrarely
changes.Entrepreneursarecommitedtoseeingthestartupthrough
tothatdestination.Everysetbackisanopportunityforlearning
howtogetwheretheywanttogo(seethechartbelow).
Inreallife,astartupisaportfolioofactivities.Alotishappening
simultaneously:theengineisrunning,acquiringnewcustomersand
servingexistingones;wearetuning,tryingtoimproveour
product,
marketing,andoperations;andwearesteering,decidingifand
whentopivot.Thechalengeofentrepreneurshipistobalanceal
theseactivities.Eventhesmaleststartupfacesthechalengeof
supportingexistingcustomerswhiletryingtoinnovate.Eventhe
mostestablishedcompanyfacestheimperativetoinvestin
innovationlestitbecomeobsolete.Ascompaniesgrow,what
changesisthemixoftheseactivitiesinthecompany’s
portfolioof
work.
Entrepreneurshipismanagement.Andyet,imagineamodern
managerwhoistaskedwithbuildinganewproductinthecontext
ofanestablishedcompany.Imaginethatshegoesbackto
her
company’schiefnancialocer(CFO)ayearlaterandsays,“We
havefailedtomeetthegrowthtargetswepredicted.Infact,we
havealmostnonewcustomersandnonewrevenue.However,we
havelearnedanincredibleamountandareonthecuspofa
breakthroughnewlineofbusiness.Alweneedisanotheryear.”
Mostofthetime,thiswouldbethelastreportthisintrapreneur
wouldgiveheremployer.Thereasonisthatingeneral
management,afailuretodeliverresultsisduetoeitherafailureto
planadequatelyorafailuretoexecuteproperly.Bothare
signicantlapses,yetnewproductdevelopmentinourmodern
economyroutinelyrequiresexactlythiskindoffailureontheway
togreatness.IntheLeanStartupmovement,wehavecometo
realizethattheseinternalinnovatorsareactualyentrepreneurs,too,
andthatentrepreneurialmanagementcanhelpthemsucceed;thisis
thesubjectofthenextchapter.
2DEFINE
WHO,EXACTLY,ISANENTREPRENEUR?
AsItraveltheworldtalkingabouttheLeanStartup,I’m
consistentlysurprisedthatImeetpeopleintheaudiencewho
seemoutofplace.Inadditiontothemoretraditional
startup
entrepreneursImeet,thesepeoplearegeneralmanagers,mostly
workinginverylargecompanies,whoaretaskedwithcreatingnew
venturesorproductinnovations.Theyareadeptatorganizational
politics:theyknowhowtoformautonomousdivisionswith
separateprotandlossstatements(P&Ls)andcanshield
controversialteamsfromcorporatemeddling.Thebiggestsurprise
isthattheyarevisionaries.LikethestartupfoundersI
haveworked
withforyears,theycanseethefutureoftheirindustriesandare
preparedtotakeboldriskstoseekoutnewandinnovative
solutionstotheproblemstheircompaniesface.
Mark,forexample,isamanagerforanextremely
largecompany
whocametooneofmylectures.Heistheleaderofadivisionthat
recentlyhadbeencharteredtobringhiscompanyintothetwenty-
rstcenturybybuildinganewsuiteofproductsdesignedtotake
advantageoftheInternet.Whenhecametotalktomeafterward,I
startedtogivehimthestandardadviceabouthowtocreate
innovationteamsinsidebigcompanies,andhestoppedmein
midstream:“Yeah,I’vereadTheInnovator’sDilemma.1
I’vegotthat
altakencareof.”Hewasalong-termemployeeofthecompany
andasuccessfulmanagertoboot,somanaginginternalpoliticswas
andasuccessfulmanagertoboot,somanaginginternalpoliticswas
theleastofhisproblems.Ishouldhaveknown;hissuccesswasa
testamenttohisabilitytonavigatethecompany’scorporate
policies,personnel,andprocessestogetthingsdone.
Next,Itriedtogivehimsomeadviceaboutthefuture,aboutcool
newhighlyleveragedproductdevelopmenttechnologies.He
interruptedmeagain:“Right.IknowalabouttheInternet,andI
haveavisionforhowourcompanyneedstoadapttoitordie.”
Markhasaltheentrepreneurialprerequisites
nailed—proper
teamstructure,goodpersonnel,astrongvisionforthefuture,and
anappetiteforrisktaking—andsoitnalyoccurredtometoask
whyhewascomingtomeforadvice.Hesaid,“It’sasifwehaveal
oftherawmaterials:kindling,wood,paper,int,evensome
sparks.Butwhere’sthere?”Thetheoriesofmanagementthat
Markhadstudiedtreatinnovationlikea“blackbox”byfocusingon
thestructurescompaniesneedtoputinplacetoform
internal
startupteams.ButMarkfoundhimselfworkinginsidetheblack
box—andinneedofguidance.
WhatMarkwasmissingwasaprocessforconvertingtheraw
materialsofinnovationinto
real-worldbreakthroughsuccesses.
Onceateamissetup,whatshoulditdo?Whatprocessshouldit
use?Howshoulditbeheldaccountabletoperformance
milestones?ThesearequestionstheLeanStartupmethodologyis
designedtoanswer.
Mypoint?MarkisanentrepreneurjustlikeaSiliconValeyhigh-
techfounderwithagaragestartup.Heneedstheprinciplesofthe
LeanStartupjustasmuchasthefolksIthoughtofasclassic
entrepreneursdo.
Entrepreneurswhooperateinsideanestablishedorganization
sometimesarecaled“intrapreneurs”becauseofthespecial
circumstancesthatatendbuildingastartupwithinalarger
company.AsIhaveappliedLeanStartupideasinanever-widening
varietyofcompaniesandindustries,Ihavecometobelievethat
intrapreneurshavemuchmoreincommonwiththerestofthe
communityofentrepreneursthanmostpeoplebelieve.
Thus,when
Iusethetermentrepreneur,Iamreferringtothewholestartup
ecosystemregardlessofcompanysize,sector,orstageof
ecosystemregardlessofcompanysize,sector,orstageof
development.
Thisbookisforentrepreneursofalstripes:fromyoung
visionarieswithlitlebackingbutgreatideastoseasoned
visionarieswithinlargercompaniessuchasMark—andthepeople
whoholdthemaccountable.
IFI’MANENTREPRENEUR,WHAT’SASTARTUP?
TheLeanStartupisasetofpracticesforhelpingentrepreneurs
increasetheiroddsofbuildingasuccessfulstartup.Tosettherecord
straight,it’simportanttodefinewhatastartupis:
Astartupisahumaninstitutiondesignedtocreateanew
productorserviceunderconditionsofextremeuncertainty.
I’vecometorealizethatthemostimportantpartofthis
denitioniswhatitomits.Itsaysnothingaboutsizeofthe
company,theindustry,orthesectoroftheeconomy.Anyonewhois
creatinganewproductorbusinessunderconditionsofextreme
uncertaintyisanentrepreneurwhetherheorsheknowsitornot
andwhetherworkinginagovernmentagency,a
venture-backed
company,anonprot,oradecidedlyfor-protcompanywith
financialinvestors.
Let’stakealookateachofthepieces.Thewordinstitution
connotesbureaucracy,process,evenlethargy.How
canthatbepart
ofastartup?Yetsuccessfulstartupsarefulofactivitiesassociated
withbuildinganinstitution:hiringcreativeemployees,coordinating
theiractivities,andcreatingacompanyculturethatdeliversresults.
Weoftenlosesightofthefactthatastartupisnotjustabouta
product,atechnologicalbreakthrough,orevenabriliantidea.A
startupisgreaterthanthesumofitsparts;itisanacutelyhuman
enterprise.
Thefactthatastartup’sproductorserviceisanewinnovationis
alsoanessentialpartofthedenitionandatrickyparttoo.Iprefer
tousethebroadestdenitionofproduct,onethatencompassesany
tousethebroadestdenitionofproduct,onethat
encompassesany
sourceofvalueforthepeoplewhobecomecustomers.Anything
thosecustomersexperiencefromtheirinteractionwithacompany
shouldbeconsideredpartofthatcompany’sproduct.Thisistrueof
agrocerystore,ane-commercewebsite,aconsultingservice,anda
nonprotsocialserviceagency.Ineverycase,theorganizationis
dedicatedtouncoveringanewsourceofvalueforcustomersand
caresabouttheimpactofitsproductonthosecustomers.
It’salsoimportantthatthewordinnovationbeunderstood
broadly.Startupsusemanykindsofinnovation:novelscientic
discoveries,repurposinganexistingtechnologyforanewuse,
devisinganewbusinessmodelthatunlocksvaluethat
washidden,
orsimplybringingaproductorservicetoanewlocationora
previouslyunderservedsetofcustomers.Inalthesecases,
innovationisattheheartofthecompany’ssuccess.
Thereisonemoreimportantpartofthisdefinition:the
contextin
whichtheinnovationhappens.Mostbusinesses—largeandsmal
alike—areexcludedfromthiscontext.Startupsaredesignedto
confrontsituationsofextremeuncertainty.Toopenupanew
businessthatisanexactcloneofanexistingbusinessaltheway
downtothebusinessmodel,pricing,targetcustomer,andproduct
maybeanatractiveeconomicinvestment,butitisnotastartup
becauseitssuccessdependsonlyonexecution—somuch
sothatthis
successcanbemodeledwithhighaccuracy.(Thisiswhysomany
smalbusinessescanbenancedwithsimplebankloans;thelevel
ofriskanduncertaintyisunderstoodwelenoughthataloanocer
canassessitsprospects.)
Mosttoolsfromgeneralmanagementarenotdesignedtoourish
intheharshsoilofextremeuncertaintyinwhichstartupsthrive.
Thefutureisunpredictable,customersfaceagrowingarrayof
alternatives,andthepaceofchangeiseverincreasing.Yetmost
startups—ingaragesandenterprisesalike—stilaremanagedby
usingstandardforecasts,productmilestones,anddetailedbusiness
plans.
THESNAPTAXSTORY
In2009,astartupdecidedtotrysomethingrealyaudacious.They
wantedtoliberatetaxpayersfromexpensivetaxstoresby
automatingtheprocessofcolectinginformationtypicalyfoundon
W-2forms(theend-of-year
statementthatmostemployeesreceive
fromtheiremployerthatsummarizestheirtaxablewagesforthe
year).Thestartupquicklyranintodiculties.Eventhoughmany
consumershadaccesstoaprinter/scannerintheirhomeoroce,
fewknewhowtousethosedevices.Afternumerousconversations
withpotentialcustomers,theteamlitupontheideaofhaving
customerstakephotographsoftheformsdirectlyfromtheircel
phone.Intheprocessoftestingthisconcept,
customersasked
somethingunexpected:woulditbepossibletonishthewholetax
returnrightonthephoneitself?
Thatwasnotaneasytask.Traditionaltaxpreparationrequires
consumerstowadethrough
hundredsofquestions,manyforms,and
alotofpaperwork.Thisstartuptriedsomethingnovelbydeciding
toshipanearlyversionofitsproductthatcoulddomuchlessthan
acompletetaxpackage.Theinitialversionworkedonlyfor
consumerswithaverysimplereturntole,anditworkedonlyin
California.
Insteadofhavingconsumersloutacomplexform,they
alowedthecustomerstousethephone’scameratotakeapicture
oftheirW-2forms.From
thatsinglepicture,thecompany
developedthetechnologytocompileandlemostofthe1040EZ
taxreturn.Comparedwiththedrudgeryoftraditionaltaxling,the
newproduct—caledSnapTax—providesamagicalexperience.
Fromitsmodestbeginning,SnapTaxgrewintoasignicantstartup
successstory.Itsnationwidelaunchin2011showedthatcustomers
lovedit,tothetuneofmorethan350,000downloadsintherst
threeweeks.
Thisisthekindofamazinginnovationyou’dexpectfromanew
startup.
However,thenameofthiscompanymaysurpriseyou.SnapTax
wasdevelopedbyIntuit,America’slargestproducerofnance,tax,
andaccountingtoolsforindividualsandsmalbusinesses.With
andaccountingtoolsforindividualsandsmalbusinesses.With
morethan7,700employeesandannualrevenuesinthebilions,
Intuitisnotatypicalstartup.2
TheteamthatbuiltSnapTaxdoesn’tlookmuchlikethe
archetypalimageofentrepreneurseither.Theydon’tworkina
garageoreatramennoodles.Theircompanydoesn’tlackfor
resources.Theyarepaidafulsalaryandbenefits.Theycomeintoa
regularoficeeveryday.Yettheyareentrepreneurs.
Storieslikethisonearenotnearlyascommoninsidelarge
corporationsastheyshouldbe.Afteral,SnapTaxcompetesdirectly
withoneofIntuit’sagshipproducts:thefulyfeaturedTurboTax
desktopsoftware.Usualy,companieslikeIntuitfalintothetrap
describedinClaytonChristensten’sTheInnovator’sDilemma:they
areverygoodatcreatingincrementalimprovementstoexisting
productsandservingexistingcustomers,which
Christensencaled
sustaininginnovation,butstruggletocreatebreakthroughnew
products—disruptiveinnovation—thatcancreatenewsustainable
sourcesofgrowth.
OneremarkablepartoftheSnapTaxstoryiswhatthe
team
leaderssaidwhenIaskedthemtoaccountfortheirunlikelysuccess.
Didtheyhiresuperstarentrepreneursfromoutsidethecompany?
No,theyassembledateamfromwithinIntuit.Didtheyface
constantmeddlingfromseniormanagement,whichisthebaneof
innovationteamsinmanycompanies?No,theirexecutivesponsors
createdan“islandoffreedom”wheretheycouldexperimentas
necessary.Didtheyhaveahugeteam,alargebudget,
andlotsof
marketingdolars?Nope,theystartedwithateamoffive.
WhatalowedtheSnapTaxteamtoinnovatewasnottheirgenes,
destiny,orastrologicalsignsbutaprocessdeliberatelyfacilitatedby
Intuit’sseniormanagement.Innovationisabotoms-up,
decentralized,andunpredictablething,butthatdoesn’tmeanit
cannotbemanaged.Itcan,buttodosorequiresanew
managementdiscipline,onethatneedstobemasterednotjustby
practicingentrepreneursseekingtobuildthenextbigthingbutalso
bythepeoplewhosupportthem,nurturethem,andholdthem
accountable.Inotherwords,cultivatingentrepreneurshipisthe
accountable.Inotherwords,cultivatingentrepreneurship
isthe
responsibilityofseniormanagement.Today,acuting-edge
companysuchasIntuitcanpointtosuccessstorieslikeSnapTax
becauseithasrecognizedtheneedforanewmanagement
paradigm.Thisisa
realizationthatwasyearsinthemaking.3
ASEVEN-THOUSAND-PERSONLEANSTARTUP
In1983,Intuit’sfounder,thelegendaryentrepreneurScotCook,
hadtheradicalnotion(withcofounderTomProulx)thatpersonal
accountingshouldhappenbycomputer.Theirsuccesswasfarfrom
inevitable;theyfacednumerouscompetitors,anuncertainfuture,
andaninitialytinymarket.Adecadelater,thecompanywent
publicandsubsequentlyfendedowel-publicizedat
acksfrom
largerincumbents,includingthesoftwarebehemothMicrosoft.
PartlywiththehelpoffamedventurecapitalistJohnDoerr,Intuit
becameafulydiversiedenterprise,amemberoftheFortune
1000thatnowprovidesdozensofmarket-leadingproductsacross
itsmajordivisions.
Thisisthekindofentrepreneurialsuccesswe’reusedtohearing
about:aragtagteamofunderdogswhoeventualyachievefame,
acclaim,andsignificantriches.
Flash-forwardto2002.Cookwasfrustrated.Hehadjusttabulated
tenyearsofdataonalofIntuit’snewproductintroductionsand
hadconcludedthatthecompanywasgetingameaslyreturnonits
massiveinvestments.Simplyput,toomanyofitsnewproducts
werefailing.Bytraditionalstandards,Intuitisanextremelywel-
managedcompany,butasScotdugintotherootcausesofthose
failures,hecametoadicultconclusion:theprevailing
managementparadigmheandhiscompanyhadbeenpracticing
wasinadequatetotheproblemofcontinuousinnovationinthe
moderneconomy.
Byfal2009,CookhadbeenworkingtochangeIntuit’s
managementculturefor
severalyears.Hecameacrossmyearly
workontheLeanStartupandaskedmetogiveatalkatIntuit.In
SiliconValeythisisnotthekindofinvitationyouturndown.I
SiliconValeythisisnotthekindofinvitationyouturndown.I
admitIwascurious.IwasstilatthebeginningofmyLeanStartup
journeyanddidn’thavemuchappreciationforthechalengesfaced
byaFortune1000companylikehis.
MyconversationswithCookandIntuitchiefexecutiveocer
(CEO)BradSmithweremyinitiationintothethinkingofmodern
generalmanagers,whostrugglewithentrepreneurshipeverybitas
muchasdoventurecapitalistsandfoundersinagarage.Tocombat
thesechalenges,ScotandBradaregoingbackto
Intuit’sroots.
Theyareworkingtobuildentrepreneurshipandrisktakingintoal
theirdivisions.
Forexample,consideroneofIntuit’sagshipproducts.Because
TurboTaxdoesmostofitssalesaroundtaxseasoninthe
United
States,itusedtohaveanextremelyconservativeculture.Overthe
courseoftheyear,themarketingandproductteamswould
conceiveonemajorinitiativethatwouldberoledoutjustintime
fortaxseason.Nowtheytestovervehundreddierentchangesin
atwo-and-a-half-monthtaxseason.They’rerunninguptoseventy
dierenttestsperweek.Theteamcanmakeachangeliveonits
websiteonThursday,runitovertheweekend,readthe
resultson
Monday,andcometoconclusionsstartingTuesday;thenthey
rebuildnewtestsonThursdayandlaunchthenextsetonThursday
night.
AsScotputit,“Boy,theamountoflearningtheyget
isjust
immensenow.Andwhatitdoesisdevelopentrepreneurs,because
whenyouhaveonlyonetest,youdon’thaveentrepreneurs,you
havepoliticians,becauseyouhavetosel.Outofahundredgood
ideas,you’vegottoselyouridea.Soyoubuildupasocietyof
politiciansandsalespeople.Whenyouhavevehundredtests
you’rerunning,theneverybody’sideascanrun.Andthenyoucreate
entrepreneurswhorunandlearnandcanretestand
relearnas
opposedtoasocietyofpoliticians.Sowe’retryingtodrivethat
throughoutourorganization,usingexampleswhichhavenothingto
dowithhightech,likethewebsiteexample.Everybusinesstoday
hasawebsite.Youdon’thavetobehightechtousefast-cycle
testing.”
Thiskindofchangeishard.Afteral,thecompanyhasa
Thiskindofchangeishard.Afteral,thecompanyhasa
signicantnumberofexistingcustomerswhocontinueto
demand
exceptionalserviceandinvestorswhoexpectsteady,growing
returns.
Scotsays,
Itgoesagainstthegrainofwhatpeoplehavebeentaughtin
businessandwhatleadershavebeentaught.Theproblem
isn’twiththeteamsortheentrepreneurs.Theylovethe
chancetoquicklygettheirbabyoutintothemarket.They
lovethechancetohavethecustomervoteinsteadofthe
suitsvoting.Therealissueiswiththeleadersandthe
middlemanagers.Therearemanybusinessleaderswho
havebeensuccessfulbecauseofanalysis.Theythinkthey’re
analysts,andtheirjobistodogreatplanningandanalyzing
andhaveaplan.
Theamountoftimeacompanycancountonholdingonto
marketleadershiptoexploititsearlierinnovationsisshrinking,and
thiscreatesanimperativeforeventhemostentrenchedcompanies
toinvestininnovation.Infact,Ibelieveacompany’sonly
sustainablepathtolong-termeconomicgrowthistobuildan
“innovationfactory”thatusesLeanStartuptechniquestocreate
disruptiveinnovationsonacontinuousbasis.Inother
words,
establishedcompaniesneedtogureouthowtoaccomplishwhat
ScotCookdidin1983,butonanindustrialscaleandwithan
establishedcohortofmanagerssteepedintraditionalmanagement
culture.
Everthemaverick,Cookaskedmetoputtheseideastothetest,
andsoIgaveatalkthatwassimulcasttoalseventhousand–plus
IntuitemployeesduringwhichIexplainedthetheoryoftheLean
Startup,repeatingmydenition:anorganizationdesignedtocreate
newproductsandservicesunderconditionsofextremeuncertainty.
Whathappenednextisetchedinmymemory.CEOBradSmith
hadbeensitingnexttomeasIspoke.WhenIwasdone,he
gotup
andsaidbeforealofIntuit’semployees,“Folks,listenup.You
andsaidbeforealofIntuit’semployees,“Folks,listenup.You
heardEric’sdefinitionofastartup.Ithasthreeparts,andwehereat
Intuitmatchalthreepartsofthatdefinition.”
ScotandBradareleaderswhorealizethatsomethingnewis
neededinmanagementthinking.Intuitisproofthatthiskindof
thinkingcanworkinestablishedcompanies.Bradexplainedtome
howtheyholdthemselvesaccountablefortheirnewinnovation
eortsbymeasuringtwothings:thenumberofcustomersusing
productsthatdidn’texistthreeyearsagoandthepercentageof
revenuecomingfromoferingsthatdidnotexistthree
yearsago.
Undertheoldmodel,ittookanaverageof5.5yearsfora
successfulnewproducttostartgenerating$50milioninrevenue.
Bradexplainedtome,“We’vegenerated$50milioninoerings
thatdidnotexisttwelve
monthsagointhelastyear.Nowit’snot
oneparticularoering.It’sacombinationofawholebunchof
innovationhappening,butthat’sthekindofstuthat’screating
someenergyforus,thatwethinkwecantrulyshort-circuitthe
rampbykilingthingsthatdon’tmakesensefastanddoubling
downontheonesthatdo.”ForacompanyaslargeasIntuit,these
aremodestresultsandearlydays.Theyhavedecadesoflegacy
systemsandlegacythinkingtoovercome.However,their
leadership
inadoptingentrepreneurialmanagementisstartingtopayof.
Leadershiprequirescreatingconditionsthatenableemployeesto
dothekindsofexperimentationthatentrepreneurshiprequires.For
example,changesinTurboTaxenabledtheIntuitteamtodevelop
vehundredexperimentspertaxseason.Beforethat,marketers
withgreatideascouldn’thavedonethosetestsevenifthey’d
wantedto,becausetheydidn’thaveasysteminplace
through
whichtochangethewebsiterapidly.Intuitinvestedinsystemsthat
increasedthespeedatwhichtestscouldbebuilt,deployed,and
analyzed.
AsCooksays,“Developingtheseexperimentation
systemsisthe
responsibilityofseniormanagement;theyhavetobeputinbythe
leadership.It’smovingleadersfromplayingCaesarwiththeir
thumbsupanddownoneveryideato—instead—putinginthe
cultureandthesystemssothatteamscanmoveandinnovateatthe
speedoftheexperimentationsystem.”
3LEARN
Asanentrepreneur,nothingplaguedmemorethanthequestion
ofwhethermycompanywas
makingprogresstowardcreatinga
successfulbusiness.Asanengineerandlaterasamanager,Iwas
accustomedtomeasuringprogressbymakingsureourwork
proceededaccordingtoplan,washighquality,andcostaboutwhat
wehadprojected.
Aftermanyyearsasanentrepreneur,Istartedtoworryabout
measuringprogressinthisway.Whatifwefoundourselves
buildingsomethingthatnobodywanted?Inthatcasewhatdidit
materifwediditontimeandonbudget?WhenIwenthomeat
theendofaday’swork,theonlythingsIknewforsurewerethatI
hadkeptpeoplebusyandspentmoneythatday.Ihopedthatmy
team’seortstookusclosertoourgoal.Ifwewoundup
takinga
wrongturn,I’dhavetotakecomfortinthefactthatatleastwe’d
learnedsomethingimportant.
Unfortunately,“learning”istheoldestexcuseinthebookfora
failureofexecution.It’swhatmanagersfalbackonwhen
theyfail
toachievetheresultswepromised.Entrepreneurs,underpressure
tosucceed,arewildlycreativewhenitcomestodemonstrating
whatwehavelearned.Wecanaltelagoodstorywhenourjob,
career,orreputationdependsonit.
However,learningiscoldcomforttoemployeeswhoare
folowinganentrepreneurintotheunknown.Itiscoldcomfortto
theinvestorswhoalocatepreciousmoney,time,andenergyto
entrepreneurialteams.Itiscoldcomforttotheorganizations—large
entrepreneurialteams.Itiscoldcomforttotheorganizations—large
andsmal—thatdependonentrepreneurialinnovationtosurvive.
Youcan’ttakelearningtothebank;youcan’tspenditor
investit.
Youcannotgiveittocustomersandcannotreturnittolimited
partners.Isitanywonderthatlearninghasabadnamein
entrepreneurialandmanagerialcircles?
Yetifthefundamentalgoal
ofentrepreneurshipistoengagein
organizationbuildingunderconditionsofextremeuncertainty,its
mostvitalfunctionislearning.Wemustlearnthetruthaboutwhich
elementsofourstrategyareworkingtorealizeourvisionand
whicharejustcrazy.Wemustlearnwhatcustomersrealywant,not
whattheysaytheywantorwhatwethinktheyshouldwant.We
mustdiscoverwhetherweareonapaththatwilleadtogrowinga
sustainablebusiness.
IntheLeanStartupmodel,wearerehabilitatinglearningwitha
conceptIcalvalidatedlearning.Validatedlearningisnotafter-the-
factrationalizationoragoodstorydesignedtohidefailure.Itisa
rigorousmethodfordemonstratingprogresswhen
oneisembedded
inthesoilofextremeuncertaintyinwhichstartupsgrow.Validated
learningistheprocessofdemonstratingempiricalythatateamhas
discoveredvaluabletruthsaboutastartup’spresentandfuture
businessprospects.Itismoreconcrete,moreaccurate,andfaster
thanmarketforecastingorclassicalbusinessplanning.Itisthe
principalantidotetothelethalproblemofachievingfailure:
successfulyexecutingaplanthatleadsnowhere.
VALIDATEDLEARNINGATIMVU
Letmeilustratethiswithanexamplefrommycareer.Many
audienceshaveheardmerecountthestoryofIMVU’sfoundingand
themanymistakeswemadeindevelopingourrstproduct.I’l
nowelaborateononeofthosemistakestoilustratevalidated
learningclearly.
ThoseofusinvolvedinthefoundingofIMVUaspiredtobe
seriousstrategicthinkers.Eachofushadparticipatedinprevious
venturesthathadfailed,andwewereloathtorepeatthat
venturesthathadfailed,andwewereloathtorepeatthat
experience.Ourmainconcernsintheearlydaysdealtwiththe
folowingquestions:Whatshouldwebuildandforwhom?What
marketcouldweenteranddominate?Howcouldwebuilddurable
valuethatwouldnotbesubjecttoerosionbycompetition?1
BriliantStrategy
Wedecidedtoentertheinstantmessaging(IM)market.In2004,
thatmarkethadhundredsofmilionsofconsumersactively
participatingworldwide.However,themajorityofthecustomers
whowereusingIMproductswerenotpayingfortheprivilege.
Instead,largemediaandportalcompaniessuchas
AOL,Microsoft,
andYahoo!operatedtheirIMnetworksasalossleaderforother
serviceswhilemakingmodestamountsofmoneythrough
advertising.
IMisanexampleofamarketthatinvolvesstrongnetwork
efects.Likemostcommunicationnetworks,IMisthoughttofolow
Metcalfe’slaw:thevalueofanetworkasawholeisproportionalto
thesquareofthenumberofparticipants.Inotherwords,themore
peopleinthenetwork,themorevaluablethenetwork.
Thismakes
intuitivesense:thevaluetoeachparticipantisdrivenprimarilyby
howmanyotherpeopleheorshecancommunicatewith.Imagine
aworldinwhichyouowntheonlytelephone;itwouldhaveno
value.Onlywhenotherpeoplealsohaveatelephonedoesit
becomevaluable.
In2004,theIMmarketwaslockedupbyahandfulof
incumbents.Thetopthreenetworkscontroledmorethan80
percentoftheoveralusage
andwereintheprocessof
consolidatingtheirgainsinmarketshareattheexpenseofa
numberofsmalerplayers.2Thecommonwisdomwasthatitwas
moreorlessimpossibletobringanewIMnetworktomarket
withoutspendinganextraordinaryamountofmoneyonmarketing.
Thereasonforthatwisdomissimple.Becauseofthepowerof
networkeects,IMproductshavehighswitchingcosts.Toswitch
fromonenetworktoanother,customerswouldhaveto
convince
fromonenetworktoanother,customerswouldhavetoconvince
theirfriendsandcoleaguestoswitchwiththem.Thisextrawork
forcustomerscreatesabarriertoentryintheIMmarket:withal
consumerslockedintoanincumbent’sproduct,thereareno
customersleftwithwhomtoestablishabeachhead.
AtIMVUwesetledonastrategyofbuildingaproductthat
wouldcombinethelargemassappealoftraditionalIMwiththe
highrevenuepercustomerofthree-dimensional(3D)videogames
andvirtualworlds.Becauseofthenearimpossibilityofbringinga
newIMnetworktomarket,wedecidedtobuildanIMadd-on
productthatwouldinteroperatewiththeexisting
networks.Thus,
customerswouldbeabletoadopttheIMVUvirtualgoodsand
avatarcommunicationtechnologywithouthavingtoswitchIM
providers,learnanewuserinterface,and—mostimportant—bring
theirfriendswiththem.
Infact,wethoughtthislastpointwasessential.Fortheadd-on
producttobeuseful,customerswouldhavetouseitwiththeir
existingfriends.Everycommunicationwouldcomeembeddedwith
aninvitationtojoinIMVU.Ourproductwouldbeinherentlyviral,
spreadingthroughouttheexistingIMnetworkslikeanepidemic.To
achievethatviralgrowth,itwasimportantthatouradd-onproduct
supportasmanyoftheexistingIMnetworksas
possibleandwork
onalkindsofcomputers.
SixMonthstoLaunch
Withthisstrategyinplace,mycofoundersandIbeganaperiodof
intensework.Asthechieftechnologyocer,itwasmy
responsibility,amongother
things,towritethesoftwarethatwould
supportIMinteroperabilityacrossnetworks.MycofoundersandI
workedformonths,putingincrazyhoursstrugglingtogetourrst
productreleased.Wegaveourselvesaharddeadlineofsixmonths
—180days—tolaunchtheproductandatractourrstpaying
customers.Itwasagruelingschedule,butweweredeterminedto
launchontime.
Theadd-onproductwassolargeandcomplexandhadsomany
Theadd-onproductwassolargeandcomplexandhadsomany
movingpartsthatwehadtocutalotofcornerstogetitdoneon
time.Iwon’tmincewords:therstversionwasterrible.Wespent
endlesshoursarguingaboutwhichbugstoxandwhich
wecould
livewith,whichfeaturestocutandwhichtotrytocramin.Itwasa
wonderfulandterrifyingtime:wewerefulofhopeaboutthe
possibilitiesforsuccessandfuloffearabouttheconsequencesof
shippingabadproduct.
Personaly,Iwasworriedthatthelowqualityoftheproduct
wouldtarnishmyreputationasanengineer.PeoplewouldthinkI
didn’tknowhowtobuildaqualityproduct.Alofusfeared
tarnishingtheIMVUbrand;
afteral,wewerechargingpeople
moneyforaproductthatdidn’tworkverywel.Wealenvisioned
thedamningnewspaperheadlines:“IneptEntrepreneursBuild
DreadfulProduct.”
Aslaunchdayapproached,
ourfearsescalated.Inoursituation,
manyentrepreneurialteamsgiveintofearandpostponethelaunch
date.AlthoughIunderstandthisimpulse,Iamgladwepersevered,
sincedelaypreventsmanystartupsfromgetingthefeedbackthey
need.Ourpreviousfailuresmadeusmoreafraidofanother,even
worse,outcomethanshippingabadproduct:buildingsomething
thatnobodywants.Andso,teethclenchedandapologiesatthe
ready,wereleasedourproducttothepublic.
Launch
Andthen—nothinghappened!Itturnedoutthatourfearswere
unfounded,becausenobodyeventriedourproduct.AtrstIwas
relievedbecauseatleastnobodywasndingouthowbadthe
productwas,butsoonthatgavewaytoseriousfrustration.Afteral
thehourswehadspentarguingaboutwhichfeaturestoincludeand
whichbugstox,ourvaluepropositionwassofarothat
customersweren’tgetingfarenoughintotheexperiencetondout
howbadourdesignchoiceswere.Customerswouldn’teven
downloadourproduct.
Overtheensuingweeksandmonths,welaboredtomakethe
Overtheensuingweeksandmonths,welaboredtomakethe
productbeter.Webroughtinasteadyowofcustomersthrough
ouronlineregistrationanddownloadprocess.Wetreatedeach
day’scustomersasabrand-newreportcardtoletusknowhowwe
weredoing.Weeventualylearnedhowtochangethe
product’s
positioningsothatcustomersatleastwoulddownloadit.Wewere
makingimprovementstotheunderlyingproductcontinuously,
shippingbugxesandnewchangesdaily.However,despiteour
besteorts,wewereabletopersuadeonlyapatheticalysmal
numberofpeopletobuytheproduct.
Inretrospect,onegooddecisionwemadewastosetclear
revenuetargetsforthoseearlydays.Intherstmonthweintended
tomake$300intotalrevenue,andwedid—barely.Manyfriends
andfamilymemberswereasked(okay,begged).Eachmonthour
smalrevenuetargetsincreased,rstto$350andthento$400.As
theyrose,ourstrugglesincreased.Wesoonranoutof
friendsand
family;ourfrustrationescalated.Weweremakingtheproduct
betereveryday,yetourcustomers’behaviorremainedunchanged:
theystilwouldn’tuseit.
Ourfailuretomovethenumbersproddedusto
accelerateour
eortstobringcustomersintoouroceforin-personinterviews
andusabilitytests.Thequantitativetargetscreatedthemotivation
toengageinqualitativeinquiryandguidedusinthequestionswe
asked;thisisapaternwe’lseethroughoutthisbook.
IwishIcouldsaythatIwastheonetorealizeourmistakeand
suggestthesolution,butintruth,Iwasthelasttoadmitthe
problem.Inshort,ourentirestrategicanalysisofthemarketwas
uterlywrong.Weguredthisoutempiricaly,through
experimentation,ratherthanthroughfocusgroupsormarket
research.Customerscouldnotteluswhattheywanted;most,after
al,hadneverheardof3Davatars.Instead,theyrevealedthetruth
throughtheiractionorinactionaswestruggledtomakethe
productbeter.
TalkingtoCustomers
Outofdesperation,wedecidedtotalktosomepotentialcustomers.
Webroughtthemintoouroce,andsaid,“Trythisnew
product;
it’sIMVU.”Ifthepersonwasateenager,aheavyuserofIM,ora
techearlyadopter,heorshewouldengagewithus.Inconstrast,if
itwasamainstreamperson,theresponsewas,“Right.Soexactly
whatwouldyoulikemetodo?”We’dgetnowherewiththe
mainstreamgroup;theythoughtIMVUwastooweird.
Imagineaseventeen-year-oldgirlsitingdownwithustolookat
thisproduct.Shechoosesheravatarandsays,“Oh,thisis
realy
fun.”She’scustomizingtheavatar,decidinghowshe’sgoingtolook.
Thenwesay,“Alright,it’stimetodownloadtheinstantmessaging
add-on,”andsheresponds,“What’sthat?”
“Wel,it’sthisthingthat
interoperateswiththeinstantmessaging
client.”She’slookingatusandthinking,“I’veneverheardofthat,
myfriendshaveneverheardofthat,whydoyouwantmetodo
that?”Itrequiredalotofexplanation;aninstantmessagingadd-on
wasnotaproductcategorythatexistedinhermind.
Butsinceshewasintheroomwithus,wewereabletotalkher
intodoingit.Shedownloadstheproduct,andthenwesay,“Okay,
inviteoneofyourfriendstochat.”Andshesays,“Noway!”Wesay,
“Whynot?”Andshesays,“Wel,Idon’tknowifthisthingiscool
yet.Youwantmetoriskinvitingoneofmyfriends?Whatarethey
goingtothinkofme?Ifitsucks,they’regoingtothinkIsuck,
right?”Andwesay,“No,no,it’sgoingtobesomuchfun
onceyou
getthepersoninthere;it’sasocialproduct.”Shelooksatus,her
faceledwithdoubt;youcanseethatthisisadealbreaker.Of
course,thersttimeIhadthatexperience,Isaid,“It’salright,it’s
justthisoneperson,sendherawayandgetmeanewone.”Then
thesecondcustomercomesinandsaysthesamething.Thenthe
thirdcustomercomesin,andit’sthesamething.Youstarttosee
paterns,andnomaterhowstubbornyouare,there’s
obviously
somethingwrong.
Customerskeptsaying,“Iwanttouseitbymyself.Iwanttotryit
outrsttoseeifit’srealycoolbeforeIinviteafriend.”Ourteam
wasfromthevideogameindustry,soweunderstood
whatthat
meant:single-playermode.Sowebuiltasingle-playerversion.
meant:single-playermode.Sowebuiltasingle-playerversion.
We’dbringnewcustomersintoouroce.They’dcustomizethe
avataranddownloadtheproductlikebefore.Thentheywouldgo
intosingle-playermode,andwe’dsay,“Playwithyouravatarand
dressitup;checkoutthecoolmovesitcanmake.”Folowedby,
“Okay,youdidthatbyyourself;nowit’stimeto
inviteoneofyour
friends.”Youcanseewhat’scoming.They’dsay,“Noway!Thisisn’t
cool.”Andwe’dsay,“Wel,wetoldyouitwasn’tgoingtobecool!
Whatisthepointofasingle-playerexperienceforasocial
product?”See,wethoughtwe
shouldgetagoldstarjustfor
listeningtoourcustomers.Exceptourcustomersstildidn’tlikethe
product.Theywouldlookatusandsay,“Listen,oldman,youdon’t
understand.Whatisthedealwiththiscrazybusinessofinviting
friendsbeforeIknowifit’scool?”Itwasatotaldealbreaker.
Outoffurtherdesperation,weintroducedafeaturecaled
ChatNowthatalowsyoutopushabutonandberandomly
matchedwithsomebodyelseanywhereintheworld.The
only
thingyouhaveincommonisthatyoubothpushedthebutonat
thesametime.Alofasudden,inourcustomerservicetests,people
weresaying,“Oh,thisisfun!”
Sowe’dbringthemin,
they’duseChatNow,andmaybethey
wouldmeetsomebodytheythoughtwascool.They’dsay,“Hey,
thatguywasneat;Iwanttoaddhimtomybuddylist.Where’smy
buddylist?”Andwe’dsay,“Oh,no,youdon’twantanewbuddy
list;youwanttouseyourregularAOLbuddylist.”Remember,this
washowweplannedtoharnesstheinteroperabilitythatwould
leadtonetworkeectsandviralgrowth.Picturethecustomer
lookingatus,asking,“Whatdoyouwantmetodo
exactly?”And
we’dsay,“Wel,justgivethestrangeryourAIMscreennamesoyou
canputhimonyourbuddylist.”Youcouldseetheireyesgowide,
andthey’dsay,“Areyoukiddingme?AstrangeronmyAIMbuddy
list?”Towhichwe’drespond,“Yes;otherwiseyou’dhaveto
downloadawholenewIMclientwithanewbuddylist.”And
they’dsay,“DoyouhaveanyideahowmanyIMclientsIalready
run?”
“No.Oneortwo,maybe?”That’showmanyclientseachofusin
theoceused.Towhichtheteenagerwouldsay,“Duh!Irun
theoceused.Towhichtheteenagerwouldsay,“Duh!Irun
eight.”Wehadnoideahowmanyinstantmessaging
clientsthere
wereintheworld.
Wehadtheincorrectpreconceptionthatit’sachalengetolearn
newsoftwareandit’strickytomoveyourfriendsovertoanew
buddylist.Ourcustomersrevealedthatthiswas
nonsense.We
wantedtodrawdiagramsonthewhiteboardthatshowedwhyour
strategywasbriliant,butourcustomersdidn’tunderstandconcepts
likenetworkeectsandswitchingcosts.Ifwetriedtoexplainwhy
theyshouldbehavethewaywepredicted,they’djustshaketheir
headsatus,bewildered.
Wehadamentalmodelforhowpeopleusedsoftwarethatwas
yearsoutofdate,andsoeventualy,painfuly,afterdozensof
meetingslikethat,itstartedtodawnonusthattheIMadd-on
conceptwasfundamentalyflawed.3
OurcustomersdidnotwantanIMadd-on;theywantedastand-
aloneIMnetwork.Theydidnotconsiderhavingtolearnhowto
useanewIMprogramabarrier;onthecontrary,ourearlyadopters
usedmanydierentIMprogramssimultaneously.Ourcustomers
werenotintimidatedbytheideaofhavingtotaketheirfriends
withthemtoanewIMnetwork;itturnedoutthat
theyenjoyed
thatchalenge.Evenmoresurprising,ourassumptionthatcustomers
wouldwanttouseavatar-basedIMprimarilywiththeirexisting
friendswasalsowrong.Theywantedtomakenewfriends,an
activitythat3Davatarsareparticularlywelsuitedtofacilitating.
Bitbybit,customerstoreapartourseeminglybriliantinitial
strategy.
ThrowingMyWorkAway
Perhapsyoucansympathizewithoursituationandforgive
my
obstinacy.Afteral,itwasmyworkoverthepriormonthsthat
neededtobethrownaway.Ihadslavedoverthesoftwarethatwas
requiredtomakeourIMprograminteroperatewithother
networks,whichwasattheheartofouroriginalstrategy.Whenit
cametimetopivotandabandonthatoriginalstrategy,almostalof
cametimetopivotandabandonthatoriginalstrategy,almostalof
mywork—thousandsoflinesofcode—wasthrownout.I
felt
betrayed.Iwasadevoteeofthelatestinsoftwaredevelopment
methods(knowncolectivelyasagiledevelopment),which
promisedtohelpdrivewasteoutofproductdevelopment.
However,despitethat,Ihadcommitedthebiggestwaste
ofal:
buildingaproductthatourcustomersrefusedtouse.Thatwas
realydepressing.
Iwondered:inlightofthefactthatmyworkturnedouttobea
wasteoftimeandenergy,wouldthecompanyhave
beenjustas
weloifIhadspentthelastsixmonthsonabeachsipping
umbreladrinks?HadIrealybeenneeded?Wouldithavebeen
beterifIhadnotdoneanyworkatal?
Thereis,asImentionedat
thebeginningofthischapter,always
onelastrefugeforpeopleachingtojustifytheirownfailure.I
consoledmyselfthatifwehadn’tbuiltthisrstproduct—mistakes
andal—weneverwouldhavelearnedtheseimportantinsights
aboutcustomers.Weneverwouldhavelearnedthatourstrategy
wasawed.Thereistruthinthisexcuse:whatwelearnedduring
thosecriticalearlymonthssetIMVUonapaththatwouldleadto
oureventualbreakoutsuccess.
Foratime,this“learning”consolationmademefeelbeter,but
myreliefwasshort-lived.Here’sthequestionthatbotheredme
mostofal:ifthegoalofthosemonthswastolearntheseimportant
insightsaboutcustomers,whydidittakesolong?How
muchofour
eortcontributedtotheessentiallessonsweneededtolearn?
CouldwehavelearnedthoselessonsearlierifIhadn’tbeenso
focusedonmakingtheproduct“beter”byaddingfeaturesand
fixingbugs?
VALUEVS.WASTE
Inotherwords,whichofoureortsarevalue-creatingandwhich
arewasteful?Thisquestionisattheheartoftheleanmanufacturing
revolution;itistherstquestionanylean
manufacturingadherent
istrainedtoask.Learningtoseewasteandthensystematicaly
istrainedtoask.Learningtoseewasteandthensystematicaly
eliminateithasalowedleancompaniessuchasToyotato
dominateentireindustries.In
theworldofsoftware,theagile
developmentmethodologiesIhadpracticeduntilthattimehad
theiroriginsinleanthinking.Theyweredesignedtoeliminate
wastetoo.
Yetthosemethodshadled
medownaroadinwhichthemajority
ofmyteam’sefortswerewasted.Why?
Theanswercametomeslowlyoverthesubsequentyears.Lean
thinkingdenesvalueasprovidingbenettothecustomer;
anythingelseiswaste.Inamanufacturingbusiness,customersdon’t
carehowtheproductisassembled,onlythatitworkscorrectly.But
inastartup,whothecustomerisandwhatthecustomermightnd
valuableareunknown,partoftheveryuncertaintythatisan
essentialpartofthedenitionofastartup.Irealizedthatasa
startup,weneededanewdenitionofvalue.Therealprogresswe
hadmadeatIMVUwaswhatwehadlearnedoverthoserst
monthsaboutwhatcreatesvalueforcustomers.
Anythingwehaddoneduringthosemonthsthatdidnot
contributetoourlearningwasaformofwaste.Wouldithavebeen
possibletolearnthesamethingswithlesseort?Clearly,the
answerisyes.
Foronething,thinkofalthe
debateandprioritizationofeort
thatwentintofeaturesthatcustomerswouldneverdiscover.Ifwe
hadshippedsooner,wecouldhaveavoidedthatwaste.Also
consideralthewastecausedbyourincorrectstrategicassumptions.
IhadbuiltinteroperabilityformorethanadozendierentIM
clientsandnetworks.Wasthisrealynecessarytotestour
assumptions?Couldwehavegotenthesamefeedbackfromour
customerswithhalfasmanynetworks?Withonlythree?
Withonly
one?SincethecustomersofalIMnetworksfoundourproduct
equalyunatractive,theleveloflearningwouldhavebeenthe
same,butourefortwouldhavebeendramaticalyless.
Here’sthethoughtthatkept
meupnights:didwehaveto
supportanynetworksatal?Isitpossiblethatwecouldhave
discoveredhowawedourassumptionswerewithoutbuilding
anything?Forexample,whatifwesimplyhadoeredcustomers
anything?Forexample,whatifwesimplyhadoeredcustomers
theopportunitytodownloadtheproductfromussolelyonthe
basisofitsproposedfeaturesbeforebuildinganything?Remember,
almostnocustomerswerewilingtouseouroriginal
product,so
wewouldn’thavehadtodomuchapologizingwhenwefailedto
deliver.(Notethatthisisdierentfromaskingcustomerswhatthey
want.Mostofthetimecustomersdon’tknowwhattheywantin
advance.)Wecouldhaveconductedanexperiment,oering
customersthechancetotrysomethingandthenmeasuringtheir
behavior.
Suchthoughtexperimentswereextremelydisturbingtome
becausetheyunderminedmyjobdescription.Astheheadof
productdevelopment,Ithoughtmyjobwastoensurethetimely
deliveryofhigh-qualityproductsandfeatures.Butifmanyofthose
featureswereawasteoftime,whatshouldIbedoing
instead?How
couldweavoidthiswaste?
I’vecometobelievethatlearningistheessentialunitofprogress
forstartups.Theeortthatisnotabsolutelynecessaryforlearning
whatcustomerswantcanbeeliminated.Icalthis
validated
learningbecauseitisalwaysdemonstratedbypositive
improvementsinthestartup’scoremetrics.Aswe’veseen,it’seasy
tokidyourselfaboutwhatyouthinkcustomerswant.It’salsoeasy
tolearnthingsthatare
completelyirrelevant.Thus,validated
learningisbackedupbyempiricaldatacolectedfromreal
customers.
WHEREDOYOUFINDVALIDATION?
AsIcanatest,anybodywhofailsinastartupcanclaim
thatheor
shehaslearnedalotfromtheexperience.Theycantela
compelingstory.Infact,inthestoryofIMVUsofar,youmight
havenoticedsomethingmissing.Despitemyclaimsthatwelearned
alotinthoseearlymonths,
lessonsthatledtooureventualsuccess,
Ihaven’toeredanyevidencetobackthatup.Inhindsight,it’seasy
tomakesuchclaimsandsoundcredible(andyou’lseesome
evidencelaterinthebook),butimagineusinIMVU’searlymonths
evidencelaterinthebook),butimagineusinIMVU’searlymonths
tryingtoconvinceinvestors,employees,familymembers,andmost
ofalourselvesthatwehadnotsquanderedourtimeandresources.
Whatevidencedidwehave?
Certainlyourstoriesoffailurewereentertaining,andwehad
fascinatingtheoriesaboutwhatwehaddonewrongandwhatwe
neededtodotocreateamoresuccessfulproduct.However,the
proofdidnotcomeuntilweputthosetheoriesinto
practiceand
builtsubsequentversionsoftheproductthatshowedsuperior
resultswithactualcustomers.
ThenextfewmonthsarewherethetruestoryofIMVUbegins,
notwithourbriliantassumptionsandstrategies
andwhiteboard
gamesmanshipbutwiththehardworkofdiscoveringwhat
customersrealywantedandadjustingourproductandstrategyto
meetthosedesires.Weadoptedtheviewthatourjobwastonda
synthesisbetweenourvisionandwhatcustomerswouldaccept;it
wasn’ttocapitulatetowhatcustomersthoughttheywantedorto
telcustomerswhattheyoughttowant.
Aswecametounderstandourcustomersbeter,wewereableto
improveourproducts.Aswedidthat,thefundamentalmetricsof
ourbusinesschanged.Intheearlydays,despiteoureortsto
improvetheproduct,ourmetricswerestubbornlyat.Wetreated
eachday’scustomersasanewreportcard.We’dpayat
entionto
thepercentageofnewcustomerswhoexhibitedproductbehaviors
suchasdownloadingandbuyingourproduct.Eachday,roughlythe
samenumberofcustomerswouldbuytheproduct,andthatnumber
waspretyclosetozerodespitethemanyimprovements.
However,oncewepivotedawayfromtheoriginalstrategy,things
startedtochange.Alignedwithasuperiorstrategy,ourproduct
developmenteortsbecamemagicalymoreproductive—
not
becausewewereworkingharderbutbecausewewereworking
smarter,alignedwithourcustomers’realneeds.Positivechangesin
metricsbecamethequantitativevalidationthatourlearningwas
real.Thiswascriticalyimportantbecausewecouldshowour
stakeholders—employees,investors,andourselves—thatwewere
makinggenuineprogress,notdeludingourselves.Itisalsotheright
waytothinkaboutproductivityinastartup:not
intermsofhow
waytothinkaboutproductivityinastartup:notintermsofhow
muchstuwearebuildingbutintermsofhowmuchvalidated
learningwe’regetingforoureforts.4
Forexample,inoneearly
experiment,wechangedourentire
website,homepage,andproductregistrationowtoreplace
“avatarchat”with“3Dinstantmessaging.”Newcustomerswere
splitautomaticalybetweenthesetwoversionsofthesite;halfsaw
one,andhalfsawtheother.Wewereabletomeasurethe
dierenceinbehaviorbetweenthetwogroups.Notonlywerethe
peopleintheexperimentalgroupmorelikelytosignupforthe
product,theyweremorelikelytobecomelong-termpaying
customers.
Wehadplentyoffailedexperimentstoo.Duringoneperiodin
whichwebelievedthatcustomersweren’tusingtheproduct
becausetheydidn’tunderstanditsmanybenets,wewentsofaras
topaycustomerserviceagentstoactasvirtualtourguidesfornew
customers.Unfortunately,customerswhogotthatVIPtreatment
werenomorelikelytobecomeactiveorpayingcustomers.
EvenafterditchingtheIMadd-onstrategy,itstiltook
monthsto
understandwhyithadn’tworked.Afterourpivotandmanyfailed
experiments,wenalyguredoutthisinsight:customerswanted
touseIMVUtomakenewfriendsonline.Ourcustomersintuitively
graspedsomethingthatwewereslowtorealize.Altheexisting
socialproductsonlinewerecenteredoncustomers’real-life
identity.IMVU’savatartechnology,however,wasuniquelywel
suitedtohelppeoplegettoknoweachotheronline
without
compromisingsafetyoropeningthemselvesuptoidentitytheft.
Onceweformedthishypothesis,ourexperimentsbecamemuch
morelikelytoproducepositiveresults.Wheneverwewouldchange
theproducttomakeiteasierforpeopletondandkeepnew
friends,wediscoveredthatcustomersweremorelikelytoengage.
Thisistruestartupproductivity:systematicalyguringouttheright
thingstobuild.
Thesewerejustafewexperimentsamonghundredsthatweran
weekinandweekoutaswestartedtolearnwhichcustomers
wouldusetheproductandwhy.Eachbitofknowledgewe
wouldusetheproductandwhy.Eachbitofknowledge
we
gatheredsuggestednewexperimentstorun,whichmovedour
metricscloserandclosertoourgoal.
THEAUDACITYOFZERO
DespiteIMVU’searlysuccess,ourgrossnumberswerestilprety
smal.Unfortunately,becauseofthetraditionalwaybusinessesare
evaluated,thisisadangeroussituation.Theironyisthatitisoften
easiertoraisemoneyoracquireotherresourceswhenyouhave
zerorevenue,zerocustomers,andzerotractionthanwhen
youhave
asmalamount.Zeroinvitesimagination,butsmalnumbersinvite
questionsaboutwhetherlargenumberswilevermaterialize.
Everyoneknows(orthinksheorsheknows)storiesofproductsthat
achievedbreakthroughsuccessovernight.Aslongasnothinghas
beenreleasedandnodatahavebeencolected,itisstilpossibleto
imagineovernightsuccessinthefuture.Smalnumberspourcold
wateronthathope.
Thisphenomenoncreatesabrutalincentive:postponegetingany
datauntilyouarecertainofsuccess.Ofcourse,aswe’lsee,such
delayshavetheunfortunateeectofincreasingtheamountof
wastedwork,decreasingessentialfeedback,and
dramaticaly
increasingtheriskthatastartupwilbuildsomethingnobody
wants.
However,releasingaproductandhopingforthebestisnota
goodplaneither,becausethisincentiveisreal.Whenwe
launched
IMVU,wewereignorantofthisproblem.Ourearliestinvestorsand
advisersthoughtitwasquaintthatwehada$300-per-month
revenueplanatrst.Butafterseveralmonthswithourrevenue
hoveringaround$500permonth,somebegantolosefaith,asdid
someofouradvisers,employees,andevenspouses.Infact,atone
point,someinvestorswereseriouslyrecommendingthatwepul
theproductoutofthemarketandreturntostealthmode.
Fortunately,aswepivotedandexperimented,incorporatingwhat
welearnedintoourproductdevelopmentandmarketingeorts,
ournumbersstartedtoimprove.
ournumbersstartedtoimprove.
Butnotbymuch!Ontheonehand,wewereluckytoseea
growthpaternthatstartedtolooklikethefamoushockeystick
graph.Ontheotherhand,thegraphwentuponlytoafew
thousanddolarspermonth.Theseearlygraphs,although
promising,werenotby
themselvessucienttocombatthelossof
faithcausedbyourearlyfailure,andwelackedthelanguageof
validatedlearningtoprovideanalternativeconcepttoralyaround.
Wewerequitefortunatethatsomeofourearlyinvestors
understooditsimportanceandwerewilingtolookbeyondour
smalgrossnumberstoseetherealprogressweweremaking.
(You’lseetheexactsamegraphstheydidinChapter7.)
Thus,wecanmitigatethewastethathappensbecauseofthe
audacityofzerowithvalidatedlearning.Whatweneededto
demonstratewasthatourproductdevelopmenteortswereleading
ustowardmassivesuccesswithoutgivingintothetemptationto
falbackonvanitymetricsand“successtheater”—the
workwedo
tomakeourselveslooksuccessful.Wecouldhavetriedmarketing
gimmicks,boughtaSuperBowlad,ortriedamboyantpublic
relations(PR)asawayofjuicingourgrossnumbers.Thatwould
havegiveninvestorstheilusionoftraction,butonlyforashort
time.Eventualy,thefundamentalsofthebusinesswouldwinout
andthePRbumpwouldpass.Becausewewouldhavesquandered
preciousresourcesontheatricsinsteadofprogress,
wewouldhave
beeninrealtrouble.
Sixtymilionavatarslater,IMVUisstilgoingstrong.Itslegacyis
notjustagreatproduct,anamazingteam,andpromisingnancial
resultsbutawholenewwayofmeasuringtheprogressof
startups.
LESSONSBEYONDIMVU
IhavehadmanyopportunitiestoteachtheIMVUstoryasa
businesscaseeversinceStanford’sGraduateSchoolofBusiness
wroteanocialstudyaboutIMVU’searlyyears.5The
caseisnow
partoftheentrepreneurshipcurriculumatseveralbusinessschools,
includingHarvardBusinessSchool,whereIserveasan
includingHarvardBusinessSchool,whereIserveasan
entrepreneurinresidence.I’vealsotoldthesestoriesat
countless
workshops,lectures,andconferences.
EverytimeIteachtheIMVUstory,studentshavean
overwhelmingtemptationtofocusonthetacticsitilustrates:
launchingalow-qualityearlyprototype,charging
customersfrom
dayone,andusinglow-volumerevenuetargetsasawaytodrive
accountability.Theseareusefultechniques,buttheyarenotthe
moralofthestory.Therearetoomanyexceptions.Noteverykind
ofcustomerwilacceptalow-qualityprototype,forexample.Ifthe
studentsaremoreskeptical,theymayarguethatthetechniquesdo
notapplytotheirindustryorsituation,butworkonlybecause
IMVUisasoftwarecompany,aconsumer
Internetbusiness,ora
non-mission-criticalapplication.
Noneofthesetakeawaysisespecialyuseful.TheLeanStartupis
notacolectionofindividualtactics.Itisaprincipledapproachto
newproductdevelopment.
Theonlywaytomakesenseofits
recommendationsistounderstandtheunderlyingprinciplesthat
makethemwork.Aswe’lseeinlaterchapters,theLeanStartup
modelhasbeenappliedtoawidevarietyofbusinessesand
industries:manufacturing,cleantech,restaurants,andevenlaundry.
ThetacticsfromtheIMVUstorymayormaynotmakesensein
yourparticularbusiness.
Instead,thewayforwardistolearntoseeeverystartupinany
industryasagrandexperiment.Thequestionisnot“Canthis
productbebuilt?”Inthemoderneconomy,almostanyproductthat
canbeimaginedcanbebuilt.Themorepertinentquestionsare
“Shouldthisproductbebuilt?”and“Canwebuilda
sustainable
businessaroundthissetofproductsandservices?”Toanswerthose
questions,weneedamethodforsystematicalybreakingdowna
businessplanintoitscomponentpartsandtestingeachpart
empiricaly.
Inotherwords,weneedthescienticmethod.IntheLean
Startupmodel,everyproduct,everyfeature,everymarketing
campaign—everythingastartupdoes—isunderstoodtobean
experimentdesignedto
achievevalidatedlearning.This
experimentalapproachworksacrossindustriesandsectors,aswe’l
experimentalapproachworksacrossindustriesandsectors,aswe’l
seeinChapter4.
4EXPERIMENT
Icomeacrossmanystartupsthatarestrugglingtoanswerthe
folowingquestions:Whichcustomeropinionsshouldwelistento,
ifany?Howshouldweprioritizeacrossthemanyfeatureswe
couldbuild?Whichfeaturesareessentialtotheproduct’s
success
andwhichareancilary?Whatcanbechangedsafely,andwhat
mightangercustomers?Whatmightpleasetoday’scustomersatthe
expenseoftomorrow’s?Whatshouldweworkonnext?
Thesearesomeofthequestionsteamsstruggletoanswerifthey
havefolowedthe“let’sjustshipaproductandseewhathappens”
plan.Icalthisthe“justdoit”schoolofentrepreneurshipafter
Nike’sfamousslogan.1Unfortunately,iftheplanis
toseewhat
happens,ateamisguaranteedtosucceed—atseeingwhathappens
—butwon’tnecessarilygainvalidatedlearning.Thisisoneofthe
mostimportantlessonsofthescienticmethod:ifyoucannotfail,
youcannotlearn.
FROMALCHEMYTOSCIENCE
TheLeanStartupmethodologyreconceivesastartup’seortsas
experimentsthattestitsstrategytoseewhichpartsarebriliantand
whicharecrazy.Atrue
experimentfolowsthescienticmethod.It
beginswithaclearhypothesisthatmakespredictionsaboutwhatis
supposedtohappen.Itthenteststhosepredictionsempiricaly.Just
asscienticexperimentationisinformedbytheory,startup
experimentationisguidedbythestartup’svision.Thegoalofevery
experimentationisguidedbythestartup’svision.Thegoalofevery
startupexperimentistodiscoverhowtobuildasustainable
businessaroundthatvision.
ThinkBig,StartSmal
Zapposistheworld’slargestonlineshoestore,withannualgross
salesinexcessof$1bilion.Itisknownasoneofthemost
successful,customer-friendlye-commercebusinessesintheworld,
butitdidnotstartthatway.
FounderNickSwinmurnwasfrustratedbecausetherewasno
centralonlinesitewithagreatselectionofshoes.Heenvisioneda
newandsuperiorretailexperience.Swinmurncouldhavewaiteda
longtime,insistingontestinghiscompletevisioncompletewith
warehouses,distributionpartners,andthepromiseofsignicant
sales.Manyearlye-commercepioneersdidjustthat,including
infamousdot-comfailuressuchasWebvanand
Pets.com.
Instead,hestartedbyrunninganexperiment.Hishypothesiswas
thatcustomerswerereadyandwilingtobuyshoesonline.Totest
it,hebeganbyaskinglocalshoestoresifhecouldtakepicturesof
theirinventory.Inexchangeforpermissiontotakethepictures,he
wouldpostthepicturesonlineandcomebacktobuytheshoesat
fulpriceifacustomerboughtthemonline.
Zapposbeganwithatiny,simpleproduct.Itwasdesignedto
answeronequestionaboveal:istherealreadysucientdemand
forasuperioronlineshoppingexperienceforshoes?However,a
wel-designedstartupexperimentliketheoneZapposbeganwith
doesmorethantestasingleaspectofabusinessplan.In
thecourse
oftestingthisrstassumption,manyotherassumptionsweretested
aswel.Toseltheshoes,Zapposhadtointeractwithcustomers:
takingpayment,handlingreturns,anddealingwithcustomer
support.Thisisdecidedlydierentfrommarketresearch.IfZappos
hadreliedonexistingmarketresearchorconductedasurvey,it
couldhaveaskedwhatcustomersthoughttheywanted.Bybuilding
aproductinstead,albeitasimpleone,thecompany
learnedmuch
more:
1.Ithadmoreaccuratedataaboutcustomerdemandbecauseit
wasobservingrealcustomerbehavior,notaskinghypothetical
questions.
2.Itputitselfinapositiontointeractwithrealcustomersand
learnabouttheirneeds.Forexample,thebusinessplanmight
calfordiscountedpricing,buthowarecustomerperceptions
oftheproductafectedbythediscountingstrategy?
3.Italoweditselftobesurprisedwhencustomersbehavedin
unexpectedways,revealinginformationZapposmightnothave
knowntoaskabout.Forexample,whatifcustomersreturned
theshoes?
Zappos’initialexperimentprovidedaclear,quantiable
outcome:eitherasucientnumberofcustomerswouldbuythe
shoesortheywouldnot.Italsoputthecompanyinapositionto
observe,interactwith,andlearnfromrealcustomersandpartners.
Thisqualitativelearningisanecessarycompaniontoquantitative
testing.Althoughtheearlyeortsweredecidedlysmal-scale,that
didnotpreventthehugeZapposvisionfrombeingrealized.Infact,
in2009Zapposwasacquiredbythee-commercegiant
Amazon.comforareported$1.2bilion.2
ForLong-TermChange,ExperimentImmediately
CarolineBarlerinisadirectorintheglobalsocialinnovation
divisionatHewlet-Packard(HP),amultinationalcompanywith
morethanthreehundredthousandemployeesandmorethan$100
bilioninannualsales.Caroline,wholeadsglobalcommunity
involvement,isasocialentrepreneurworkingtogetmoreofHP’s
employeestotakeadvantageofthecompany’spolicyon
volunteering.
Corporateguidelinesencourageeveryemployeetospendupto
fourhoursamonthofcompanytimevolunteeringinhisorher
community;thatvolunteerworkcouldtaketheformofany
philanthropiceort:paintingfences,buildinghouses,orevenusing
philanthropiceort:paintingfences,buildinghouses,orevenusing
probonoorwork-basedskilsoutsidethecompany.Encouraging
thelatertypeofvolunteeringwasCaroline’spriority.
Becauseofits
talentandvalues,HP’scombinedworkforcehasthepotentialto
haveamonumentalpositiveimpact.Adesignercouldhelpa
nonprotwithanewwebsitedesign.Ateamofengineerscould
wireaschoolforInternetaccess.
Caroline’sprojectisjustbeginning,andmostemployeesdonot
knowthatthisvolunteeringpolicyexists,andonlyatinyfraction
takeadvantageofit.Mostofthevolunteeringhasbeenofthelow-
impactvariety,involvingmanuallabor,evenwhenthevolunteers
werehighlytrainedexperts.Barlerin’svisionistotakethehundreds
ofthousandsofemployeesinthecompanyandtransformtheminto
aforceforsocialgood.
Thisisthekindofcorporateinitiativeundertakeneverydayat
companiesaroundtheworld.Itdoesn’tlooklikeastartupbythe
conventionaldenitionorwhatweseeinthemovies.Onthe
surfaceitseemstobesuitedtotraditionalmanagement
and
planning.However,IhopethediscussioninChapter2has
promptedyoutobealitlesuspicious.Here’showwemight
analyzethisprojectusingtheLeanStartupframework.
Caroline’sprojectfaces
extremeuncertainty:therehadneverbeen
avolunteercampaignofthismagnitudeatHPbefore.How
condentshouldshebethatsheknowstherealreasonspeople
aren’tvolunteering?Mostimportant,howmuchdoessherealy
knowabouthowtochangethebehaviorofhundredsofthousand
peopleinmorethan170countries?Barlerin’sgoalistoinspireher
coleaguestomaketheworldabeterplace.Lookedatthatway,
herplanseemsfulofuntestedassumptions—anda
lotofvision.
Inaccordancewithtraditionalmanagementpractices,Barlerinis
spendingtimeplanning,getingbuy-infromvariousdepartments
andothermanagers,andpreparingaroadmapofinitiativesforthe
rsteighteenmonthsofherproject.Shealsohasastrong
accountabilityframeworkwithmetricsfortheimpactherproject
shouldhaveonthecompanyoverthenextfouryears.Likemany
entrepreneurs,shehasabusinessplanthatlaysoutherintentions
nicely.Yetdespitealthatwork,sheis—sofar—creatingone-o
nicely.Yetdespitealthatwork,sheis—sofar—creatingone-o
winsandnoclosertoknowingifhervisionwilbeabletoscale.
Oneassumption,forexample,mightbethatthe
company’slong-
standingvaluesincludedacommitmenttoimprovingthe
communitybutthatrecenteconomictroublehadresultedinan
increasedcompanywidestrategicfocusonshort-termprotability.
Perhapslongtimeemployeeswouldfeeladesiretorearmtheir
valuesofgivingbacktothecommunitybyvolunteering.Asecond
assumptioncouldbethattheywouldnditmoresatisfyingand
thereforemoresustainabletousetheiractualworkplace
skilsina
volunteercapacity,whichwouldhaveagreaterimpactonbehalfof
theorganizationstowhichtheydonatedtheirtime.Alsolurking
withinCaroline’splansaremanypracticalassumptionsabout
employees’wilingnesstotakethetimetovolunteer,theirlevelof
commitmentanddesire,andthewaytobestreachthemwithher
message.
TheLeanStartupmodeloersawaytotestthesehypotheses
rigorously,immediately,andthoroughly.Strategicplanningtakes
monthstocomplete;theseexperimentscouldbeginimmediately.
Bystartingsmal,Carolinecouldpreventatremendousamountof
wastedowntheroadwithoutcompromisingheroveral
vision.
Here’swhatitmightlooklikeifCarolineweretotreatherproject
asanexperiment.
BreakItDown
Therststepwouldbetobreakdownthegrandvisionintoits
componentparts.Thetwomostimportantassumptions
entrepreneursmakearewhatIcalthevaluehypothesisandthe
growthhypothesis.
Thevaluehypothesistestswhetheraproductorservicerealy
deliversvaluetocustomers
oncetheyareusingit.What’sagood
indicatorthatemployeesnddonatingtheirtimevaluable?We
couldsurveythemtogettheiropinion,butthatwouldnotbevery
accuratebecausemostpeoplehaveahardtimeassessingtheir
feelingsobjectively.
feelingsobjectively.
Experimentsprovideamoreaccurategauge.Whatcouldwesee
inrealtimethatwouldserveasaproxyforthevalueparticipants
weregainingfromvolunteering?Wecouldnd
opportunitiesfora
smalnumberofemployeestovolunteerandthenlookatthe
retentionrateofthoseemployees.Howmanyofthemsignupto
volunteeragain?Whenanemployeevoluntarilyinveststheirtime
andatentioninthisprogram,
thatisastrongindicatorthatthey
finditvaluable.
Forthegrowthhypothesis,whichtestshownewcustomerswil
discoveraproductorservice,wecandoasimilaranalysis.Oncethe
programisupandrunning,
howwilitspreadamongthe
employees,frominitialearlyadopterstomassadoptionthroughout
thecompany?Alikelywaythisprogramcouldexpandisthrough
viralgrowth.Ifthatistrue,themostimportantthingtomeasureis
behavior:wouldtheearlyparticipantsactivelyspreadthewordto
otheremployees?
Inthiscase,asimpleexperimentwouldinvolvetakingavery
smalnumber—adozen,perhaps—ofexistinglong-termemployees
andprovidinganexceptionalvolunteeropportunityforthem.
BecauseCaroline’shypothesiswasthatemployeeswouldbe
motivatedbytheirdesiretoliveuptoHP’shistoricalcommitment
tocommunityservice,theexperimentwouldtarget
employeeswho
feltthegreatestsenseofdisconnectbetweentheirdailyroutineand
thecompany’sexpressedvalues.Thepointisnottondthe
averagecustomerbuttondearlyadopters:thecustomerswhofeel
theneedfortheproductmostacutely.Thosecustomerstendtobe
moreforgivingofmistakesandareespecialyeagertogive
feedback.
Next,usingatechniqueIcaltheconciergeminimumviable
product(describedindetailinChapter6),Carolinecouldmake
surethefirstfewparticipantshadanexperiencethatwasasgoodas
shecouldmakeit,completelyalignedwithhervision.Unlikeina
focusgroup,hergoalwouldbetomeasurewhatthe
customers
actualydid.Forexample,howmanyoftherstvolunteersactualy
completetheirvolunteerassignments?Howmanyvolunteera
secondtime?Howmanyarewilingtorecruitacoleagueto
secondtime?Howmanyarewilingtorecruitacoleagueto
participateinasubsequentvolunteeractivity?
Additionalexperimentscanexpandonthisearlyfeedbackand
learning.Forexample,ifthegrowthmodelrequiresthatacertain
percentageofparticipantssharetheirexperienceswithcoleagues
andencouragetheirparticipation,thedegreetowhichthattakes
placecanbetestedevenwithaverysmalsampleofpeople.Iften
peoplecompletetherstexperiment,howmanydowe
expectto
volunteeragain?Iftheyareaskedtorecruitacoleague,howmany
doweexpectwildoso?Rememberthatthesearesupposedtobe
thekindsofearlyadopterswiththemosttogainfromtheprogram.
Putanotherway,whatifaltenearlyadoptersdeclineto
volunteeragain?Thatwouldbeahighlysignicant—andvery
negative—result.Ifthenumbersfromsuchearlyexperimentsdon’t
lookpromising,thereisclearlyaproblemwiththestrategy.That
doesn’tmeanit’stimetogiveup;onthecontrary,itmeansit’stime
togetsomeimmediatequalitativefeedbackabouthowtoimprove
theprogram.Here’swherethiskindofexperimentationhasan
advantageovertraditionalmarketresearch.Wedon’t
haveto
commissionasurveyorndnewpeopletointerview.Wealready
haveacohortofpeopletotalktoaswelasknowledgeabouttheir
actualbehavior:theparticipantsintheinitialexperiment.
Thisentireexperimentcouldbeconductedinamaterofweeks,
lessthanone-tenththetimeofthetraditionalstrategicplanning
process.Also,itcanhappeninparalelwithstrategicplanning
whiletheplanisstilbeingformulated.Evenwhen
experiments
produceanegativeresult,thosefailuresproveinstructiveandcan
inuencethestrategy.Forexample,whatifnovolunteerscanbe
foundwhoareexperiencingtheconictofvalueswithinthe
organizationthatwassuchanimportantassumptioninthebusiness
plan?Ifso,congratulations:it’stimetopivot(aconceptthatis
exploredinmoredetailinChapter8).3
ANEXPERIMENTISAPRODUCT
IntheLeanStartupmodel,anexperimentismorethanjusta
IntheLeanStartupmodel,anexperimentismorethanjusta
theoreticalinquiry;itisalsoarstproduct.Ifthisoranyother
experimentissuccessful,italowsthemanagertoget
startedwith
hisorhercampaign:enlistingearlyadopters,addingemployeesto
eachfurtherexperimentoriteration,andeventualystartingto
buildaproduct.Bythetimethatproductisreadytobedistributed
widely,itwilalreadyhaveestablishedcustomers.Itwilhave
solvedrealproblemsandoerdetailedspecicationsforwhat
needstobebuilt.Unlikeatraditionalstrategicplanningormarket
researchprocess,thisspecicationwilberootedin
feedbackon
whatisworkingtodayratherthaninanticipationofwhatmight
worktomorrow.
Toseethisinaction,consideranexamplefromKodak.Kodak’s
historyisboundupwithcamerasandlm,buttodayit
also
operatesasubstantialonlinebusinesscaledKodakGalery.Mark
CookisKodakGalery’svicepresidentofproducts,andheis
workingtochangeKodakGalery’scultureofdevelopmentto
embraceexperimentation.
Markexplained,“Traditionaly,theproductmanagersays,‘Ijust
wantthis.’Inresponse,theengineersays,‘I’mgoingtobuildit.’
Instead,Itrytopushmyteamtofirstanswerfourquestions:
1.Doconsumersrecognizethattheyhavetheproblemyouare
tryingtosolve?
2.Iftherewasasolution,wouldtheybuyit?
3.Wouldtheybuyitfromus?
4.Canwebuildasolutionforthatproblem?”
Thecommontendencyofproductdevelopmentistoskipstraight
tothefourthquestionandbuildasolutionbeforeconrmingthat
customershavetheproblem.Forexample,KodakGaleryoered
weddingcardswithgildedtextandgraphicsonitssite.
Those
designswerepopularwithcustomerswhoweregetingmarried,
andsotheteamredesignedthecardstobeusedatotherspecial
occasions,suchasforholidays.Themarketresearchanddesign
processindicatedthatcustomerswouldlikethenewcards,andthat
processindicatedthatcustomerswouldlikethenewcards,andthat
findingjustifiedthesignificantefortthatwentintocreatingthem.
Daysbeforethelaunch,theteamrealizedthecardswere
too
diculttounderstandfromtheirdepictiononthewebsite;people
couldn’tseehowbeautifultheywere.Theywerealsohardto
produce.Cookrealizedthattheyhaddonetheworkbackward.He
explained,“Untilwecouldgureouthowtoselandmakethe
product,itwasn’tworthspendinganyengineeringtimeon.”
Learningfromthatexperience,Cooktookadierentapproach
whenheledhisteamthroughthedevelopmentofanewset
of
featuresforaproductthatmakesiteasiertosharephotostakenat
anevent.Theybelievedthatanonline“eventalbum”would
provideawayforpeoplewhoatendedawedding,aconference,or
anothergatheringtosharephotoswithotheratendees.Unlike
otheronlinephotosharingservices,KodakGalery’seventalbum
wouldhavestrongprivacycontrols,assuringthatthephotoswould
besharedonlywithpeoplewhoatendedthesameevent.
Inabreakwiththepast,Cookledthegroupthroughaprocessof
identifyingrisksandassumptionsbeforebuildinganythingandthen
testingthoseassumptionsexperimentaly.
Thereweretwomainhypothesesunderlyingtheproposedevent
album:
1.Theteamassumedthatcustomerswouldwanttocreatethe
albumsinthefirstplace.
2.Itassumedthateventparticipantswoulduploadphotosto
eventalbumscreatedbyfriendsorcoleagues.
TheKodakGaleryteambuiltasimpleprototypeoftheevent
album.Itlackedmanyfeatures—somany,infact,thattheteamwas
reluctanttoshowittocustomers.However,evenatthatearlystage,
alowingcustomerstousetheprototypehelpedtheteam
refute
theirhypotheses.First,creatinganalbumwasnotaseasyasthe
teamhadpredicted;noneoftheearlycustomerswereabletocreate
one.Further,customerscomplainedthattheearlyproductversion
lackedessentialfeatures.
Thosenegativeresultsdemoralizedtheteam.Theusability
Thosenegativeresultsdemoralizedtheteam.Theusability
problemsfrustratedthem,asdidcustomercomplainsaboutmissing
features,manyofwhichmatchedtheoriginalroadmap.Cook
explainedthateventhoughtheproductwasmissingfeatures,the
projectwasnotafailure.Theinitialproduct—awsandal—
conrmedthatusersdidhavethedesiretocreateevent
albums,
whichwasextremelyvaluableinformation.Wherecustomers
complainedaboutmissingfeatures,thissuggestedthattheteamwas
ontherighttrack.Theteamnowhadearlyevidencethatthose
featureswereinfactimportant.Whataboutfeaturesthatwereon
theroadmapbutthatcustomersdidn’tcomplainabout?Maybe
thosefeaturesweren’tasimportantastheyinitialyseemed.
Throughabetalaunchtheteamcontinuedtolearnand
iterate.
Whiletheearlyuserswereenthusiasticandthenumberswere
promising,theteammadeamajordiscovery.Throughtheuseof
onlinesurveyingtoolKISSinsights,theteamlearnedthatmany
customerswantedtobeabletoarrangetheorderofpicturesbefore
theywouldinviteotherstocontribute.Knowingtheyweren’tready
tolaunch,Cookheldohisdivision’sgeneralmanagerby
explaininghowiteratingandexperimentingbefore
beginningthe
marketingcampaignwouldyieldfarbeterresults.Inaworld
wheremarketinglaunchdateswereoftensetmonthsinadvance,
waitinguntiltheteamhadrealysolvedtheproblemwasabreak
fromthepast.
ThisprocessrepresentedadramaticchangeforKodakGalery;
employeeswereusedtobeingmeasuredontheirprogressat
completingtasks.AsCooksays,“Successisnotdeliveringafeature;
successislearninghowtosolvethecustomer’sproblem.”4
THEVILLAGELAUNDRYSERVICE
InIndia,duetothecostofawashingmachine,lessthanseven
percentofthepopulationhaveoneintheirhomes.Mostpeople
eitherhandwashtheirclothingathomeorpayaDhobitodoitfor
them.Dhobistaketheclothestothenearestriver,washtheminthe
riverwater,bangthemagainstrockstogetthemclean,andhang
riverwater,bangthemagainstrockstogetthem
clean,andhang
themtodry,whichtakestwotosevendays.Theresult?Clothesare
returnedinabouttendaysandareprobablynotthatclean.
AkshayMehrahadbeenworkingatProcter&GambleSingapore
foreightyearswhenhesensedanopportunity.Asthebrand
manageroftheTideandPantenebrandsforIndiaandASEAN
countries,hethoughthecouldmakelaundryservicesavailableto
peoplewhopreviouslycouldnotaordthem.Returningto
India,
AkshayjoinedtheVilageLaundryServices(VLS),createdby
InnosightVentures.VLSbeganaseriesofexperimentstotestits
businessassumptions.
Fortheirrstexperiment,VLSmountedaconsumer-grade
laundrymachineonthebackofapickuptruckparkedonastreet
cornerinBangalore.Theexperimentcostlessthan$8,000andhad
thesimplegoalofprovingthatpeoplewouldhandovertheir
laundryandpaytohaveitcleaned.Theentrepreneurs
didnotclean
thelaundryonthetruck,whichwasmoreformarketingandshow,
buttookito-sitetobecleanedandbroughtitbacktotheir
customersbytheendoftheday.
TheVLSteamcontinuedthe
experimentforaweek,parkingthe
truckondierentstreetcorners,diggingdeepertodiscoveralthey
couldabouttheirpotentialcustomers.Theywantedtoknowhow
theycouldencouragepeopletocometothetruck.Didcleaning
speedmater?Wascleanlinessaconcern?Whatwerepeopleasking
forwhentheylefttheirlaundrywiththem?Theydiscoveredthat
customerswerehappytogivethemtheirlaundrytoclean.
However,thosecustomersweresuspiciousofthewashingmachine
mountedonthebackofthetruck,concernedthatVLSwouldtake
theirlaundryandrun.Toaddressthatconcern,VLScreateda
slightlymoresubstantialmobilecartthatlookedmorelikeakiosk.
VLSalsoexperimentedwithparkingthecartsinfrontofa
local
minimarketchain.FurtheriterationshelpedVLSgureoutwhich
servicespeopleweremostinterestedinandwhatpricetheywere
wilingtopay.Theydiscoveredthatcustomersoftenwantedtheir
clothesironedandwerewilingtopaydoublethepricetogettheir
laundrybackinfourhoursratherthantwenty-fourhours.
Asaresultofthoseearlyexperiments,VLScreatedanend
Asaresultofthoseearlyexperiments,VLScreatedan
end
productthatwasathree-footbyfour-footmobilekioskthat
includedanenergy-ecient,consumer-gradewashingmachine,a
dryer,andanextra-longextensioncord.ThekioskusedWestern
detergentsandwassupplieddailywithfreshcleanwaterdelivered
byVLS.
Sincethen,theVilageLaundryServicehasgrownsubstantialy,
withfourteenlocationsoperationalinBangalore,Mysore,and
Mumbai.AsCEOAkshayMehrasharedwithme,“Wehaveserviced
116,000kgs.in2010(vs.30,600kg.in2009).Andalmost60
percentofthebusinessiscomingfromrepeatcustomers.Wehave
servicedmorethan10,000customersinthepastyear
aloneacross
altheoutlets.”5
ALEANSTARTUPINGOVERNMENT?
OnJuly21,2010,PresidentObamasignedtheDodd–FrankWal
StreetReformandConsumerProtectionActintolaw.Oneofits
landmarkprovisionscreatedanewfederalagency,theConsumer
FederalProtectionBureau(CFPB).Thisagencyistaskedwith
protectingAmericancitizensfrompredatorylendingbynancial
servicescompaniessuchascreditcardcompanies,
studentlenders,
andpaydayloanoces.Theplancalsforittoaccomplishthisby
setingupacalcenterwheretrainedcaseworkerswileldcals
directlyfromthepublic.
Lefttoitsowndevices,anewgovernmentagencywould
probablyhirealargestawithalargebudgettodevelopaplan
thatisexpensiveandtime-consuming.However,theCFPBis
consideringdoingthingsdiferently.Despiteits$500milionbudget
andhigh-profileorigins,theCPFBisrealyastartup.
PresidentObamataskedhischieftechnologyocer,Aneesh
Chopra,withcolectingideasforhowtosetupthenewstartup
agency,andthatishowIcametobeinvolved.OnoneofChopra’s
visitstoSiliconValey,heinvitedanumberof
entrepreneursto
makesuggestionsforwaystocultivateastartupmentalityinthe
newagency.Inparticular,hisfocuswasonleveragingtechnology
newagency.Inparticular,hisfocuswasonleveragingtechnology
andinnovationtomaketheagencymoreecient,cost-eective,
andthorough.
Mysuggestionwasdrawnstraightfromtheprinciplesofthis
chapter:treattheCFPBasanexperiment,identifytheelementsof
theplanthatareassumptionsratherthanfacts,andgureoutways
totestthem.Usingtheseinsights,wecouldbuildaminimumviable
productandhavetheagencyupandrunning—onamicroscale—
longbeforetheoficialplanwassetinmotion.
Thenumberoneassumptionunderlyingthecurrentplanisthat
onceAmericansknowtheycancaltheCFPBforhelpwith
nancialfraudandabuse,therewilbeasignicantvolumeof
citizenswhodothat.Thissoundsreasonable,asitis
basedon
marketresearchabouttheamountoffraudthataectsAmericans
eachyear.However,despitealthatresearch,itisstilan
assumption.Iftheactualcalvolumediersmarkedlyfromthatin
theplan,itwilrequiresigni
cantrevision.WhatifAmericanswho
aresubjectedtonancialabusedon’tviewthemselvesasvictims
andthereforedon’tseekhelp?Whatiftheyhaveverydierent
notionsofwhatproblemsareimportant?Whatiftheycalthe
agencyseekinghelpforproblemsthatareoutsideitspurview?
Oncetheagencyisupandrunningwitha$500milionbudget
andacorrespondinglylargesta,alteringtheplanwilbe
expensiveandtime-consuming,butwhywaittogetfeedback?To
startexperimentingimmediately,theagencycouldstartwiththe
creationofasimplehotlinenumber,usingoneofthenewbreedof
low-costandfastsetupplatformssuchasTwilio.Withafewhours’
work,theycouldaddsimplevoiceprompts,oeringcalers
amenu
ofnancialproblemstochoosefrom.Intherstversion,the
promptscouldbedrawnstraightfromtheexistingresearch.Instead
ofacaseworkerontheline,eachpromptcouldoerthecaler
usefulinformationabouthowtosolveherorhisproblem.
Insteadofmarketingthishotlinetothewholecountry,theagency
couldruntheexperimentinamuchmorelimitedway:startwitha
smalgeographicarea,perhapsassmalasafewcityblocks,and
insteadofpayingforexpensivetelevisionorradioadvertisingtolet
peopleknowabouttheservice,usehighlytargetedadvertising.
peopleknowabouttheservice,usehighlytargetedadvertising.
Flyersonbilboards,newspaperadvertisementsto
thoseblocks,or
specialytargetedonlineadswouldbeagoodstart.Sincethetarget
areaissosmal,theycouldaordtopayapremiumtocreatea
highlevelofawarenessinthetargetzone.Thetotalcostwould
remainquitesmal.
Asacomprehensivesolutiontotheproblemofnancialabuse,
thisminimumviableproductisnotverygoodcomparedwithwhat
a$500milionagencycouldaccomplish.Butitisalsonotvery
expensive.Thisproductcouldbebuiltinamaterofdaysorweeks,
andthewholeexperimentprobablywouldcostonlyafew
thousanddolars.
Whatwewouldlearnfromthisexperimentwouldbeinvaluable.
Onthebasisoftheselectionsofthoserstcalers,theagencycould
immediatelystarttogetasenseofwhatkindsofproblems
Americansbelievetheyhave,notjustwhatthey“should”have.The
agencycouldbegintotestmarketingmessages:What
motivates
peopletocal?Itcouldstarttoextrapolatereal-worldtrends:What
percentageofpeopleinthetargetareaactualycal?The
extrapolationwouldnotbeperfect,butitwouldestablisha
baselinebehaviorthatwould
befarmoreaccuratethanmarket
research.
Mostimportant,thisproductwouldserveasaseedthatcould
germinateintoamuchmoreelaborateservice.Withthisbeginning,
theagencycouldengageina
continuousprocessofimprovement,
slowlybutsurelyaddingmoreandbetersolutions.Eventualy,it
wouldstathehotlinewithcaseworkers,perhapsatrstaddressing
onlyonecategoryofproblems,togivethecaseworkersthebest
chanceofsuccess.Bythetimetheocialplanwasreadyfor
implementation,thisearlyservicecouldserveasareal-world
template.
TheCFPBisjustgetingstarted,butalreadytheyareshowing
signsoffolowinganexperimentalapproach.Forexample,instead
ofdoingageographicalylimitedrolout,theyaresegmentingtheir
rstproductsbyusecase.Theyhaveestablishedapreliminary
orderofnancialproductstoprovideconsumerservices
for,with
creditcardscomingrst.Astheirrstexperimentunfolds,theywil
creditcardscomingrst.Astheirrstexperimentunfolds,theywil
havetheopportunitytocloselymonitoraloftheothercomplaints
andconsumerfeedbacktheyreceive.Thisdatawilinuencethe
depth,breadth,andsequenceoffutureoferings.
AsDavidForrest,theCFPB’schieftechnologyocer,toldme,
“OurgoalistogiveAmericancitizensaneasywaytotelusabout
theproblemstheyseeoutthereintheconsumernancial
marketplace.Wehaveanopportunitytocloselymonitorwhatthe
publicistelingusandreacttonewinformation.Marketschange
althetimeandourjobistochangewiththem.”6
Theentrepreneursandmanagersproledinthisbookaresmart,
capable,andextremelyresults-oriented.Inmanycases,theyarein
themidstofbuildinganorganizationinawayconsistentwiththe
bestpracticesofcurrentmanagementthinking.They
facethesame
chalengesinboththepublicandprivatesectors,regardlessof
industry.Aswe’veseen,eventheseasonedmanagersandexecutives
attheworld’sbest-runcompaniesstruggletoconsistentlydevelop
andlaunchinnovativenewproducts.
Theirchalengeistoovercometheprevailingmanagement
thinkingthatputsitsfaithinwel-researchedplans.Remember,
planningisatoolthatonlyworksinthepresenceofalongand
stableoperatinghistory.Andyet,doanyofusfeelthattheworld
aroundusisgetingmoreandmorestableeveryday?Changing
suchamind-setishardbutcriticaltostartupsuccess.Myhopeis
thatthisbookwilhelpmanagersandentrepreneurs
makethis
change.
PartTwo
PartTwo
STEER
HowVisionLeadstoSteering
Atitsheart,astartupisacatalystthattransformsideasinto
products.Ascustomersinteractwiththoseproducts,
theygenerate
feedbackanddata.Thefeedbackisbothqualitative(suchaswhat
theylikeanddon’tlike)andquantitative(suchashowmany
peopleuseitandnditvaluable).AswesawinPartOne,the
productsastartupbuildsarerealyexperiments;thelearningabout
howtobuildasustainablebusinessistheoutcomeofthose
experiments.Forstartups,thatinformationismuchmoreimportant
thandolars,awards,ormentionsinthepress,
becauseitcan
influenceandreshapethenextsetofideas.
Wecanvisualizethisthree-stepprocesswiththissimplediagram:
ThisBuild-Measure-LearnfeedbackloopisatthecoreoftheLean
Startupmodel.InPartTwo,wewilexamineitingreatdetail.
Manypeoplehaveprofessionaltrainingthatemphasizesone
elementofthisfeedbackloop.Forengineers,it’s
learningtobuild
thingsasecientlyaspossible.Somemanagersareexpertsat
strategizingandlearningatthewhiteboard.Plentyofentrepreneurs
focustheirenergiesontheindividualnouns:havingthebest
productideaorthebest-designedinitialproductorobsessingover
dataandmetrics.Thetruthisthatnoneoftheseactivitiesbyitselfis
ofparamountimportance.Instead,weneedtofocusourenergies
onminimizingthetotaltimethroughthisfeedbackloop.
Thisisthe
essenceofsteeringastartupandisthesubjectofPartTwo.Wewil
walkthroughacompleteturnoftheBuild-Measure-Learnfeedback
loop,discussingeachofthecomponentsindetail.
ThepurposeofPartOnewas
toexploretheimportanceof
learningasthemeasureofprogressforastartup.AsIhopeis
learningasthemeasureofprogressforastartup.AsIhopeis
evidentbynow,byfocusingourenergiesonvalidatedlearning,we
canavoidmuchofthewastethatplaguesstartupstoday.Asinlean
manufacturing,learningwhereandwhentoinvestenergyresultsin
savingtimeandmoney.
Toapplythescienticmethodtoastartup,weneedtoidentify
whichhypothesestotest.Icaltheriskiestelementsofastartup’s
plan,thepartsonwhicheverythingdepends,leap-of-faith
assumptions.Thetwomostimportantassumptionsarethevalue
hypothesisandthegrowthhypothesis.Thesegiveriseto
tuning
variablesthatcontrolastartup’sengineofgrowth.Eachiterationof
astartupisanatempttorevthisenginetoseeifitwilturn.Once
itisrunning,theprocessrepeats,shiftingintohigherandhigher
gears.
Onceclearontheseleap-of-faithassumptions,therststepisto
entertheBuildphaseasquicklyaspossiblewithaminimumviable
product(MVP).TheMVPisthatversionoftheproductthatenables
afulturnoftheBuild-Measure-Learnloopwithaminimum
amountofeortandtheleastamountofdevelopmenttime.The
minimumviableproductlacksmanyfeaturesthatmayprove
essentiallateron.However,insomeways,creatinga
MVPrequires
extrawork:wemustbeabletomeasureitsimpact.Forexample,it
isinadequatetobuildaprototypethatisevaluatedsolelyfor
internalqualitybyengineersanddesigners.Wealsoneedtogetit
infrontofpotentialcustomerstogaugetheirreactions.Wemay
evenneedtotryselingthemtheprototype,aswe’lsoonsee.
WhenweentertheMeasurephase,thebiggestchalengewilbe
determiningwhethertheproductdevelopmenteorts
areleading
torealprogress.Remember,ifwe’rebuildingsomethingthat
nobodywants,itdoesn’tmuchmaterifwe’redoingitontimeand
onbudget.ThemethodIrecommendiscaledinnovation
accounting,aquantitativeapproachthatalowsustoseewhether
ourengine-tuningeortsarebearingfruit.Italsoalowsustocreate
learningmilestones,whichareanalternativetotraditionalbusiness
andproductmilestones.Learningmilestonesare
usefulfor
entrepreneursasawayofassessingtheirprogressaccuratelyand
objectively;theyarealsoinvaluabletomanagersandinvestorswho
objectively;theyarealsoinvaluabletomanagersandinvestorswho
mustholdentrepreneursaccountable.However,notalmetricsare
createdequal,andinChapter7I’lclarifythedangerofvanity
metricsincontrasttothenuts-and-boltsusefulnessofactionable
metrics,whichhelptoanalyzecustomerbehaviorin
waysthat
supportinnovationaccounting.
Finaly,andmostimportant,there’sthepivot.Uponcompleting
theBuild-Measure-Learnloop,weconfrontthemostdicult
questionanyentrepreneur
faces:whethertopivottheoriginal
strategyorpersevere.Ifwe’vediscoveredthatoneofour
hypothesesisfalse,itistimetomakeamajorchangetoanew
strategichypothesis.
TheLeanStartupmethod
buildscapital-ecientcompanies
becauseitalowsstartupstorecognizethatit’stimetopivotsooner,
creatinglesswasteoftimeandmoney.Althoughwewritethe
feedbackloopasBuild-Measure-Learnbecausetheactivitieshappen
inthatorder,ourplanningrealyworksinthereverseorder:we
gureoutwhatweneedtolearn,useinnovationaccountingto
gureoutwhatweneedtomeasuretoknowifwearegaining
validatedlearning,andthengureoutwhatproductwe
needto
buildtorunthatexperimentandgetthatmeasurement.Alofthe
techniquesinPartTwoaredesignedtominimizethetotaltime
throughtheBuild-Measure-Learnfeedbackloop.
5LEAP
In2004,threecolegesophomoresarrivedinSiliconValeywith
theiredglingcolegesocialnetwork.Itwasliveonahandfulof
colegecampuses.Itwasnotthemarket-leadingsocialnetworkor
eventherstcolegesocialnetwork;othercompanies
hadlaunched
soonerandwithmorefeatures.With150,000registeredusers,it
madeverylitlerevenue,yetthatsummertheyraisedtheirrst
$500,000inventurecapital.Lessthanayearlater,theyraisedan
additional$12.7milion.
Ofcourse,bynowyou’veguessedthatthesethreecolege
sophomoreswereMarkZuckerberg,DustinMoskovitz,andChris
HughesofFacebook.Theirstoryisnowworldfamous.Manythings
aboutitareremarkable,butI’dliketofocusononlyone:how
Facebookwasabletoraisesomuchmoneywhenitsactualusage
wassosmal.1
Byalaccounts,whatimpressedinvestorsthemostweretwofacts
aboutFacebook’searlygrowth.Thefirstfactwastherawamountof
timeFacebook’sactiveusersspentonthesite.Morethanhalfofthe
userscamebacktothesiteeverysingleday.2Thisisanexampleof
howacompanycanvalidateitsvaluehypothesis—that
customers
ndtheproductvaluable.Thesecondimpressivethingabout
Facebook’searlytractionwastherateatwhichithadtakenoverits
rstfewcolegecampuses.Therateofgrowthwasstaggering:
FacebooklaunchedonFebruary4,2004,andbytheendofthat
monthalmostthree-quartersofHarvard’sundergraduateswere
usingit,withoutadolarofmarketingoradvertisinghavingbeen
usingit,withoutadolarofmarketingoradvertising
havingbeen
spent.Inotherwords,Facebookalsohadvalidateditsgrowth
hypothesis.Thesetwohypothesesrepresenttwoofthemost
importantleap-of-faithquestionsanynewstartupfaces.3
Atthetime,IheardmanypeoplecriticizeFacebook’searly
investors,claimingthatFacebookhad“nobusinessmodel”andonly
modestrevenuesrelativetothevaluationoeredbyitsinvestors.
TheysawinFacebookareturntotheexcessesofthe
dot-comera,
whencompanieswithlitlerevenueraisedmassiveamountsofcash
topursueastrategyof“atractingeyebals”and“getingbigfast.”
Manydot-com-erastartupsplannedtomakemoneylaterbyseling
theeyebalstheyhadboughttootheradvertisers.Intruth,those
dot-comfailureswerelitlemorethanmiddlemen,eectively
payingmoneytoacquirecustomers’atentionandthenplanningto
reselittoothers.Facebookwasdierent,becauseit
employeda
dierentengineofgrowth.Itpaidnothingforcustomeracquisition,
anditshighengagementmeantthatitwasaccumulatingmassive
amountsofcustomeratentioneveryday.Therewasneverany
questionthatatentionwouldbevaluabletoadvertisers;theonly
questionwashowmuchtheywouldpay.
ManyentrepreneursareatemptingtobuildthenextFacebook,
yetwhentheytrytoapplythelessonsofFacebookandother
famousstartupsuccessstories,theyquicklygetconfused.Isthe
lessonofFacebookthatstartupsshouldnotchargecustomers
moneyintheearlydays?Orisitthatstartupsshouldneverspend
moneyonmarketing?Thesequestionscannotbeanswered
inthe
abstract;thereareanalmostinfinitenumberofcounterexamplesfor
anytechnique.Instead,aswesawinPartOne,startupsneedto
conductexperimentsthathelpdeterminewhattechniqueswil
workintheiruniquecircumstances.Forstartups,theroleofstrategy
istohelpfigureouttherightquestionstoask.
STRATEGYISBASEDONASSUMPTIONS
Everybusinessplanbeginswithasetofassumptions.Itlaysouta
strategythattakesthoseassumptionsasagivenandproceedsto
strategythattakesthoseassumptionsasagivenandproceedsto
showhowtoachievethecompany’svision.Becausethe
assumptionshaven’tbeenprovedtobetrue(theyare
assumptions,
afteral)andinfactareoftenerroneous,thegoalofastartup’s
earlyefortsshouldbetotestthemasquicklyaspossible.
Whattraditionalbusinessstrategyexcelsatishelpingmanagers
identifyclearlywhat
assumptionsarebeingmadeinaparticular
business.Therstchalengeforanentrepreneuristobuildan
organizationthatcantesttheseassumptionssystematicaly.The
secondchalenge,asinalentrepreneurialsituations,istoperform
thatrigoroustestingwithoutlosingsightofthecompany’soveral
vision.
Manyassumptionsinatypicalbusinessplanareunexceptional.
Thesearewel-establishedfactsdrawnfrompastindustry
experienceorstraightforwarddeductions.InFacebook’scase,itwas
clearthatadvertiserswouldpayforcustomers’atention.Hidden
amongthesemundanedetailsareahandfulofassumptionsthat
requiremorecouragetostate—inthepresenttense—with
astraight
face:weassumethatcustomershaveasignicantdesiretousea
productlikeours,orweassumethatsupermarketswilcarryour
product.Actingasiftheseassumptionsaretrueisaclassic
entrepreneursuperpower.Theyarecaledleapsoffaithprecisely
becausethesuccessoftheentireventurerestsonthem.Iftheyare
true,tremendousopportunityawaits.Iftheyarefalse,thestartup
riskstotalfailure.
Mostleapsoffaithtaketheformofanargumentbyanalogy.For
example,onebusinessplanIrememberarguedasfolows:“Justas
thedevelopmentofprogressiveimageloadingalowedthe
widespreaduseoftheWorldWideWeboverdial-up,so
tooour
progressiverenderingtechnologywilalowourproducttorunon
low-endpersonalcomputers.”Youprobablyhavenoideawhat
progressiveimageloadingorrenderingis,anditdoesn’tmuch
mater.Butyouknowtheargument(perhapsyou’veevenusedit):
PrevioustechnologyXwasusedtowinmarketYbecauseof
atributeZ.WehaveanewtechnologyX2thatwilenable
ustowinmarketY2becausewetoohaveatributeZ.
Theproblemwithanalogieslikethisisthattheyobscurethetrue
leapoffaith.Thatistheirgoal:tomakethebusinessseemless
risky.Theyareusedtopersuadeinvestors,employees,orpartners
tosignon.Mostentrepreneurswouldcringe
toseetheirleapof
faithwritenthisway:
Largenumbersofpeoplealreadywantedaccesstothe
WorldWideWeb.Theyknewwhatitwas,theycouldaord
it,buttheycouldnotgetaccesstoitbecausethetimeit
tooktoloadimageswastoolong.Whenprogressiveimage
loadingwasintroduced,italowedpeopletogetontothe
WorldWideWebandteltheirfriendsaboutit.Thus,
companyXwonmarketY.
Similarly,thereisalreadyalargenumberofpotential
customerswhowantaccesstoourproductrightnow.They
knowtheywantit,theycanaordit,buttheycannotaccess
itbecausetherenderingistooslow.Whenwedebutour
productwithprogressiverenderingtechnology,theywil
ocktooursoftwareandteltheirfriends,andwewilwin
marketY2.
Thereareseveralthingstonoticeinthisrevisedstatement.First,
it’simportanttoidentifythefactsclearly.Isitrealytruethat
progressiveimageloading
causedtheadoptionoftheWorldWide
Web,orwasthisjustonefactoramongmany?Moreimportant,isit
realytruethattherearelargenumbersofpotentialcustomersout
therewhowantoursolutionrightnow?Theearlieranalogywas
designedtoconvincestakeholdersthatareasonablerststepisto
buildthenewstartup’stechnologyandseeifcustomerswiluseit.
Therestatedapproachshouldmakeclearthatwhatisneededisto
dosomeempiricaltestingrst:let’smakesurethatthere
realyare
hungrycustomersoutthereeagertoembraceournewtechnology.
AnalogsandAntilogs
Thereisnothingintrinsicalywrongwithbasingstrategyon
comparisonstoothercompaniesandindustries.In
fact,that
approachcanhelpyoudiscoverassumptionsthatarenotrealy
leapsoffaith.Forexample,theventurecapitalistRandyKomisar,
whosebookGetingtoPlanBdiscussedtheconceptofleapsof
faithingreatdetail,usesaframeworkof“analogs”and“antilogs”to
plotstrategy.
Heexplainstheanalog-antilogconceptbyusingtheiPodasan
example.“Ifyouwerelookingforanalogs,youwouldhavetolook
attheWalkman,”hesays.“ItsolvedacriticalquestionthatSteve
Jobsneverhadtoaskhimself:Wilpeoplelistentomusicina
publicplaceusingearphones?Wethinkofthatasanonsense
questiontoday,butitisfundamental.WhenSony
askedthe
question,theydidnothavetheanswer.SteveJobshad[theanswer]
intheanalog[version]”Sony’sWalkmanwastheanalog.Jobsthen
hadtofacethefactthatalthoughpeoplewerewilingtodownload
music,theywerenotwilingtopayforit.“Napsterwasanantilog.
Thatantiloghadtoleadhimtoaddresshisbusinessinaparticular
way,”Komisarsays.“Outoftheseanalogsandantilogscomea
seriesofunique,unansweredquestions.Thoseareleapsof
faiththat
I,asanentrepreneur,amtakingifIgothroughwiththisbusiness
venture.Theyaregoingtomakeorbreakmybusiness.IntheiPod
business,oneofthoseleapsoffaithwasthatpeoplewouldpayfor
music.”Ofcoursethatleapoffaithturnedouttobecorrect.4
Beyond“TheRightPlaceattheRightTime”
Thereareanynumberoffamousentrepreneurswhomademilions
becausetheyseemedtobeintherightplaceattherighttime.
However,foreverysuccessfulentrepreneurwhowasintheright
placeintherighttime,therearemanymorewhowerethere,too,
inthatrightplaceattherighttimebutstilmanagedtofail.Henry
Fordwasjoinedbynearlyvehundredotherentrepreneurs
inthe
earlytwentiethcentury.Imaginebeinganautomobileentrepreneur,
trainedinstate-of-the-artengineering,onthegroundoorofoneof
trainedinstate-of-the-artengineering,onthegroundoorofoneof
thebiggestmarketopportunitiesinhistory.Yetthevastmajority
managedtomakenomoneyatal.5Wesawthesamephenomenon
withFacebook,whichfacedearlycompetitionfromothercolege-
basedsocialnetworkswhoseheadstartprovedirrelevant.
Whatdierentiatesthesuccessstoriesfromthefailuresisthatthe
successfulentrepreneurshadtheforesight,theability,andthetools
todiscoverwhichpartsoftheirplanswereworkingbriliantlyand
whichweremisguided,andadapttheirstrategies
accordingly.
ValueandGrowth
AswesawintheFacebookstory,twoleapsoffaithstandaboveal
others:thevaluecreationhypothesisandthegrowthhypothesis.
Therststepinunderstandinganewproductorserviceisto
gure
outifitisfundamentalyvalue-creatingorvalue-destroying.Iuse
thelanguageofeconomicsinreferringtovalueratherthanprot,
becauseentrepreneursincludepeoplewhostartnot-for-protsocial
ventures,thoseinpublicsectorstartups,andinternalchangeagents
whodonotjudgetheirsuccessbyprotalone.Evenmore
confusing,therearemanyorganizationsthatarewildlyprofitablein
theshorttermbutultimatelyvalue-destroying,suchasthe
organizersofPonzischemes,andfraudulentormisguided
companies(e.g.,EnronandLehmanBrothers).
Asimilarthingistrueforgrowth.Aswithvalue,it’sessentialthat
entrepreneursunderstandthereasonsbehindastartup’sgrowth.
Therearemanyvalue-destroyingkindsofgrowththatshouldbe
avoided.Anexamplewouldbeabusinessthatgrowsthrough
continuousfund-raisingfrominvestorsandlotsofpaidadvertising
butdoesnotdevelopavalue-creatingproduct.
SuchbusinessesareengagedinwhatIcalsuccesstheater,using
theappearanceofgrowthtomakeitseemthattheyaresuccessful.
Oneofthegoalsofinnovationaccounting,whichisdiscussedin
depthinChapter7,istohelpdierentiatethesefalse
startupsfrom
trueinnovators.Traditionalaccountingjudgesnewventuresbythe
samestandardsitusesforestablishedcompanies,butthese
samestandardsitusesforestablishedcompanies,butthese
indicationsarenotreliablepredictorsofastartup’sfuture
prospects.ConsidercompaniessuchasAmazon.comthatrackedup
hugelossesontheirwaytobreakthroughsuccess.
Likeitstraditionalcounterpart,innovationaccountingrequires
thatastartuphaveandmaintainaquantitativenancialmodelthat
canbeusedtoevaluateprogressrigorously.However,inastartup’s
earliestdays,thereisnotenoughdatatomakeaninformedguess
aboutwhatthismodelmightlooklike.Astartup’searliest
strategic
plansarelikelytobehunch-orintuition-guided,andthatisagood
thing.Totranslatethoseinstinctsintodata,entrepreneursmust,in
SteveBlank’sfamousphrase,“getoutofthebuilding”andstart
learning.
GENCHIGEMBUTSU
Theimportanceofbasingstrategicdecisionsonrsthand
understandingofcustomersisoneofthecoreprinciplesthat
underliestheToyotaProductionSystem.AtToyota,thisgoesbythe
Japanesetermgenchigembutsu,whichisoneofthemost
importantphrasesintheleanmanufacturingvocabulary.InEnglish,
itisusualytranslatedasadirectiveto“goandseeforyourself”so
thatbusinessdecisionscanbebasedondeeprsthand
knowledge.
JereyLiker,whohasextensivelydocumentedthe“ToyotaWay,”
explainsitthisway:
InmyToyotainterviews,whenIaskedwhatdistinguishes
theToyotaWayfromothermanagementapproaches,the
mostcommonrstresponsewasgenchigembutsu
—whetherIwasinmanufacturing,productdevelopment,
sales,distribution,orpublicaairs.Youcannotbesureyou
realyunderstandanypartofanybusinessproblemunless
yougoandseeforyourself
rsthand.Itisunacceptableto
takeanythingforgrantedortorelyonthereportsof
others.6
Todemonstrate,takealookatthedevelopmentofToyota’s
Siennaminivanforthe2004modelyear.AtToyota,themanager
responsibleforthedesignanddevelopmentofanewmodelis
caledthechiefengineer,across-functionalleaderwhooverseesthe
entireprocessfromconcepttoproduction.The2004Siennawas
assignedtoYujiYokoya,whohadverylitle
experienceinNorth
America,whichwastheSienna’sprimarymarket.Togureout
howtoimprovetheminivan,heproposedanaudacious
entrepreneurialundertaking:aroadtripspanningalftyU.S.
states,althirteenprovinces
andterritoriesofCanada,andalparts
ofMexico.Inal,heloggedmorethan53,000milesofdriving.In
smaltownsandlargecities,Yokoyawouldrentacurrent-model
Sienna,drivingitinadditiontotalkingtoandobservingreal
customers.Fromthosersthandobservations,Yokoyawasableto
starttestinghiscriticalassumptionsaboutwhatNorthAmerican
consumerswantedinaminivan.
Itiscommontothinkofselingtoconsumersaseasierthan
selingtoenterprises,becausecustomerslackthecomplexityof
multipledepartmentsanddierentpeopleplayingdierentroles
inthepurchasingprocess.Yokoyadiscoveredthiswasuntrueforhis
customers:“Theparentsandgrandparentsmayownthe
minivan.
Butit’sthekidswhoruleit.It’sthekidswhooccupythereartwo-
thirdsofthevehicle.Andit’sthekidswhoarethemostcritical—
andthemostappreciativeoftheirenvironment.IfIlearned
anythinginmytravels,itwasthenewSiennawouldneedkid
appeal.”7Identifyingtheseassumptionshelpedguidethecar’s
development.Forexample,Yokoyaspentanunusualamountofthe
Sienna’sdevelopmentbudgetoninternalcomfortfeatures,
which
arecriticaltoalong-distancefamilyroadtrip(suchtripsaremuch
morecommoninAmericathaninJapan).
Theresultswereimpressive,boostingtheSienna’smarketshare
dramaticaly.The2004
model’ssaleswere60percenthigherthan
thosein2003.Ofcourse,aproductliketheSiennaisaclassic
sustaininginnovation,thekindthattheworld’sbest-managed
establishedcompanies,suchasToyota,excelat.Entrepreneursface
establishedcompanies,suchasToyota,excelat.Entrepreneursface
adierentsetofchalengesbecausetheyoperatewithmuchhigher
uncertainty.Whileacompanyworkingonasustaininginnovation
knowsenoughaboutwhoandwheretheircustomersareto
use
genchigembutsutodiscoverwhatcustomerswant,startups’early
contactwithpotentialcustomersmerelyrevealswhatassumptions
requirethemosturgenttesting.
GETOUTOFTHE
BUILDING
Numberstelacompelingstory,butIalwaysremindentrepreneurs
thatmetricsarepeople,too.Nomaterhowmanyintermediaries
liebetweenacompanyanditscustomers,attheendoftheday,
customersarebreathing,thinking,buyingindividuals.Their
behaviorismeasurableandchangeable.Evenwhenoneisselingto
largeinstitutions,asinabusiness-to-businessmodel,ithelpsto
rememberthatthosebusinessesaremadeupof
individuals.Al
successfulsalesmodelsdependonbreakingdownthemonolithic
viewoforganizationsintothedisparatepeoplethatmakethemup.
AsSteveBlankhasbeenteachingentrepreneursforyears,the
factsthatweneedtogatheraboutcustomers,markets,suppliers,
andchannelsexistonly“outsidethebuilding.”Startupsneed
extensivecontactwithpotentialcustomerstounderstandthem,so
getoutofyourchairandgettoknowthem.
Therststepinthisprocessistoconrmthatyourleap-of-faith
questionsarebasedinreality,thatthecustomerhasasignicant
problemworthsolving.8WhenScotCookconceivedIntuitin1982,
hehadavision—atthattimequiteradical—thatsomeday
consumerswouldusepersonalcomputerstopaybilsandkeep
trackofexpenses.WhenCooklefthisconsultingjobtotakethe
entrepreneurialplunge,hedidn’tstartwithstacksofmarket
researchorin-depthanalysisatthewhiteboard.Instead,he
picked
uptwophonebooks:oneforPaloAlto,California,wherehewas
livingatthetime,andtheotherforWinnetka,Ilinois.
Calingpeopleatrandom,heinquiredifhecouldaskthemafew
questionsaboutthewaythey
managedtheirnances.Thoseearly
questionsaboutthewaytheymanagedtheirnances.Thoseearly
conversationsweredesignedtoanswerthisleap-of-faithquestion:
dopeoplenditfrustratingtopaybilsbyhand?Itturnedoutthat
theydid,andthisearlyvalidationgaveCooktheconrmationhe
neededtogetstartedonasolution.9
Thoseearlyconversationsdidnotdelveintotheproductfeatures
ofaproposedsolution;thatatemptwouldhavebeenfoolish.The
averageconsumersatthattimewerenotconversantenoughwith
personalcomputerstohaveanopinionaboutwhetherthey’dwant
tousetheminanewway.Thoseearlyconversationswerewith
mainstreamcustomers,notearlyadopters.Stil,the
conversations
yieldedafundamentalinsight:ifIntuitcouldndawaytosolve
thisproblem,therecouldbealargemainstreamaudienceonwhich
itcouldbuildasignificantbusiness.
DesignandtheCustomer
Archetype
Thegoalofsuchearlycontactwithcustomersisnottogain
definitiveanswers.Instead,itistoclarifyatabasic,coarselevelthat
weunderstandourpotentialcustomerandwhatproblemsthey
have.Withthatunderstanding,wecancraftacustomerarchetype,a
briefdocumentthatseekstohumanizetheproposedtarget
customer.Thisarchetypeisanessentialguideforproduct
developmentandensuresthatthedailyprioritizationdecisionsthat
everyproductteammustmakearealignedwiththecustomerto
whomthecompanyaimstoappeal.
Therearemanytechniquesforbuildinganaccuratecustomer
archetypethathavebeendevelopedoverlongyearsofpracticein
thedesigncommunity.Traditionalapproachessuchasinteraction
designordesignthinkingareenormouslyhelpful.Tome,ithas
alwaysseemedironicthatmanyoftheseapproachesarehighly
experimentalanditerative,usingtechniquessuchas
rapid
prototypingandin-personcustomerobservationstoguide
designers’work.Yetbecauseofthewaydesignagencies
traditionalyhavebeencompensated,althisworkculminatesina
monolithicdeliverabletothe
client.Alofasudden,therapid
monolithicdeliverabletotheclient.Alofasudden,therapid
learningandexperimentationstops;theassumptionisthatthe
designershavelearnedalthereistoknow.Forstartups,thisisan
unworkablemodel.Noamountofdesigncananticipatethemany
complexitiesofbringingaproducttolifeintherealworld.
Infact,anewbreedofdesignersisdevelopingbrand-new
techniquesunderthebannerofLeanUserExperience
(LeanUX).
Theyrecognizethatthecustomerarchetypeisahypothesis,nota
fact.Thecustomerproleshouldbeconsideredprovisionaluntil
thestrategyhasshownviavalidatedlearningthatwecanservethis
typeofcustomerinasustainableway.10
ANALYSISPARALYSIS
Therearetwoever-presentdangerswhenentrepreneursconduct
marketresearchandtalktocustomers.Folowersofthejust-do-it
schoolofentrepreneurship
areimpatienttogetstartedanddon’t
wanttospendtimeanalyzingtheirstrategy.They’dratherstart
buildingimmediately,oftenafterjustafewcursorycustomer
conversations.Unfortunately,becausecustomersdon’trealyknow
whattheywant,it’seasyfortheseentrepreneurstodelude
themselvesthattheyareontherightpath.
Otherentrepreneurscanfalvictimtoanalysisparalysis,endlessly
reningtheirplans.Inthiscase,talkingtocustomers,reading
researchreports,andwhiteboardstrategizingarealequaly
unhelpful.Theproblemwithmostentrepreneurs’plansisgeneraly
notthattheydon’tfolowsoundstrategicprinciplesbutthatthe
factsuponwhichtheyarebasedarewrong.
Unfortunately,mostof
theseerrorscannotbedetectedatthewhiteboardbecausethey
dependonthesubtleinteractionsbetweenproductsandcustomers.
Iftoomuchanalysisisdangerousbutnonecanleadtofailure,
howdoentrepreneursknowwhentostopanalyzingandstart
building?Theanswerisaconceptcaledtheminimumviable
product,thesubjectofChapter6.
6TEST
Grouponisoneofthefastest-
growingcompaniesofaltime.Its
namecomesfrom“groupcoupons,”aningeniousideathathas
spawnedanentireindustryofsocialcommerceimitators.
However,itdidn’tstartoutsuccessful.Whencustomerstook
Grouponuponitsrstdeal,awhoppingtwentypeoplebought
two-for-onepizzainarestaurantontherstoorofthecompany’s
Chicagoofices—hardlyaworld-changingevent.
Infact,Grouponwasn’toriginalymeanttobeaboutcommerce
atal.Thefounder,AndrewMason,intendedhiscompanyto
becomea“colectiveactivismplatform”caledThePoint.Itsgoal
wastobringpeopletogethertosolveproblemstheycouldn’tsolve
ontheirown,suchasfund-raisingforacauseorboycot
inga
certainretailer.ThePoint’searlyresultsweredisappointing,
however,andattheendof2008thefoundersdecidedtotry
somethingnew.Althoughtheystilhadgrandambitions,theywere
determinedtokeepthenewproductsimple.Theybuiltaminimum
viableproduct.Doesthissoundlikeabilion-dolarcompanyto
you?Masontelsthestory:
WetookaWordPressBlogandweskinnedittosay
Grouponandtheneveryday
wewoulddoanewpost.It
wastotalygheto.WewouldselT-shirtsontherst
versionofGroupon.We’dsayinthewrite-up,“ThisT-shirt
wilcomeinthecolorred,sizelarge.Ifyouwantadiferent
colororsize,e-mailthatto
us.”Wedidn’thaveaformto
colororsize,e-mailthattous.”Wedidn’thaveaformto
addthatstuf.Itwasjustsocobbledtogether.
Itwasenoughtoprovetheconceptandshowthatitwas
somethingthatpeoplerealyliked.Theactualcoupon
generationthatweweredoingwasalFileMaker.We
wouldrunascriptthatwoulde-mailthecouponPDFto
people.Itgottothepointwherewe’dsel500sushi
couponsinaday,andwe’dsend500PDFstopeoplewith
AppleMailatthesametime.RealyuntilJulyoftherst
yearitwasjustascramblingtograbthetigerbythetail.It
wastryingtocatchupandreasonablypiecetogethera
product.1
HandmadePDFs,apizzacoupon,andasimpleblogwereenough
tolaunchGrouponintorecord-breakingsuccess;itis
onpaceto
becomethefastestcompanyinhistorytoachieve$1bilioninsales.
Itisrevolutionizingthewaylocalbusinessesndnewcustomers,
oeringspecialdealstoconsumersinmorethan375cities
worldwide.2
Aminimumviableproduct(MVP)helpsentrepreneursstartthe
processoflearningasquicklyaspossible.3Itisnotnecessarilythe
smalestproductimaginable,though;itissimplythefastestwayto
getthroughtheBuild-Measure-Learnfeedbackloopwiththe
minimumamountofefort.
Contrarytotraditionalproductdevelopment,whichusualy
involvesalong,thoughtfulincubationperiodandstrivesfor
productperfection,thegoaloftheMVPistobegintheprocessof
learning,notendit.Unlikeaprototypeorconcepttest,anMVPis
designednotjusttoanswerproductdesignortechnicalquestions.
Itsgoalistotestfundamentalbusinesshypotheses.
WHYFIRSTPRODUCTSAREN’TMEANTTOBEPERFECT
AtIMVU,whenwewereraisingmoneyfromventureinvestors,we
AtIMVU,whenwewereraisingmoneyfromventureinvestors,we
wereembarrassed.Firstofal,ourproductwasstilbuggy
andlow-
quality.Second,althoughwewereproudofourbusinessresults,
theyweren’texactlyearth-shatering.Thegoodnewswasthatwe
wereonahockey-stick-shapedgrowthcurve.Thebadnewswas
thatthehockeystickwentuptoonlyabout$8,000permonthof
revenue.Thesenumbersweresolowthatwe’doftenhaveinvestors
askus,“Whataretheunitsonthesecharts?Arethosenumbersin
thousands?”We’dhavetoreply,“No,sir,thosearein
ones.”
However,thoseearlyresultswereextremelysignicantin
predictingIMVU’sfuturepath.Asyou’lseeinChapter7,wewere
abletovalidatetwoofourleap-of-faithassumptions:IMVUwas
providingvaluefor
customers,andwehadaworkingengineof
growth.Thegrossnumbersweresmalbecausewewereselingthe
producttovisionaryearlycustomerscaledearlyadopters.Before
newproductscanbesoldsuccessfulytothemassmarket,theyhave
tobesoldtoearlyadopters.Thesepeopleareaspecialbreedof
customer.Theyaccept—infactprefer—an80percentsolution;you
don’tneedaperfectsolutiontocapturetheirinterest.4
EarlytechnologyadopterslineduparoundtheblockforApple’s
originaliPhoneeventhoughitlackedbasicfeaturessuchascopy
andpaste,3GInternetspeed,andsupportforcorporatee-mail.
Google’soriginalsearchenginecouldanswerqueriesabout
specializedtopicssuchasStanfordUniversityandthe
Linux
operatingsystem,butitwouldbeyearsbeforeitcould“organize
theworld’sinformation.”However,thisdidnotstopearlyadopters
fromsingingitspraises.
Earlyadoptersusetheirimaginationtolinwhata
productis
missing.Theypreferthatstateofaairs,becausewhattheycare
aboutabovealisbeingthersttouseoradoptanewproductor
technology.Inconsumerproducts,it’softenthethrilofbeingthe
rstoneontheblocktoshowoanewbasketbalshoe,music
player,orcoolphone.Inenterpriseproducts,it’softenabout
gainingacompetitiveadvantagebytakingariskwithsomething
newthatcompetitorsdon’thaveyet.Earlyadoptersare
suspicious
ofsomethingthatistoopolished:ifit’sreadyforeveryonetoadopt,
ofsomethingthatistoopolished:ifit’sreadyforeveryonetoadopt,
howmuchadvantagecanonegetbybeingearly?Asaresult,
additionalfeaturesorpolishbeyondwhatearlyadoptersdemandis
aformofwastedresourcesandtime.
Thisisahardtruthformanyentrepreneurstoaccept.Afteral,
thevisionentrepreneurskeepintheirheadsisofahigh-quality
mainstreamproductthatwilchangetheworld,notoneusedbya
smalnicheofpeoplewhoarewilingtogiveitashotbeforeit’s
ready.Thatworld-changingproductispolished,slick,andreadyfor
primetime.Itwinsawardsattradeshowsand,mostofal,
is
somethingyoucanproudlyshowMomandDad.Anearly,buggy,
incompleteproductfeelslikeanunacceptablecompromise.How
manyofuswereraisedwiththeexpectationthatwewouldputour
bestworkforward?Asonemanagerputittomerecently,“Iknow
forme,theMVPfeelsalitledangerous—inagoodway—sinceI
havealwaysbeensuchaperfectionist.”
Minimumviableproductsrangeincomplexityfromextremely
simplesmoketests(litlemorethananadvertisement)toactual
earlyprototypescompletewithproblemsandmissingfeatures.
DecidingexactlyhowcomplexanMVPneedstobecannotbedone
formulaicaly.Itrequiresjudgment.Luckily,this
judgmentisnot
diculttodevelop:mostentrepreneursandproductdevelopment
peopledramaticalyoverestimatehowmanyfeaturesareneededin
anMVP.Whenindoubt,simplify.
Forexample,considera
servicesoldwithaone-monthfreetrial.
Beforeacustomercanusetheservice,heorshehastosignupfor
thetrial.Oneobviousassumption,then,ofthebusinessmodelis
thatcustomerswilsignupforafreetrialoncetheyhaveacertain
amountofinformationabouttheservice.Acriticalquestionto
consideriswhethercustomerswilinfactsignupforthefreetrial
givenacertainnumberofpromisedfeatures(thevaluehypothesis).
Somewhereinthebusinessmodel,probablyburiedina
single
celinaspreadsheet,itspeciesthe“percentageofcustomerswho
seethefreetrialoerwhothensignup.”Maybeinourprojections
wesaythatthisnumbershouldbe10percent.Ifyouthinkaboutit,
thisisaleap-of-faithquestion.Itrealyshouldberepresentedin
giantletersinaboldredfont:WEASSUME10PERCENTOFCUSTOMERSWILLSIGNUP.
giantletersinaboldredfont:WEASSUME10PERCENTOFCUSTOMERSWILLSIGNUP.
Mostentrepreneursapproachaquestionlikethisbybuildingthe
productandthencheckingtoseehowcustomersreacttoit.I
considerthistobeexactlybackwardbecauseitcanleadtoalotof
waste.First,ifitturnsoutthatwe’rebuilding
somethingnobody
wants,thewholeexercisewilbeanavoidableexpenseoftimeand
money.Ifcustomerswon’tsignupforthefreetrial,they’lneverget
toexperiencetheamazingfeaturesthatawaitthem.Eveniftheydo
signup,therearemanyotheropportunitiesforwaste.Forexample,
howmanyfeaturesdowerealyneedtoincludetoappealtoearly
adopters?Everyextrafeatureisaformofwaste,andifwedelaythe
testfortheseextrafeatures,itcomeswithatremendous
potential
costintermsoflearningandcycletime.
ThelessonoftheMVPisthatanyadditionalworkbeyondwhat
wasrequiredtostartlearningiswaste,nomaterhowimportantit
mighthaveseemedatthe
time.
Todemonstrate,I’lshareseveralMVPtechniquesfromactual
LeanStartups.Ineachcase,you’lwitnessentrepreneursavoiding
thetemptationtooverbuildandoverpromise.
THEVIDEOMINIMUM
VIABLEPRODUCT
DrewHoustonistheCEOofDropbox,aSiliconValeycompany
thatmakesanextremelyeasy-to-usele-sharingtool.Instalits
application,andaDropboxfolderappearsonyourcomputer
desktop.Anythingyoudragintothatfolderisuploaded
automaticalytotheDropboxserviceandtheninstantlyreplicated
acrossalyourcomputersanddevices.
Thefoundingteamwasmadeupofengineers,astheproduct
demandedsignicanttechnicalexpertisetobuild.Itrequired,for
example,integrationwithavarietyofcomputerplatformsand
operatingsystems:Windows,Macintosh,iPhone,Android,andso
on.Eachoftheseimplementationshappensata
deeplevelofthe
systemandrequiresspecializedknow-howtomaketheuser
experienceexceptional.Infact,oneofDropbox’sbiggest
competitiveadvantagesisthattheproductworksinsuchaseamless
competitiveadvantagesisthattheproductworksinsuchaseamless
waythatthecompetitionstrugglestoemulateit.
Thesearenotthekindofpeopleonewouldthinkofas
marketinggeniuses.Infact,noneofthemhadeverworkedina
marketingjob.Theyhadprominentventurecapitalbackersand
couldhavebeenexpectedtoapplythestandardengineering
thinkingtobuildingthebusiness:builditandtheywilcome.But
Dropboxdidsomethingdiferent.
Inparalelwiththeirproductdevelopmenteorts,thefounders
wantedfeedbackfromcustomersaboutwhatrealymateredto
them.Inparticular,Dropboxneededtotestitsleap-of-faith
question:ifwecanprovideasuperiorcustomer
experience,wil
peoplegiveourproductatry?Theybelieved—rightly,asitturned
out—thatlesynchronizationwasaproblemthatmostpeople
didn’tknowtheyhad.Onceyouexperiencethesolution,youcan’t
imaginehowyoueverlivedwithoutit.
Thisisnotthekindofentrepreneurialquestionyoucanaskor
expectananswertoinafocusgroup.Customersoftendon’tknow
whattheywant,andtheyoftenhadahardtimeunderstanding
Dropboxwhentheconceptwasexplained.Houstonlearnedthisthe
hardwaywhenhetriedtoraiseventurecapital.Inmeetingafter
meeting,investorswouldexplainthatthis“marketspace”was
crowdedwithexistingproducts,noneofthemhad
madeverymuch
money,andtheproblemwasn’taveryimportantone.Drewwould
ask:“Haveyoupersonalytriedthoseotherproducts?”Whenthey
wouldsayyes,he’dask:“Didtheyworkseamlesslyforyou?”The
answerwasalmostalwaysno.Yetinmeetingaftermeeting,the
venturecapitalistscouldnotimagineaworldinlinewithDrew’s
vision.Drew,incontrast,believedthatifthesoftware“justworked
likemagic,”customerswouldflocktoit.
Thechalengewasthatitwasimpossibletodemonstratethe
workingsoftwareinaprototypeform.Theproductrequiredthat
theyovercomesignicanttechnicalhurdles;italsohadanonline
servicecomponentthatrequiredhighreliabilityand
availability.To
avoidtheriskofwakingupafteryearsofdevelopmentwitha
productnobodywanted,Drewdidsomethingunexpectedlyeasy:he
madeavideo.
madeavideo.
Thevideoisbanal,asimplethree-minutedemonstrationofthe
technologyasitismeanttowork,butitwastargetedata
communityoftechnologyearlyadopters.Drewnarratesthevideo
personaly,andashe’snarrating,thevieweriswatchinghisscreen.
Ashedescribesthekindsofleshe’dliketosynchronize,the
viewercanwatchhismousemanipulatehiscomputer.Ofcourse,if
you’repayingatention,youstarttonoticethatthefileshe’smoving
aroundarefulofin-jokesandhumorousreferencesthat
were
appreciatedbythiscommunityofearlyadopters.Drewrecounted,
“Itdrovehundredsofthousandsofpeopletothewebsite.Ourbeta
waitinglistwentfrom5,000peopleto75,000peopleliteraly
overnight.Ittotalyblewusaway.”Today,Dropboxisoneof
SiliconValey’shotestcompanies,rumoredtobeworthmorethan
$1bilion.5
Inthiscase,thevideowastheminimumviableproduct.The
MVPvalidatedDrew’sleap-of-faithassumptionthatcustomers
wantedtheproducthewasdevelopingnotbecausetheysaidsoina
focusgrouporbecauseofahopefulanalogytoanotherbusiness,
butbecausetheyactualysignedup.
THECONCIERGEMINIMUMVIABLEPRODUCT
ConsideranotherkindofMVPtechnique:theconciergeMVP.To
understandhowthistechniqueworks,meetManuelRosso,theCEO
ofanAustin,Texas–basedstartupcaledFoodonthe
Table.Food
ontheTablecreatesweeklymealplansandgroceryliststhatare
basedonfoodyouandyourfamilyenjoy,thenhooksintoyour
localgrocerystorestofindthebestdealsontheingredients.
Afteryousignupforthesite,youwalkthroughalitlesetupin
whichyouidentifyyourmaingrocerystoreandcheckothefoods
yourfamilylikes.Later,youcanpickanothernearbystoreifyou
wanttocompareprices.Next,you’representedwitha
listofitems
thatarebasedonyourpreferencesandasked:“Whatareyouinthe
moodforthisweek?”Makeyourchoices,selectthenumberof
mealsyou’rereadytoplan,andchoosewhatyoucareaboutmost
mealsyou’rereadytoplan,andchoosewhatyoucareaboutmost
intermsoftime,money,health,orvariety.Atthispoint,thesite
searchesthroughrecipesthatmatchyourneeds,pricesoutthecost
ofthemealforyou,andletsyouprintoutyourshopping
list.6
Clearly,thisisanelaborateservice.Behindthescenes,ateamof
professionalchefsdeviserecipesthattakeadvantageofitemsthat
areonsaleatlocalgrocerystoresaroundthecountry.Thoserecipes
arematchedviacomputeralgorithmtoeachfamily’suniqueneeds
andpreferences.Trytovisualizetheworkinvolved:databasesof
almosteverygrocerystoreinthecountrymustbemaintained,
includingwhat’sonsaleateachonethisweek.Those
groceries
havetobematchedtoappropriaterecipesandthenappropriately
customized,tagged,andsorted.Ifarecipecalsforbroccolirabe,is
thatthesameingredientasthebroccolionsaleatthelocalmarket?
Afterreadingthatdescription,youmightbesurprisedtolearn
thatFoodontheTable(FotT)beganlifewithasinglecustomer.
Insteadofsupportingthousandsofgrocerystoresaroundthe
countryasitdoestoday,FotTsupportedjustone.Howdid
the
companychoosewhichstoretosupport?Thefoundersdidn’t—until
theyhadtheirrstcustomer.Similarly,theybeganlifewithno
recipeswhatsoever—untiltheirrstcustomerwasreadytobegin
hermealplanning.Infact,thecompanyserveditsrstcustomer
withoutbuildinganysoftware,withoutsigninganybusiness
developmentpartnerships,andwithouthiringanychefs.
Manuel,alongwithVPofproductSteveSanderson,wenttolocal
supermarketsandmoms’groupsinhishometownofAustin.Partof
theirmissionwasthetypicalobservationofcustomersthatisapart
ofdesignthinkingandotherideationtechniques.However,Manuel
andhisteamwerealsoonthehuntforsomethingelse:their
rst
customer.
Astheymetpotentialcustomersinthosesetings,theywould
interviewthemthewayanygoodmarketresearcherwould,butat
theendofeachinterviewtheywouldatempttomakea
sale.
They’ddescribethebenetsofFotT,nameaweeklysubscription
fee,andinvitethecustomertosignup.Mosttimestheywere
rejected.Afteral,mostpeoplearenotearlyadoptersandwilnot
rejected.Afteral,mostpeoplearenotearlyadoptersandwilnot
signupforanewservicesightunseen.Buteventualysomeonedid.
Thatoneearlyadoptergottheconciergetreatment.Insteadof
interactingwiththeFotTproductviaimpersonal
software,shegota
personalvisiteachweekfromtheCEOofthecompany.Heandthe
VPofproductwouldreviewwhatwasonsaleatherpreferred
grocerystoreandcarefulyselectrecipesonthebasisofher
preferences,goingsofarastolearnherfavoriterecipesforitems
sheregularlycookedforherfamily.Eachweektheywouldhand
her—inperson—apreparedpacketcontainingashoppinglistand
relevantrecipes,solicitherfeedback,andmakechanges
as
necessary.Mostimportant,eachweektheywouldcolectacheck
for$9.95.
Talkaboutinecient!Measuredaccordingtotraditionalcriteria,
thisisaterriblesystem,entirelynonscalableanda
completewaste
oftime.TheCEOandVPofproduct,insteadofbuildingtheir
business,areengagedinthedrudgeryofsolvingjustonecustomer’s
problem.Insteadofmarketingthemselvestomilions,theysold
themselvestoone.Worstofal,theireortsdidn’tappeartobe
leadingtoanythingtangible.Theyhadnoproduct,nomeaningful
revenue,nodatabasesofrecipes,notevenalastingorganization.
However,viewedthroughthelensoftheLeanStartup,they
were
makingmonumentalprogress.Eachweektheywerelearningmore
andmoreaboutwhatwasrequiredtomaketheirproductasuccess.
Afterafewweekstheywerereadyforanothercustomer.Each
customertheybroughtonmadeiteasiertogetthenextone,
becauseFotTcouldfocusonthesamegrocerystore,getingto
knowitsproductsandthekindsofpeoplewhoshoppedtherewel.
Eachnewcustomergottheconciergetreatment:personal
in-home
visits,theworks.Butafterafewmorecustomers,theoverheadof
servingthemone-on-onestartedtoincrease.
Onlyatthepointwherethefoundersweretoobusytobringon
additionalcustomersdid
Manuelandhisteamstarttoinvestin
automationintheformofproductdevelopment.Eachiterationof
theirminimumviableproductalowedthemtosavealitlemore
timeandserveafewmorecustomers:deliveringtherecipesand
shoppinglistviae-mailinsteadofviaanin-homevisit,startingto
shoppinglistviae-mailinsteadofviaanin-homevisit,startingto
parselistsofwhatwasonsaleautomaticalyviasoftwareinsteadof
byhand,eveneventualytakingcreditcardpayments
onlineinstead
ofahandwritencheck.
Beforelong,theyhadbuiltasubstantialserviceoering,rstin
theAustinareaandeventualynationwide.Butalongtheway,their
productdevelopmentteamwasalwaysfocusedon
scaling
somethingthatwasworkingratherthantryingtoinventsomething
thatmightworkinthefuture.Asaresult,theirdevelopmenteorts
involvedfarlesswastethanistypicalforaventureofthiskind.
Itisimportanttocontrastthiswiththecaseofasmalbusiness,
inwhichitisroutinetoseetheCEO,founder,president,andowner
servingcustomersdirectly,oneatatime.InaconciergeMVP,this
personalizedserviceisnottheproductbutalearning
activity
designedtotesttheleap-of-faithassumptionsinthecompany’s
growthmodel.Infact,acommonoutcomeofaconciergeMVPisto
invalidatethecompany’sproposedgrowthmodel,makingitclear
thatadierentapproachisneeded.Thiscanhappenevenifthe
initialMVPisprotableforthecompany.Withoutaformalgrowth
model,manycompaniesgetcaughtinthetrapofbeingsatised
withasmalprotablebusinesswhenapivot
(changeincourseor
strategy)mightleadtomoresignicantgrowth.Theonlywayto
knowistohavetestedthegrowthmodelsystematicalywithreal
customers.
PAYNOATTENTIONTOTHEEIGHTPEOPLE
BEHINDTHE
CURTAIN
MeetMaxVentilaandDamonHorowitz,technologistswitha
visiontobuildanewtypeofsearchsoftwaredesignedtoanswer
thekindsofquestionsthatbefuddlestate-of-the-art
companiessuch
asGoogle.Googlebefuddled?Thinkaboutit.Googleanditspeers
excelatansweringfactualquestions:Whatisthetalestmountainin
theworld?Whowasthetwenty-thirdpresidentoftheUnited
States?Butformoresubjectivequestions,Googlestruggles.Ask,
“What’sagoodplacetogooutforadrinkafterthebalgameinmy
“What’sagoodplacetogooutforadrinkafterthebalgameinmy
city?”andthetechnologyflails.What’sinteresting
aboutthisclassof
queriesisthattheyarerelativelyeasyforapersontoanswer.
Imaginebeingatacocktailpartysurroundedbyfriends.Howlikely
wouldyoubetogetahigh-qualityanswertoyoursubjective
question?Youalmostcertainlywouldgetone.Unlikefactual
queries,becausethesesubjectivequestionshavenosingleright
answer,today’stechnologystrugglestoanswerthem.Suchquestions
dependonthepersonansweringthem,hisorher
personal
experience,taste,andassessmentofwhatyou’relookingfor.
Tosolvethisproblem,MaxandDamoncreatedaproductcaled
Aardvark.Withtheirdeeptechnicalknowledgeandindustry
experience,itwouldhavebeenreasonabletoexpectthemtodive
inandstartprogramming.Instead,theytooksixmonthstogure
outwhattheyshouldbebuilding.Buttheydidn’tspendthatyearat
thewhiteboardstrategizingorengageinalengthymarket
research
project.
Instead,theybuiltaseriesoffunctioningproducts,eachdesigned
totestawayofsolvingthisproblemfortheircustomers.Each
productwasthenoeredtobetatesters,whosebehavior
wasused
tovalidateorrefuteeachspecichypothesis(seeexamplesin
sidebar).
ThefolowinglistofprojectsareexamplesfromAardvark’s
ideationperiod.7
Rekkit.Aservicetocolectyourratingsfromacrosstheweb
andgivebeterrecommendationstoyou.
Ninjapa.Awaythatyoucouldopenaccountsinvarious
applicationsthroughasinglewebsiteandmanageyourdata
acrossmultiplesites.
TheWebb.Acentralnumberthatyoucouldcalandtalktoa
personwhocoulddoanythingforyouthatyoucoulddo
online.
WebMacros.Awaytorecordsequencesofstepson
websitesso
thatyoucouldrepeatcommonactions,evenacrosssites,and
share“recipes”forhowyouaccomplishedonlinetasks.
InternetButonCompany.Awaytopackagestepstakenona
websiteandsmartform-l
functionality.Peoplecouldencode
butonsandsharebutonsàlasocialbookmarking.
MaxandDamonhadavisionthatcomputerscouldbeusedto
createavirtualpersonalassistanttowhichtheircustomerscould
askquestions.Becausetheassistantwasdesignedforsubjective
questions,theanswersrequiredhumanjudgment.Thus,theearly
Aardvarkexperimentstriedmanyvariationsonthistheme,building
aseriesofprototypesforwayscustomerscould
interactwiththe
virtualassistantandgettheirquestionsanswered.Altheearly
prototypesfailedtoengagethecustomers.
AsMaxdescribesit,“Weself-fundedthecompanyandreleased
verycheapprototypestotest.
WhatbecameAardvarkwasthesixth
prototype.Eachprototypewasatwo-tofour-weekeort.Weused
humanstoreplicatethebackendasmuchaspossible.Weinvited
onehundredtotwohundredfriendstotrytheprototypesand
measuredhowmanyofthemcameback.Theresultswere
unambiguouslynegativeuntilAardvark.”
Becauseoftheshorttimeline,noneoftheprototypesinvolved
advancedtechnology.Instead,theywereMVPsdesignedtotesta
moreimportantquestion:whatwouldberequiredtogetcustomers
toengagewiththeproductandteltheirfriendsaboutit?
“OncewechoseAardvark,”Ventilasays,“wecontinuedtorun
“OncewechoseAardvark,”Ventilasays,“wecontinuedtorun
withhumansreplicatingpiecesofthebackendforninemonths.We
hiredeightpeopletomanagequeries,classifyconversations,etc.
WeactualyraisedourseedandseriesAroundsbeforethesystem
wasautomated—theassumptionwasthatthelines
betweenhumans
andarticialinteligencewouldcross,andweatleastprovedthat
wewerebuildingstufpeoplewouldrespondto.
“Aswerenedtheproduct,wewouldbringinsixtotwelve
peopleweeklytoreactto
mockups,prototypes,orsimulationsthat
wewereworkingon.Itwasamixofexistingusersandpeoplewho
neversawtheproductbefore.Wehadourengineersjoinformany
ofthesesessions,bothsothattheycouldmakemodicationsinreal
time,butalsosowecouldalexperiencethepainofausernot
knowingwhattodo.”8
TheAardvarkproducttheysetledonworkedviainstant
messaging(IM).CustomerscouldsendAardvarkaquestionviaIM,
andAardvarkwouldgetthem
ananswerthatwasdrawnfromthe
customer’ssocialnetwork:thesystemwouldseekoutthe
customer’sfriendsandfriendsoffriendsandposethequestionto
them.Onceitgotasuitableanswer,itwouldreportbacktothe
initialcustomer.
Ofcourse,aproductlikethatrequiresaveryimportant
algorithm:givenaquestionaboutacertaintopic,whoisthebest
personinthecustomer’ssocialnetworktoanswerthatquestion?
Forexample,aquestion
aboutrestaurantsinSanFrancisco
shouldn’tberoutedtosomeoneinSeatle.Morechalengingstil,a
questionaboutcomputerprogrammingprobablyshouldn’tbe
routedtoanartstudent.
Throughouttheirtesting
process,MaxandDamonencountered
manydiculttechnologicalproblemslikethese.Eachtime,they
emphaticalyrefusedtosolvethematthatearlystage.Instead,they
usedWizardofOztestingtofakeit.InaWizardofOztest,
customersbelievetheyareinteractingwiththeactualproduct,but
behindthesceneshumanbeingsaredoingthework.Likethe
conciergeMVP,thisapproachisincrediblyinecient.Imaginea
servicethatalowedcustomerstoaskquestionsof
human
servicethatalowedcustomerstoaskquestionsofhuman
researchers—forfree—andexpectareal-timeresponse.Sucha
service(atscale)wouldlosemoney,butitiseasytobuildona
microscale.Atthatscale,italowedMaxandDamontoanswer
theseal-importantquestions:Ifwecansolvethetoughtechnical
problemsbehindthisarticialinteligenceproduct,wilpeopleuse
it?Wiltheiruseleadtothecreationofaproductthathas
real
value?
ItwasthissystemthatalowedMaxandDamontopivotoverand
overagain,rejectingconceptsthatseemedpromisingbutthat
wouldnothavebeenviable.Whentheywerereadyto
startscaling,
theyhadaready-maderoadmapofwhattobuild.Theresult:
Aardvarkwasacquiredforareported$50milion—byGoogle.9
THEROLEOFQUALITYANDDESIGNINANMVP
Oneofthemostvexing
aspectsoftheminimumviableproductis
thechalengeitposestotraditionalnotionsofquality.Thebest
professionalsandcraftspersonsalikeaspiretobuildquality
products;itisapointofpride.
Modernproductionprocessesrelyonhighqualityasawayto
boosteciency.TheyoperateusingW.EdwardsDeming’sfamous
dictumthatthecustomeristhemostimportantpartofthe
productionprocess.Thismeansthatwemustfocusour
energies
exclusivelyonproducingoutcomesthatthecustomerperceivesas
valuable.Alowingsloppyworkintoourprocessinevitablyleadsto
excessivevariation.Variationinprocessyieldsproductsofvarying
qualityintheeyesofthecustomerthatatbestrequirereworkand
atworstleadtoalostcustomer.Mostmodernbusinessand
engineeringphilosophiesfocusonproducinghigh-quality
experiencesforcustomersasaprimaryprinciple;itisthe
foundationofSixSigma,leanmanufacturing,designthinking,
extremeprogramming,andthesoftwarecraftsmanshipmovement.
Thesediscussionsofqualitypresupposethatthecompanyalready
knowswhatatributesoftheproductthecustomerwil
perceiveas
worthwhile.Inastartup,thisisariskyassumptiontomake.Often
worthwhile.Inastartup,thisisariskyassumptiontomake.Often
wearenotevensurewhothecustomeris.Thus,forstartups,I
believeinthefolowingqualityprinciple:
Ifwedonotknowwhothecustomeris,wedonotknow
whatqualityis.
Evena“low-quality”MVPcanactinserviceofbuildingagreat
high-qualityproduct.Yes,MVPssometimesare
perceivedaslow-
qualitybycustomers.Ifso,weshouldusethisasanopportunityto
learnwhatatributescustomerscareabout.Thisisinnitelybeter
thanmerespeculationorwhiteboardstrategizing,becauseit
providesasolidempiricalfoundationonwhichtobuildfuture
products.
Sometimes,however,customersreactquitedierently.Many
famousproductswerereleasedina“low-quality”state,and
customerslovedthem.ImagineifCraigNewmark,intheearlydays
ofCraigslist,hadrefusedtopublishhishumblee-mailnewsleter
becauseitlackedsucienthighdesign.Whatifthefoundersof
Grouponhadfelt“twopizzasforthepriceofone”was
beneath
them?
Ihavehadmanysimilarexperiences.IntheearlydaysofIMVU,
ouravatarswerelockedinoneplace,unabletomovearoundthe
screen.Thereason?WewerebuildinganMVPandhadnot
yet
tackledthediculttaskofcreatingthetechnologythatwould
alowavatarstowalkaroundthevirtualenvironmentstheyinhabit.
Inthevideogameindustry,thestandardisthat3Davatarsshould
moveuidlyastheywalk,avoidobstaclesintheirpath,andtake
aninteligentroutetowardtheirdestination.Famousbest-seling
gamessuchasElectronicArts’TheSimsworkonthisprinciple.We
didn’twanttoshipalow-qualityversionofthis
feature,sowe
optedinsteadtoshipwithstationaryavatars.
Feedbackfromthecustomerswasveryconsistent:theywanted
theabilitytomovetheiravatarsaroundtheenvironment.Wetook
thisasbadnewsbecauseit
meantwewouldhavetospend
considerableamountsoftimeandmoneyonahigh-qualitysolution
similartoTheSims.Butbeforewecommitedourselvestothat
similartoTheSims.Butbeforewecommitedourselvestothat
path,wedecidedtotryanotherMVP.Weusedasimplehack,
whichfeltalmostlikecheating.Wechangedtheproductsothat
customerscouldclickwheretheywantedtheiravatartogo,andthe
avatarwouldteleportthereinstantly.Nowalking,no
obstacle
avoidance.Theavatardisappearedandthenreappearedaninstant
laterinthenewplace.Wecouldn’tevenaordfancyteleportation
graphicsorsoundeects.Wefeltlameshippingthisfeature,butit
wasalwecouldaford.
Youcanimagineoursurprisewhenwestartedtogetpositive
customerfeedback.Weneveraskedaboutthemovementfeature
directly(weweretooembarrassed).Butwhenaskedtonamethe
topthingsaboutIMVUtheylikedbest,customersconsistentlylisted
avatar“teleportation”amongthetopthree(unbelievably,they
oftenspecicalydescribeditas“moreadvancedthanTheSims”).
Thisinexpensivecompromiseoutperformed
manyfeaturesofthe
productweweremostproudof,featuresthathadtakenmuchmore
timeandmoneytoproduce.
Customersdon’tcarehowmuchtimesomethingtakestobuild.
Theycareonlyifitservestheirneeds.Ourcustomers
preferredthe
quickteleportationfeaturebecauseitalowedthemtogetwhere
theywantedtogoasfastaspossible.Inretrospect,thismakes
sense.Wouldn’twealliketogetwhereverwe’regoinginan
instant?Nolines,nohoursonaplaneorsitingonthetarmac,no
connections,nocabsorsubways.Beammeup,Scoty.Our
expensive“real-world”approachwasbeatenhandilybyacool
fantasy-worldfeaturethatcostmuchlessbutthatour
customers
preferred.
Sowhichversionoftheproductislow-quality,again?
MVPsrequirethecouragetoputone’sassumptionstothetest.If
customersreactthewayweexpect,wecantakethatas
conrmationthatourassumptionsarecorrect.Ifwereleasea
poorlydesignedproductandcustomers(evenearlyadopters)
cannotgureouthowtouseit,thatwilconrmourneedtoinvest
insuperiordesign.Butwemustalwaysask:whatifthey
don’tcare
aboutdesigninthesamewaywedo?
Thus,theLeanStartupmethodisnotopposedtobuildinghigh-
Thus,theLeanStartupmethodisnotopposedtobuildinghigh-
qualityproducts,butonlyin
serviceofthegoalofwinningover
customers.Wemustbewilingtosetasideourtraditional
professionalstandardstostarttheprocessofvalidatedlearningas
soonaspossible.Butonceagain,thisdoesnotmeanoperatingina
sloppyorundisciplinedway.(Thisisanimportantcaveat.Thereis
acategoryofqualityproblemsthathavetheneteectofslowing
downtheBuild-Measure-Learnfeedbackloop.Defectsmakeitmore
diculttoevolvetheproduct.Theyactualyinterferewith
our
abilitytolearnandsoaredangeroustotolerateinanyproduction
process.Wewilconsidermethodsforguringoutwhentomake
investmentsinpreventingthesekindsofproblemsinPartThree.)
Asyouconsiderbuildingyourownminimumviableproduct,let
thissimplerulesuce:removeanyfeature,process,oreortthat
doesnotcontributedirectlytothelearningyouseek.
SPEEDBUMPSINBUILDINGANMVP
BuildinganMVPisnotwithoutrisks,bothrealandimagined.Both
canderailastartupeortunlesstheyareunderstoodaheadoftime.
Themostcommonspeedbumpsarelegalissues,fearsabout
competitors,brandingrisks,andtheimpactonmorale.
Forstartupsthatrelyonpatentprotection,therearespecial
chalengeswithreleasinganearlyproduct.Insomejurisdictions,
thewindowforlingapatentbeginswhentheproductisreleased
tothegeneralpublic,anddependingonthewaythe
MVPis
structured,releasingitmaystartthisclock.Evenifyourstartupis
notinoneofthosejurisdictions,youmaywantinternationalpatent
protectionandmaywinduphavingtoabidebythesemore
stringentrequirements.(Inmyopinion,issueslikethisareoneof
themanywaysinwhichcurrentpatentlawinhibitsinnovationand
shouldberemediedasamaterofpublicpolicy.)
Inmanyindustries,patentsareusedprimarilyfordefensive
purposes,asadeterrenttoholdcompetitorsatbay.Insuchcases,
thepatentrisksofanMVPareminorcomparedwiththelearning
benets.However,inindustriesinwhichanewscientic
benets.However,inindustriesinwhichanew
scientic
breakthroughisattheheartofacompany’scompetitiveadvantage,
theserisksneedtobebalancedmorecarefuly.Inalcases,
entrepreneursshouldseeklegalcounseltoensurethatthey
understandtherisksfuly.
Legalrisksmaybedaunting,butyoumaybesurprisedtolearn
thatthemostcommonobjectionIhaveheardovertheyearsto
buildinganMVPisfearofcompetitors—especialylargeestablished
companies—stealingastartup’sideas.Ifonlyitweresoeasyto
haveagoodideastolen!Partofthespecialchalengeofbeinga
startupisthenearimpossibilityofhavingyouridea,company,or
productbenoticedbyanyone,letalonea
competitor.Infact,Ihave
oftengivenentrepreneursfearfulofthisissuethefolowing
assignment:takeoneofyourideas(oneofyourlesserinsights,
perhaps),ndthenameoftherelevantproductmanageratan
establishedcompanywhohasresponsibilityforthatarea,andtryto
getthatcompanytostealyouridea.Calthemup,writethema
memo,sendthemapressrelease—goahead,tryit.Thetruthisthat
mostmanagersinmostcompaniesarealready
overwhelmedwith
goodideas.Theirchalengeliesinprioritizationandexecution,and
itisthosechalengesthatgiveastartuphopeofsurviving.10
Ifacompetitorcanoutexecuteastartuponcetheideaisknown,
thestartupisdoomedanyway.Thereasontobuildanewteamto
pursueanideaisthatyoubelieveyoucanacceleratethroughthe
Build-Measure-Learnfeedbackloopfasterthananyoneelsecan.If
that’strue,itmakesnodierencewhatthecompetition
knows.If
it’snottrue,astartuphasmuchbiggerproblems,andsecrecywon’t
xthem.Soonerorlater,asuccessfulstartupwilfacecompetition
fromfastfolowers.Aheadstartisrarelylargeenoughtomater,
andtimespentinstealthmode—awayfromcustomers—isunlikely
toprovideaheadstart.Theonlywaytowinistolearnfasterthan
anyoneelse.
Manystartupsplantoinvestinbuildingagreatbrand,andan
MVPcanseemlikeadangerousbrandingrisk.Similarly,
entrepreneursinexistingorganizationsoftenareconstrainedbythe
fearofdamagingtheparentcompany’sestablishedbrand.Ineither
fearofdamagingtheparentcompany’sestablishedbrand.
Ineither
ofthesecases,thereisaneasysolution:launchtheMVPundera
dierentbrandname.Inaddition,along-termreputationisonlyat
riskwhencompaniesengageinvocallaunchactivitiessuchasPR
andbuildinghype.Whenaproductfailstoliveuptothose
pronouncements,reallong-termdamagecanhappentoacorporate
brand.Butstartupshavetheadvantageofbeingobscure,havinga
patheticalysmalnumberofcustomers,andnothaving
much
exposure.Ratherthanlamentingthem,usetheseadvantagesto
experimentundertheradarandthendoapublicmarketinglaunch
oncetheproducthasproveditselfwithrealcustomers.11
Finaly,ithelpstoprepare
forthefactthatMVPsoftenresultin
badnews.Unliketraditionalconcepttestsorprototypes,theyare
designedtospeaktothefulrangeofbusinessquestions,notjust
designortechnicalones,andtheyoftenprovideaneededdoseof
reality.Infact,piercingtherealitydistortioneldisquite
uncomfortable.Visionariesareespecialyafraidofafalsenegative:
thatcustomerswilrejectaawedMVPthatistoosmalortoo
limited.Itispreciselythisatitudethatoneseeswhencompanies
launchfulyformedproductswithoutpriortesting.Theysimply
couldn’tbeartotesttheminanythinglessthantheirfulsplendor.
Yetthereiswisdominthevisionary’sfear.Teamssteepedin
traditionalproductdevelopmentmethodsare
trainedtomake
go/kildecisionsonaregularbasis.Thatistheessenceofthe
waterfalorstage-gatedevelopmentmodel.IfanMVPfails,teams
areliabletogiveuphopeandabandontheprojectaltogether.But
thisisasolvableproblem.
FROMTHEMVPTOINNOVATIONACCOUNTING
Thesolutiontothisdilemmaisacommitmenttoiteration.You
havetocommittoalocked-inagreement—aheadoftime—thatno
materwhatcomesoftestingtheMVP,youwilnotgiveuphope.
Successfulentrepreneursdonotgiveupattherstsignoftrouble,
nordotheyperseveretheplanerightintotheground.Instead,they
possessauniquecombinationofperseveranceandexibility.
The
possessauniquecombinationofperseveranceandexibility.The
MVPisjusttherststeponajourneyoflearning.Downthatroad
—aftermanyiterations—youmaylearnthatsomeelementofyour
productorstrategyisawedanddecideitistimetomakea
change,whichIcalapivot,toadierentmethodforachieving
yourvision.
Startupsareespecialyatriskwhenoutsidestakeholdersand
investors(especialycorporateCFOsforinternalprojects)havea
crisisofcondence.Whentheprojectwasauthorizedorthe
investmentmade,theentrepreneurpromisedthatthenewproduct
wouldbeworld-changing.Customersweresupposedtoocktoit
inrecordnumbers.Whyaresofewactualydoingso?
Intraditionalmanagement,amanagerwhopromisestodeliver
somethingandfailstodosoisintrouble.Thereareonlytwo
possibleexplanations:afailureofexecutionorafailuretoplan
appropriately.Bothareequalyinexcusable.Entrepreneurial
managersfaceadicultproblem:becausetheplansand
projectionswemakearefulofuncertainty,howcanweclaim
successwhenweinevitablyfailtodeliverwhatwe
promised?Put
anotherway,howdoestheCFOorVCknowthatwe’refailing
becausewelearnedsomethingcriticalandnotbecausewewere
goofingoformisguided?
Thesolutiontothisproblemresidesattheheartofthe
Lean
Startupmodel.Wealneedadisciplined,systematicapproachto
guringoutifwe’remakingprogressanddiscoveringifwe’re
actualyachievingvalidatedlearning.Icalthissysteminnovation
accounting,analternativetotraditionalaccountingdesigned
specificalyforstartups.ItisthesubjectofChapter7.
7MEASURE
Atthebeginning,astartupislitlemorethanamodelonapiece
ofpaper.Thenancialsinthe
businessplanincludeprojections
ofhowmanycustomersthecompanyexpectstoatract,how
muchitwilspend,andhowmuchrevenueandprotthatwil
leadto.It’sanidealthat’susualyfarfromwherethestartupisin
itsearlydays.
Astartup’sjobisto(1)rigorouslymeasurewhereitisrightnow,
confrontingthehardtruthsthatassessmentreveals,andthen(2)
deviseexperimentstolearnhowtomovetherealnumberscloserto
theidealreflectedinthebusinessplan.
Mostproducts—eventheonesthatfail—donothavezero
traction.Mostproductshavesomecustomers,somegrowth,and
somepositiveresults.Oneofthemostdangerousoutcomesfora
startupistobumblealonginthelandofthelivingdead.Employees
andentrepreneurstendtobeoptimisticbynature.Wewanttokeep
believinginourideasevenwhenthewritingisonthewal.Thisis
whythemythofperseveranceissodangerous.
Wealknowstories
ofepicentrepreneurswhomanagedtopuloutavictorywhen
thingsseemedincrediblybleak.Unfortunately,wedon’thearstories
aboutthecountlessnamelessotherswhoperseveredtoolong,
leadingtheircompaniestofailure.
WHYSOMETHINGASSEEMINGLYDULLASACCOUNTINGWILL
CHANGEYOURLIFE
Peopleareaccustomedtothinkingofaccountingasdryandboring,
anecessaryevilused
primarilytopreparenancialreportsand
surviveaudits,butthatisbecauseaccountingissomethingthathas
becometakenforgranted.Historicaly,undertheleadershipof
peoplesuchasAlfredSloanatGeneralMotors,accountingbecame
anessentialpartofthemethodofexertingcentralizedcontrolover
far-ungdivisions.AccountingalowedGMtosetclearmilestones
foreachofitsdivisionsandthenholdeachmanageraccountablefor
hisorherdivision’ssuccessinreachingthosegoals.Al
modern
corporationsusesomevariationofthatapproach.Accountingisthe
keytotheirsuccess.
Unfortunately,standardaccountingisnothelpfulinevaluating
entrepreneurs.Startupsaretoounpredictablefor
forecastsand
milestonestobeaccurate.
Irecentlymetwithaphenomenalstartupteam.Theyarewel
nanced,havesignicantcustomertraction,andaregrowing
rapidly.Theirproductisaleaderinanemerging
categoryof
enterprisesoftwarethatusesconsumermarketingtechniquestosel
intolargecompanies.Forexample,theyrelyonemployee-to-
employeeviraladoptionratherthanatraditionalsalesprocess,
whichmighttargetthechiefinformationocerortheheadof
informationtechnology(IT).Asaresult,theyhavetheopportunity
tousecuting-edgeexperimentaltechniquesastheyconstantly
revisetheirproduct.Duringthemeeting,Iaskedtheteam
asimple
questionthatImakeahabitofaskingstartupswheneverwemeet:
areyoumakingyourproductbeter?Theyalwayssayyes.ThenI
ask:howdoyouknow?Iinvariablygetthisanswer:wel,wearein
engineeringandwemadeanumberofchangeslastmonth,andour
customersseemtolikethem,andouroveralnumbersarehigher
thismonth.Wemustbeontherighttrack.
Thisisthekindofstorytelingthattakesplaceatmoststartup
boardmeetings.Mostmilestonesarebuiltthesameway:hita
certainproductmilestone,maybetalktoafewcustomers,andseeif
thenumbersgoup.Unfortunately,thisisnotagoodindicatorof
whetherastartupismakingprogress.Howdoweknow
thatthe
changeswe’vemadearerelatedtotheresultswe’reseeing?More
changeswe’vemadearerelatedtotheresultswe’reseeing?More
important,howdoweknowthatwearedrawingtherightlessons
fromthosechanges?
Toanswerthesekindsofquestions,startupshaveastrongneed
foranewkindofaccountinggearedspecicalytodisruptive
innovation.That’swhatinnovationaccountingis.
AnAccountability
FrameworkThatWorksAcrossIndustries
Innovationaccountingenablesstartupstoproveobjectivelythat
theyarelearninghowtogrowasustainablebusiness.Innovation
accountingbeginsbyturningtheleap-of-faithassumptionsdiscussed
inChapter5intoaquantitativenancialmodel.Everybusiness
planhassomekindofmodelassociatedwithit,evenifit’swriten
onthebackofanapkin.Thatmodelprovidesassumptionsabout
whatthebusinesswillooklikeatasuccessfulpointin
thefuture.
Forexample,thebusinessplanforanestablishedmanufacturing
companywouldshowitgrowinginproportiontoitssalesvolume.
Astheprotsfromthesalesofgoodsarereinvestedinmarketing
andpromotions,thecompanygainsnewcustomers.Therateof
growthdependsprimarilyonthreethings:theprotabilityofeach
customer,thecostofacquiringnewcustomers,andtherepeat
purchaserateofexistingcustomers.Thehigherthese
valuesare,the
fasterthecompanywilgrowandthemoreprotableitwilbe.
Thesearethedriversofthecompany’sgrowthmodel.
Bycontrast,amarketplacecompanythatmatchesbuyersand
selerssuchaseBaywilhave
adierentgrowthmodel.Itssuccess
dependsprimarilyonthenetworkeectsthatmakeitthepremier
destinationforbothbuyersandselerstotransactbusiness.Selers
wantthemarketplacewiththehighestnumberofpotential
customers.Buyerswantthemarketplacewiththemostcompetition
amongselers,whichleadstothegreatestavailabilityofproducts
andthelowestprices.(Ineconomics,thissometimesiscaled
supply-sideincreasingreturnsanddemand-side
increasingreturns.)
Forthiskindofstartup,theimportantthingtomeasureisthatthe
networkeectsareworking,asevidencedbythehighretentionrate
networkeectsareworking,asevidencedbythehighretentionrate
ofnewbuyersandselers.Ifpeoplestickwiththeproductwith
verylitleatrition,themarketplacewilgrownomaterhowthe
companyacquiresnewcustomers.Thegrowthcurvewillooklike
acompoundinginteresttable,withtherateofgrowth
dependingon
the“interestrate”ofnewcustomerscomingtotheproduct.
Thoughthesetwobusinesseshaveverydierentdriversof
growth,wecanstiluseacommonframeworktoholdtheirleaders
accountable.Thisframework
supportsaccountabilityevenwhenthe
modelchanges.
HOWINNOVATIONACCOUNTINGWORKS—THREELEARNING
MILESTONES
Innovationaccountingworksinthreesteps:rst,useaminimum
viableproducttoestablishrealdataonwherethecompanyisright
now.Withoutaclear-eyedpictureofyourcurrentstatus—nomater
howfarfromthegoalyoumaybe—youcannotbegintotrackyour
progress.
Second,startupsmustatempttotunetheenginefromthe
baselinetowardtheideal.Thismaytakemanyatempts.Afterthe
startuphasmadealthemicrochangesandproductoptimizationsit
cantomoveitsbaselinetowardtheideal,the
companyreachesa
decisionpoint.Thatisthethirdstep:pivotorpersevere.
Ifthecompanyismakinggoodprogresstowardtheideal,that
meansit’slearningappropriatelyandusingthatlearningefectively,
inwhichcaseitmakessense
tocontinue.Ifnot,themanagement
teameventualymustconcludethatitscurrentproductstrategyis
awedandneedsaseriouschange.Whenacompanypivots,it
startstheprocessaloveragain,reestablishinganewbaselineand
thentuningtheenginefromthere.Thesignofasuccessfulpivotis
thattheseengine-tuningactivitiesaremoreproductiveafterthe
pivotthanbefore.
EstablishtheBaseline
Forexample,astartupmightcreateacompleteprototype
ofits
productandoertoselittorealcustomersthroughitsmain
marketingchannel.ThissingleMVPwouldtestmostofthestartup’s
assumptionsandestablishbaselinemetricsforeachassumption
simultaneously.Alternatively,astartupmightprefertobuild
separateMVPsthatareaimedatgetingfeedbackonone
assumptionatatime.Beforebuildingtheprototype,thecompany
mightperformasmoketestwithitsmarketingmaterials.
Thisisan
olddirectmarketingtechniqueinwhichcustomersaregiventhe
opportunitytopreorderaproductthathasnotyetbeenbuilt.A
smoketestmeasuresonlyonething:whethercustomersare
interestedintryingaproduct.Byitself,thisisinsucientto
validateanentiregrowthmodel.Nonetheless,itcanbeveryuseful
togetfeedbackonthisassumptionbeforecommitingmoremoney
andotherresourcestotheproduct.
TheseMVPsprovidethefirstexampleofalearningmilestone.An
MVPalowsastartuptolinrealbaselinedatainitsgrowth
model—conversionrates,sign-upandtrialrates,customerlifetime
value,andsoon—andthisisvaluableasthefoundation
forlearning
aboutcustomersandtheirreactionstoaproductevenifthat
foundationbeginswithextremelybadnews.
Whenoneischoosingamongthemanyassumptionsinabusiness
plan,itmakessensetotest
theriskiestassumptionsfirst.Ifyoucan’t
ndawaytomitigatetheseriskstowardtheidealthatisrequired
forasustainablebusiness,thereisnopointintestingtheothers.For
example,amediabusinessthatisselingadvertisinghastwobasic
assumptionsthattaketheformofquestions:Canitcapturethe
atentionofadenedcustomersegmentonanongoingbasis?and
canitselthatatentiontoadvertisers?Inabusinessinwhichthe
advertisingratesforaparticularcustomersegment
arewelknown,
thefarriskierassumptionistheabilitytocaptureatention.
Therefore,therstexperimentsshouldinvolvecontentproduction
ratherthanadvertisingsales.Perhapsthecompanywilproducea
pilotepisodeorissuetoseehowcustomersengage.
TuningtheEngine
Oncethebaselinehasbeenestablished,thestartupcanwork
towardthesecondlearningmilestone:tuningtheengine.Every
productdevelopment,
marketing,orotherinitiativethatastartup
undertakesshouldbetargetedatimprovingoneofthedriversofits
growthmodel.Forexample,acompanymightspendtime
improvingthedesignofitsproducttomakeiteasierfornew
customerstouse.Thispresupposesthattheactivationrateofnew
customersisadriverofgrowthandthatitsbaselineislowerthan
thecompanywouldlike.Todemonstratevalidatedlearning,the
designchangesmustimprovetheactivationrateofnew
customers.
Iftheydonot,thenewdesignshouldbejudgedafailure.Thisisan
importantrule:agooddesignisonethatchangescustomer
behaviorforthebeter.
Comparetwostartups.Therstcompanysetsoutwithaclear
baselinemetric,ahypothesisaboutwhatwilimprovethatmetric,
andasetofexperimentsdesignedtotestthathypothesis.The
secondteamsitsarounddebatingwhatwouldimprovetheproduct,
implementsseveralofthosechangesatonce,and
celebratesifthere
isanypositiveincreaseinanyofthenumbers.Whichstartupis
morelikelytobedoingeectiveworkandachievinglasting
results?
PivotorPersevere
Overtime,ateamthatislearningitswaytowardasustainable
businesswilseethenumbersinitsmodelrisefromthehorrible
baselineestablishedbytheMVPandconvergetosomethinglikethe
idealoneestablishedinthebusinessplan.Astartupthat
failstodo
sowilseethatidealrecedeeverfartherintothedistance.When
thisisdoneright,eventhemostpowerfulrealitydistortioneld
won’tbeabletocoverupthissimplefact:ifwe’renotmovingthe
driversofourbusinessmodel,we’renotmakingprogress.That
becomesasuresignthatit’stimetopivot.
INNOVATIONACCOUNTINGATIMVU
Here’swhatinnovationaccountinglookedlikeforusintheearly
daysofIMVU.Ourminimumviableproducthadmanydefectsand,
whenwerstreleasedit,extremelylowsales.Wenaturaly
assumedthatthelackofsaleswasrelatedtothelowqualityofthe
product,soweekafterweekweworkedonimprovingthe
quality
oftheproduct,trustingthatoureortswereworthwhile.Attheend
ofeachmonth,wewouldhaveaboardmeetingatwhichwewould
presenttheresults.Thenightbeforetheboardmeeting,we’drun
ourstandardanalytics,measuringconversionrates,customercounts,
andrevenuetoshowwhatagoodjobwehaddone.Forseveral
meetingsinarow,thiscausedalast-minutepanicbecausethe
qualityimprovementswerenotyieldinganychangein
customer
behavior.Thisledtosomefrustratingboardmeetingsatwhichwe
couldshowgreatproduct“progress”butnotmuchinthewayof
businessresults.Afterawhile,ratherthanleaveittothelast
minute,webegantotrackourmetricsmorefrequently,tightening
thefeedbackloopwithproductdevelopment.Thiswasevenmore
depressing.Weekin,weekout,ourproductchangeswerehaving
noefect.
ImprovingaProductonFiveDolarsaDay
Wetrackedthe“funnelmetrics”behaviorsthatwerecriticaltoour
engineofgrowth:customerregistration,thedownloadofour
application,trial,repeatusage,andpurchase.Tohaveenoughdata
tolearn,weneededjustenoughcustomersusingourproducttoget
realnumbersforeachbehavior.Wealocatedabudgetofve
dolarsperday:enoughtobuyclicksonthethen-newGoogle
AdWordssystem.Inthosedays,theminimumyoucould
bidfora
clickwas5cents,buttherewasnooveralminimumtoyour
spending.Thus,wecouldaordtoopenanaccountandgetstarted
eventhoughwehadverylitlemoney.1
eventhoughwehadverylit
lemoney.
Fivedolarsboughtusahundredclicks—everyday.Froma
marketingpointofviewthiswasnotverysignicant,butfor
learningitwaspriceless.Everysingledaywewereabletomeasure
ourproduct’sperformancewithabrandnewsetofcustomers.Also,
eachtimewerevisedtheproduct,wegotabrandnewreportcard
onhowweweredoingtheverynextday.
Forexample,onedaywewoulddebutanewmarketingmessage
aimedatrst-timecustomers.Thenextdaywemightchangethe
waynewcustomerswereinitiatedintotheproduct.Otherdays,we
wouldaddnewfeatures,xbugs,roloutanewvisualdesign,or
tryanewlayoutforourwebsite.Everytime,wetold
ourselveswe
weremakingtheproductbeter,butthatsubjectivecondencewas
puttotheacidtestofrealnumbers.
Dayinanddayoutwewereperformingrandomtrials.Eachday
wasanewexperiment.Each
day’scustomerswereindependentof
thoseofthedaybefore.Mostimportant,eventhoughourgross
numbersweregrowing,itbecameclearthatourfunnelmetrics
werenotchanging.
Hereisagraphfromoneof
IMVU’searlyboardmeetings:
Thisgraphrepresentsapproximatelysevenmonthsofwork.Over
thatperiod,weweremakingconstantimprovementstotheIMVU
product,releasingnewfeaturesonadailybasis.Wewere
conductingalotofin-personcustomerinterviews,andourproduct
developmentteamwasworkingextremelyhard.
CohortAnalysis
Toreadthegraph,youneedtounderstandsomethingcaledcohort
analysis.Thisisoneofthemostimportanttoolsofstartupanalytics.
Althoughitsoundscomplex,itisbasedonasimplepremise.
Althoughitsoundscomplex,itisbasedonasimplepremise.
Insteadoflookingatcumulativetotalsorgrossnumberssuchas
totalrevenueandtotalnumberofcustomers,onelooksatthe
performanceofeachgroupofcustomersthatcomesinto
contact
withtheproductindependently.Eachgroupiscaledacohort.The
graphshowstheconversionratestoIMVUofnewcustomerswho
joinedineachindicatedmonth.Eachconversionrateshowsthe
percentageofcustomerwhoregisteredinthatmonthwho
subsequentlywentontotaketheindicatedaction.Thus,amongal
thecustomerswhojoinedIMVUinFebruary2005,about60
percentofthemloggedintoourproductatleastonetime.
Managerswithanenterprisesalesbackgroundwilrecognizethis
funnelanalysisasthetraditionalsalesfunnelthatisusedtomanage
prospectsontheirwaytobecomingcustomers.LeanStartupsuseit
inproductdevelopment,too.Thistechniqueisusefulin
manytypes
ofbusiness,becauseeverycompanydependsforitssurvivalon
sequencesofcustomerbehaviorcaledows.Customerows
governtheinteractionofcustomerswithacompany’sproducts.
Theyalowustounderstandabusinessquantitativelyandhave
muchmorepredictivepowerthandotraditionalgrossmetrics.
Ifyoulookclosely,you’lseethatthegraphshowssomeclear
trends.Someproductimprovementsarehelping—a
litle.The
percentageofnewcustomerswhogoontousetheproductatleast
vetimeshasgrownfromlessthan5percenttoalmost20percent.
Yetdespitethisfourfoldincrease,thepercentageofnewcustomers
whopaymoneyforIMVUisstuckataround1percent.Think
aboutthatforamoment.Aftermonthsandmonthsofwork,
thousandsofindividualimprovements,focusgroups,design
sessions,andusabilitytests,thepercentageofnew
customerswho
subsequentlypaymoneyisexactlythesameasitwasattheonset
eventhoughmanymorecustomersaregetingachancetotrythe
product.
Thankstothepowerofcohortanalysis,wecouldnot
blamethis
failureonthelegacyofpreviouscustomerswhowereresistantto
change,externalmarketconditions,oranyotherexcuse.Each
cohortrepresentedanindependentreportcard,andtryaswe
might,weweregetingstraightC’s.Thishelpedusrealizewehada
might,weweregetingstraightC’s.Thishelpedusrealizewehada
problem.
Iwasinchargeoftheproductdevelopmentteam,smalthoughit
wasinthosedays,andsharedwithmycofoundersthesensethatthe
problemhadtobewithmyteam’seorts.Iworkedharder,triedto
focusonhigher-andhigher-qualityfeatures,andlostalotofsleep.
Ourfrustrationgrew.WhenIcouldthinkofnothingelseto
do,I
wasnalyreadytoturntothelastresort:talkingtocustomers.
Armedwithourfailuretomakeprogresstuningourengineof
growth,Iwasreadytoasktherightquestions.
Beforethisfailure,inthe
company’searliestdays,itwaseasyto
talktopotentialcustomersandcomeawayconvincedwewereon
therighttrack.Infact,whenwewouldinvitecustomersintothe
oceforin-personinterviewsandusabilitytests,itwaseasyto
dismissnegativefeedback.Iftheydidn’twanttousetheproduct,I
assumedtheywerenotinourtargetmarket.“Firethatcustomer,”
I’dsaytothepersonresponsibleforrecruitingforourtests.“Find
mesomeoneinourtargetdemographic.”Ifthenext
customerwas
morepositive,IwouldtakeitasconfirmationthatIwasrightinmy
targeting.Ifnot,I’dfireanothercustomerandtryagain.
Bycontrast,onceIhaddatainhand,myinteractionswith
customerschanged.Suddenly
Ihadurgentquestionsthatneeded
answering:Whyaren’tcustomersrespondingtoourproduct
“improvements”?Whyisn’tourhardworkpayingo?For
example,wekeptmakingiteasierandeasierforcustomerstouse
IMVUwiththeirexistingfriends.Unfortunately,customersdidn’t
wanttoengageinthatbehavior.Makingiteasiertousewastotaly
besidethepoint.Onceweknewwhattolookfor,genuine
understandingcamemuchfaster.Aswasdescribedin
Chapter3,
thiseventualyledtoacriticalyimportantpivot:awayfromanIM
add-onusedwithexistingfriendsandtowardastand-alonenetwork
onecanusetomakenewfriends.Suddenly,ourworriesabout
productivityvanished.Onceoureortswerealignedwithwhat
customersrealywanted,ourexperimentsweremuchmorelikely
tochangetheirbehaviorforthebeter.
Thispaternwouldrepeattimeandagain,fromthedayswhen
weweremakinglessthanathousanddolarsinrevenuepermonth
weweremakinglessthanathousanddolarsinrevenuepermonth
althewayuptothetimeweweremakingmilions.Infact,thisis
thesignofasuccessfulpivot:thenewexperimentsyourun
are
overalmoreproductivethantheexperimentsyouwererunning
before.
Thisisthepatern:poorquantitativeresultsforceustodeclare
failureandcreatethemotivation,context,and
spaceformore
qualitativeresearch.Theseinvestigationsproducenewideas—new
hypotheses—tobetested,leadingtoapossiblepivot.Eachpivot
unlocksnewopportunitiesforfurtherexperimentation,andthe
cyclerepeats.Eachtimewerepeatthissimplerhythm:establishthe
baseline,tunetheengine,andmakeadecisiontopivotor
persevere.
OPTIMIZATIONVERSUSLEARNING
Engineers,designers,andmarketersarealskiledat
optimization.
Forexample,directmarketersareexperiencedatsplittestingvalue
propositionsbysendingadierentoertotwosimilargroupsof
customerssothattheycanmeasuredierencesintheresponserates
ofthetwogroups.Engineers,ofcourse,areskiledatimprovinga
product’sperformance,justasdesignersaretalentedatmaking
productseasiertouse.Altheseactivitiesinawel-runtraditional
organizationoerincrementalbenetforincrementaleort.
As
longasweareexecutingtheplanwel,hardworkyieldsresults.
However,thesetoolsforproductimprovementdonotworkthe
samewayforstartups.Ifyouarebuildingthewrongthing,
optimizingtheproductorits
marketingwilnotyieldsignicant
results.Astartuphastomeasureprogressagainstahighbar:
evidencethatasustainablebusinesscanbebuiltarounditsproducts
orservices.That’sastandardthatcanbeassessedonlyifastartup
hasmadeclear,tangiblepredictionsaheadoftime.
Intheabsenceofthosepredictions,productandstrategydecisions
arefarmoredicultandtime-consuming.Ioftenseethisinmy
consultingpractice.I’vebeencaledinmanytimestohelpastartup
thatfeelsthatitsengineeringteam“isn’tworkinghardenough.”
thatfeelsthatitsengineeringteam“isn’tworkinghardenough.”
WhenImeetwiththoseteams,therearealwaysimprovementsto
bemadeandIrecommendthem,butinvariablythereal
problemis
notalackofdevelopmenttalent,energy,oreort.Cycleaftercycle,
theteamisworkinghard,butthebusinessisnotseeingresults.
Managerstrainedinatraditionalmodeldrawthelogical
conclusion:ourteamisnotworkinghard,notworkingeectively,
ornotworkingeficiently.
Thusthedownwardcyclebegins:theproductdevelopmentteam
valiantlytriestobuildaproductaccordingtothespecicationsitis
receivingfromthecreativeorbusinessleadership.Whengood
resultsarenotforthcoming,businessleadersassumethatany
discrepancybetweenwhatwasplannedandwhatwasbuiltisthe
causeandtrytospecifythenextiterationingreater
detail.Asthe
specicationsgetmoredetailed,theplanningprocessslowsdown,
batchsizeincreases,andfeedbackisdelayed.Ifaboardofdirectors
orCFOisinvolvedasastakeholder,itdoesn’ttakelongfor
personnelchangestofolow.
Afewyearsago,ateamthatselsproductstolargemedia
companiesinvitedmetohelpthemasaconsultantbecausethey
wereconcernedthattheirengineerswerenotworkinghardenough.
However,thefaultwasnotin
theengineers;itwasintheprocess
thewholecompanywasusingtomakedecisions.Theyhad
customersbutdidnotknowthemverywel.Theyweredeluged
withfeaturerequestsfromcustomers,theinternalsalesteam,and
thebusinessleadership.Everynewinsightbecameanemergency
thathadtobeaddressedimmediately.Asaresult,long-term
projectswerehamperedbyconstantinterruptions.Evenworse,the
teamhadnoclearsenseofwhetheranyofthechanges
theywere
makingmateredtocustomers.Despitetheconstanttuningand
tweaking,thebusinessresultswereconsistentlymediocre.
Learningmilestonespreventthisnegativespiralbyemphasizinga
morelikelypossibility:the
companyisexecuting—withdiscipline!
—aplanthatdoesnotmakesense.Theinnovationaccounting
frameworkmakesitclearwhenthecompanyisstuckandneedsto
changedirection.
Intheexampleabove,early
inthecompany’slife,theproduct
Intheexampleabove,earlyinthecompany’slife,theproduct
developmentteamwasincrediblyproductivebecausethe
company’sfoundershadidentiedalargeunmetneedinthetarget
market.Theinitialproduct,whileawed,waspopularwithearly
adopters.Addingthemajorfeaturesthatcustomersaskedfor
seemedtoworkwonders,astheearlyadoptersspreadtheword
abouttheinnovationfarandwide.Butunaskedand
unanswered
wereotherlurkingquestions:Didthecompanyhaveaworking
engineofgrowth?Wasthisearlysuccessrelatedtothedailywork
oftheproductdevelopmentteam?Inmostcases,theanswerwas
no;successwasdrivenbydecisionstheteamhadmadeinthepast.
Noneofitscurrentinitiativeswerehavinganyimpact.Butthiswas
obscuredbecausethecompany’sgrossmetricswereal“upandto
theright.”
Aswe’lseeinamoment,thisisacommondanger.Companies
ofanysizethathaveaworkingengineofgrowthcancometorely
onthewrongkindofmetricstoguidetheiractions.Thisiswhat
temptsmanagerstoresorttotheusualbagofsuccess
theatertricks:
last-minuteadbuys,channelstung,andwhiz-bangdemos,ina
desperateatempttomakethegrossnumberslookbeter.Energy
investedinsuccesstheaterisenergythatcouldhavebeenusedto
helpbuildasustainablebusiness.Icalthetraditionalnumbers
usedtojudgestartups“vanitymetrics,”andinnovationaccounting
requiresustoavoidthetemptationtousethem.
VANITYMETRICS:AWORDOFCAUTION
Toseethedangerofvanitymetricsclearly,let’sreturnoncemoreto
theearlydaysofIMVU.Takealookatthefolowinggraph,which
isfromthesameerainIMVU’shistoryasthatshownearlierinthis
chapter.Itcoversthesametimeperiodasthecohort-stylegraphon
thispage;infact,itisfromthesameboardpresentation.
ThisgraphshowsthetraditionalgrossmetricsforIMVUsofar:
totalregisteredusersandtotalpayingcustomers(thegrossrevenue
graphlooksalmostthesame).Fromthisviewpoint,thingslook
muchmoreexciting.That’swhyIcalthesevanitymetrics:theygive
muchmoreexciting.That’swhyIcalthesevanitymetrics:theygive
therosiestpossiblepicture.You’lseeatraditionalhockeystick
graph(theidealinarapid-growthcompany).Aslongasyoufocus
onthetop-linenumbers(signingupmorecustomers,anincreasein
overalrevenue),you’lbeforgivenforthinkingthis
product
developmentteamismakinggreatprogress.Thecompany’sgrowth
engineisworking.Eachmonthitisabletoacquirecustomersand
hasapositivereturnoninvestment.Theexcessrevenuefromthose
customersisreinvestedthenextmonthinacquiringmore.That’s
wherethegrowthiscomingfrom.
Butthinkbacktothesamedatapresentedinacohortstyle.
IMVUisaddingnewcustomers,butitisnotimprovingtheyieldon
eachnewgroup.Theengineisturning,buttheeortstotunethe
eachnewgroup.Theengineisturning,buttheeortstotunethe
enginearenotbearingmuchfruit.Fromthetraditionalgraph
alone,youcannottelwhetherIMVUisonpacetobuilda
sustainablebusiness;youcertainlycan’ttelanythingaboutthe
eficacyoftheentrepreneurialteambehindit.
Innovationaccountingwilnotworkifastartupisbeingmisled
bythesekindsofvanitymetrics:grossnumberof
customersandso
on.Thealternativeisthekindofmetricsweusetojudgeour
businessandourlearningmilestones,whatIcalactionablemetrics.
ACTIONABLEMETRICSVERSUSVANITYMETRICS
Togetabetersenseoftheimportanceofgoodmetrics,let’slookat
acompanycaledGrockit.Itsfounder,FarboodNivi,spentadecade
workingasateacherattwolargefor-proteducationcompanies,
PrincetonReviewandKaplan,helpingstudents
preparefor
standardizedtestssuchastheGMAT,LSAT,andSAT.Hisengaging
classroomstylewonaccoladesfromhisstudentsandpromotions
fromhissuperiors;hewashonoredwithPrincetonReview’s
NationalTeacheroftheYearaward.ButFarbwasfrustratedwith
thetraditionalteachingmethodsusedbythosecompanies.Teaching
sixtoninehoursperdaytothousandsofstudents,hehadmany
opportunitiestoexperimentwithnewapproaches.2
Overtime,Farbconcludedthatthetraditionallecturemodelof
education,withitsone-to-manyinstructionalapproach,was
inadequateforhisstudents.Hesetouttodevelopasuperior
approach,usingacombinationofteacher-led
lectures,individual
homework,andgroupstudy.Inparticular,Farbwasfascinatedby
howeectivethestudent-to-studentpeer-drivenlearningmethod
wasforhisstudents.Whenstudentscouldhelpeachother,they
benetedintwoways.First,theycouldgetcustomizedinstruction
fromapeerwhowasmuchlessintimidatingthanateacher.
Second,theycouldreinforcetheirlearningbyteachingittoothers.
Overtime,Farb’sclassesbecameincreasinglysocial—
andsuccessful.
Asthisunfolded,Farbfeltmoreandmorethathisphysical
presenceintheclassroomwaslessimportant.Hemadean
presenceintheclassroomwaslessimportant.Hemadean
importantconnection:“Ihavethissociallearningmodelinmy
classroom.There’salthissocialstugoingonontheweb.”His
ideawastobringsocialpeer-to-peerlearningtopeoplewhocould
notaordanexpensiveclassfromKaplanorPrinceton
Reviewor
anevenmoreexpensiveprivatetutor.FromthisinsightGrockitwas
born.
Farbexplains,“Whetheryou’restudyingfortheSAToryou’re
studyingforalgebra,youstudyinoneofthreeways.
Youspend
sometimewithexperts,youspendsometimeonyourown,and
youspendsometimewithyourpeers.Grockitoersthesethree
sameformatsofstudying.Whatwedoisweapplytechnologyand
algorithmstooptimizethosethreeforms.”
Farbistheclassicentrepreneurialvisionary.Herecountshis
originalinsightthisway:“Let’sforgeteducationaldesignupuntil
now,let’sforgetwhat’spossibleandjustredesignlearningwith
today’sstudentsandtoday’stechnologyinmind.Therewereplenty
ofmulti-bilion-dolarorganizationsintheeducationspace,andI
don’tthinktheywereinnovatinginthewaythatweneededthem
toandIdidn’tthinkweneededthemanymore.To
me,it’srealyal
aboutthestudentsandIdidn’tfeellikethestudentswerebeing
servedaswelastheycould.”
TodayGrockitoersmanydierenteducationalproducts,butin
thebeginningFarbfolowedaleanapproach.Grockitbuilt
a
minimumviableproduct,whichwassimplyFarbteachingtestprep
viathepopularonlinewebconferencingtoolWebEx.Hebuiltno
customsoftware,nonewtechnology.Hesimplyatemptedtobring
hisnewteachingapproachtostudentsviatheInternet.Newsabout
anewkindofprivatetutoringspreadquickly,andwithinafew
monthsFarbwasmakingadecentlivingteachingonline,with
monthlyrevenuesof$10,000to$15,000.Butlikemany
entrepreneurswithambition,Farbdidn’tbuildhisMVPjustto
makealiving.Hehadavisionofamorecolaborative,more
eectivekindofteachingforstudentseverywhere.Withhisinitial
traction,hewasabletoraisemoneyfromsomeofthe
most
prestigiousinvestorsinSiliconValey.
WhenIrstmetFarb,hiscompanywasalreadyonthefasttrack
WhenIrstmetFarb,hiscompanywasalreadyonthefasttrack
tosuccess.Theyhadraised
venturecapitalfromwel-regarded
investors,hadbuiltanawesometeam,andwerefreshoan
impressivedebutatoneofSiliconValey’sfamousstartup
competitions.
Theywereextremely
process-orientedanddisciplined.Their
productdevelopmentfolowedarigorousversionoftheagile
developmentmethodologyknownasExtremeProgramming
(describedbelow),thankstotheirpartnershipwithaSan
Francisco–basedcompanycaledPivotalLabs.Theirearlyproduct
washailedbythepressasabreakthrough.
Therewasonlyoneproblem:theywerenotseeingsucient
growthintheuseoftheproductbycustomers.Grockitisan
excelentcasestudybecauseitsproblemswerenotamaterof
failureofexecutionordiscipline.
Folowingstandardagilepractice,Grockit’sworkproceededina
seriesofsprints,orone-monthiterationcycles.Foreachsprint,Farb
wouldprioritizetheworktobedonethatmonthbywritingaseries
ofuserstories,atechniquetakenfromagiledevelopment.Instead
ofwritingaspecicationforanewfeaturethatdescribeditin
technicalterms,Farbwouldwriteastorythatdescribed
thefeature
fromthepointofviewofthecustomer.Thatstoryhelpedkeepthe
engineersfocusedonthecustomer’sperspectivethroughoutthe
developmentprocess.
Eachfeaturewasexpressedinplainlanguageinterms
everyone
couldunderstandwhethertheyhadatechnicalbackgroundornot.
Againfolowingstandardagilepractice,Farbwasfreeto
reprioritizethesestoriesatanytime.Ashelearnedmoreabout
whatcustomerswanted,hecouldmovethingsaroundinthe
productbacklog,thequeueofstoriesyettobebuilt.Theonlylimit
onthisabilitytochangedirectionswasthathecouldnotinterrupt
anytaskthatwasinprogress.Fortunately,thestorieswere
writen
insuchawaythatthebatchsizeofwork(whichI’ldiscussinmore
detailinChapter9)wasonlyadayortwo.
Thissystemiscaledagiledevelopmentforagoodreason:teams
thatemployitareableto
changedirectionquickly,staylighton
theirfeet,andbehighlyresponsivetochangesinthebusiness
theirfeet,andbehighlyresponsivetochangesinthebusiness
requirementsoftheproductowner(themanageroftheprocess—in
thiscaseFarb—whoisresponsibleforprioritizingthestories).
Howdidtheteamfeelattheendofeachsprint?They
consistentlydeliverednewproductfeatures.Theywouldcolect
feedbackfromcustomersintheformofanecdotesandinterviews
thatindicatedthatatleastsomecustomerslikedthenewfeatures.
Therewasalwaysacertainamountofdatathatshowed
improvement:perhapsthetotalnumberofcustomerswas
increasing,thetotalnumberofquestionsansweredbystudentswas
goingup,orthenumberofreturningcustomerswasincreasing.
However,IsensedthatFarbandhisteamwereleftwithlingering
doubtsaboutthecompany’soveralprogress.Wastheincreasein
theirnumbersactualycausedbytheirdevelopmenteorts?
Or
coulditbeduetootherfactors,suchasmentionsofGrockitinthe
press?WhenImettheteam,Iaskedthemthissimplequestion:
HowdoyouknowthattheprioritizationdecisionsthatFarbis
makingactualymakesense?
Theiranswer:“That’snotourdepartment.Farbmakesthe
decisions;weexecutethem.”
AtthattimeGrockitwasfocusedonjustonecustomersegment:
prospectivebusinessschoolstudentswhowerestudyingforthe
GMAT.Theproductalowedstudentstoengageinonlinestudy
sessionswithfelowstudentswhowerestudyingforthesameexam.
Theproductwasworking:thestudentswhocompletedtheir
studyingviaGrockitachievedsignicantlyhigherscoresthanthey
hadbefore.ButtheGrockitteamwasstrugglingwiththeage-old
startupproblems:Howdoweknowwhichfeaturestoprioritize?
Howcanwegetmorecustomerstosignupandpay?Howcanwe
getoutthewordaboutourproduct?
IputthisquestiontoFarb:“Howcondentareyouthatyouare
makingtherightdecisionsintermsofestablishingpriorities?”Like
moststartupfounders,hewaslookingattheavailabledataand
makingthebesteducatedguesseshecould.Butthis
leftalotof
roomforambiguityanddoubt.
Farbbelievedinhisvisionthoroughlyandcompletely,yethewas
startingtoquestionwhetherhiscompanywasonpacetorealize
startingtoquestionwhether
hiscompanywasonpacetorealize
thatvision.Theproductimprovedeveryday,butFarbwantedto
makesurethoseimprovementsmateredtocustomers.Ibelievehe
deservesalotofcreditforrealizingthis.Unlikemanyvisionaries,
whoclingtotheiroriginalvisionnomaterwhat,Farbwaswiling
toputhisvisiontothetest.
Farbworkedhardtosustainhisteam’sbeliefthatGrockitwas
destinedforsuccess.Hewasworriedthatmoralewouldsuerif
anyonethoughtthatthepersonsteeringtheshipwasuncertain
aboutwhichdirectiontogo.Farbhimselfwasn’tsureifhisteam
wouldembraceatruelearningculture.Afteral,thiswaspartof
thegrandbargainofagiledevelopment:engineers
agreetoadapt
theproducttothebusiness’sconstantlychangingrequirementsbut
arenotresponsibleforthequalityofthosebusinessdecisions.
Agileisanecientsystemofdevelopmentfromthepointof
viewofthedevelopers.Italowsthemtostayfocusedoncreating
featuresandtechnicaldesigns.Anatempttointroducetheneedto
learnintothatprocesscouldundermineproductivity.
(Leanmanufacturingfacedsimilarproblemswhenitwas
introducedinfactories.Managerswereusedtofocusingonthe
utilizationrateofeachmachine.Factoriesweredesignedtokeep
machinesrunningatfulcapacityasmuchofthetimeaspossible.
Viewedfromtheperspectiveofthemachine,thatise
cient,but
fromthepointofviewoftheproductivityoftheentirefactory,itis
wildlyinecientattimes.Astheysayinsystemstheory,thatwhich
optimizesonepartofthesystemnecessarilyunderminesthesystem
asawhole.)
WhatFarbandhisteamdidn’trealizewasthatGrockit’sprogress
wasbeingmeasuredbyvanitymetrics:thetotalnumberof
customersandthetotalnumberofquestionsanswered.Thatwas
whatwascausinghisteamtospinitswheels;thosemetricsgavethe
teamthesensationofforwardmotioneventhoughthecompany
wasmakinglitleprogress.What’sinterestingishowcloselyFarb’s
methodfolowedsupercialaspectsoftheLeanStartup
learning
milestones:theyshippedanearlyproductandestablishedsome
baselinemetrics.Theyhadrelativelyshortiterations,eachofwhich
wasjudgedbyitsabilitytoimprovecustomermetrics.
wasjudgedbyitsabilityto
improvecustomermetrics.
However,becauseGrockitwasusingthewrongkindsofmetrics,
thestartupwasnotgenuinelyimproving.Farbwasfrustratedinhis
eortstolearnfromcustomerfeedback.Ineverycycle,thetypeof
metricshisteamwasfocusedonwouldchange:onemonththey
wouldlookatgrossusagenumbers,anothermonthregistration
numbers,andsoon.Thosemetricswouldgoupanddown
seeminglyontheirown.Hecouldn’tdrawclearcause-
and-eect
inferences.Prioritizingworkcorrectlyinsuchanenvironmentis
extremelychalenging.
Farbcouldhaveaskedhisdataanalysttoinvestigateaparticular
question.Forexample,whenweshippedfeatureX,didita
ect
customerbehavior?Butthatwouldhaverequiredtremendoustime
andeort.When,exactly,didfeatureXship?Whichcustomers
wereexposedtoit?Wasanythingelselaunchedaroundthatsame
time?Werethereseasonalfactorsthatmightbeskewingthedata?
Findingtheseanswerswouldhaverequiredparsingreamsand
reamsofdata.Theansweroftenwouldcomeweeksafterthe
questionhadbeenasked.Inthemeantime,theteam
wouldhave
movedontonewprioritiesandnewquestionsthatneededurgent
atention.
Comparedtoalotofstartups,theGrockitteamhadahuge
advantage:theyweretremendouslydisciplined.Adisciplinedteam
mayapplythewrongmethodologybutcanshiftgearsquicklyonce
itdiscoversitserror.Mostimportant,adisciplinedteamcan
experimentwithitsownworkingstyleanddrawmeaningful
conclusions.
CohortsandSplit-tests
Grockitchangedthemetricstheyusedtoevaluatesuccessintwo
ways.Insteadoflookingatgrossmetrics,Grockitswitchedto
cohort-basedmetrics,andinsteadoflookingforcause-and-eect
relationshipsafterthefact,Grockitwouldlauncheachnewfeature
asatruesplit-testexperiment.
Asplit-testexperimentisoneinwhichdierentversionsofa
Asplit-testexperimentisoneinwhichdierentversionsofa
productareoeredtocustomersatthesametime.Byobservingthe
changesinbehaviorbetweenthetwogroups,onecanmake
inferencesabouttheimpactofthedierentvariations.This
techniquewaspioneeredbydirectmailadvertisers.Forexample,
consideracompanythatsendscustomersacatalogofproductsto
buy,suchasLands’EndorCrate&Barrel.Ifyouwantedtotesta
catalogdesign,youcouldsendanewversionofitto50percentof
thecustomersandsendtheoldstandardcatalogtothe
other50
percent.Toassureascienticresult,bothcatalogswouldcontain
identicalproducts;theonlydierencewouldbethechangestothe
design.Togureoutifthenewdesignwaseective,alyouwould
havetodowaskeeptrackofthesalesguresforbothgroupsof
customers.(ThistechniqueissometimescaledA/Btestingafterthe
practiceofassigningleternamestoeachvariation.)Althoughsplit
testingoftenisthoughtofasamarketing-specic(oreven
adirect
marketing–specic)practice,LeanStartupsincorporateitdirectly
intoproductdevelopment.
ThesechangesledtoanimmediatechangeinFarb’s
understandingofthebusiness.Splittestingoftenuncoverssurprising
things.Forexample,manyfeaturesthatmaketheproductbeterin
theeyesofengineersanddesignershavenoimpactoncustomer
behavior.ThiswasthecaseatGrockit,asithasbeeninevery
companyIhaveseenadoptthistechnique.Although
workingwith
splittestsseemstobemoredicultbecauseitrequiresextra
accountingandmetricstokeeptrackofeachvariation,italmost
alwayssavestremendousamountsoftimeinthelongrunby
eliminatingworkthatdoesn’tmatertocustomers.
Splittestingalsohelpsteamsrenetheirunderstandingofwhat
customerswantanddon’twant.Grockit’steamconstantlyadded
newwaysfortheircustomerstointeractwitheachotherinthe
hopethatthosesocialcommunicationtoolswouldincreasethe
product’svalue.Inherentinthoseeortswasthebeliefthat
customersdesiredmorecommunicationduringtheirstudying.
Whensplittestingrevealedthattheextrafeaturesdidnot
change
customerbehavior,itcaledthatbeliefintoquestion.
Thequestioninginspiredtheteamtoseekadeeper
Thequestioninginspiredtheteamtoseekadeeper
understandingofwhatcustomersrealywanted.Theybrainstormed
newideasforproductexperimentsthatmighthavemoreimpact.In
fact,manyoftheseideaswerenotnew.Theyhadsimplybeen
overlookedbecausethecompanywasfocusedonbuildingsocial
tools.Asaresult,Grockittestedanintensivesolo-
studyingmode,
completewithquestsandgamelikelevels,sothatstudentscould
havethechoiceofstudyingbythemselvesorwithothers.Justasin
Farb’soriginalclassroom,thisprovedextremelyeective.Without
thedisciplineofsplittesting,thecompanymightnothavehadthis
realization.Infact,overtime,throughdozensoftests,itbecame
clearthatthekeytostudentengagementwastooerthema
combinationofsocialandsolofeatures.Students
preferredhavinga
choiceofhowtostudy.
Kanban
Folowingtheleanmanufacturingprincipleofkanban,orcapacity
constraint,Grockitchangedtheproductprioritizationprocess.
Underthenewsystem,userstorieswerenotconsideredcomplete
untiltheyledtovalidatedlearning.Thus,storiescouldbecataloged
asbeinginoneoffourstatesofdevelopment:intheproduct
backlog,activelybeingbuilt,done(featurecompletefrom
a
technicalpointofview),orintheprocessofbeingvalidated.
Validatedwasdenedas“knowingwhetherthestorywasagood
ideatohavebeendoneintherstplace.”Thisvalidationusualy
wouldcomeintheformofasplittestshowingachangein
customerbehaviorbutalsomightincludecustomerinterviewsor
surveys.
Thekanbanrulepermitedonlysomanystoriesineachofthe
fourstates.Asstoriesow
fromonestatetotheother,thebuckets
lup.Onceabucketbecomesful,itcannotacceptmorestories.
Onlywhenastoryhasbeenvalidatedcanitberemovedfromthe
kanbanboard.Ifthevalidationfailsanditturnsoutthestoryisa
badidea,therelevantfeatureisremovedfromtheproduct(seethe
chartonthispage).
KANBANDIAGRAMOFWORKASITPROGRESSES
FROMSTAGETOSTAGE
(Nobucketcancontainmorethanthreeprojectsatatime.)
WorkonAbegins.DandEareindevelopment.Fawaitsvalidation.
Fisvalidated.DandEawaitvalidation.G,H,Iarenewtaskstobeundertaken.BandC
arebeingbuilt.Acompletesdevelopment.
BandChavebeenbuilt,butunderkanban,cannotbemovedtothenextbucketfor
validationuntilA,D,Ehave
beenvalidated.WorkcannotbeginonHandIuntilspace
opensupinthebucketsahead.
Ihaveimplementedthissystemwithseveralteams,andthe
Ihaveimplementedthissystemwithseveralteams,andthe
initialresultisalwaysfrustrating:eachbucketlsup,startingwith
the“validated”bucketandmovingontothe“done”bucket,until
it’snotpossibletostartanymorework.Teamsthatareusedto
measuringtheirproductivitynarrowly,bythenumberof
stories
theyaredelivering,feelstuck.Theonlywaytostartworkonnew
featuresistoinvestigatesomeofthestoriesthataredonebut
haven’tbeenvalidated.Thatoftenrequiresnonengineeringeorts:
talkingtocustomers,lookingatsplit-testdata,andthelike.
Pretysooneveryonegetsthehangofit.Thisprogressoccursin
tsandstartsatrst.Engineeringmaynishabigbatchofwork,
folowedbyextensivetestingandvalidation.Asengineerslookfor
waystoincreasetheirproductivity,theystarttorealizethatifthey
includethevalidationexercisefromthebeginning,thewholeteam
canbemoreproductive.
Forexample,whybuildanewfeaturethatisnotpartofasplit-
testexperiment?Itmaysaveyoutimeintheshortrun,butitwil
takemoretimelatertotest,duringthevalidationphase.Thesame
logicappliestoastorythatanengineerdoesn’tunderstand.Under
theoldsystem,heorshewouldjustbuilditandndout
laterwhat
itwasfor.Inthenewsystem,thatbehaviorisclearly
counterproductive:withoutaclearhypothesis,howcanastoryever
bevalidated?WesawthisbehavioratIMVU,too.Ioncesawa
juniorengineerfacedowna
seniorexecutiveoverarelatively
minorchange.Theengineerinsistedthatthenewfeaturebesplit-
tested,justlikeanyother.Hispeersbackedhimup;itwas
consideredabsolutelyobviousthatalfeaturesshouldberoutinely
tested,nomaterwhowascommissioningthem.(Embarrassingly,
altoooftenIwastheexecutiveinquestion.)Asolidprocesslays
thefoundationforahealthyculture,onewhereideasareevaluated
bymeritandnotbyjobtitle.
Mostimportant,teamsworkinginthissystembegintomeasure
theirproductivityaccordingtovalidatedlearning,notintermsof
theproductionofnewfeatures.
HypothesisTestingatGrockit
WhenGrockitmadethistransition,theresultsweredramatic.In
onecase,theydecidedtotestoneoftheirmajorfeatures,caled
lazyregistration,toseeifitwasworththeheavyinvestmentthey
weremakinginongoingsupport.Theywerecondent
inthis
featurebecauselazyregistrationisconsideredoneofthedesignbest
practicesforonlineservices.Inthissystem,customersdonothave
toregisterfortheserviceupfront.Instead,theyimmediatelybegin
usingtheserviceandareaskedtoregisteronlyaftertheyhavehad
achancetoexperiencetheservice’sbenefit.
Forastudent,lazyregistrationworkslikethis:whenyoucometo
theGrockitwebsite,you’reimmediatelyplacedinastudysession
withotherstudentsworkingonthesametest.Youdon’thaveto
giveyourname,e-mailaddress,orcreditcardnumber.Thereis
nothingtopreventyoufromjumpinginandgetingstarted
immediately.ForGrockit,thiswasessentialtotesting
oneofits
coreassumptions:thatcustomerswouldbewilingtoadoptthis
newwayoflearningonlyiftheycouldseeproofthatitwas
workingearlyon.
Asaresultofthishypothesis,Grockit’sdesignrequired
thatit
managethreeclassesofusers:unregisteredguests,registered(trial)
guests,andcustomerswhohadpaidforthepremiumversionofthe
product.Thisdesignrequiredsignicantextraworktobuildand
maintain:themoreclassesofusersthereare,themoreworkis
requiredtokeeptrackofthem,andthemoremarketingeortis
requiredtocreatetherightincentivestoenticecustomersto
upgradetothenextclass.Grockithadundertakenthis
extraeort
becauselazyregistrationwasconsideredanindustrybestpractice.
Iencouragedtheteamtotryasimplesplit-test.Theytookone
cohortofcustomersandrequiredthattheyregisterimmediately,
basedonnothingmorethanGrockit’smarketingmaterials.Totheir
surprise,thiscohort’sbehaviorwasexactlythesameasthatofthe
lazyregistrationgroup:theyhadthesamerateofregistration,
activation,andsubsequentretention.Inotherwords,the
extraeort
oflazyregistrationwasacompletewasteeventhoughitwas
consideredanindustrybestpractice.
consideredanindustrybestpractice.
Evenmoreimportantthanreducingwastewasthe
insightthat
thistestsuggested:customerswerebasingtheirdecisionabout
Grockitonsomethingotherthantheiruseoftheproduct.
Thinkaboutthis.Thinkaboutthecohortofcustomerswhowere
requiredtoregisterforthe
productbeforeenteringastudysession
withotherstudents.Theyhadverylitleinformationaboutthe
product,nothingmorethanwaspresentedonGrockit’shomepage
andregistrationpage.Bycontrast,thelazyregistrationgrouphada
tremendousamountofinformationabouttheproductbecausethey
hadusedit.Yetdespitethisinformationdisparity,customer
behaviorwasexactlythesame.
ThissuggestedthatimprovingGrockit’spositioningand
marketingmighthaveamoresignicantimpactonatractingnew
customersthanwouldaddingnewfeatures.Thiswasjusttherstof
manyimportantexperimentsGrockitwasabletorun.Sincethose
earlydays,theyhaveexpandedtheircustomerbase
dramaticaly:
theynowoertestprepfornumerousstandardizedtests,including
theGMAT,SAT,ACT,andGRE,aswelasonlinemathandEnglish
coursesforstudentsingrades7through12.
Grockitcontinuestoevolve
itsprocess,seekingcontinuous
improvementateveryturn.Withmorethantwentyemployeesin
itsSanFranciscooce,Grockitcontinuestooperatewiththesame
deliberate,disciplinedapproachthathasbeentheirhalmarkal
along.Theyhavehelpedclosetoamilionstudentsandaresureto
helpmilionsmore.
THEVALUEOFTHETHREEA’S
TheseexamplesfromGrockitdemonstrateeachofthethreeA’sof
metrics:actionable,
accessible,andauditable.
Actionable
Forareporttobeconsideredactionable,itmustdemonstrateclear
Forareporttobeconsideredactionable,itmustdemonstrateclear
causeandeect.Otherwise,itisavanitymetric.The
reportsthat
Grockit’steambegantousetojudgetheirlearningmilestonesmade
itextremelyclearwhatactionswouldbenecessarytoreplicatethe
results.
Bycontrast,vanitymetricsfailthiscriterion.Takethe
numberof
hitstoacompanywebsite.Let’ssaywehave40,000hitsthismonth
—anewrecord.Whatdoweneedtodotogetmorehits?Wel,that
depends.Wherearethenewhitscomingfrom?Isitfrom40,000
newcustomersorfromoneguywithanextremelyactiveweb
browser?ArethehitstheresultofanewmarketingcampaignorPR
push?Whatisahit,anyway?Doeseachpageinthebrowsercount
asonehit,ordoaltheembeddedimagesand
multimediacontent
countaswel?Thosewhohavesatinameetingdebatingtheunits
ofmeasurementinareportwilrecognizethisproblem.
Vanitymetricswreakhavocbecausetheypreyonaweaknessof
thehumanmind.Inmy
experience,whenthenumbersgoup,
peoplethinktheimprovementwascausedbytheiractions,by
whatevertheywereworkingonatthetime.Thatiswhyit’sso
commontohaveameetinginwhichmarketingthinksthenumbers
wentupbecauseofanewPRormarketingeortandengineering
thinksthebeternumbersaretheresultofthenewfeaturesit
added.Findingoutwhatisactualygoingonisextremelycostly,
andsomostmanagerssimplymoveon,doingthebestthey
canto
formtheirownjudgmentonthebasisoftheirexperienceandthe
colectiveinteligenceintheroom.
Unfortunately,whenthenumbersgodown,itresultsinavery
dierentreaction:nowit’s
somebodyelse’sfault.Thus,mostteam
membersordepartmentsliveinaworldwheretheirdepartmentis
constantlymakingthingsbeter,onlytohavetheirhardwork
sabotagedbyotherdepartmentsthatjustdon’tgetit.Isitany
wonderthesedepartmentsdeveloptheirowndistinctlanguage,
jargon,culture,anddefensemechanismsagainstthebozosworking
downthehal?
Actionablemetricsaretheantidotetothisproblem.Whencause
andeectisclearlyunderstood,peoplearebeterabletolearn
fromtheiractions.Humanbeingsareinnatelytalentedlearners
fromtheiractions.Humanbeingsareinnatelytalentedlearners
whengivenaclearandobjectiveassessment.
Accessible
Altoomanyreportsarenotunderstoodbytheemployeesand
managerswhoaresupposedtousethemtoguidetheirdecision
making.Unfortunately,mostmanagersdonotrespondtothis
complexitybyworkinghandinhandwiththedatawarehousing
teamtosimplifythereportssothattheycanunderstandthem
beter.Departmentstoooftenspendtheirenergylearninghowto
usedatatogetwhattheywantratherthanasgenuine
feedbackto
guidetheirfutureactions.
Thereisanantidotetothismisuseofdata.First,makethereports
assimpleaspossiblesothateveryoneunderstandsthem.
Rememberthesaying“Metricsarepeople,too.”Theeasiestwayto
makereportscomprehensibleistousetangible,concreteunits.
Whatisawebsitehit?Nobodyisrealysure,buteveryoneknows
whatapersonvisitingthewebsiteis:onecanpracticalypicture
thosepeoplesitingattheircomputers.
Thisiswhycohort-basedreportsarethegoldstandardoflearning
metrics:theyturncomplexactionsintopeople-basedreports.Each
cohortanalysissays:amongthepeoplewhousedourproductin
thisperiod,here’showmanyofthemexhibitedeachofthe
behaviorswecareabout.IntheIMVUexample,wesawfour
behaviors:downloadingtheproduct,loggingintotheproductfrom
one’scomputer,engaginginachatwithothercustomers,and
upgradingtothepaidversionoftheproduct.Inother
words,the
reportdealswithpeopleandtheiractions,whicharefarmore
usefulthanpilesofdatapoints.Forexample,thinkabouthowhard
itwouldhavebeentotelifIMVUwasbeingsuccessfulifwehad
reportedonlyonthetotalnumberofperson-to-person
conversations.Let’ssaywehave10,000conversationsinaperiod.Is
thatgood?Isthatonepersonbeingvery,verysocial,orisit10,000
peopleeachtryingtheproductonetimeandthengivingup?
There’snowaytoknowwithoutcreatingamoredetailedreport.
There’snowaytoknowwithoutcreatingamoredetailedreport.
Asthegrossnumbersgetlarger,accessibilitybecomesmoreand
moreimportant.Itishardtovisualizewhatitmeansifthe
number
ofwebsitehitsgoesdownfrom250,000inonemonthto200,000
thenextmonth,butmostpeopleunderstandimmediatelywhatit
meanstolose50,000customers.That’spracticalyawholestadium
fulofpeoplewhoareabandoningtheproduct.
Accessibilityalsoreferstowidespreadaccesstothereports.
Grockitdidthisespecialywel.Everydaytheirsystem
automaticalygeneratedadocumentcontainingthelatestdatafor
everysingleoneoftheirsplit-testexperimentsandotherleap-of-
faithmetrics.Thisdocumentwasmailedtoeveryemployeeofthe
company:theyalalwayshadafreshcopyintheire-mailin-boxes.
Thereportswerewellaidoutandeasytoread,witheach
experimentanditsresultsexplainedinplainEnglish.
Anotherwaytomakereportsaccessibleistouseatechniquewe
developedatIMVU.Insteadofhousingtheanalyticsordataina
separatesystem,ourreportingdataanditsinfrastructurewere
consideredpartoftheproductitselfandwereownedbythe
productdevelopmentteam.Thereportswereavailableonour
website,accessibletoanyonewithanemployeeaccount.
Eachemployeecouldlogintothesystematanytime,choose
fromalistofalcurrentandpastexperiments,andseeasimple
one-pagesummaryoftheresults.Overtime,thoseone-page
summariesbecamethedefactostandardforsetlingproduct
argumentsthroughouttheorganization.Whenpeople
needed
evidencetosupportsomethingtheyhadlearned,theywouldbring
aprintoutwiththemtotherelevantmeeting,condentthat
everyonetheyshowedittowouldunderstanditsmeaning.
Auditable
Wheninformedthattheirpetprojectisafailure,mostofusare
temptedtoblamethemessenger,thedata,themanager,thegods,
oranythingelsewecanthinkof.That’swhythethirdAofgood
metrics,“auditable,”issoessential.Wemustensurethatthedatais
metrics,“auditable,”issoessential.Wemustensurethatthedatais
credibletoemployees.
TheemployeesatIMVUwouldbrandishone-pagereportsto
demonstratewhattheyhadlearnedtosetlearguments,butthe
processoftenwasn’tsosmooth.Mostofthetime,whenamanager,
developer,orteamwasconfrontedwithresultsthatwouldkila
petproject,theloseroftheargumentwouldchalenge
theveracity
ofthedata.
Suchchalengesaremorecommonthanmostmanagerswould
admit,andunfortunately,mostdatareportingsystemsarenot
designedtoanswerthemsuccessfuly.Sometimesthis
istheresultof
awel-intentionedbutmisplaceddesiretoprotecttheprivacyof
customers.Moreoften,thelackofsuchsupportingdocumentation
issimplyamaterofneglect.Mostdatareportingsystemsarenot
builtbyproductdevelopmentteams,whosejobistoprioritizeand
buildproductfeatures.Theyarebuiltbybusinessmanagersand
analysts.Managerswhomustusethesesystemscanonlycheckto
seeifthereportsaremutualyconsistent.Theyaltoo
oftenlacka
waytotestifthedataisconsistentwithreality.
Thesolution?First,rememberthat“Metricsarepeople,too.”We
needtobeabletotestthedatabyhand,inthemessyrealworld,by
talkingtocustomers.Thisis
theonlywaytobeabletocheckifthe
reportscontaintruefacts.Managersneedtheabilitytospotcheck
thedatawithrealcustomers.Italsohasasecondbenet:systems
thatprovidethislevelofauditabilitygivemanagersand
entrepreneurstheopportunitytogaininsightsintowhycustomers
arebehavingthewaythedataindicate.
Second,thosebuildingreportsmustmakesurethemechanisms
thatgeneratethereportsarenottoocomplex.Wheneverpossible,
reportsshouldbedrawndirectlyfromthemasterdata,ratherthan
fromanintermediatesystem,whichreducesopportunitiesforerror.
Ihavenoticedthateverytimeateamhasoneofitsjudgmentsor
assumptionsoverturnedasaresultofatechnicalproblem
withthe
data,itsconfidence,morale,anddisciplineareundermined.
Whenwewatchentrepreneurssucceedinthemythmakingworldof
Holywood,books,andmagazines,thestoryisalwaysstructuredthe
sameway.First,weseethepluckyprotagonisthavinganepiphany,
hatchingagreatnewidea.Welearnabouthisorhercharacterand
personality,howheorshecametobeintherightplaceattheright
time,andhowheorshetookthedramaticleaptostarta
business.
Thenthephotomontagebegins.It’susualyshort,justafew
minutesoftime-lapsephotographyornarrative.Weseethe
protagonistbuildingateam,maybeworkinginalab,writingon
whiteboards,closingsales,poundingonafewkeyboards.Atthe
endofthemontage,thefoundersaresuccessful,andthestorycan
moveontomoreinterestingfare:howtosplitthespoilsoftheir
success,whowilappearonmagazinecovers,whosues
whom,and
implicationsforthefuture.
Unfortunately,therealworkthatdeterminesthesuccessof
startupshappensduringthephotomontage.Itdoesn’tmakethecut
intermsofthebigstorybecauseitistooboring.Only
5percentof
entrepreneurshipisthebigidea,thebusinessmodel,the
whiteboardstrategizing,andthesplitingupofthespoils.Theother
95percentisthegrityworkthatismeasuredbyinnovation
accounting:product
prioritizationdecisions,decidingwhich
customerstotargetorlistento,andhavingthecouragetosubjecta
grandvisiontoconstanttestingandfeedback.
Onedecisionstandsoutabovealothersasthemostdicult,the
mosttime-consuming,andthebiggestsourceofwasteformost
startups.Wealmustfacethisfundamentaltest:decidingwhento
pivotandwhentopersevere.Tounderstandwhathappensduring
thephotomontage,wehavetounderstandhowtopivot,
andthatis
thesubjectofChapter8.
8PIVOT(ORPERSEVERE)
Everyentrepreneureventualyfacesanoverridingchalengein
developingasuccessfulproduct:decidingwhentopivotand
whentopersevere.Everythingthathasbeendiscussedsofarisa
preludetoaseeminglysimplequestion:arewemakingsucient
progresstobelievethatouroriginalstrategichypothesisiscorrect,
ordoweneedtomakeamajorchange?Thatchangeis
caleda
pivot:astructuredcoursecorrectiondesignedtotestanew
fundamentalhypothesisabouttheproduct,strategy,andengineof
growth.
Becauseofthescienticmethodologythatunderlies
theLean
Startup,thereisoftenamisconceptionthatitoersarigidclinical
formulaformakingpivotorperseveredecisions.Thisisnottrue.
Thereisnowaytoremovethehumanelement—vision,intuition,
judgment—fromthepracticeofentrepreneurship,norwouldthat
bedesirable.
Mygoalinadvocatingascienticapproachtothecreationof
startupsistochannelhumancreativityintoitsmostproductive
form,andthereisnobiggerdestroyerofcreativepotentialthanthe
misguideddecisiontopersevere.Companiesthatcannotbring
themselvestopivottoanewdirectiononthebasisoffeedback
fromthemarketplacecangetstuckinthelandoftheliving
dead,
neithergrowingenoughnordying,consumingresourcesand
commitmentfromemployeesandotherstakeholdersbutnot
movingahead.
Thereisgoodnewsaboutourrelianceonjudgment,
though.We
Thereisgoodnewsaboutourrelianceonjudgment,though.We
areabletolearn,weareinnatelycreative,andwehavea
remarkableabilitytoseethesignalinthenoise.Infact,weareso
goodatthisthatsometimesweseesignalsthataren’tthere.The
heartofthescienticmethodistherealizationthatalthoughhuman
judgmentmaybefaulty,wecanimproveourjudgmentby
subjectingourtheoriestorepeatedtesting.
Startupproductivityisnotaboutcrankingoutmorewidgetsor
features.Itisaboutaligningoureortswithabusinessandproduct
thatareworkingtocreatevalueanddrivegrowth.Inotherwords,
successfulpivotsputusonapathtowardgrowinga
sustainable
business.
INNOVATIONACCOUNTINGLEADSTOFASTERPIVOTS
Toseethisprocessinaction,meetDavidBineti,theCEOof
Votizen.Davidhashadalongcareerhelpingtobring
theAmerican
politicalprocessintothetwenty-rstcentury.Intheearly1990s,he
helpedbuildUSA.gov,therstportalforthefederalgovernment.
He’salsoexperiencedsomeclassicstartupfailures.Whenitcame
timetobuildVotizen,Davidwasdeterminedtoavoidbetingthe
farmonhisvision.
Davidwantedtotackletheproblemofcivicparticipationinthe
politicalprocess.Hisrstproductconceptwasasocialnetworkof
veriedvoters,aplacewherepeoplepassionateaboutciviccauses
couldgettogether,shareideas,andrecruittheirfriends.Davidbuilt
hisfirstminimumviableproductforjustover$1,200inaboutthree
monthsandlaunchedit.
Davidwasn’tbuildingsomethingthatnobodywanted.Infact,
fromitsearliestdays,Votizenwasabletoatractearlyadopters
wholovedthecoreconcept.Likealentrepreneurs,Davidhadto
renehisproductandbusinessmodel.Whatmade
David’s
chalengeespecialyhardwasthathehadtomakethosepivotsin
thefaceofmoderatesuccess.
David’sinitialconceptinvolvedfourbigleapsoffaith:
1.Customerswouldbeinterestedenoughinthe
socialnetworkto
signup.(Registration)
2.Votizenwouldbeabletoverifythemasregisteredvoters.
(Activation)
3.Customerswhowereveriedvoterswouldengagewiththe
site’sactivismtoolsovertime.(Retention)
4.Engagedcustomerswouldteltheirfriendsabouttheservice
andrecruitthemintociviccauses.(Referral)
Threemonthsand$1,200later,David’srstMVPwasin
customers’hands.Inthe
initialcohorts,5percentsignedupforthe
serviceand17percentveriedtheirregisteredvoterstatus(seethe
chartbelow).Thenumbersweresolowthattherewasn’tenough
datatotelwhatsortofengagementorreferralwouldoccur.Itwas
timetostartiterating.
INITIALMVP
Registration
5%
Activation
17%
Retention
Toolow
Referral
Toolow
Davidspentthenexttwomonthsandanother$5,000splittesting
newproductfeatures,messaging,andimprovingtheproduct’s
designtomakeiteasierto
use.Thosetestsshoweddramatic
improvements,goingfroma5percentregistrationrateto17
percentandfroma17percentactivationratetoover90percent.
Suchisthepowerofsplittesting.ThisoptimizationgaveDavida
criticalmassofcustomerswithwhichtomeasurethenexttwoleaps
offaith.However,asshowninthechartbelow,thosenumbers
provedtobeevenmorediscouraging:Davidachievedareferralrate
ofonly4percentandaretentionrateof5percent.
INITIALMVP
AFTEROPTIMIZATION
Registration
5%
17%
Activation
17%
90%
Retention
Toolow
5%
Referral
Toolow
4%
Davidknewhehadtodo
moredevelopmentandtesting.Forthe
nextthreemonthshecontinuedtooptimize,splittest,andrene
hispitch.Hetalkedtocustomers,heldfocusgroups,anddid
countlessA/Bexperiments.AswasexplainedinChapter7,ina
splittest,dierentversionsofaproductareoeredtodierent
customersatthesametime.Byobservingthechangesinbehavior
betweenthetwogroups,onecanmakeinferencesabouttheimpact
ofthedierentvariations.Asshowninthechartbelow,the
referral
ratenudgedupslightlyto6percentandtheretentionratewentup
to8percent.AdisappointedDavidhadspenteightmonthsand
$20,000tobuildaproductthatwasn’tlivinguptothegrowth
modelhe’dhopedfor.
BEFOREOPTIMIZATIONAFTEROPTIMIZATION
Registration
17%
17%
Activation
90%
90%
Retention
5%
8%
Referral
4%
6%
Davidfacedthedicultchal
engeofdecidingwhethertopivot
orpersevere.Thisisoneofthehardestdecisionsentrepreneursface.
Thegoalofcreatinglearningmilestonesisnottomakethedecision
Thegoalofcreatinglearningmilestonesisnottomakethedecision
easy;itistomakesurethatthereisrelevantdataintheroomwhen
itcomestimetodecide.
Remember,atthispointDavidhashadmanycustomer
conversations.Hehasplentyoflearningthathecanuseto
rationalizethefailurehehas
experiencedwiththecurrentproduct.
That’sexactlywhatmanyentrepreneursdo.InSiliconValey,we
calthisexperiencegetingstuckinthelandofthelivingdead.It
happenswhenacompanyhasachievedamodicumofsuccess—just
enoughtostayalive—butisnotlivinguptotheexpectationsofits
foundersandinvestors.Suchcompaniesareaterribledrainof
humanenergy.Outofloyalty,theemployeesandfoundersdon’t
wanttogivein;theyfeelthatsuccessmightbejustaround
the
corner.
Davidhadtwoadvantagesthathelpedhimavoidthisfate:
1.Despitebeingcommitedtoasignicantvision,hehaddone
hisbesttolaunchearlyanditerate.Thus,hewasfacinga
pivot
orperseveremomentjusteightmonthsintothelifeofhis
company.Themoremoney,time,andcreativeenergythathas
beensunkintoanidea,theharderitistopivot.Davidhad
doneweltoavoidthattrap.
2.Davidhadidentiedhisleap-of-faithquestionsexplicitlyatthe
outsetand,moreimportant,hadmadequantitativepredictions
abouteachofthem.Itwouldnothavebeendicultforhimto
declaresuccessretroactivelyfromthatinitialventure.Afteral,
someofhismetrics,suchasactivation,weredoingquitewel.
Intermsofgrossmetricssuchastotalusage,thecompanyhad
positivegrowth.ItisonlybecauseDavidfocusedon
actionable
metricsforeachofhisleap-of-faithquestionsthathewasable
toacceptthathiscompanywasfailing.Inaddition,because
DavidhadnotwastedenergyonprematurePR,hewasableto
makethisdeterminationwithoutpublicembarrassmentor
distraction.
Failureisaprerequisitetolearning.Theproblemwiththenotion
ofshippingaproductandthenseeingwhathappensisthatyouare
ofshippingaproductandthenseeingwhathappensisthatyouare
guaranteedtosucceed—atseeingwhathappens.Butthenwhat?As
soonasyouhaveahandfulofcustomers,you’relikelytohaveve
opinionsaboutwhattodonext.Whichshouldyou
listento?
Votizen’sresultswereokay,buttheywerenotgoodenough.
Davidfeltthatalthoughhisoptimizationwasimprovingthe
metrics,theywerenottrendingtowardamodelthatwouldsustain
thebusinessoveral.Butlikealgoodentrepreneurs,hedidnot
giveupprematurely.Daviddecidedtopivotandtestanew
hypothesis.Apivotrequiresthatwekeeponefootrootedinwhat
we’velearnedsofar,whilemakingafundamental
changein
strategyinordertoseekevengreatervalidatedlearning.Inthis
case,David’sdirectcontactwithcustomersprovedessential.
Hehadheardthreerecurringbitsoffeedbackinhistesting:
1.“Ialwayswantedtogetmoreinvolved;thismakesitsomuch
easier.”
2.“ThefactthatyouproveI’mavotermaters.”
3.“There’snoonehere.What’sthepointofcomingback?”1
Daviddecidedtoundertake
whatIcalazoom-inpivot,
refocusingtheproductonwhatpreviouslyhadbeenconsideredjust
onefeatureofalargerwhole.Thinkofthecustomercomments
above:customersliketheconcept,theylikethevoterregistration
technology,buttheyaren’tgetingvalueoutofthesocial
networkingpartoftheproduct.
DaviddecidedtochangeVotizenintoaproductcaled@2gov,a
“sociallobbyingplatform.”Ratherthangetcustomersintegratedin
acivicsocialnetwork,@2govalowsthemtocontacttheirelected
representativesquicklyandeasilyviaexistingsocialnetworkssuch
asTwiter.Thecustomerengagesdigitaly,but@2govtranslates
thatdigitalcontactintopaperform.MembersofCongress
receive
old-fashionedprintedletersandpetitionsasaresult.Inother
words,@2govtranslatesthehigh-techworldofitscustomersinto
thelow-techworldofpolitics.
@2govhadaslightlydierent
setofleap-of-faithquestionsto
@2govhadaslightlydierentsetofleap-of-faithquestionsto
answer.Itstildependedoncustomerssigningup,verifyingtheir
voterstatus,andreferringtheirfriends,butthegrowthmodel
changed.Insteadofrelyingonanengagement-drivenbusiness
(“sticky”growth),@2govwasmoretransactional.David’s
hypothesiswasthatpassionateactivistswouldbewilingtopay
moneytohave@2govfacilitatecontactsonbehalf
ofvoterswho
caredabouttheirissues.
David’snewMVPtookfourmonthsandanother$30,000.He’d
nowspentagrandtotalof$50,000andworkedfortwelvemonths.
Buttheresultsfromhisnextroundoftestingwere
dramatic:
registrationrate42percent,activation83percent,retention21
percent,andreferralawhopping54percent.However,thenumber
ofactivistswilingtopaywaslessthan1percent.Thevalueofeach
transactionwasfartoolowtosustainaprotablebusinesseven
afterDavidhaddonehisbesttooptimizeit.
BeforewegettoDavid’snextpivot,noticehowconvincinglyhe
wasabletodemonstratevalidatedlearning.Hehopedthatwiththis
newproduct,hewouldbeabletoimprovehisleap-of-faithmetrics
dramaticaly,andhedid(seethechartbelow).
BEFOREPIVOT
AFTERPIVOT
Engineofgrowth
Sticky
Paid
Registrationrate
17%
42%
Activation
90%
83%
Retention
8%
21%
Referral
6%
54%
Revenue
n/a
1%
Lifetimevalue(LTV)
n/a
Minimal
Hedidthisnotbyworkingharderbutbyworkingsmarter,taking
Hedidthisnotbyworkingharderbutbyworking
smarter,taking
hisproductdevelopmentresourcesandapplyingthemtoanew
anddierentproduct.Comparedwiththepreviousfourmonthsof
optimization,thenewfourmonthsofpivotinghadresultedina
dramaticalyhigherreturnoninvestment,butDavidwasstilstuck
inanage-oldentrepreneurialtrap.Hismetricsandproductwere
improving,butnotfastenough.
Davidpivotedagain.Thistime,ratherthanrelyonactiviststo
paymoneytodrivecontacts,hewenttolargeorganizations,
professionalfund-raisers,andbigcompanies,whichalhavea
professionalorbusinessinterestinpoliticalcampaigning.The
companiesseemedextremelyeagertouseandpayfor
David’s
service,andDavidquicklysignedletersofintenttobuildthe
functionalitytheyneeded.Inthispivot,DaviddidwhatIcala
customersegmentpivot,keepingthefunctionalityoftheproduct
thesamebutchangingtheaudiencefocus.Hefocusedonwhopays:
fromconsumerstobusinessesandnonprotorganizations.Inother
words,Davidwentfrombeingabusiness-to-consumer(B2C)
companytobeingabusiness-to-business(B2B)company.
Inthe
processhechangedhisplannedgrowthmodel,asweltoone
wherehewouldbeabletofundgrowthoutoftheprotsgenerated
fromeachB2Bsale.
Threemonthslater,Davidhadbuiltthefunctionalityhe
had
promised,basedonthoseearlyletersofintent.Butwhenhewent
backtocompaniestocolecthischecks,hediscoveredmore
problems.Companyaftercompanyprocrastinated,delayed,and
ultimatelypasseduptheopportunity.Althoughtheyhadbeen
excitedenoughtosignaleterofintent,closingarealsalewas
muchmoredicult.Itturnedoutthatthosecompanieswerenot
earlyadopters.
Onthebasisoftheletersofintent,Davidhadincreasedhishead
count,takingonadditionalsalesstaandengineersinanticipation
ofhavingtoservicehigher-marginbusiness-to-businessaccounts.
Whenthesalesdidn’tmaterialize,thewholeteam
hadtowork
hardertotrytondrevenueelsewhere.Yetnomaterhowmany
salescalstheywentonandnomaterhowmuchoptimizationthey
didtotheproduct,themodelwasn’tworking.Returningtohis
didtotheproduct,themodelwasn’tworking.Returningtohis
leap-of-faithquestions,Davidconcludedthattheresultsrefutedhis
business-to-businesshypothesis,andsohedecidedtopivotonce
again.
Althistime,Davidwaslearningandgainingfeedbackfromhis
potentialcustomers,buthewasinanunsustainablesituation.You
can’tpaystawithwhatyou’velearned,andraisingmoneyatthat
juncturewouldhaveescalatedtheproblem.
Raisingmoneywithout
earlytractionisnotacertainthing.Ifhehadbeenabletoraise
money,hecouldhavekeptthecompanygoingbutwouldhave
beenpouringmoneyintoavalue-destroyingengineofgrowth.He
wouldbeinahigh-pressuresituation:useinvestor’scashtomake
theengineofgrowthworkorriskhavingtoshutdownthe
company(orbereplaced).
Daviddecidedtoreducestaandpivotagain,thistime
atemptingwhatIcalaplatformpivot.Insteadofsel
ingan
applicationtoonecustomeratatime,Davidenvisionedanew
growthmodelinspiredbyGoogle’sAdWordsplatform.Hebuilta
self-servesalesplatformwhereanyonecouldbecomeacustomer
withjustacreditcard.Thus,nomaterwhatcauseyouwere
passionateabout,youcouldgoto@2gov’swebsiteand@2gov
wouldhelpyoundnewpeopletogetinvolved.Asalways,the
newpeoplewereveriedregisteredvoters,andsotheir
opinions
carriedweightwithelectedoficials.
Thenewproducttookonlyoneadditionalmonthtobuildand
immediatelyshowedresults:51percentsign-uprate,92percent
activationrate,28percent
retentionrate,64percentreferralrate
(seethechartbelow).Mostimportant,11percentofthese
customerswerewilingtopay20centspermessage.Most
important,thiswasthebeginningofanactualgrowthmodelthat
couldwork.Receiving20centspermessagemightnotsoundlike
much,butthehighreferralratemeantthat@2govcouldgrowits
tracwithoutspendingsignicantmarketingmoney(thisisthe
viralengineofgrowth).
BEFOREPIVOT
AFTERPIVOT
Engineofgrowth
Paid
Viral
Registrationrate
42%
51%
Activation
83%
92%
Retention
21%
28%
Referral
54%
64%
Revenue
1%
11%
Lifetimevalue(LTV)
Minimal
$0.20permessage
Votizen’sstoryexhibitssomecommonpaterns.Oneofthemost
importanttonoteistheaccelerationofMVPs.TherstMVPtook
eightmonths,thenextfourmonths,thenthree,thenone.Eachtime
Davidwasabletovalidateor
refutehisnexthypothesisfasterthan
before.
Howcanoneexplainthisacceleration?Itistemptingtocreditit
totheproductdevelopmentworkthathadbeengoingon.Many
featureshadbeencreated,
andwiththemafairamountof
infrastructure.Therefore,eachtimethecompanypivoted,itdidn’t
havetostartfromscratch.Butthisisnotthewholestory.Forone
thing,muchoftheproducthadtobediscardedbetweenpivots.
Worse,theproductthatremainedwasclassifiedasalegacyproduct,
onethatwasnolongersuitedtothegoalsofthecompany.Asis
usualythecase,theeortrequiredtoreformalegacyproducttook
extrawork.Counteractingtheseforceswerethehard-
wonlessons
Davidhadlearnedthrougheachmilestone.Votizenacceleratedits
MVPprocessbecauseitwaslearningcriticalthingsaboutits
customers,market,andstrategy.
Today,twoyearsafterits
inception,Votizenisdoingwel.They
recentlyraised$1.5milionfromFacebook’sinitialinvestorPeter
Thiel,oneoftheveryfewconsumerInternetinvestmentshehas
madeinrecentyears.Votizen’ssystemnowcanprocessvoter
identityinrealtimeforforty-sevenstatesrepresenting94percentof
identityinrealtimeforforty-sevenstatesrepresenting94percentof
theU.S.populationandhasdeliveredtensofthousandsofmessages
toCongress.TheStartupVisacampaignused
Votizen’stoolsto
introducetheStartupVisaAct(S.565),whichistherstlegislation
introducedintotheSenatesolelyasaresultofsociallobbying.
Theseactivitieshaveatractedtheatentionofestablished
WashingtonconsultantswhoareseekingtoemployVotizen’stools
infuturepoliticalcampaigns.
DavidBinetisumsuphisexperiencebuildingaLeanStartup:
In2003Istartedacompanyinroughlythesamespaceas
I’mintoday.Ihadroughly
thesamedomainexpertiseand
industrycredibility,freshotheUSA.govsuccess.Butback
thenmycompanywasatotalfailure(despiteconsuming
signicantlygreaterinvestment),whilenowIhavea
businessmakingmoneyandclosingdeals.BackthenIdid
thetraditionallinearproductdevelopmentmodel,releasing
anamazingproduct(itrealywas)after12monthsof
development,onlytondthatnoonewouldbuyit.This
timeIproducedfourversions
intwelveweeksand
generatedmyrstsalerelativelysoonafterthat.Anditisn’t
justmarkettiming—twoothercompaniesthatlaunchedin
asimilarspacein2003subsequentlysoldfortensof
milionsofdolars,and
othersin2010folowedalinear
modelstraighttothedeadpool.
ASTARTUP’SRUNWAYISTHENUMBEROFPIVOTSITCAN
STILLMAKE
Seasonedentrepreneursoftenspeakoftherunwaythattheir
startup
hasleft:theamountoftimeremaininginwhichastartupmust
eitherachievelift-oorfail.Thisusualyisdenedasthe
remainingcashinthebankdividedbythemonthlyburnrate,ornet
drainonthataccount
balance.Forexample,astartupwith$1
milioninthebankthatisspending$100,000permonthhasa
projectedrunwayoftenmonths.
projectedrunwayoftenmonths.
Whenstartupsstarttorun
lowoncash,theycanextendthe
runwaytwoways:bycutingcostsorbyraisingadditionalfunds.
Butwhenentrepreneurscutcostsindiscriminately,theyareasliable
tocutthecoststhatarealowingthecompanytogetthroughits
Build-Measure-Learnfeedbackloopastheyaretocutwaste.Ifthe
cutsresultinaslowdowntothisfeedbackloop,altheyhave
accomplishedistohelpthestartupgooutofbusinessmoreslowly.
Thetruemeasureofrunwayishowmanypivotsastartup
has
left:thenumberofopportunitiesithastomakeafundamental
changetoitsbusinessstrategy.Measuringrunwaythroughthelens
ofpivotsratherthanthatoftimesuggestsanotherwaytoextend
thatrunway:gettoeachpivotfaster.Inotherwords,thestartuphas
tondwaystoachievethesameamountofvalidatedlearningat
lowercostorinashortertime.AlthetechniquesintheLean
Startupmodelthathavebeendiscussedsofarhavethisas
their
overarchinggoal.
PIVOTSREQUIRECOURAGE
Askmostentrepreneurswhohavedecidedtopivotandtheywil
telyouthattheywishtheyhadmadethedecisionsooner.Ibelieve
therearethreereasonswhythishappens.
First,vanitymetricscanalowentrepreneurstoformfalse
conclusionsandliveintheirownprivatereality.Thisisparticularly
damagingtothedecisiontopivotbecauseitrobsteamsofthe
beliefthatitisnecessarytochange.Whenpeopleareforcedto
changeagainsttheirbeterjudgment,theprocessisharder,takes
longer,andleadstoalessdecisiveoutcome.
Second,whenanentrepreneurhasanunclearhypothesis,it’s
almostimpossibletoexperiencecompletefailure,andwithout
failurethereisusualynoimpetustoembarkontheradicalchange
apivotrequires.AsImentionedearlier,thefailureofthe“launchit
andseewhathappens”approachshouldnowbe
evident:youwil
alwayssucceed—inseeingwhathappens.Exceptinrarecases,the
earlyresultswilbeambiguous,andyouwon’tknowwhetherto
earlyresultswilbeambiguous,andyouwon’tknowwhetherto
pivotorpersevere,whethertochangedirectionorstaythecourse.
Third,manyentrepreneursareafraid.Acknowledgingfailurecan
leadtodangerouslylowmorale.Mostentrepreneurs’biggestfearis
notthattheirvisionwilprovetobewrong.Moreterrifying
isthe
thoughtthatthevisionmightbedeemedwrongwithouthaving
beengivenarealchancetoproveitself.Thisfeardrivesmuchof
theresistancetotheminimumviableproduct,splittesting,and
othertechniquestotesthypotheses.Ironicaly,thisfeardrivesup
theriskbecausetestingdoesn’toccuruntilthevisionisfuly
represented.However,bythattimeitisoftentoolatetopivot
becausefundingisrunningout.Toavoidthisfate,
entrepreneurs
needtofacetheirfearsandbewilingtofail,ofteninapublicway.
Infact,entrepreneurswhohaveahighprole,eitherbecauseof
personalfameorbecausetheyareoperatingaspartofafamous
brand,faceanextremeversionofthisproblem.
AnewstartupinSiliconValeycaledPathwasstartedby
experiencedentrepreneurs:DaveMorin,whopreviouslyhad
overseenFacebook’splatforminitiative;DustinMierau,product
designerandcocreatorofMacster;andShawnFanningofNapster
fame.Theydecidedtoreleaseaminimumviableproductin2010.
Becauseofthehigh-prolenatureofitsfounders,theMVPatracted
signicantpressatention,especialyfromtechnology
andstartup
blogs.Unfortunately,theirproductwasnottargetedattechnology
earlyadopters,andasaresult,theearlybloggerreactionwasquite
negative.(Manyentrepreneursfailtolaunchbecausetheyareafraid
ofthiskindofreaction,worryingthatitwilharmthemoraleofthe
entirecompany.Thealureofpositivepress,especialyinour
“home”industry,isquitestrong.)
Luckily,thePathteamhadthecouragetoignorethisfearand
focusonwhattheircustomerssaid.Asaresult,theywereableto
getessentialearlyfeedbackfromactualcustomers.Path’sgoalisto
createamorepersonalsocialnetworkthatmaintainsitsquality
overtime.Manypeoplehavehadtheexperienceofbeing
overconnectedonexistingsocialnetworks,sharingwithpast
coworkers,highschoolfriends,relatives,andcoleagues.Suchbroad
groupsmakeithardtoshareintimatemoments.Pathtookan
groupsmakeithardtoshareintimatemoments.Pathtook
an
unusualapproach.Forexample,itlimitedthenumberof
connectionstofty,basedonbrainresearchbytheanthropologist
RobinDunbaratOxford.Hisresearchsuggeststhatftyisroughly
thenumberofpersonalrelationshipsinanyperson’slifeatany
giventime.
Formembersofthetechpress(andmanytechearlyadopters)
this“articial”constraintonthenumberofconnectionswas
anathema.Theyroutinelyusenewsocialnetworkingproductswith
thousandsofconnections.Fiftyseemedwaytoosmal.Asaresult,
Pathenduredalotofpubliccriticism,whichwashardtoignore.
Butcustomersockedtotheplatform,andtheirfeedback
was
decidedlydierentfromthenegativityinthepress.Customersliked
theintimatemomentsandconsistentlywantedfeaturesthatwere
notontheoriginalproductroadmap,suchastheabilitytoshare
howfriends’picturesmadethemfeelandtheabilitytoshare“video
moments.”
DaveMorinsummeduphisexperiencethisway:
Therealityofourteamandourbackgroundsbuiltupa
massivewalofexpectations.Idon’tthinkitwouldhave
materedwhatwewouldhavereleased;wewouldhave
beenmetwithexpectationsthatarehardtoliveupto.But
tousitjustmeantweneededtogetourproductandour
visionoutintothemarketbroadlyinordertogetfeedback
andtobeginiteration.Wehumblytestourtheoriesandour
approachtoseewhatthemarketthinks.Listentofeedback
honestly.Andcontinuetoinnovateinthedirectionswe
thinkwilcreatemeaningintheworld.
Path’sstoryisjustbeginning,butalreadytheircourageinfacing
downcriticsispayingo.Ifandwhentheyneedtopivot,they
won’tbehamperedbyfear.Theyrecentlyraised$8.5milionin
venturecapitalinaroundledbyKleinerPerkinsCaueld&
Byers.
Indoingso,Pathreportedlyturneddownanacquisitionoerfor
$100milionfromGoogle.2
THEPIVOTORPERSEVEREMEETING
Thedecisiontopivotrequiresaclear-eyedandobjectivemind-set.
We’vediscussedtheteltalesignsoftheneedtopivot:the
decreasingeectivenessofproductexperimentsandthegeneral
feelingthatproductdevelopmentshouldbemoreproductive.
Wheneveryouseethosesymptoms,considerapivot.
Thedecisiontopivotisemotionalychargedforanystartupand
hastobeaddressedinastructuredway.Onewaytomitigatethis
chalengeistoschedulethemeetinginadvance.Irecommendthat
everystartuphavearegular“pivotorpersevere”meeting.
Inmy
experience,lessthanafewweeksbetweenmeetingsistoooften
andmorethanafewmonthsistooinfrequent.However,each
startupneedstofinditsownpace.
Eachpivotorpersevere
meetingrequirestheparticipationof
boththeproductdevelopmentandbusinessleadershipteams.At
IMVU,wealsoaddedtheperspectivesofoutsideadviserswho
couldhelpusseepastourpreconceptionsandinterpretdatain
newways.Theproductdevelopmentteammustbringacomplete
reportoftheresultsofitsproductoptimizationeortsovertime
(notjustthepastperiod)aswelasacomparisonofhowthose
resultsstackupagainstexpectations(again,over
time).The
businessleadershipshouldbringdetailedaccountsoftheir
conversationswithcurrentandpotentialcustomers.
Let’stakealookatthisprocessinactioninadramaticpivot
donebyacompanycaled
Wealthfront.Thatcompanywasfounded
in2007byDanCarrolandaddedAndyRachleasCEOshortly
thereafter.Andyisawel-knowngureinSiliconValey:heisa
cofounderandformergeneralpartneroftheventurecapitalrm
BenchmarkCapitalandisonthefacultyoftheStanfordGraduate
SchoolofBusiness,whereheteachesavarietyofcourseson
technologyentrepreneurship.I
rstmetAndywhenhe
commissionedacasestudy
onIMVUtoteachhisstudentsaboutthe
processwehadusedtobuildthecompany.
Wealthfront’smissionistodisruptthemutualfundindustryby
bringinggreatertransparency,access,andvaluetoretailinvestors.
bringinggreatertransparency,access,andvaluetoretailinvestors.
WhatmakesWealthfront’sstoryunusual,however,isnotwhereitis
todaybuthowitbegan:asanonlinegame.
InWealthfront’soriginalincarnationitwascaledkaChingand
wasconceivedasakindoffantasyleagueforamateurinvestors.It
alowedanyonetoopenavirtualtradingaccountandbuilda
portfoliothatwasbasedonrealmarketdatawithouthavingto
investrealmoney.Theideawastoidentifydiamondsin
therough:
amateurtraderswholackedtheresourcestobecomefundmanagers
butwhopossessedmarketinsight.Wealthfront’sfoundersdidnot
wanttobeintheonlinegamingbusinessperse;kaChingwaspart
ofasophisticatedstrategyintheserviceoftheirlargervision.Any
studentofdisruptiveinnovationwouldhavelookedon
approvingly:theywerefolowingthatsystemperfectlybyinitialy
servingcustomerswhowereunabletoparticipateinthe
mainstreammarket.Overtime,theybelieved,theproductwould
becomemoreandmoresophisticated,eventualyalowingusersto
serve(anddisrupt)existingprofessionalfundmanagers.
Toidentifythebestamateurtradingsavants,Wealthfrontbuilt
sophisticatedtechnologytoratetheskilofeachfundmanager,
usingtechniquesemployedbythemostsophisticatedevaluatorsof
moneymanagers,thepremierU.S.universityendowments.Those
methodsalowedthemtoevaluatenotjustthereturns
themanagers
generatedbutalsotheamountofrisktheyhadtakenalongwith
howconsistenttheyperformedrelativetotheirdeclaredinvestment
strategy.Thus,fundmanagerswhoachievedgreatreturnsthrough
recklessgambles(i.e.,investmentsoutsidetheirareaofexpertise)
wouldberankedlowerthanthosewhohadfiguredouthowtobeat
themarketthroughskil.
WithitskaChinggame,Wealthfronthopedtotesttwoleap-of-
faithassumptions:
1.Asignicantpercentageofthegameplayerswould
demonstrateenoughtalentasvirtualfundmanagerstoprove
themselvessuitabletobecomemanagersofrealassets(the
valuehypothesis).
valuehypothesis).
2.Thegamewouldgrowusingtheviralengineofgrowthand
generatevalueusingafreemiumbusinessmodel.Thegame
wasfreetoplay,buttheteamhopedthatapercentageofthe
playerswouldrealizethattheywerelousytradersand
thereforewanttoconverttopayingcustomersonce
Wealthfrontstartedoeringrealassetmanagementservices
(thegrowthhypothesis).
kaChingwasahugeearlysuccess,atractingmorethan
450,000
gamersinitsinitiallaunch.Bynow,youshouldbesuspiciousof
thiskindofvanitymetric.Manylessdisciplinedcompanieswould
havecelebratedthatsuccessandfelttheirfuturewassecure,but
Wealthfronthadidentieditsassumptionsclearlyandwasableto
thinkmorerigorously.BythetimeWealthfrontwasreadytolaunch
itspaidnancialproduct,onlysevenamateurmanagershad
qualiedasworthyofmanagingotherpeople’smoney,farlessthan
theidealmodelhadanticipated.Afterthepaidproductlaunched,
theywereabletomeasuretheconversionrateofgamersinto
payingcustomers.Heretoothenumberswerediscouraging:the
conversionratewasclosetozero.Theirmodelhad
predictedthat
hundredsofcustomerswouldsignup,butonlyfourteendid.
Theteamworkedvaliantlytondwaystoimprovetheproduct,
butnoneshowedanyparticularpromise.Itwastimeforapivotor
perseveremeeting.
Ifthedatawehavediscussedsofarwasalthatwasavailableat
thatcriticalmeeting,Wealthfrontwouldhavebeenintrouble.They
wouldhaveknownthattheircurrentstrategywasn’tworkingbut
notwhattodotoxit.Thatiswhyitwascriticalthatthey
folowedtherecommendationearlierinthischaptertoinvestigate
alternativepossibilities.Inthiscase,Wealthfronthadpursuedtwo
importantlinesofinquiry.
Therstwasaseriesof
conversationswithprofessionalmoney
managers,beginningwithJohnPowers,theheadofStanford
University’sendowment,whoreactedsurprisinglypositively.
Wealthfront’sstrategywaspremisedontheassumptionthat
professionalmoneymanagerswouldbereluctanttojointhesystem
professionalmoneymanagerswouldbereluctanttojointhesystem
becausetheincreasedtransparencywouldthreatentheirsenseof
authority.Powershadnosuchconcerns.CEOAndy
Rachlethen
beganaseriesofconversationswithotherprofessionalinvestment
managersandbroughttheresultsbacktothecompany.Hisinsights
wereasfolows:
1.Successfulprofessionalmoneymanagersfeltthey
hadnothing
tofearfromtransparency,sincetheybelieveditwouldvalidate
theirskils.
2.Moneymanagersfacedsignicantchalengesinmanagingand
scalingtheirownbusinesses.Theywerehamperedbythe
dicultyofservicingtheirownaccountsandthereforehadto
requirehighminimuminvestmentsasawaytoscreennew
clients.
ThesecondproblemwassoseverethatWealthfrontwaselding
coldcalsfromprofessionalmanagersaskingoutofthebluetojoin
theplatform.Thesewereclassicearlyadopterswhohadthevision
toseepastthecurrentproducttosomethingtheycoulduseto
achieveacompetitiveadvantage.
Thesecondcriticalqualitativeinformationcameoutof
conversationswithconsumers.Itturnedoutthattheyfoundthe
blendingofvirtualandrealportfoliomanagementonthekaChing
websiteconfusing.Farfrombeingacleverwayof
acquiring
customers,thefreemiumstrategywasgetinginthewayby
promotingconfusionaboutthecompany’spositioning.
Thisdatainformedthepivotorperseveremeeting.With
everyonepresent,theteamdebatedwhattodowithits
future.The
currentstrategywasn’tworking,butmanyemployeeswerenervous
aboutabandoningtheonlinegame.Afteral,itwasanimportant
partofwhattheyhadsignedontobuild.Theyhadinvested
signicanttimeandenergybuildingandsupportingthose
customers.Itwaspainful—asitalwaysis—torealizethatthat
energyhadbeenwasted.
Wealthfrontdecideditcouldnotpersevereasitexisted.The
companychoseinsteadto
celebratewhatithadlearned.Ifithad
companychoseinsteadtocelebratewhatithadlearned.Ifithad
notlauncheditscurrentproduct,theteamneverwouldhave
learnedwhatitneededtoknowtopivot.Infact,theexperience
taughtthemsomethingessentialabouttheirvision.AsAndysays,
“Whatwerealywantedtochangewasnotwhomanagesthe
moneybutwhohasaccesstothebestpossibletalent.We’d
originalythoughtwe’dneedtobuildasignicantbusinesswith
amateurmanagerstogetprofessionalstocomeonboard,but
fortunatelyitturnsoutthatwasn’tnecessary.”
Thecompanypivoted,abandoningthegamingcustomers
altogetherandfocusingonprovidingaservicethatalowed
customerstoinvestwithprofessionalmanagers.Onthesurface,the
pivotseemsquitedramaticinthatthecompanychangedits
positioning,itsname,anditspartnerstrategy.Itevenjetisoneda
largeproportionofthefeaturesithadbuilt.Butatitscore,a
surprisingamountstayedthesame.Themostvaluableworkthe
companyhaddonewasbuildingtechnologytoevaluatemanagers’
eectiveness,andthisbecamethekernelaroundwhichthenew
businesswasbuilt.Thisisalsocommonwithpivots;it
isnot
necessarytothrowouteverythingthatcamebeforeandstartover.
Instead,it’saboutrepurposingwhathasbeenbuiltandwhathas
beenlearnedtofindamorepositivedirection.
Today,Wealthfrontis
prosperingasaresultofitspivot,with
over$180milioninvestedontheplatformandmorethanforty
professionalmanagers.3ItrecentlywasnamedoneofFast
Company’stenmostinnovativecompaniesinnance.4The
companycontinuestooperatewithagility,scalinginlinewiththe
growthprinciplesoutlinedinChapter12.Wealthfrontisalsoa
leadingadvocateofthedevelopmenttechniqueknownas
continuousdeployment,whichwe’ldiscussin
Chapter9.
FAILURETOPIVOT
Thedecisiontopivotissodicultthatmanycompaniesfailto
makeit.IwishIcouldsaythateverytimeIwasconfrontedwith
theneedtopivot,Ihandleditwel,butthisisfarfromtrue.
I
theneedtopivot,Ihandleditwel,butthisisfarfromtrue.I
rememberonefailuretopivotespecialywel.
AfewyearsafterIMVU’sfounding,thecompanywashaving
tremendoussuccess.The
businesshadgrowntoover$1milionper
monthinrevenue;wehadcreatedmorethantwentymilionavatars
forourcustomers.Wemanagedtoraisesignicantnewroundsof
nancing,andliketheglobaleconomy,wewereridinghigh.But
dangerlurkedaroundthecorner.
Unknowingly,wehadfalenintoaclassicstartuptrap.Wehad
beensosuccessfulwithourearlyeortsthatwewereignoringthe
principlesbehindthem.Asaresult,wemissedtheneedtopivot
evenasitstaredusintheface.
Wehadbuiltanorganizationthatexceledatthekindsof
activitiesdescribedinearlierchapters:creatingminimumviable
productstotestnewideasandrunningexperimentstotunethe
engineofgrowth.Beforewehadbeguntoenjoysuccess,many
peoplehadadvisedagainstour“low-quality”minimumviable
productandexperimentalapproach,urgingustoslowdown.They
wantedustodothingsrightandfocusonqualityinstead
ofspeed.
Weignoredthatadvice,mostlybecausewewantedtoclaimthe
advantagesofspeed.Afterourapproachwasvindicated,theadvice
wereceivedchanged.Nowmostoftheadviceweheardwasthat
“youcan’targuewithsuccess,”urgingustostaythecourse.We
likedthisadvicebeter,butitwasequalywrong.
Rememberthattherationaleforbuildinglow-qualityMVPsis
thatdevelopinganyfeaturesbeyondwhatearlyadoptersrequireis
aformofwaste.However,thelogicofthistakesyouonlysofar.
Onceyouhavefoundsuccesswithearlyadopters,youwanttosel
tomainstreamcustomers.Mainstreamcustomershavedierent
requirementsandaremuchmoredemanding.
Thekindofpivotweneedediscaledacustomersegmentpivot.
Inthispivot,thecompanyrealizesthattheproductit’sbuilding
solvesarealproblemforrealcustomersbutthattheyarenotthe
customersitoriginalyplannedtoserve.Inother
words,theproduct
hypothesisisconrmedonlypartialy.(Thischapterdescribedsuch
apivotintheVotizenstory,above.)
Acustomersegmentpivotisanespecialytrickypivottoexecute
Acustomersegmentpivotis
anespecialytrickypivottoexecute
because,aswelearnedthehardwayatIMVU,theveryactionsthat
madeussuccessfulwithearlyadopterswerediametricalyopposed
totheactionswe’dhavetomastertobesuccessfulwithmainstream
customers.Welackedaclearunderstandingofhowourengineof
growthoperated.Wehadbeguntotrustourvanitymetrics.Wehad
stoppedusinglearningmilestonestoholdourselvesaccountable.
Instead,itwasmuchmoreconvenienttofocusonthe
ever-larger
grossmetricsthatweresoexciting:breakingnewrecordsinsigning
uppayingcustomersandactiveusers,monitoringourcustomer
retentionrate—younameit.Underthesurface,itshouldhavebeen
clearthatourefortsattuningtheenginewerereachingdiminishing
returns,theclassicsignoftheneedtopivot.
Forexample,wespentmonthstryingtoimprovetheproduct’s
activationrate(therateatwhichnewcustomersbecomeactive
consumersoftheproduct),whichremainedstubbornlylow.Wedid
countlessexperiments:usabilityimprovements,newpersuasion
techniques,incentiveprograms,customerquests,andothergame-
likefeatures.Individualy,manyofthesenewfeatures
andnew
marketingtoolsweresuccessful.Wemeasuredthemrigorously,
usingA/Bexperimentation.Buttakeninaggregate,overthecourse
ofmanymonths,wewereseeingnegligiblechangesintheoveral
driversofourengineofgrowth.Evenouractivationrate,whichhad
beenthecenterofourfocus,edgeduponlyafewpercentage
points.
Weignoredthesignsbecausethecompanywasstilgrowing,
deliveringmonthaftermonthof“upandtotheright”results.But
wewerequicklyexhaustingourearlyadoptermarket.Itwas
getingharderandhardertondcustomerswecouldacquireatthe
priceswewereaccustomedtopaying.Aswedroveour
marketing
teamtondmorecustomers,theywereforcedtoreachoutmoreto
mainstreamcustomers,butmainstreamcustomersarelessforgiving
ofanearlyproduct.Theactivationandmonetizationratesofnew
customersstartedtogodown,drivingupthecostofacquiringnew
customers.Pretysoon,ourgrowthwasatliningandourengine
sputeredandstaled.
Ittookusfartoolongtomakethechangesnecessarytoxthis
Ittookusfartoolongtomakethechangesnecessarytoxthis
situation.Aswithalpivots,wehadtogetbacktobasicsandstart
theinnovationaccountingcycleover.Itfeltlikethecompany’s
secondfounding.Wehadgotenrealygoodatoptimizing,
tuning,
anditerating,butintheprocesswehadlostsightofthepurposeof
thoseactivities:testingaclearhypothesisintheserviceofthe
company’svision.Instead,wewerechasinggrowth,revenue,and
profitswhereverwecouldfindthem.
Weneededtoreacquaintourselveswithournewmainstream
customers.Ourinteractiondesignersledthewaybydevelopinga
clearcustomerarchetypethatwasbasedonextensivein-person
conversationsandobservation.Next,weneededtoinvestheavilyin
amajorproductoverhauldesignedtomaketheproduct
dramaticalyeasiertouse.Becauseofouroverfocusonne-tuning,
wehadstoppedmakinglargeinvestmentslikethese,preferringto
investinlower-riskandlower-yieldtestingexperiments.
However,investinginquality,design,andlargerprojectsdidnot
requirethatweabandonourexperimentalroots.Onthecontrary,
oncewerealizedourmistakeandexecutedthepivot,those
skils
serveduswel.Wecreatedasandboxforexperimentationlikethe
onedescribedinChapter12andhadacross-functionalteamwork
exclusivelyonthismajorredesign.Astheybuilt,theycontinuously
testedtheirnewdesignheadtoheadagainsttheoldone.Initialy,
thenewdesignperformedworsethantheoldone,asisusualythe
case.Itlackedthefeaturesandfunctionalityoftheolddesignand
hadmanynewmistakesaswel.Buttheteam
relentlesslyimproved
thedesignuntil,monthslater,itperformedbeter.Thisnewdesign
laidthefoundationforourfuturegrowth.
Thisfoundationhaspaidohandsomely.By2009,revenuehad
morethandoubledtoover
$25milionannualy.Butwemight
haveenjoyedthatsuccessearlierifwehadpivotedsooner.5
ACATALOGOFPIVOTS
Pivotscomeindierentavors.Thewordpivotsometimesisused
incorrectlyasasynonymfor
change.Apivotisaspecialkindof
incorrectlyasasynonymforchange.Apivotisaspecialkindof
changedesignedtotestanewfundamentalhypothesisaboutthe
product,businessmodel,andengineofgrowth.
Zoom-inPivot
Inthiscase,whatpreviouslywasconsideredasinglefeatureina
productbecomesthewholeproduct.Thisisthetypeofpivot
Votizenmadewhenitpivotedawayfromafulsocialnetworkand
towardasimplevotercontactproduct.
Zoom-outPivot
Inthereversesituation,sometimesasinglefeatureisinsucientto
supportawholeproduct.Inthistypeofpivot,whatwasconsidered
thewholeproductbecomesa
singlefeatureofamuchlarger
product.
CustomerSegmentPivot
Inthispivot,thecompanyrealizesthattheproductitisbuilding
solvesarealproblemforrealcustomersbutthattheyarenotthe
typeofcustomersitoriginalyplannedtoserve.Inotherwords,the
producthypothesisispartialyconrmed,solvingtheright
problem,butforadiferentcustomerthanoriginalyanticipated.
CustomerNeedPivot
Asaresultofgetingtoknow
customersextremelywel,it
sometimesbecomesclearthattheproblemwe’retryingtosolvefor
themisnotveryimportant.However,becauseofthiscustomer
intimacy,weoftendiscoverotherrelatedproblemsthatare
importantandcanbesolvedbyourteam.Inmanycases,these
relatedproblemsmayrequirelitlemorethanrepositioningthe
relatedproblemsmayrequirelitlemorethanrepositioningthe
existingproduct.Inothercases,itmayrequirea
completelynew
product.Again,thisacasewheretheproducthypothesisispartialy
conrmed;thetargetcustomerhasaproblemworthsolving,just
nottheonethatwasoriginalyanticipated.
Afamousexampleisthe
chainPotbelySandwichShop,which
todayhasovertwohundredstores.Itbeganasanantiquestorein
1977;theownersstartedtoselsandwichesasawaytobolster
tractotheirstores.Pretysoontheyhadpivotedtheirwayintoan
entirelydiferentlineofbusiness.
PlatformPivot
Aplatformpivotreferstoachangefromanapplicationtoa
platformorviceversa.Mostcommonly,startupsthataspireto
createanewplatformbegin
lifebyselingasingleapplication,the
so-caledkilerapp,fortheirplatform.Onlylaterdoestheplatform
emergeasavehicleforthirdpartiestoleverageasawaytocreate
theirownrelatedproducts.However,thisorderisnotalwayssetin
stone,andsomecompanieshavetoexecutethispivotmultiple
times.
BusinessArchitecturePivot
ThispivotborrowsaconceptfromGeoreyMoore,whoobserved
thatcompaniesgeneralyfolowoneoftwomajorbusiness
architectures:highmargin,lowvolume(complexsystemsmodel)or
lowmargin,highvolume(volumeoperationsmodel).6Theformer
commonlyisassociatedwithbusinesstobusiness(B2B)or
enterprisesalescycles,andthelaterwithconsumerproducts(there
arenotableexceptions).Inabusinessarchitecturepivot,astartup
switchesarchitectures.Somecompanieschangefromhighmargin,
lowvolumebygoingmassmarket(e.g.,Google’ssearch
“appliance”);others,originalydesignedforthemassmarket,
turnedouttorequirelongandexpensivesalescycles.
ValueCapturePivot
Therearemanywaystocapturethevalueacompanycreates.These
methodsarereferredtocommonlyasmonetizationorrevenue
models.Thesetermsare
muchtoolimiting.Implicitintheideaof
monetizationisthatitisaseparate“feature”ofaproductthatcan
beaddedorremovedatwil.Inreality,capturingvalueisan
intrinsicpartoftheproducthypothesis.Often,changestothewaya
companycapturesvaluecanhavefar-reachingconsequencesforthe
restofthebusiness,product,andmarketingstrategies.
EngineofGrowthPivot
Aswe’lseeinChapter10,therearethreeprimaryenginesof
growththatpowerstartups:
theviral,sticky,andpaidgrowth
models.Inthistypeofpivot,acompanychangesitsgrowthstrategy
toseekfasterormoreprofitablegrowth.Commonlybutnotalways,
theengineofgrowthalsorequiresachangeinthewayvalueis
captured.
ChannelPivot
Intraditionalsalesterminology,themechanismbywhicha
companydeliversitsproducttocustomersiscaledthesales
channelordistributionchannel.Forexample,
consumerpackaged
goodsaresoldinagrocerystore,carsaresoldindealerships,and
muchenterprisesoftwareissold(withextensivecustomization)by
consultingandprofessionalservicesrms.Often,therequirements
ofthechanneldeterminetheprice,features,andcompetitive
landscapeofaproduct.Achannelpivotisarecognitionthatthe
samebasicsolutioncouldbedeliveredthroughadierentchannel
withgreatereectiveness.Wheneveracompany
abandonsa
previouslycomplexsalesprocessto“seldirect”toitsendusers,a
channelpivotisinprogress.
channelpivotisinprogress.
Itispreciselybecauseofitsdestructiveeectonsaleschannels
thattheInternethashadsuchadisruptiveinuenceinindustries
thatpreviouslyrequiredcomplexsalesanddistributionchannels,
suchasnewspaper,magazine,andbookpublishing.
TechnologyPivot
Occasionaly,acompanydiscoversawaytoachievethesame
solutionbyusingacompletelydierenttechnology.Technology
pivotsaremuchmorecommoninestablishedbusinesses.Inother
words,theyareasustaininginnovation,anincremental
improvementdesignedtoappealtoandretainanexistingcustomer
base.Establishedcompaniesexcelatthiskindofpivotbecauseso
muchisnotchanging.Thecustomersegmentisthesame,the
customer’sproblemisthesame,thevalue-capture
modelisthe
same,andthechannelpartnersarethesame.Theonlyquestionis
whetherthenewtechnologycanprovidesuperiorpriceand/or
performancecomparedwiththeexistingtechnology.
APIVOTISASTRATEGIC
HYPOTHESIS
Althoughthepivotsidentiedabovewilbefamiliartostudentsof
businessstrategy,theabilitytopivotisnosubstituteforsound
strategicthinking.Theproblemwithprovidingfamousexamplesof
pivotsisthatmostpeoplearefamiliaronlywiththesuccessfulend
strategiesoffamouscompanies.MostreadersknowthatSouthwest
orWalmartisanexampleofalow-costdisruptionintheirmarkets,
thatMicrosoftanexampleofaplatformmonopoly,and
that
Starbuckshasleveragedapowerfulpremiumbrand.Whatis
generalylesswelknownarethepivotsthatwererequiredto
discoverthosestrategies.Companieshaveastrongincentivetoalign
theirPRstoriesaroundtheheroicfounderandmakeitseemthat
theirsuccesswastheinevitableresultofagoodidea.
Thus,althoughstartupsoftenpivotintoastrategythatseems
Thus,althoughstartupsoften
pivotintoastrategythatseems
similartothatofasuccessfulcompany,itisimportantnottoput
toomuchstockintheseanalogies.It’sextremelydiculttoknowif
theanalogyhasbeendrawnproperly.Havewecopiedtheessential
featuresorjustsupercialones?Wilwhatworkedinthatindustry
workinours?Wilwhathasworkedinthepastworktoday?A
pivotisbeterunderstoodasanewstrategichypothesisthatwil
requireanewminimumviableproducttotest.
Pivotsareapermanentfactoflifeforanygrowingbusiness.Even
afteracompanyachievesinitialsuccess,itmustcontinuetopivot.
Thosefamiliarwiththetechnologylifecycleideasoftheoristssuch
asGeoreyMooreknowcertainlater-stagepivotsby
thenameshe
hasgiventhem:theChasm,theTornado,theBowlingAley.
Readersofthedisruptiveinnovationliteraturespearheadedby
Harvard’sClaytonChristensenwilbefamiliarwithestablished
companiesthatfailtopivotwhentheyshould.Thecriticalskilfor
managerstodayistomatchthosetheoriestotheirpresentsituation
sothattheyapplytherightadviceattherighttime.
Modernmanagerscannothaveescapedthedelugeofrecent
bookscalingonthemtoadapt,change,reinvent,orupendtheir
existingbusinesses.Manyoftheworksinthiscategoryarelongon
exhortationsandshortonspecifics.
Apivotisnotjustanexhortationtochange.Remember,itisa
specialkindofstructuredchangedesignedtotestanew
fundamentalhypothesisabouttheproduct,businessmodel,and
engineofgrowth.ItistheheartoftheLeanStartupmethod.Itis
whatmakesthecompaniesthatfolowLeanStartupresilientinthe
faceofmistakes:ifwetakeawrongturn,wehavethetoolswe
needtorealizeitandtheagilitytofindanotherpath.
InPartTwo,wehavelookedatastartupideafromitsinitialleaps
offaith,testeditwithaminimumviableproduct,usedinnovation
accountingandactionablemetricstoevaluatetheresults,andmade
thedecisiontopivotorpersevere.
Ihavetreatedthesesubjectsingreatdetailtoprepareforwhat
Ihavetreatedthesesubjectsingreatdetailtoprepareforwhat
comesnext.Onthepage,theseprocessesmayseemclinical,slow,
andsimple.Intherealworld,somethingdierentisneeded.We
havelearnedtosteerwhenmovingslowly.Nowwemustlearnto
race.Layingasolidfoundationisonlytherststep
towardourreal
destination:acceleration.
PartThree
PartThree
ACCELERATE
StartYourEngines
Mostofthedecisionsstartupsfacearenotclear-cut.Howoften
shouldyoureleaseaproduct?Isthereareasontoreleaseweekly
ratherthandailyorquarterlyorannualy?Productreleasesincur
overhead,andsofromaneciencypointofview,releasingoften
leaveslesstimetodevotetobuildingtheproduct.However,
waitingtoolongtoreleasecanleadtotheultimate
waste:making
somethingthatnobodywants.
Howmuchtimeandenergyshouldcompaniesinvestin
infrastructureandplanningearlyoninanticipationofsuccess?
Spendtoomuchandyouwasteprecioustimethat
couldhavebeen
spentlearning.Spendtoolitleandyoumayfailtotakeadvantage
ofearlysuccessandcedemarketleadershiptoafastfolower.
Whatshouldemployeesspendtheirdaysdoing?Howdowehold
peopleaccountableforlearningatanorganizationallevel?
Traditionaldepartmentscreateincentivestructuresthatkeep
peoplefocusedonexcelenceintheirspecialties:marketing,sales,
productdevelopment.Butwhatifthecompany’sbest
interestsare
servedbycross-functionalcolaboration?Startupsneed
organizationalstructuresthatcombattheextremeuncertaintythatis
astartup’schiefenemy.
Theleanmanufacturingmovementfacedsimilarquestionsonthe
factoryoor.Theiranswersarerelevantforstartupsaswel,with
somemodifications.
Thecriticalrstquestionforanyleantransformationis:which
Thecriticalrstquestionforanyleantransformationis:which
activitiescreatevalueandwhichareaformofwaste?Onceyou
understandthisdistinction,youcanbeginusingleantechniquesto
driveoutwasteandincreasetheeciencyofthevalue-creating
activities.Forthesetechniquestobeusedina
startup,theymustbe
adaptedtotheuniquecircumstancesofentrepreneurship.Recal
fromChapter3thatvalueinastartupisnotthecreationofstu,
butrathervalidatedlearningabouthowtobuildasustainable
business.Whatproductsdocustomersrealywant?Howwilour
businessgrow?Whoisourcustomer?Whichcustomersshouldwe
listentoandwhichshouldweignore?Thesearethequestionsthat
needansweringasquicklyaspossibletomaximizea
startup’s
chancesofsuccess.Thatiswhatcreatesvalueforastartup.
InPartThree,wewildeveloptechniquesthatalowLean
Startupstogrowwithoutsacricingthespeedandagilitythatare
thelifebloodofeverystartup.Contrarytocommonbelief,lethargy
andbureaucracyarenottheinevitablefateofcompaniesasthey
achievematurity.Ibelievethatwiththeproperfoundation,Lean
Startupscangrowtobecomeleanenterprisesthatmaintain
their
agility,learningorientation,andcultureofinnovationevenasthey
scale.
InChapter9,wewilseehowLeanStartupstakeadvantageof
thecounterintuitivepowerofsmalbatches.Justaslean
manufacturinghaspursuedajust-in-timeapproachtobuilding
products,reducingtheneedforin-processinventory,LeanStartups
practicejust-in-timescalability,conductingproductexperiments
withoutmakingmassiveup-frontinvestmentsinplanning
and
design.
Chapter10wilexplorethemetricsstartupsshoulduseto
understandtheirgrowthastheyaddnewcustomersanddiscover
newmarkets.Sustainablegrowthfolowsoneofthreeenginesof
growth:paid,viral,orsticky.Byidentifyingwhichengineofgrowth
astartupisusing,itcanthendirectenergywhereitwilbemost
eectiveingrowingthebusiness.Eachenginerequiresafocuson
uniquemetricstoevaluatethesuccessofnewproducts
and
prioritizenewexperiments.Whenusedwiththeinnovation
accountingmethoddescribedinPartTwo,thesemetricsalow
accountingmethoddescribedinPartTwo,thesemetricsalow
startupstogureoutwhentheirgrowthisatriskofrunningout
andpivotaccordingly.
Chapter11showshowtobuildanadaptiveorganizationby
investingintherightamountofprocesstokeepteamsnimbleas
theygrow.Wewilseehowtechniquesfromthetoolkitoflean
manufacturing,suchastheFiveWhys,helpstartupteamsgrow
withoutbecomingbureaucraticordysfunctional.Wealsowilsee
howleandisciplinessetthe
stageforastartuptotransitionintoan
establishedcompanydrivenbyoperationalexcelence.
InChapter12,we’lcomefulcircle.Asstartupsgrowinto
establishedcompanies,theyfacethesamepressuresthatmakeit
necessaryfortoday’s
enterprisestondnewwaystoinvestin
disruptiveinnovation.Infact,we’lseethatanadvantageofa
successfulstartup’srapidgrowthisthatthecompanycankeepits
entrepreneurialDNAevenasitmatures.Today’scompaniesmust
learntomasteramanagementportfolioofsustainableand
disruptiveinnovation.Itisanobsoleteviewthatseesstartupsas
goingthroughdiscretephasesthatleaveearlierkindsofwork—
suchasinnovation—behind.Rather,moderncompanies
mustexcel
atdoingmultiplekindsofworkinparalel.Todoso,we’lexplore
techniquesforincubatinginnovationteamswithinthecontextofan
establishedcompany.
Ihaveincludedanepiloguecaled“WasteNot”inwhich
I
considersomeofthebroaderimplicationsofthesuccessoftheLean
Startupmovement,placeitinhistoricalcontext(including
cautionarylessonsfrompastmovements),andmakesuggestionsfor
itsfuturedirection.
9BATCH
InthebookLeanThinking,JamesWomackandDanielJones
recountastoryofstungnewsletersintoenvelopeswiththe
assistanceofoneoftheauthor’stwoyoungchildren.Every
envelopehadtobeaddressed,stamped,ledwithaleter,and
sealed.Thedaughters,agesixandnine,knewhowtheyshouldgo
aboutcompletingtheproject:“Daddy,rstyoushouldfoldalof
thenewsleters.Thenyoushouldatachtheseal.Then
youshould
putonthestamps.”Theirfatherwantedtodoitthe
counterintuitiveway:completeeachenvelopeoneatatime.They
—likemostofus—thoughtthatwasbackward,explainingtohim
“thatwouldn’tbeeficient!”
Heandhisdaughterseachtookhalfthe
envelopesandcompetedtoseewhowouldfinishfirst.
Thefatherwontherace,andnotjustbecauseheisanadult.It
happenedbecausetheoneenvelopeatatimeapproachisafaster
wayofgetingthejobdoneeventhoughitseemsinecient.This
hasbeenconrmedinmanystudies,includingonethatwas
recordedonvideo.1
Theoneenvelopeatatimeapproachiscaled“single-piece
ow”inleanmanufacturing.Itworksbecauseofthesurprising
powerofsmalbatches.Whenwedoworkthatproceedsinstages,
the“batchsize”referstohowmuchworkmovesfromonestageto
thenextatatime.Forexample,ifwewerestung
onehundred
envelopes,theintuitivewaytodoit—foldingonehundredletersat
atime—wouldhaveabatchsizeofonehundred.Single-pieceow
issonamedbecauseithasabatchsizeofone.
issonamedbecauseithasa
batchsizeofone.
Whydoesstungoneenvelopeatatimegetthejobdonefaster
eventhoughitseemslikeitwouldbeslower?Becauseourintuition
doesn’ttakeintoaccounttheextratimerequiredtosort,stack,and
movearoundthelargepilesofhalf-completeenvelopeswhenit’s
donetheotherway.2Itseemsmoreeficienttorepeatthesametask
overandover,inpartbecauseweexpectthatwewilgetbeterat
thissimpletaskthemorewedoit.Unfortunately,in
process-
orientedworklikethis,individualperformanceisnotnearlyas
importantastheoveralperformanceofthesystem.
Eveniftheamountoftimethateachprocesstookwasexactlythe
same,thesmalbatch
productionapproachstilwouldbesuperior,
andforevenmorecounterintuitivereasons.Forexample,imagine
thattheletersdidn’ttintheenvelopes.Withthelarge-batch
approach,wewouldn’tndthatoutuntilnearlytheend.With
smalbatches,we’dknowalmostimmediately.Whatifthe
envelopesaredefectiveandwon’tseal?Inthelarge-batch
approach,we’dhavetounstualtheenvelopes,getnewones,and
restuthem.Inthesmal-batchapproach,we’dndthisout
immediatelyandhavenoreworkrequired.
Altheseissuesarevisibleinaprocessassimpleasstung
envelopes,buttheyareofrealandmuchgreaterconsequenceinthe
workofeverycompany,largeorsmal.Thesmal-batchapproach
producesanishedproducteveryfewseconds,whereasthelarge-
batchapproachmustdeliveraltheproductsatonce,attheend.
Imaginewhatthismightlooklikeifthetimehorizonwashours,
days,orweeks.Whatifitturnsoutthatthecustomers
havedecided
theydon’twanttheproduct?Whichprocesswouldalowa
companytofindthisoutsooner?
Leanmanufacturersdiscoveredthebenetsofsmalbatches
decadesago.Inthepost–WorldWarIeconomy,
Japanese
carmakerssuchasToyotacouldnotcompetewithhugeAmerican
factoriesthatusedthelatestmassproductiontechniques.Folowing
theintuitivelyecientwayofbuilding,massproductionfactories
builtcarsbyusingever-largerbatchsizes.Theywouldspendhuge
amountsofmoneybuyingmachinesthatcouldproducecarpartsby
amountsofmoneybuyingmachinesthatcouldproducecarpartsby
thetens,hundreds,orthousands.Bykeepingthose
machines
runningatpeakspeed,theycoulddrivedowntheunitcostofeach
partandproducecarsthatwereincrediblyinexpensivesolongas
theywerecompletelyuniform.
TheJapanesecarmarketwas
fartoosmalforcompaniessuchas
Toyotatoemploythoseeconomiesofscale;thus,Japanese
companiesfacedintensepressurefrommassproduction.Also,in
thewar-ravagedJapaneseeconomy,capitalwasnotavailablefor
massiveinvestmentsinlargemachines.
ItwasagainstthisbackdropthatinnovatorssuchasTaichiOhno,
ShigeoShingo,andothersfoundawaytosucceedbyusingsmal
batches.Insteadofbuyinglargespecializedmachinesthatcould
producethousandsofpartsatatime,Toyotausedsmalergeneral-
purposemachinesthatcouldproduceawidevarietyofpartsin
smalbatches.Thisrequiredguringoutwaystorecongureeach
machinerapidlytomaketherightpartattherighttime.
By
focusingonthis“changeovertime,”Toyotawasabletoproduce
entireautomobilesbyusingsmalbatchesthroughouttheprocess.
Thisrapidchangingofmachineswasnoeasyfeat.Asinanylean
transformation,existingsystemsandtoolsoftenneedtobe
reinventedtosupportworkinginsmalerbatches.ShigeoShingo
createdtheconceptofSMED(Single-MinuteExchangeofDie)in
ordertoenableasmalerbatchsizeofworkinearly
Toyota
factories.Hewassorelentlessinrethinkingthewaymachineswere
operatedthathewasabletoreducechangeovertimesthat
previouslytookhourstolessthantenminutes.Hedidthis,notby
askingworkerstowork
faster,butbyreimaginingandrestructuring
theworkthatneededtobedone.Everyinvestmentinbetertools
andprocesshadacorrespondingbenetintermsofshrinkingthe
batchsizeofwork.
Becauseofitssmalerbatch
size,Toyotawasabletoproducea
muchgreaterdiversityofproducts.Itwasnolongernecessarythat
eachproductbeexactlythesametogaintheeconomiesofscale
thatpoweredmassproduction.Thus,Toyotacouldserveitssmaler,
morefragmentedmarketsandstilcompetewiththemass
producers.Overtime,thatcapabilityalowedToyotatomove
producers.Overtime,thatcapabilityalowedToyotatomove
successfulyintolargerandlargermarketsuntilit
becamethe
world’slargestautomakerin2008.
Thebiggestadvantageofworkinginsmalbatchesisthatquality
problemscanbeidentiedmuchsooner.Thisistheoriginof
Toyota’sfamousandoncord,
whichalowsanyworkertoaskfor
helpassoonastheynoticeanyproblem,suchasadefectina
physicalpart,stoppingtheentireproductionlineifitcannotbe
correctedimmediately.Thisisanotherverycounterintuitive
practice.Anassemblylineworksbestwhenitisfunctioning
smoothly,rolingcaraftercarotheendoftheline.Theandon
cordcaninterruptthiscarefulowasthelineishaltedrepeatedly.
However,thebenetsofndingandxingproblems
faster
outweighthiscost.Thisprocessofcontinuouslydrivingoutdefects
hasbeenawin-winforToyotaanditscustomers.Itistherootcause
ofToyota’shistorichighqualityratingsandlowcosts.
SMALLBATCHESIN
ENTREPRENEURSHIP
WhenIteachentrepreneursthismethod,Ioftenbeginwithstories
aboutmanufacturing.Beforelong,Icanseethequestioninglooks:
whatdoesthishavetodowithmystartup?Thetheorythatisthe
foundationofToyota’ssuccesscanbeusedtodramaticalyimprove
thespeedatwhichstartupsfindvalidatedlearning.
Toyotadiscoveredthatsmalbatchesmadetheirfactoriesmore
ecient.Incontrast,intheLeanStartupthegoalisnottoproduce
morestueciently.Itisto—asquicklyaspossible—learnhowto
buildasustainablebusiness.
Thinkbacktotheexampleofenvelopestung.Whatifitturns
outthatthecustomerdoesn’twanttheproductwe’rebuilding?
Althoughthisisnevergoodnewsforanentrepreneur,ndingout
soonerismuchbeterthanndingoutlater.Workinginsmal
batchesensuresthatastartupcanminimizetheexpenditureof
time,money,andeortthatultimatelyturnsouttohave
been
wasted.
SmalBatchesatIMVU
AtIMVU,weappliedtheselessonsfrommanufacturingtotheway
wework.Normaly,newversionsofproductslikeoursarereleased
tocustomersonamonthly,quarterly,oryearlycycle.
Takealookatyourcelphone.Oddsare,itisnottheveryrst
versionofitskind.EveninnovativecompaniessuchasApple
produceanewversionoftheiragshipphonesaboutonceayear.
Bundledupinthatproductreleasearedozensofnewfeatures(at
thereleaseofiPhone4,Appleboastedmorethan1,500changes).
Ironicaly,manyhigh-techproductsaremanufacturedin
advancedfacilitiesthatfolowthelatestinleanthinking,including
smalbatchesandsingle-pieceow.However,theprocessthatis
usedtodesigntheproductisstuckintheeraofmassproduction.
Thinkofalthechangesthataremadetoaproductsuchasthe
iPhone;al1,500ofthemarereleasedtocustomersinone
giant
batch.
Behindthescenes,inthedevelopmentanddesignoftheproduct
itself,largebatchesarestiltherule.Theworkthatgoesintothe
developmentofanewproductproceedsonavirtual
assemblyline.
Productmanagersgureoutwhatfeaturesarelikelytoplease
customers;productdesignersthengureouthowthosefeatures
shouldlookandfeel.Thesedesignsarepassedtoengineering,
whichbuildssomethingnewormodiesanexistingproductand,
oncethisisdone,handsitotosomebodyresponsibleforverifying
thatthenewproductworksthewaytheproductmanagersand
designersintended.ForaproductsuchastheiPhone,
theseinternal
handofsmayhappenonamonthlyorquarterlybasis.
Thinkbackonemoretimetotheenvelope-stungexercise.
Whatisthemosteficientwaytodothiswork?
AtIMVU,weatemptedtodesign,develop,andshipournew
featuresoneatatime,takingadvantageofthepowerofsmal
batches.Here’swhatitlookedlike.
Insteadofworkinginseparatedepartments,engineersand
designerswouldworktogethersidebysideononefeatureata
time.Wheneverthatfeaturewasreadytobetestedwithcustomers,
time.Wheneverthatfeaturewasreadytobetestedwithcustomers,
theyimmediatelywouldreleaseanewversionoftheproduct,
whichwouldgoliveonourwebsiteforarelativelysmal
number
ofpeople.Theteamwouldbeableimmediatelytoassessthe
impactoftheirwork,evaluateitseectoncustomers,anddecide
whattodonext.Fortinychanges,thewholeprocessmightbe
repeatedseveraltimesperday.Infact,intheaggregate,IMVU
makesaboutftychangestoitsproduct(onaverage)everysingle
day.
JustaswiththeToyotaProductionSystem,thekeytobeingable
tooperatethisquicklyistocheckfordefectsimmediately,thus
preventingbiggerproblemslater.Forexample,wehadan
extensivesetofautomatedteststhatassuredthataftereverychange
ourproductstilworkedasdesigned.Let’ssayanengineer
accidentalyremovedanimportantfeature,suchasthecheckout
butonononeofoure-commercepages.Withoutthisbuton,
customersnolongercouldbuyanythingfromIMVU.It’sasifour
businessinstantlybecameahobby.Analogouslytothe
Toyota
andoncord,IMVUusedanelaboratesetofdefensemechanisms
thatpreventedengineersfromaccidentalybreakingsomething
important.
Wecaledthisourproduct’simmunesystembecause
those
automaticprotectionswentbeyondcheckingthattheproduct
behavedasexpected.Wealsocontinuouslymonitoredthehealthof
ourbusinessitselfsothatmistakeswerefoundandremoved
automaticaly.
Goingbacktoourbusiness-to-hobbyexampleofthemissing
checkoutbuton,let’smaketheproblemalitlemoreinteresting.
Imaginethatinsteadofremovingthebutonaltogether,anengineer
makesamistakeandchangesthebuton’scolorsothatitisnow
whiteonawhitebackground.Fromthepointofviewofautomated
functionaltests,thebutonisstilthereandeverythingisworking
normaly;fromthecustomer’spointofview,the
butonisgone,
andsonobodycanbuyanything.Thisclassofproblemsishardto
detectsolelywithautomationbutisstilcatastrophicfroma
businesspointofview.AtIMVU,ourimmunesystemis
programmedtodetectthese
businessconsequencesand
programmedtodetectthesebusinessconsequencesand
automaticalyinvokeourequivalentoftheandoncord.
Whenourimmunesystemdetectsaproblem,anumberofthings
happenimmediately:
1.Thedefectivechangeisremovedimmediatelyand
automaticaly.
2.Everyoneontherelevantteamisnotifiedoftheproblem.
3.Theteamisblockedfromintroducinganyfurtherchanges,
preventingtheproblemfrom
beingcompoundedbyfuture
mistakes…
4.…untiltherootcauseoftheproblemisfoundandxed.(This
rootcauseanalysisisdiscussedingreaterdetailinChapter11.)
AtIMVU,wecaledthiscontinuousdeployment,and
eveninthe
fast-movingworldofsoftwaredevelopmentitisstilconsidered
controversial.3AstheLeanStartupmovementhasgainedtraction,it
hascometobeembracedbymoreandmorestartups,eventhose
thatoperatemission-criticalapplications.Amongthemostcuting
edgeexamplesisWealthfront,whosepivotwasdescribedin
Chapter8.Thecompanypracticestruecontinuousdeployment—
includingmorethanadozenreleasestocustomersevery
day—inan
SEC-regulatedenvironment.4
ContinuousDeploymentBeyondSoftware
WhenItelthisstorytopeoplewhoworkinaslower-moving
industry,theythinkIamdescribingsomethingfuturistic.But
increasingly,moreandmoreindustriesareseeingtheirdesign
processacceleratedbythesameunderlyingforcesthatmakethis
kindofrapiditerationpossibleinthesoftwareindustry.Thereare
threewaysinwhichthisishappening:
1.Hardwarebecomingsoftware.Thinkaboutwhathashappened
inconsumerelectronics.Thelatestphonesandtabletcomputersare
inconsumerelectronics.Thelatestphonesandtabletcomputersare
litlemorethanascreenconnectedtotheInternet.
Almostalof
theirvalueisdeterminedbytheirsoftware.Evenold-school
productssuchasautomobilesareseeingever-largerpartsoftheir
valuebeinggeneratedbythesoftwaretheycarryinside,which
controlseverythingfromtheentertainmentsystemtotuningthe
enginetocontrolingthebrakes.Whatcanbebuiltoutofsoftware
canbemodiedmuchfasterthanaphysicalormechanicaldevice
can.
2.Fastproductionchanges.Becauseofthesuccessofthelean
manufacturingmovement,manyassemblylinesaresetuptoalow
eachnewproductthatcomesothelinetobecustomized
completelywithoutsacricingqualityorcost-eectiveness.
Historicaly,thishasbeenusedtooerthecustomermanychoices
ofproduct,butinthefuture,thiscapabilitywilalowthedesigners
ofproductstogetmuchfasterfeedbackaboutnewversions.When
thedesignchanges,thereisnoexcessinventoryofthe
oldversionto
slowthingsdown.Sincemachinesaredesignedforrapid
changeovers,assoonasthenewdesignisready,newversionscan
beproducedquickly.
3.3Dprintingandrapidprototypingtools.Asjustone
example,
mostproductsandpartsthataremadeoutofplastictodayaremass
producedusingatechniquecaledinjectionmolding.Thisprocess
isextremelyexpensiveandtime-consumingtosetup,butonceitis
upandrunning,itcanreproducehundredsofthousandsofidentical
individualitemsatanextremelylowcost.Itisaclassiclarge-batch
productionprocess.Thishasputentrepreneurswhowantto
developanewphysicalproductatadisadvantage,
sinceingeneral
onlylargecompaniescanaordtheselargeproductionrunsfora
newproduct.However,newtechnologiesarealowing
entrepreneurstobuildsmalbatchesofproductsthatareofthe
samequalityasproducts
madewithinjectionmolding,butatmuch
lowercostandmuch,muchfaster.
Theessentiallessonisnotthateveryoneshouldbeshippingfty
timesperdaybutthatbyreducingbatchsize,wecangetthrough
theBuild-Measure-Learnfeedbackloopmorequicklythanour
competitorscan.Theabilitytolearnfasterfromcustomersisthe
essentialcompetitiveadvantagethatstartupsmustpossess.
SMALLBATCHESINACTION
Toseethisprocessinaction,letmeintroduceyoutoacompanyin
Boise,Idaho,caledSGWDesignworks.SGW’sspecialtyisrapid
productiontechniquesforphysicalproducts.Manyofitsclientsare
startups.
SGWDesignworkswasengagedbyaclientwhohadbeenasked
byamilitarycustomertobuildacomplexeldx-raysystemto
detectexplosivesandotherdestructivedevicesatbordercrossings
andinwarzones.
Conceptualy,thesystemconsistedofanadvancedheadunitthat
readx-raylm,multiplex-raylmpanels,andtheframeworkto
holdthepanelswhilethelmwasbeingexposed.Theclient
alreadyhadthetechnologyforthex-raypanelsandthe
headunit,
buttomaketheproductworkinruggedmilitarysetings,SGW
neededtodesignanddeliverthesupportingstructurethatwould
makethetechnologyusableintheeld.Theframeworkhadtobe
stabletoensureaqualityx-rayimage,durableenoughforuseina
warzone,easytodeploywithminimaltraining,andsmalenough
tocolapseintoabackpack.
Thisispreciselythekindofproductweareaccustomedto
thinkingtakesmonthsor
yearstodevelop,yetnewtechniquesare
shrinkingthattimeline.SGWimmediatelybegantogeneratethe
visualprototypesbyusing3Dcomputer-aideddesign(CAD)
software.The3Dmodelsservedasarapidcommunicationtool
betweentheclientandtheSGWteamtomakeearlydesign
decisions.
Theteamandclientsetledonadesignthatusedanadvanced
Theteamandclientsetledonadesignthatusedanadvanced
lockinghingetoprovidethecolapsibilityrequiredwithout
compromisingstability.Thedesignalsointegratedasuction
cup/pumpmechanismtoalowforfast,repeatableatachmentto
thex-raypanels.Soundscomplicated,right?
Threedayslater,theSGWteamdeliveredtherstphysical
prototypestotheclient.Theprototypesweremachinedoutof
aluminumdirectlyfromthe3Dmodel,usingatechniquecaled
computernumericalcontrol(CNC)andwerehand
assembledby
theSGWteam.
Theclientimmediatelytooktheprototypestoitsmilitarycontact
forreview.Thegeneralconceptwasacceptedwithanumberof
minordesignmodications.Inthenextvedays,another
fulcycle
ofdesigniteration,prototyping,anddesignreviewwascompleted
bytheclientandSGW.Therstproductionrunoffortycompleted
unitswasreadyfordeliverythreeandahalfweeksafterthe
initiationofthedevelopmentproject.
SGWrealizedthatthiswasawinningmodelbecausefeedbackon
designdecisionswasnearlyinstantaneous.Theteamusedthesame
processtodesignanddelivereightproducts,servingawiderange
offunctions,inatwelve-monthperiod.Halfofthoseproductsare
generatingrevenuetoday,andtherestareawaitinginitialorders,
althankstothepowerofworkinginsmalbatches.
THEPROJECTTIMELINE
Designandengineeringof
theinitialvirtualprototype
1day
Productionandassemblyofinitialhardprototypes
3days
Designiteration:twoadditionalcycles
5days
Initialproductionrunandassemblyofinitialfortyunits15days
SmalBatchesinEducation
Noteverytypeofproduct—asitexiststoday—alowsfordesign
Noteverytypeofproduct—asitexiststoday—alowsfordesign
changeinsmalbatches.Butthatisnoexcuseforstickingto
outdatedmethods.Asignicantamountofworkmaybeneededto
enableinnovatorstoexperimentinsmalbatches.Aswaspointed
outinChapter2,forestablishedcompanies
lookingtoaccelerate
theirinnovationteams,buildingthisplatformforexperimentation
istheresponsibilityofseniormanagement.
Imaginethatyouareaschoolteacherinchargeofteachingmath
tomiddleschoolstudents.
Althoughyoumayteachconceptsin
smalbatches,onedayatatime,youroveralcurriculumcannot
changeveryoften.Becauseyoumustsetupthecurriculumin
advanceandteachthesameconceptsinthesameordertoevery
studentintheclassroom,youcantryanewcurriculumatmostonly
onceayear.
Howcouldamathteacherexperimentwithsmalbatches?Under
thecurrentlarge-batchsystemforeducatingstudents,itwouldbe
quitedicult;ourcurrenteducationalsystemwasdesignedinthe
eraofmassproductionanduseslargebatchesextensively.
Anewbreedofstartupsisworkinghardtochangealthat.In
SchoolofOne,studentshavedaily“playlists”oftheir
learningtasks
thatareatunedtoeachstudent’slearningneeds,basedonthat
student’sreadinessandlearningstyle.Forexample,Juliaisway
aheadofgradelevelinmathandlearnsbestinsmalgroups,soher
playlistmightincludethreeorfourvideosmatchedtoheraptitude
level,athirty-minuteone-on-onetutoringsessionwithherteacher,
andasmalgroupactivityinwhichsheworksonamathpuzzle
withthreepeersatsimilaraptitudelevels.Thereare
assessments
builtintoeachactivitysothatdatacanbefedbacktotheteacherto
chooseappropriatetasksforthenextplaylist.Thisdatacanbe
aggregatedacrossclasses,schools,orevenwholedistricts.
Nowimaginetryingtoexperimentwithacurriculumbyusinga
toolsuchasSchoolofOne.Eachstudentisworkingathisorher
ownpace.Let’ssayyouareateacherwhohasanewsequencein
mindforhowmathconceptsshouldbetaught.Youcansee
immediatelytheimpactofthechangeonthoseofyourstudents
whoareatthatpointinthecurriculum.Ifyoujudgeittobeagood
change,youcouldrolitoutimmediatelyforeverysinglestudent;
change,youcouldrolitoutimmediatelyforeverysingle
student;
whentheygettothatpartofthecurriculum,theywilgetthenew
sequenceautomaticaly.Inotherwords,toolslikeSchoolofOne
enableteacherstoworkinmuchsmalerbatches,tothebenetof
theirstudents.(And,astoolsreachwide-scaleadoption,successful
experimentsbyindividualteacherscanberoledoutdistrict-,city-,
orevennationwide.)Thisapproachishavinganimpactand
earningaccolades.Timemagazinerecentlyincluded
SchoolofOne
inits“mostinnovativeideas”list;itwastheonlyeducational
organizationtomakethelist.5
THELARGE-BATCHDEATHSPIRAL
Smalbatchesposeachalengetomanagerssteepedin
traditional
notionsofproductivityandprogress,becausetheybelievethat
functionalspecializationismoreeficientforexpertworkers.
Imagineyou’reaproductdesigneroverseeinganewproductand
youneedtoproducethirtyindividualdesigndrawings.Itprobably
seemsthatthemostecientwaytoworkisinseclusion,by
yourself,producingthedesignsonebyone.Then,whenyou’re
donewithalofthem,youpassthedrawingsontothe
engineering
teamandletthemwork.Inotherwords,youworkinlargebatches.
Fromthepointofviewofindividualeciency,workinginlarge
batchesmakessense.Italsohasotherbenets:itpromotesskil
building,makesiteasiertoholdindividualcontributors
accountable,and,mostimportant,alowsexpertstoworkwithout
interruption.Atleastthat’sthetheory.Unfortunately,realityseldom
worksoutthatway.
Considerourhypothetical
example.Afterpassingthirtydesign
drawingstoengineering,thedesignerisfreetoturnhisorher
atentiontothenextproject.Butremembertheproblemsthatcame
upduringtheenvelope-stungexercise.Whathappenswhen
engineeringhasquestionsabouthowthedrawingsaresupposedto
work?Whatifsomeofthedrawingsareunclear?Whatif
somethinggoeswrongwhenengineeringatemptstousethe
drawings?
drawings?
Theseproblemsinevitablyturnintointerruptionsforthe
designer,andnowthoseinterruptionsareinterferingwiththenext
largebatchthedesignerissupposedtobeworkingon.Ifthe
drawingsneedtoberedone,
theengineersmaybecomeidlewhile
theywaitforthereworktobecompleted.Ifthedesignerisnot
available,theengineersmayhavetoredothedesignsthemselves.
Thisiswhysofewproductsareactualybuiltthewaytheyare
designed.
WhenIworkwithproductmanagersanddesignersincompanies
thatuselargebatches,Ioftendiscoverthattheyhavetoredotheir
workveorsixtimesforeveryrelease.OneproductmanagerI
workedwithwassoinundatedwithinterruptionsthathetookto
comingintotheoceinthemiddleofthenightsothathecould
workuninterrupted.WhenIsuggestedthathetryswitchingthe
workprocessfromlarge-batchtosingle-pieceow,he
refused—
becausethatwouldbeinecient!Sostrongistheinstincttowork
inlargebatches,thatevenwhenalarge-batchsystemis
malfunctioning,wehaveatendencytoblameourselves.
Largebatchestendtogrowovertime.Becausemoving
thebatch
forwardoftenresultsinadditionalwork,rework,delays,and
interruptions,everyonehasanincentivetodoworkinever-larger
batches,tryingtominimizethisoverhead.Thisiscaledthelarge-
batchdeathspiralbecause,unlikeinmanufacturing,thereareno
physicallimitsonthemaximumsizeofabatch.6Itispossiblefor
batchsizetokeepgrowingandgrowing.Eventualy,onebatchwil
becomethehighest-priorityproject,a“betthecompany”
new
versionoftheproduct,becausethecompanyhastakensuchalong
timesincethelastrelease.Butnowthemanagersareincentivized
toincreasebatchsizeratherthanshiptheproduct.Inlightofhow
longtheproducthasbeenindevelopment,whynotxonemore
bugoraddonemorefeature?Whorealywantstobethemanager
whoriskedthesuccessofthishugereleasebyfailingtoaddressa
potentialycriticalflaw?
Iworkedatacompanythatenteredthisdeathspiral.Wehad
beenworkingformonthsonanewversionofarealycoolproduct.
beenworkingformonthsonanewversionofarealycoolproduct.
Theoriginalversionhadbeenyearsinthemaking,and
expectationsforthenextreleasewereincrediblyhigh.Butthe
longerweworked,themoreafraidwebecameofhowcustomers
wouldreactwhentheynalysawthenewversion.Asourplans
becamemoreambitious,sotoodidthenumberofbugs,
conicts,
andproblemswehadtodealwith.Pretysoonwegotintoa
situationinwhichwecouldnotshipanything.Ourlaunchdate
seemedtorecedeintothedistance.Themoreworkwegotdone,
themoreworkwehadtodo.Thelackofabilitytoshipeventualy
precipitatedacrisisandachangeofmanagement,albecauseofthe
trapoflargebatches.
Thesemisconceptionsaboutbatchsizeareincrediblycommon.
Hospitalpharmaciesoftendeliverbigbatchesofmedicationsto
patientoorsonceadaybecauseit’secient(asingletrip,right?).
Butmanyofthosemedsgetsentbacktothepharmacywhena
patient’sordershavechangedorthepatientismovedor
discharged,
causingthepharmacystaftodolotsofreworkandreprocessing(or
trashing)ofmeds.Deliveringsmalerbatcheseveryfourhours
reducesthetotalworkloadforthepharmacyandensuresthatthe
rightmedsareattherightplacewhenneeded.
Hospitallabbloodcolectionsoftenaredoneinhourlybatches;
phlebotomistscolectbloodforanhourfrommultiplepatientsand
thensendortakealthesamplestothelab.Thisaddsto
turnaroundtimefortestresultsandcanharmtestquality.Ithas
becomecommonforhospitalstobringsmalbatches(twopatients)
orasingle-patientowofspecimenstothelabeveniftheyhaveto
hireanextraphlebotomistortwotodoso,becausethe
totalsystem
costislower.7
PULL,DON’TPUSH
Let’ssayyouareoutforadrive,ponderingthemeritsofsmal
batches,andndyourselfaccidentalyputingadentinyournew
2011blueToyotaCamry.Youtakeitintothedealershipforrepair
andwaittohearthebadnews.Therepairtechniciantelsyouthat
andwaittohearthebadnews.Therepairtechniciantelsyouthat
youneedtohavethebumperreplaced.Hegoestocheck
their
inventorylevelsandtelsyouhehasanewbumperinstockand
theycancompleteyourrepairimmediately.Thisisgoodnewsfor
everyone—youbecauseyougetyourcarbacksoonerandthe
dealershipbecausetheyhaveahappycustomeranddon’trunthe
riskofyourtakingthecarsomewhereelseforrepair.Also,they
don’thavetostoreyourcarorgiveyoualoanerwhiletheywaitfor
theparttocomein.
Intraditionalmassproduction,thewaytoavoidstockouts—not
havingtheproductthecustomerwants—istokeepalarge
inventoryofsparesjustincase.Itmaybethattheblue2011Camry
bumperisquitepopular,butwhataboutlastyear’smodel
orthe
modelfromveyearsago?Themoreinventoryyoukeep,the
greaterthelikelihoodyouwilhavetherightproductinstockfor
everycustomer.Butlargeinventoriesareexpensivebecausethey
havetobetransported,stored,andtracked.Whatifthe2011
bumperturnsouttohaveadefect?Althesparesinalthe
warehousesinstantlybecomewaste.
Leanproductionsolvestheproblemofstockoutswitha
techniquecaledpul.When
youbringacarintothedealershipfor
repair,oneblue2011Camrybumpergetsused.Thiscreatesa
“hole”inthedealer’sinventory,whichautomaticalycausesasignal
tobesenttoalocalrestockingfacilitycaledtheToyotaParts
DistributionCenter(PDC).ThePDCsendsthedealeranewbumper,
whichcreatesanotherholeininventory.Thissendsasimilarsignal
toaregionalwarehousecaledtheToyotaPartsRedistribution
Center(PRC),wherealpartssuppliersshiptheirproducts.
That
warehousesignalsthefactorywherethebumpersaremadeto
produceonemorebumper,whichismanufacturedandshippedto
thePRC.
Theidealgoalistoachievesmalbatchesalthewaydown
to
single-pieceowalongtheentiresupplychain.Eachstepinthe
linepulsthepartsitneedsfromthepreviousstep.Thisisthe
famousToyotajust-in-timeproductionmethod.8
Whencompaniesswitchto
thiskindofproduction,their
Whencompaniesswitchtothiskindofproduction,their
warehousesimmediatelyshrink,astheamountofjust-in-case
inventory[caledwork-in-progress(WIP)inventory]isreduced
dramaticaly.Thisalmost
magicalshrinkageofWIPiswherelean
manufacturinggetsitsname.It’sasifthewholesupplychain
suddenlywentonadiet.
Startupsstruggletoseetheirwork-in-progressinventory.When
factorieshaveexcessWIP,it
literalypilesuponthefactoryoor.
Becausemoststartupworkisintangible,it’snotnearlyasvisible.
Forexample,altheworkthatgoesintodesigningtheminimum
viableproductis—untilthemomentthatproductisshipped—just
WIPinventory.Incompletedesigns,not-yet-validatedassumptions,
andmostbusinessplansareWIP.AlmosteveryLeanStartup
techniquewe’vediscussedsofarworksitsmagicintwoways:by
convertingpushmethodstopulandreducingbatchsize.
Bothhave
thenetefectofreducingWIP.
Inmanufacturing,pulisusedprimarilytomakesureproduction
processesaretunedtolevelsofcustomerdemand.Withoutthis,
factoriescanwindupmaking
muchmore—ormuchless—ofa
productthancustomersrealywant.However,applyingthis
approachtodevelopingnewproductsisnotstraightforward.Some
peoplemisunderstandtheLeanStartupmodelassimplyapplying
pultocustomerwants.Thisassumesthatcustomerscouldtelus
whatproductstobuildandthatthiswouldactasthepulsignalto
productdevelopmenttomakethem.9
Aswasmentionedearlier,thisisnotthewaytheLeanStartup
modelworks,becausecustomersoftendon’tknowwhattheywant.
Ourgoalinbuildingproductsistobeabletorunexperimentsthat
wilhelpuslearnhowtobuildasustainablebusiness.Thus,the
rightwaytothinkabouttheproductdevelopmentprocess
ina
LeanStartupisthatitisrespondingtopulrequestsintheformof
experimentsthatneedtoberun.
Assoonasweformulateahypothesisthatwewanttotest,the
productdevelopmentteam
shouldbeengineeredtodesignandrun
thisexperimentasquicklyaspossible,usingthesmalestbatchsize
thatwilgetthejobdone.Rememberthatalthoughwewritethe
thatwilgetthejobdone.Rememberthatalthoughwewritethe
feedbackloopasBuild-Measure-Learnbecausetheactivitieshappen
inthatorder,ourplanningrealyworksinthereverseorder:we
gureoutwhatweneedtolearnandthenworkbackwardstosee
whatproductwilworkasanexperimenttogetthat
learning.Thus,
itisnotthecustomer,butratherourhypothesisaboutthecustomer,
thatpulsworkfromproductdevelopmentandotherfunctions.Any
otherworkiswaste.
HypothesisPulinCleanTech
Toseethisinaction,let’stakealookatBerkeley-basedstartup
AlphabetEnergy.Anymachineorprocessthatgeneratespower,
whetheritisamotorinafactoryoracoal-burningpowerplant,
generatesheatasaby-product.AlphabetEnergyhas
developeda
productthatcangenerateelectricityfromthiswasteheat,usinga
newkindofmaterialcaledathermoelectric.AlphabetEnergy’s
thermoelectricmaterialwasdevelopedovertenyearsbyscientists
attheLawrenceBerkeleyNationalLaboratories.
Aswithmanycleantechnologyproducts,therearehuge
chalengesinbringingaproductlikethistomarket.Whileworking
throughitsleap-of-faithassumptions,Alphabetguredoutearly
thatdevelopingasolutionforwastethermoelectricityrequired
buildingaheatexchangerandagenericdevicetotransferheatfrom
onemediumtoanotheraswelasdoingproject-specic
engineering.Forinstance,ifAlphabetwantedtobuildasolutionfor
autilitysuchasPacicGasandElectric,theheatexchangerwould
havetobecongured,shaped,andinstaledtocapturetheheat
fromapowerplant’sexhaustsystem.
WhatmakesAlphabetEnergyuniqueisthatthecompanymadea
savvydecisionearlyonintheresearchprocess.Insteadofusing
relativelyrareelementsasmaterials,theydecidedtobasetheir
researchonsiliconwafers,thesamephysicalsubstancethat
computercentralprocessingunits(CPUs)aremadefrom.
AsCEO
MathewSculinexplains,“Ourthermoelectricistheonlyonethat
canuselow-costsemiconductorinfrastructureformanufacturing.”
canuselow-costsemiconductorinfrastructureformanufacturing.”
ThishasenabledAlphabetEnergytodesignandbuilditsproducts
insmalbatches.
Bycontrast,mostsuccessfulcleantechnologystartupshavehadto
makesubstantialearlyinvestments.Thesolarpanelprovider
SunPowerhadtobuildinfactoriestomanufactureitspanelsand
partnerwithinstalersbeforebecomingfulyoperational.Similarly,
BrightSourceraised$291miliontobuildandoperatelarge-scale
solarplantswithoutdeliveringawattoasingle
customer.
Insteadofhavingtoinvesttimeandmoneyinexpensive
fabricationfacilities,Alphabetisabletotakeadvantageofthe
massiveexistinginfrastructurethatproducessiliconwafersfor
computerelectronics.Asa
result,Alphabetcangofromaproduct
concepttoholdingaphysicalversioninitshandinjustsixweeks
fromendtoend.Alphabet’schalengehasbeentondthe
combinationofperformance,price,andphysicalshapethatisa
matchforearlycustomers.Althoughitstechnologyhas
revolutionarypotential,earlyadopterswildeployitonlyifthey
canseeaclearreturnoninvestment.
ItmightseemthatthemostobviousmarketforAlphabet’s
technologywouldbepowerplants,andindeed,thatwastheteam’s
initialhypothesis.Alphabethypothesizedthatsimplecyclegas
turbineswouldbeanidealapplication;theseturbines,whichare
similartojetenginesstrappedtotheground,are
usedbypower
generatorstoprovideenergyforpeakdemand.Alphabetbelieved
thatatachingitssemiconductorstothoseturbineswouldbesimple
andcheap.
Thecompanywentabouttestingthishypothesisin
smalbatches
bybuildingsmal-scalesolutionsforitscustomersasawayof
learning.Aswithmanyinitialideas,theirhypothesiswasdisproved
quickly.Powercompanieshavealowtoleranceforrisk,making
themunlikelytobecomeearlyadopters.Becauseitwasn’tweighed
downbyalarge-batchapproach,Alphabetwasreadytopivotafter
justthreemonthsofinvestigation.
Alphabethaseliminatedmanyotherpotentialmarketsaswel,
leadingtoaseriesofcustomersegmentpivots.Thecompany’s
currenteortsarefocusedonmanufacturingrms,whichhavethe
currenteortsarefocusedonmanufacturingrms,whichhavethe
abilitytoexperimentwith
newtechnologiesinseparatepartsof
theirfactory;thisalowsearlyadopterstoevaluatethereal-world
benetsbeforecommitingtoalargerdeployment.Theseearly
deploymentsareputingmoreofAlphabet’sassumptionstothe
test.Unlikeinthecomputerhardwarebusiness,customersarenot
wilingtopaytopdolarformaximumperformance.Thishas
requiredsignicantchangesinAlphabet’sproduct,conguringitto
achievethelowestcostperwatpossible.
Althisexperimentationhascostthecompanyatinyfractionof
whatotherenergystartupshaveconsumed.Todate,Alphabethas
raisedapproximately$1milion.Onlytimewilteliftheywil
prevail,butthankstothepowerofsmalbatches,they
wilbeable
todiscoverthetruthmuchfaster.10
TheToyotaProductionSystemisprobablythemostadvanced
systemofmanagementintheworld,butevenmoreimpressiveis
thefactthatToyotahasbuilt
themostadvancedlearning
organizationinhistory.Ithasdemonstratedanabilitytounleashthe
creativityofitsemployees,achieveconsistentgrowth,andproduce
innovativenewproductsrelentlesslyoverthecourseofnearlya
century.11
Thisisthekindoflong-termsuccesstowhichentrepreneurs
shouldaspire.Althoughleanproductiontechniquesarepowerful,
theyareonlyamanifestationofahigh-functioningorganizationthat
iscommitedtoachievingmaximumperformancebyemployingthe
rightmeasuresofprogressoverthelongterm.Processisonlythe
foundationuponwhichagreatcompanyculturecandevelop.But
withoutthisfoundation,eortstoencouragelearning,
creativity,
andinnovationwilfalat—asmanydisilusioneddirectorsofHR
canatest.
TheLeanStartupworksonlyifweareabletobuildan
organizationasadaptableandfastasthechalengesitfaces.This
requirestacklingthehumanchalengesinherentinthisnewwayof
working;thatisthesubjectoftheremainderofPartThree.
THESTARTUPWAY
10
GROW
Irecentlyhadtwostartupsseekmyadviceonthesameday.As
typesofbusinesses,theycouldnothavebeenmoredierent.The
rstisdevelopingamarketplacetohelptradersofcolectibles
connectwithoneanother.Thesepeoplearehard-core
fansof
movies,anime,orcomicswhostrivetoputtogethercomplete
colectionsoftoysandotherpromotionalmerchandiserelatedto
thecharacterstheylove.Thestartupaspirestocompetewithonline
marketplacessuchaseBayaswelasphysicalmarketplaces
atachedtoconventionsandothergatheringsoffans.
Thesecondstartupselsdatabasesoftwaretoenterprise
customers.Theyhaveanext-generationdatabasetechnologythat
cansupplementorreplaceoeringsfromlargecompaniessuchas
Oracle,IBM,andSAP.Theircustomersarechiefinformationoficers
(CIOs),ITmanagers,andengineersinsomeoftheworld’slargest
organizations.Thesearelong-lead-timesalesthat
require
salespeople,salesengineering,instalationsupport,and
maintenancecontracts.
Youcouldbeforgivenforthinkingthesetwocompanieshave
absolutelynothingincommon,yetbothcameto
mewiththeexact
sameproblem.Eachonehadearlycustomersandpromisingearly
revenue.Theyhadvalidatedandinvalidatedmanyhypothesesin
theirbusinessmodelsandwereexecutingagainsttheirproductroad
mapssuccessfuly.Theircustomershadprovidedahealthymixof
positivefeedbackandsuggestionsforimprovements.Both
companieshadusedtheirearlysuccesstoraisemoneyfromoutside
companieshadusedtheirearlysuccesstoraisemoney
fromoutside
investors.
Theproblemwasthatneithercompanywasgrowing.
BothCEOsbroughtmeidentical-lookinggraphsshowingthat
theirearlygrowthhadatlined.Theycouldnotunderstandwhy.
Theywereacutelyawareoftheneedtoshowprogresstotheir
employeesandinvestorsandcametomebecausetheywanted
adviceonhowtojump-starttheirgrowth.Shouldtheyinvestin
moreadvertisingormarketingprograms?Should
theyfocuson
productqualityornewfeatures?Shouldtheytrytoimprove
conversionratesorpricing?
Asitturnsout,bothcompaniesshareadeepsimilarityintheway
theirbusinessesgrow—andthereforeasimilarconfusion
about
whattodo.Bothareusingthesameengineofgrowth,thetopicof
thischapter.
WHEREDOESGROWTHCOMEFROM?
Theengineofgrowthisthemechanismthatstartupsusetoachieve
sustainablegrowth.Iusethewordsustainabletoexcludealone-
timeactivitiesthatgenerateasurgeofcustomersbuthavenolong-
termimpact,suchasasingleadvertisementorapublicitystuntthat
mightbeusedtojump-startgrowthbutcouldnotsustain
that
growthforthelongterm.
Sustainablegrowthischaracterizedbyonesimplerule:
Newcustomerscomefromtheactionsofpastcustomers.
Therearefourprimarywayspastcustomersdrivesustainable
growth:
1.Wordofmouth.Embeddedinmostproductsisanaturallevel
ofgrowththatiscausedbysatisedcustomers’enthusiasmforthe
product.Forexample,whenIboughtmyrstTiVoDVR,Icouldn’t
stoptelingmyfriendsandfamilyaboutit.Pretysoon,myentire
familywasusingone.
2.Asasideeectofproductusage.Fashionorstatus,suchas
luxurygoodsproducts,driveawarenessofthemselveswhenever
theyareused.Whenyouseesomeonedressedinthelatestclothes
ordrivingacertaincar,youmaybeinuencedtobuythatproduct.
Thisisalsotrueofso-caledviralproductssuchasFacebookand
PayPal.Whenacustomersendsmoneytoafriend
usingPayPal,the
friendisexposedautomaticalytothePayPalproduct.
3.Throughfundedadvertising.Mostbusinessesemploy
advertisingtoenticenewcustomerstousetheirproducts.Forthis
tobeasourceofsustainable
growth,theadvertisingmustbepaid
foroutofrevenue,notone-timesourcessuchasinvestmentcapital.
Aslongasthecostofacquiringanewcustomer(theso-caled
marginalcost)islessthantherevenuethatcustomergenerates(the
marginalrevenue),theexcess(themarginalprot)canbeusedto
acquiremorecustomers.Themoremarginalprot,thefasterthe
growth.
4.Throughrepeatpurchaseoruse.Someproductsaredesigned
tobepurchasedrepeatedlyeitherthroughasubscriptionplan(a
cablecompany)orthroughvoluntaryrepurchases(groceriesor
lightbulbs).Bycontrast,manyproductsandservicesare
intentionalydesignedasone-timeevents,suchas
wedding
planning.
ThesesourcesofsustainablegrowthpowerfeedbackloopsthatI
havetermedenginesofgrowth.Eachislikeacombustionengine,
turningoverandover.Thefastertheloopturns,the
fasterthe
companywilgrow.Eachenginehasanintrinsicsetofmetricsthat
determinehowfastacompanycangrowwhenusingit.
THETHREEENGINESOFGROWTH
WesawinPartTwohow
importantitisforstartupstousetheright
kindofmetrics—actionablemetrics—toevaluatetheirprogress.
However,thisleavesalargeamountofvarietyintermsofwhich
numbersoneshouldmeasure.Infact,oneofthemostexpensive
formsofpotentialwasteforastartupisspendingtimearguing
abouthowtoprioritizenewdevelopmentonceithasaproducton
themarket.Atanytime,thecompanycouldinvestitsenergyin
ndingnewcustomers,servicingexistingcustomers
beter,
improvingoveralquality,ordrivingdowncosts.Inmyexperience,
thediscussionsaboutthesekindsofprioritydecisionscanconsume
asubstantialfractionofthecompany’stime.
Enginesofgrowthare
designedtogivestartupsarelativelysmal
setofmetricsonwhichtofocustheirenergies.Asoneofmy
mentors,theventurecapitalinvestorShawnCarolan,putit,
“Startupsdon’tstarve;theydrown.”Therearealwaysazilionnew
ideasabouthowtomaketheproductbeteroatingaround,but
thehardtruthisthatmostofthoseideasmakeadierenceonlyat
themargins.Theyaremereoptimizations.Startupshavetofocuson
thebigexperimentsthatleadtovalidatedlearning.The
enginesof
growthframeworkhelpsthemstayfocusedonthemetricsthat
mater.
TheStickyEngineofGrowth
Thisbringsusbacktothetwostartupsthatkickedothischapter.
Bothareusingtheexactsameengineofgrowthdespitebeingin
verydierentindustries.Bothproductsaredesignedtoatractand
retaincustomersforthelongterm.Theunderlyingmechanismof
thatretentionisdierentinthetwocases.Forthecol
ectible
company,theideaistobecomethenumberoneshopping
destinationforfanaticalcolectors.Thesearepeoplewhoare
constantlyhuntingforthelatestitemsandthebestdeals.Ifthe
company’sproductworksasdesigned,colectorswhostartusingit
wilcheckconstantlyandrepeatedlytoseeifnewitemsareforsale
aswelaslistingtheirownitemsforsaleortrade.
aswelaslistingtheirownitemsforsaleortrade.
Thestartupdatabasevendorreliesonrepeatusageforavery
dierentreason.Databasetechnologyisusedonlyasthefoundation
foracustomer’sownproducts,suchasawebsiteorapointofsale
system.Onceyoubuildaproductontopofaparticular
database
technology,itisextremelydiculttoswitch.IntheITindustry,
suchcustomersaresaidtobelockedintothevendortheychoose.
Forsuchaproducttogrow,ithastooersuchacompelingnew
capabilitythatcustomersarewilingtoriskbeingtiedtoa
proprietaryvendorforapotentialylongtime.
Thus,bothbusinessesrelyonhavingahighcustomerretention
rate.Theyhaveanexpectationthatonceyoustartusingtheir
product,youwilcontinuetodoso.Thisisthesamedynamicasa
mobiletelephoneserviceprovider:whenacustomercancelshisor
herservice,itgeneralymeansthatheorsheisextremely
dissatisedorisswitchingtoacompetitor’sproduct.This
isin
contrastto,say,groceriesonastoreaisle.Inthegroceryretail
business,customertastesuctuate,andifacustomerbuysaPepsi
thisweekinsteadofCoke,it’snotnecessarilyabigdeal.
Therefore,companiesusingthestickyengineofgrowthtrack
theiratritionrateorchurnrateverycarefuly.Thechurnrateis
denedasthefractionofcustomersinanyperiodwhofailto
remainengagedwiththecompany’sproduct.
Therulesthatgovernthestickyengineofgrowthareprety
simple:iftherateofnewcustomeracquisitionexceedsthechurn
rate,theproductwilgrow.Thespeedofgrowthisdeterminedby
whatIcaltherateofcompounding,whichis
simplythenatural
growthrateminusthechurnrate.Likeabankaccountthatearns
compoundinginterest,havingahighrateofcompoundingwillead
toextremelyrapidgrowth—withoutadvertising,viralgrowth,or
publicitystunts.
Unfortunately,bothofthesestickystartupsweretrackingtheir
progressusinggenericindicatorssuchasthetotalnumberof
customers.Eventheactionablemetricstheywereusing,suchasthe
activationrateandrevenuepercustomer,weren’tveryhelpful
becauseinthestickyengineofgrowth,thesevariableshavelitle
impactongrowth.(Inthestickyengineofgrowth,theyarebeter
impactongrowth.(Inthestickyengineofgrowth,they
arebeter
suitedtotestingthevaluehypothesisthatwasdiscussedinChapter
5.)Afterourmeeting,oneofthetwostartupstookmeuponmy
advicetomodelitscustomerbehaviorbyusingthestickyengineof
growthasatemplate.Theresultswerestriking:a61percent
retentionrateanda39percentgrowthrateofnewcustomers.In
otherwords,itschurnrateandnewcustomeracquisitionbalanced
eachotheralmostperfectly,leadingtoacompounding
growthrate
ofjust0.02percent—almostzero.
Thisistypicalforcompaniesinanengagementbusinessthatare
strugglingtondgrowth.Aninsiderwhoworkedatthedot-com-
eracompanyPointCastonce
showedmehowthatcompany
sueredasimilardysfunction.WhenPointCastwasstrugglingto
grow,itwasnonethelessincrediblysuccessfulinnewcustomer
acquisition—justlikethisstickystartup(39percenteveryperiod).
Unfortunately,thisgrowthisbeingosetbyanequivalentamount
ofchurn.Onceitismodeledthisway,thegoodnewsshouldbe
apparent:thereareplentyofnewcustomerscominginthedoor.
Thewaytondgrowthistofocusonexistingcustomers
forthe
productevenmoreengagingtothem.Forexample,thecompany
couldfocusongetingmoreandbeterlistings.Thiswouldcreate
anincentiveforcustomerstocheckbackoften.Alternatively,the
companycoulddosomethingmoredirectsuchasmessagingthem
aboutlimited-timesalesorspecialofers.Eitherway,itsfocusneeds
tobeonimprovingcustomerretention.Thisgoesagainstthe
standardintuitioninthatifacompanylacksgrowth,it
should
investmoreinsalesandmarketing.Thiscounterintuitiveresultis
hardtoinferfromstandardvanitymetrics.
TheViralEngineofGrowth
OnlinesocialnetworksandTupperwareareexamplesofproducts
forwhichcustomersdothelion’sshareofthemarketing.Awareness
oftheproductspreadsrapidlyfrompersontopersonsimilarlyto
thewayavirusbecomesanepidemic.Thisisdistinctfromthe
thewayavirusbecomesanepidemic.Thisisdistinct
fromthe
simpleword-of-mouthgrowthdiscussedabove.Instead,products
thatexhibitviralgrowthdependonperson-to-persontransmission
asanecessaryconsequenceofnormalproductuse.Customersare
notintentionalyactingasevangelists;theyarenotnecessarily
tryingtospreadthewordabouttheproduct.Growthhappens
automaticalyasasideeectofcustomersusingtheproduct.
Virusesarenotoptional.
Forexample,oneofthemostfamousviralsuccessstoriesisa
companycaledHotmail.In1996,SabeerBhatiaandJackSmith
launchedanewweb-basede-mailservicethatoeredcustomers
freeaccounts.Atrst,growthwassluggish;withonlya
smalseed
investmentfromtheventurecapitalrmDraperFisherJurvetson,
theHotmailteamcouldnotaordanextensivemarketing
campaign.Buteverythingchangedwhentheymadeonesmal
tweaktotheproduct.They
addedtothebotomofeverysinglee-
mailthemessage“P.S.Getyourfreee-mailatHotmail”alongwith
aclickablelink.
Withinweeks,thatsmalproductchangeproducedmassive
results.Withinsixmonths,
BhatiaandSmithhadsignedupmore
than1milionnewcustomers.Fiveweekslater,theyhitthe2
milionmark.Eighteenmonthsafterlaunchingtheservice,with12
milionsubscribers,theysoldthecompanytoMicrosoftfor$400
milion.1
ThesamephenomenonisatworkinTupperware’sfamous
“houseparties,”inwhichcustomersearncommissionsbyselingthe
producttotheirfriendsandneighbors.Everysalespitchisan
opportunitynotonlytoselTupperwareproductsbutalsoto
persuadeothercustomerstobecomeTupperwarerepresentatives.
Tupperwarepartiesarestilgoingstrongdecadesaftertheystarted.
Manyothercontemporarycompanies,suchasPampered
Chef
(ownedbyWarrenBuet’sBerkshireHathaway),SouthernLiving,
andTastefulySimple,haveadoptedasimilarmodelsuccessfuly.
Liketheotherenginesofgrowth,theviralengineispoweredby
afeedbackloopthatcanbequantied.Itiscaledtheviralloop,
anditsspeedisdeterminedbyasinglemathematicaltermcaled
anditsspeedisdeterminedbyasinglemathematicaltermcaled
theviralcoecient.Thehigherthiscoecientis,the
fasterthe
productwilspread.Theviralcoecientmeasureshowmanynew
customerswiluseaproductasaconsequenceofeachnew
customerwhosignsup.Putanotherway,howmanyfriendswil
eachcustomerbringwith
himorher?Sinceeachfriendisalsoa
newcustomer,heorshehasanopportunitytorecruityetmore
friends.
Foraproductwithaviralcoecientof0.1,oneineveryten
customerswilrecruitoneofhisorherfriends.Thisisnot
a
sustainableloop.Imaginethatonehundredcustomerssignup.They
wilcausetenfriendstosignup.Thosetenfriendswilcauseone
additionalpersontosignup,buttheretheloopwilfizzleout.
Bycontrast,aviralloopwithacoecientthatisgreaterthan1.0
wilgrowexponentialy,becauseeachpersonwhosignsupwil
bring,onaverage,morethanoneotherpersonwithhimorher.
Toseetheseefectsgraphicaly,takealookatthischart:
Companiesthatrelyontheviralengineofgrowthmustfocuson
increasingtheviralcoecientmorethananythingelse,because
increasingtheviralcoecientmorethananythingelse,because
eventinychangesinthisnumberwilcausedramatic
changesin
theirfutureprospects.
Aconsequenceofthisisthatmanyviralproductsdonotcharge
customersdirectlybutrelyonindirectsourcesofrevenuesuchas
advertising.Thisisthecasebecauseviralproductscannot
aordto
haveanyfrictionimpedetheprocessofsigningcustomersupand
recruitingtheirfriends.Thiscanmaketestingthevaluehypothesis
forviralproductsespecialychalenging.
Thetruetestofthevalue
hypothesisisalwaysavoluntary
exchangeofvaluebetweencustomersandthestartupthatserves
them.Alotofconfusionstemsfromthefactthatthisexchangecan
bemonetary,asinthecaseofTupperware,ornonmonetary,asin
thecaseofFacebook.Intheviralengineofgrowth,monetary
exchangedoesnotdrivenewgrowth;itisusefulonlyasan
indicatorthatcustomersvaluetheproductenoughtopayforit.If
FacebookorHotmailhadstartedchargingcustomersintheirearly
days,itwouldhavebeenfoolish,asitwouldhaveimpededtheir
abilitytogrow.However,itisnottruethatcustomersdonotgive
thesecompaniessomethingofvalue:byinvestingtheirtimeand
atentionintheproduct,theymaketheproductvaluableto
advertisers.Companiesthatseladvertisingactualyservetwo
dierentgroupsofcustomers—consumersandadvertisers—and
exchangeadiferentcurrencyofvaluewitheach.2
Thisismarkedlydierentfromcompaniesthatactivelyuse
moneytofueltheirexpansion,suchasaretailchainthatcangrow
asfastasitcanfundtheopeningofnewstoresatsuitablelocations.
Thesecompaniesareusingadiferentengineofgrowthaltogether.
ThePaidEngineofGrowth
Imagineanotherpairofbusinesses.Therstmakes$1oneach
customeritsignsup;thesecondmakes$100,000fromeach
customeritsignsup.Topredictwhichcompanywilgrowfaster,
youneedtoknowonlyoneadditionalthing:howmuchit
coststo
signupanewcustomer.
signupanewcustomer.
ImaginethattherstcompanyusesGoogleAdWordstondnew
customersonlineandpaysanaverageof80centseachtimeanew
customerjoins.Thesecondcompanyselsheavygoodstolarge
companies.Eachsalerequiresasignicanttimeinvestmentfroma
salespersonandon-sitesalesengineeringtohelpinstaltheproduct;
thesehardcoststotalupto$80,000pernewcustomer.
Both
companieswilgrowattheexactsamerate.Eachhasthesame
proportionofrevenue(20percent)availabletoreinvestinnew
customeracquisition.Ifeithercompanywantstoincreaseitsrateof
growth,itcandosoinoneoftwoways:increasetherevenuefrom
eachcustomerordrivedownthecostofacquiringanewcustomer.
That’sthepaidengineofgrowthatwork.
InrelatingtheIMVUstoryinChapter3,Italkedabouthowwe
madeamajorearlymistakeinsetinguptheIMVUstrategy.We
ultimatelywounduphavingtomakeanengineofgrowthpivot.
WeoriginalythoughtthatourIMadd-onstrategywouldalowthe
producttogrowviraly.Unfortunately,customers
refusedtogo
alongwithourbriliantstrategy.
Ourbasicmisconceptionwasabeliefthatcustomerswouldbe
wilingtouseIMVUasanadd-ontoexistinginstantmessaging
networks.Webelievedthat
theproductwouldspreadviraly
throughthosenetworks,passedfromcustomertocustomer.The
problemwiththattheoryisthatsomekindsofproductsarenot
compatiblewithviralgrowth.
IMVU’scustomersdidn’t
wanttousetheproductwiththeir
existingfriends.Theywantedtouseittomakenewfriends.
Unfortunately,thatmeanttheydidnothaveastrongincentiveto
bringnewcustomerstotheproduct;theyviewedthatasourjob.
Fortunately,IMVUwasabletogrowbyusingpaidadvertising
becauseourcustomerswerewilingtopaymoreforourproduct
thanitcostustoreachthemviaadvertising.
Liketheotherengines,thepaidengineofgrowthispoweredby
afeedbackloop.Eachcustomerpaysacertainamountofmoneyfor
theproductoverhisorher“lifetime”asacustomer.Oncevariable
costsarededucted,thisusualyiscaledthecustomerlifetimevalue
(LTV).Thisrevenuecanbeinvestedingrowthbybuying
(LTV).Thisrevenuecanbeinvestedingrowthbybuying
advertising.
Supposeanadvertisementcosts$100andcausesftynew
customerstosignupfortheservice.Thisadhasacostper
acquisition(CPA)of$2.00.
Inthisexample,iftheproducthasan
LTVthatisgreaterthan$2,theproductwilgrow.Themargin
betweentheLTVandtheCPAdetermineshowfastthepaidengine
ofgrowthwilturn(thisiscaledthemarginalprot).Conversely,
iftheCPAremainsat$2.00buttheLTVfalsbelow$2.00,the
company’sgrowthwilslow.Itmaymakeupthedierencewith
one-timetacticssuchasusinginvestedcapitalorpublicitystunts,
butthosetacticsarenotsustainable.Thiswasthefate
ofmany
failedcompanies,includingnotabledot-comameoutsthat
erroneouslybelievedthattheycouldlosemoneyoneachcustomer
but,astheoldjokegoes,makeitupinvolume.
AlthoughIhaveexplainedthepaidengineofgrowthin
termsof
advertising,itisfarbroaderthanthat.Startupsthatemployan
outboundsalesforcearealsousingthisengine,asareretail
companiesthatrelyonfoottrac.Althesecostsshouldbe
factoredintothecostper
acquisition.
Forexample,onestartupIworkedwithbuiltcolaborationtools
forteamsandgroups.Itwentthrougharadicalpivot,switching
fromatoolthatwasusedprimarilybyhobbyistsandsmalclubsto
onethatwassoldprimarilytoenterprises,nongovernmental
organizations(NGOs),andotherextremelylargeorganizations.
However,theymadethatcustomersegmentpivotwithoutchanging
theirengineofgrowth.Previously,theyhaddone
customer
acquisitiononline,usingweb-baseddirectmarketingtechniques.I
rememberoneearlysituationinwhichthecompanyeldedacal
fromamajorNGOthatwantedtobuyitsproductandrolitout
acrossmanydivisions.Thestartuphadan“unlimited”pricingplan,
itsmostexpensive,thatcostonlyafewhundreddolarspermonth.
TheNGOliteralycouldnotmakethepurchasebecauseithadno
processinplaceforbuyingsomethingsoinexpensive.
Additionaly,
theNGOneededsubstantialhelpinmanagingtherolout,educating
itsstaonthenewtool,andtrackingtheimpactofthechange;
thosewerealservicesthecompanywasilequippedtooer.
thosewerealservicesthecompanywasilequippedtooer.
Changingcustomersegmentsrequiredthemtoswitchtohiringa
sizableoutboundsalesstathatspenttimeatendingconferences,
educatingexecutives,andauthoringwhitepapers.
Thosemuch
highercostscamewithacorrespondingreward:thecompany
switchedfrommakingonlyafewdolarspercustomertomaking
tensandthenhundredsofthousandsofdolarspermuchlarger
customer.Theirnewengineofgrowthledtosustainedsuccess.
Mostsourcesofcustomeracquisitionaresubjecttocompetition.
Forexample,primeretailstorefrontshavemorefoottracandare
thereforemorevaluable.Similarly,advertisingthatis
targetedto
moreauentcustomersgeneralycostsmorethanadvertisingthat
reachesthegeneralpublic.Whatdeterminesthesepricesisthe
averagevalueearnedinaggregatebythecompaniesthatarein
competitionforanygivencustomer’satention.Wealthyconsumers
costmoretoreachbecausetheytendtobecomemoreprotable
customers.
Overtime,anysourceofcustomeracquisitionwiltendtohave
itsCPAbidupbythiscompetition.Ifeveryoneinanindustry
makesthesameamountofmoneyoneachsale,theyalwilwind
uppayingmostoftheirmarginalprottothesourceofacquisition.
Thus,theabilitytogrowinthelongtermbyusingthe
paidengine
requiresadierentiatedabilitytomonetizeacertainsetof
customers.
IMVUisacaseinpoint.Ourcustomerswerenotconsideredvery
lucrativebyotheronlineservices:theyincludedalot
ofteenagers,
low-incomeadults,andinternationalcustomers.Otherservices
tendedtoassumethosepeoplewouldnotpayforanythingonline.
AtIMVU,wedevelopedtechniquesforcolectingonlinepayments
fromcustomerswhodidnothaveacreditcard,suchasalowing
themtobiltotheirmobilephonesorsenduscashinthemail.
Therefore,wecouldaordtopaymoretoacquirethosecustomers
thanourcompetitorscould.
ATechnicalCaveat
Technicaly,morethanoneengineofgrowthcanoperateina
businessatatime.Forexample,thereareproductsthathave
extremelyfastviralgrowthaswelasextremelylowcustomer
churnrates.Also,thereisnoreasonwhyaproductcannothave
bothhighmarginsandhighretention.However,inmyexperience,
successfulstartupsusualyfocusonjustoneengineofgrowth,
specializingineverythingthatisrequiredtomakeit
work.
Companiesthatatempttobuildadashboardthatincludesalthree
enginestendtocausealotofconfusionbecausetheoperations
expertiserequiredtomodelaltheseeectssimultaneouslyisquite
complicated.Therefore,Istronglyrecommendthatstartupsfocuson
oneengineatatime.Mostentrepreneursalreadyhaveastrong
leap-of-faithhypothesisaboutwhichengineismostlikelytowork.
Iftheydonot,timespentoutofthebuildingwith
customerswil
quicklysuggestonethatseemsprotable.Onlyafterpursuingone
enginethoroughlyshouldastartupconsiderapivottooneofthe
others.
ENGINESOFGROWTHDETERMINE
PRODUCT/MARKETFIT
MarcAndreessen,thelegendaryentrepreneurandinvestorandone
ofthefathersoftheWorldWideWeb,coinedtheterm
product/marketttodescribethemomentwhenastartupnaly
findsawidespreadsetof
customersthatresonatewithitsproduct:
Inagreatmarket—amarketwithlotsofrealpotential
customers—themarketpulsproductoutofthestartup.
Thisisthestoryofsearchkeywordadvertising,Internet
auctions,andTCP/IProuters.Conversely,inaterrible
market,youcanhavethebestproductintheworldandan
absolutelykilerteam,anditdoesn’tmater—you’regoing
tofail.3
Whenyouseeastartupthathasfoundatwithalargemarket,
it’sexhilarating.Itleavesnoroomfordoubt.ItisFord’s
ModelT
it’sexhilarating.Itleavesnoroomfordoubt.ItisFord’sModelT
yingoutofthefactoryasfastasitcouldbemade,Facebook
sweepingcolegecampusespracticalyovernight,orLotustaking
thebusinessworldbystorm,seling$54milionworthofLotus1-2-
3initsfirstyearofoperation.
Startupsoccasionalyaskmetohelpthemevaluatewhetherthey
haveachievedproduct/markett.It’seasytoanswer:ifyouare
asking,you’renotthereyet.Unfortunately,thisdoesn’thelp
companiesgureouthowtogetclosertoproduct/markett.How
canyoutelifyouareonthevergeofsuccessorhopelesslyfar
away?
AlthoughIdon’tthinkAndreessenintendedthisaspartofhis
denition,tomanyentrepreneursitimpliesthatapivotisafailure
event—“ourstartuphasfailedtoachieveproduct/markett.”It
alsoimpliestheinverse—thatonceourproducthas
achieved
product/markett,wewon’thavetopivotanymore.Both
assumptionsarewrong.
Ibelievetheconceptoftheengineofgrowthcanputtheideaof
product/markettonamorerigorousfooting.Sinceeachengineof
growthcanbedenedquantitatively,eachhasauniquesetof
metricsthatcanbeusedtoevaluatewhetherastartupisonthe
vergeofachievingproduct/markett.Astartupwithaviral
coecientof0.9ormoreisonthevergeofsuccess.Even
beter,the
metricsforeachengineofgrowthworkintandemwiththe
innovationaccountingmodeldiscussedinChapter7togive
directiontoastartup’sproductdevelopmenteorts.Forexample,if
astartupisatemptingtousetheviralengineofgrowth,itcan
focusitsdevelopmenteortsonthingsthatmightaectcustomer
behavior—ontheviralloop—andsafelyignorethosethatdonot.
Suchastartupdoesnotneedtospecializeinmarketing,
advertising,
orsalesfunctions.Conversely,acompanyusingthepaidengine
needstodevelopthosemarketingandsalesfunctionsurgently.
Astartupcanevaluatewhetheritisgetingcloserto
product/markettasittunes
itsenginebyevaluatingeachtrip
throughtheBuild-Measure-Learnfeedbackloopusinginnovation
accounting.Whatrealymatersisnottherawnumbersorvanity
metricsbutthedirectionanddegreeofprogress.
metricsbutthedirectionanddegreeofprogress.
Forexample,imaginetwostartupsthatareworkingdiligentlyto
tunethestickyengineofgrowth.Onehasacompoundingrateof
growthof5percent,andtheother10percent.Whichcompanyis
thebeterbet?Onthesurface,itmayseemthatthelargerrateof
growthisbeter,butwhatifeachcompany’sinnovationaccounting
dashboardlookslikethefolowingchart?
COMPOUNDINGGROWTHRATEASCOMPANYCOMPANY
OF
A
B
Sixmonthsago
0.1%
9.8%
Fivemonthsago
0.5%
9.6%
Fourmonthsago
2.0%
9.9%
Threemonthsago
3.2%
9.8%
Twomonthsago
4.5%
9.7%
Onemonthago
5.0%
10.0%
Evenwithnoinsightintothesetwocompanies’gross
numbers,
wecantelthatcompanyAismakingrealprogresswhereas
companyBisstuckinthemud.ThisistrueeventhoughcompanyB
isgrowingfasterthancompanyArightnow.
WHENENGINESRUNOUT
Getingastartup’sengineofgrowthupandrunningishardenough,
butthetruthisthateveryengineofgrowtheventualyrunsoutof
gas.Everyengineistiedtoagivensetofcustomersandtheir
relatedhabits,preferences,advertisingchannels,and
interconnections.Atsomepoint,thatsetofcustomerswilbe
exhausted.Thismaytakealongtimeorashorttime,dependingon
one’sindustryandtiming.
one’sindustryandtiming.
Chapter6emphasizedtheimportanceofbuildingthe
minimum
viableproductinsuchawaythatitcontainsnoadditionalfeatures
beyondwhatisrequiredbyearlyadopters.Folowingthatstrategy
successfulywilunlockanengineofgrowththatcanreachthat
targetaudience.However,makingthetransitiontomainstream
customerswilrequiretremendousadditionalwork.4Oncewehave
aproductthatisgrowingamongearlyadopters,wecouldintheory
stopworkinproductdevelopmententirely.The
productwould
continuetogrowuntilitreachedthelimitsofthatearlymarket.
Thengrowthwouldleveloorevenstopcompletely.The
chalengecomesfromthefactthatthisslowdownmighttake
monthsorevenyearstotake
place.RecalfromChapter8that
IMVUfailedthistest—atfirst—forpreciselythisreason.
Someunfortunatecompanieswindupfolowingthisstrategy
inadvertently.Becausetheyareusingvanitymetricsandtraditional
accounting,theythinktheyaremakingprogresswhentheyseetheir
numbersgrowing.Theyfalselybelievetheyaremakingtheir
productbeterwheninfacttheyarehavingnoimpactoncustomer
behavior.Thegrowthisalcomingfromanengineof
growththatis
working—runningecientlytobringinnewcustomers—notfrom
improvementsdrivenbyproductdevelopment.Thus,whenthe
growthsuddenlyslows,itprovokesacrisis.
Thisisthesameproblemthat
establishedcompaniesexperience.
Theirpastsuccesseswerebuiltonanelytunedengineofgrowth.
Ifthatenginerunsitscourseandgrowthslowsorstops,therecan
beacrisisifthecompanydoesnothavenewstartupsincubating
withinitsranksthatcanprovidenewsourcesofgrowth.
Companiesofanysizecansuerfromthisperpetualaiction.
Theyneedtomanageaportfolioofactivities,simultaneouslytuning
theirengineofgrowthanddevelopingnewsourcesof
growthfor
whenthatengineinevitablyrunsitscourse.Howtodothisisthe
subjectofChapter12.However,beforewecanmanagethat
portfolio,weneedanorganizationalstructure,culture,and
disciplinethatcanhandletheserapidandoftenunexpected
changes.Icalthisanadaptiveorganization,anditisthesubjectof
changes.Icalthisanadaptiveorganization,anditisthesubjectof
Chapter11.
11
ADAPT
WhenIwastheCTOofIMVU,IthoughtIwasdoingagoodjob
mostofthetime.Ihadbuiltanagileengineeringorganization,
andweweresuccessfulyexperimentingwiththe
techniques
thatwouldcometobeknownastheLeanStartup.However,ona
coupleofoccasionsIsuddenlyrealizedthatIwasfailingatmyjob.
Foranachievement-orientedperson,thatisincrediblydisarming.
Worstofal,youdon’tgetamemo.Ifyoudid,itwouldread
somethinglikethis:
DearEric,
Congratulations!Thejobyouusedtodoatthiscompany
isnolongeravailable.However,youhavebeentransferred
toanewjobinthecompany.Actualy,it’snotthesame
companyanymore,eventhoughithasthesamenameand
manyofthesamepeople.Andalthoughthejobhasthe
sametitle,too,andyouusedtobegoodatyouroldjob,
you’realreadyfailingatthe
newone.Thistransferis
eectiveasofsixmonthsago,sothisistoalertyouthat
you’vealreadybeenfailingatitforquitesometime.
Bestofluck!
Everytimethishappenedtome,Istruggledtogureoutwhatto
do.Iknewthatasthecompanygrew,wewouldneedadditional
processesandsystemsdesignedtocoordinatethecompany’s
operationsateachlargersize.AndyetIhadalsoseenmanystartups
operationsateachlargersize.AndyetIhadalsoseenmany
startups
becomeossiedandbureaucraticoutofamisplaceddesireto
become“professional.”
HavingnosystematalwasnotanoptionforIMVUandisnot
anoptionforyou.Therearesomanywaysforastartupto
fail.I’ve
livedthroughtheoverarchitecturefailure,inwhichatemptingto
preventalthevariouskindsofproblemsthatcouldoccurwound
updelayingthecompanyfromputingoutanyproduct.I’veseen
companiesfailtheotherwayfromtheso-caledFriendstereect,
sueringahigh-proletechnicalfailurejustwhencustomer
adoptionisgoingwild.Asadepartmentexecutive,thisoutcomeis
worstofal,becausethefailureisbothhigh-prole
andatributable
toasinglefunctionordepartment—yours.Notonlywilthe
companyfail,itwilbeyourfault.
MostoftheadviceI’veheardonthistopichassuggestedakindof
split-the-dierenceapproach
(asin,“engageinalitleplanning,but
nottoomuch”).Theproblemwiththiswily-nilyapproachisthat
it’shardtogiveanyrationaleforwhyweshouldanticipateone
particularproblembutignoreanother.Itcanfeellikethebossis
beingcapriciousorarbitrary,andthatfeedsthecommonfeeling
thatmanagement’sdecisionsconcealanulteriormotive.
Forthosebeingmanagedthisway,theirincentivesareclear.If
thebosstendstosplitthedierence,thebestwaytoinuencethe
bossandgetwhatyouwantistotakethemostextremeposition
possible.Forexample,ifonegroupisadvocatingforanextremely
lengthyreleasecycle,say,anannualnewproductintroduction,you
mightchoosetoargueforanequalyextremelyshort
releasecycle
(perhapsweeklyorevendaily),knowingthatthetwoopinionswil
beaveragedout.Then,whenthedierenceissplit,you’relikelyto
getanoutcomeclosertowhatyouactualywantedintherst
place.Unfortunately,thiskindofarmsraceescalates.Rivalsin
anothercamparelikelytodothesamething.Overtime,everyone
wiltakethemostpolarizedpositionspossible,whichmakes
splitingthedierenceevermoredicultandeverless
successful.
Managershavetotakeresponsibilityforknowinglyorinadvertently
creatingsuchincentives.Althoughitwasnottheirintentionto
rewardextremepolarization,that’sexactlywhattheyaredoing.
rewardextremepolarization,that’sexactlywhattheyaredoing.
Getingoutofthistraprequiresasignificantshiftinthinking.
BUILDINGANADAPTIVEORGANIZATION
Shouldastartupinvestinatrainingprogramfornewemployees?If
youhadaskedmeafewyearsago,Iwouldhavelaughedandsaid,
“Absolutelynot.Trainingprogramsareforbigcompaniesthatcan
aordthem.”YetatIMVUwewoundupbuildingatraining
programthatwassogood,newhireswereproductiveontheirrst
dayofemployment.Withinjustafewweeks,thoseemployees
werecontributingatahighlevel.Itrequiredahugeeortto
standardizeourworkprocessesandprepareacurriculumofthe
conceptsthatnewemployeesshouldlearn.Everynew
engineer
wouldbeassignedamentor,whowouldhelpthenewemployee
workthroughacurriculumofsystems,concepts,andtechniqueshe
orshewouldneedtobecomeproductiveatIMVU.The
performanceofthementor
andmenteewerelinked,sothementors
tookthiseducationseriously.
Whatisinteresting,lookingbackatthisexample,isthatwenever
stoppedworkanddecidedthatweneededtobuildagreattraining
program.Instead,thetraining
programevolvedorganicalyoutofa
methodicalapproachtoevolvingourownprocess.Thisprocessof
orientationwassubjecttoconstantexperimentationandrevisionso
thatitgrewmoreefective—andlessburdensome—overtime.
Icalthisbuildinganadaptiveorganization,onethat
automaticalyadjustsitsprocessandperformancetocurrent
conditions.
CanYouGoTooFast?
Sofarthisbookhasemphasizedtheimportanceofspeed.Startups
areinalife-or-deathstruggletolearnhowtobuildasustainable
businessbeforetheyrunoutofresourcesanddie.However,
focusingonspeedalonewouldbedestructive.Towork,startups
focusingonspeedalonewouldbedestructive.To
work,startups
requirebuilt-inspeedregulatorsthathelpteamsndtheiroptimal
paceofwork.
WesawanexampleofspeedregulationinChapter9withthe
useoftheandoncordinsystemssuchascontinuous
deployment.It
isepitomizedintheparadoxicalToyotaproverb,“Stopproduction
sothatproductionneverhastostop.”Thekeytotheandoncordis
thatitbringsworktoastopassoonasanuncorrectablequality
problemsurfaces—whichforcesittobeinvestigated.Thisisoneof
themostimportantdiscoveriesoftheleanmanufacturing
movement:youcannottradequalityfortime.Ifyouarecausing(or
missing)qualityproblemsnow,theresultingdefectswil
slowyou
downlater.Defectscausealotofrework,lowmorale,and
customercomplaints,alofwhichslowprogressandeatawayat
valuableresources.
SofarIhaveusedthelanguageofphysical
productstodescribe
theseproblems,butthatissimplyamaterofconvenience.Service
businesseshavethesamechalenges.Justaskanymanagerofa
training,stang,orhospitalityrmtoshowyoutheplaybookthat
specieshowemployeesaresupposedtodelivertheserviceunder
variousconditions.Whatmighthavestartedoutasasimpleguide
tendstogrowinexorablyovertime.Pretysoon,orientationis
incrediblycomplexandemployeeshaveinvesteda
lotoftimeand
energyinlearningtherules.Nowconsideranentrepreneurial
managerinthatkindofcompanytryingtoexperimentwithnew
rulesorprocedures.Thehigher-qualitytheexistingplaybookis,the
easieritwilbeforittoevolveovertime.Bycontrast,alow-quality
playbookwilbeledwithcontradictoryorambiguousrulesthat
causeconfusionwhenanythingischanged.
WhenIteachtheLeanStartupapproachtoentrepreneurswithan
engineeringbackground,thisisoneofthehardestconceptsto
grasp.Ontheonehand,thelogicofvalidatedlearningandthe
minimumviableproductsaysthatweshouldgetaproductinto
customers’handsassoonaspossibleandthatanyextra
workwedo
beyondwhatisrequiredtolearnfromcustomersiswaste.Onthe
otherhand,theBuild-Measure-Learnfeedbackloopisacontinuous
process.Wedon’tstopafteroneminimumviableproductbutuse
process.Wedon’tstopafteroneminimumviableproductbutuse
whatwehavelearnedtogettoworkimmediatelyonthenext
iteration.
Therefore,shortcutstakeninproductquality,design,or
infrastructuretodaymay
windupslowingacompanydown
tomorrow.YoucanseethisparadoxinactionatIMVU.Chapter3
recountedhowwewoundupshippingaproducttocustomersthat
wasfulofbugs,missingfeatures,andbaddesign.Thecustomers
wouldn’teventrythatproduct,andsomostofthatworkhadtobe
thrownaway.It’sagoodthingwedidn’twastealotoftimexing
thosebugsandcleaningupthatearlyversion.
However,asourlearningalowedustobuildproductsthat
customersdidwant,wefacedslowdowns.Havingalow-quality
productcaninhibitlearningwhenthedefectspreventcustomers
fromexperiencing(andgivingfeedbackon)theproduct’sbenets.
InIMVU’scase,asweoeredtheproducttomore
mainstream
customers,theyweremuchlessforgivingthanearlyadoptershad
been.Similarly,themorefeaturesweaddedtotheproduct,the
harderitbecametoaddevenmorebecauseoftheriskthatanew
featurewouldinterferewithanexistingfeature.Thesamedynamics
happeninaservicebusiness,sinceanynewrulesmayconictwith
existingrules,andthemorerules,themorepossibilitiesforconflict.
IMVUusedthetechniquesofthischaptertoachievescale
and
qualityinajust-in-timefashion.
THEWISDOMOFTHEFIVEWHYS
Toaccelerate,LeanStartupsneedaprocessthatprovidesanatural
feedbackloop.Whenyou’regoingtoofast,youcause
more
problems.Adaptiveprocessesforceyoutoslowdownandinvestin
preventingthekindsofproblemsthatarecurrentlywastingtime.
Asthosepreventiveefortspayof,younaturalyspeedupagain.
Let’sreturntothequestionofhavingatrainingprogramfornew
employees.Withoutaprogram,newemployeeswilmakemistakes
whileintheirlearningcurvethatwilrequireassistanceand
interventionfromotherteammembers,slowingeveryone
down.
interventionfromotherteammembers,slowingeveryonedown.
Howdoyoudecideiftheinvestmentintrainingisworththe
benetofspeedduetoreducedinterruptions?Figuringthisout
fromatop-downperspectiveischalenging,becauseitrequires
estimatingtwocompletelyunknownquantities:howmuchitwil
costtobuildanunknownprogramagainstanunknownbenetyou
mightreap.Evenworse,thetraditionalwaytomakethese
kindsof
decisionsisdecidedlylarge-batchthinking.Acompanyeitherhas
anelaboratetrainingprogramoritdoesnot.Untiltheycanjustify
thereturnoninvestmentfrombuildingafulprogram,most
companiesgeneralydonothing.
ThealternativeistouseasystemcaledtheFiveWhystomake
incrementalinvestmentsandevolveastartup’sprocessesgradualy.
ThecoreideaofFiveWhysistotieinvestmentsdirectlytothe
preventionofthemostproblematicsymptoms.Thesystemtakesits
namefromtheinvestigativemethodofaskingthequestion“Why?”
vetimestounderstandwhathashappened(therootcause).If
you’veeverhadtoansweraprecociouschildwhowants
toknow
“Whyistheskyblue?”andkeepsasking“Why?”aftereachanswer,
you’refamiliarwithit.Thistechniquewasdevelopedasa
systematicproblem-solvingtoolbyTaichiOhno,thefatherofthe
ToyotaProductionSystem.I
haveadapteditforuseintheLean
Startupmodelwithafewchangesdesignedspecificalyforstartups.
Attherootofeveryseeminglytechnicalproblemisahuman
problem.FiveWhysprovidesanopportunitytodiscoverwhatthat
humanproblemmightbe.TaichiOhnogivesthefolowing
example:
Whenconfrontedwithaproblem,haveyoueverstopped
andaskedwhyvetimes?Itisdiculttodoeventhoughit
soundseasy.Forexample,supposeamachinestopped
functioning:
1.Whydidthemachinestop?(Therewasanoverloadandthe
fuseblew.)
2.Whywasthereanoverload?(Thebearingwasnotsuciently
lubricated.)
lubricated.)
3.Whywasitnotlubricatedsuciently?(Thelubricationpump
wasnotpumpingsuficiently.)
4.Whywasitnotpumpingsuciently?(Theshaftofthepump
waswornandratling.)
5.Whywastheshaftwornout?(Therewasnostraineratached
andmetalscrapgotin.)
Repeating“why”vetimes,likethis,canhelpuncover
therootproblemandcorrectit.Ifthisprocedurewerenot
carriedthrough,onemightsimplyreplacethefuseorthe
pumpshaft.Inthatcase,theproblemwouldrecurwithina
fewmonths.TheToyotaproductionsystemhasbeenbuilt
onthepracticeandevolutionofthisscienticapproach.By
askingandanswering“why”vetimes,wecangettothe
realcauseoftheproblem,
whichisoftenhiddenbehind
moreobvioussymptoms.1
NotethateveninOhno’srelativelysimpleexampletheroot
causemovesawayfromatechnicalfault(ablownfuse)andtoward
ahumanerror(someoneforgottoatachastrainer).
Thisis
completelytypicalofmostproblemsthatstartupsfacenomater
whatindustrytheyarein.Goingbacktoourservicebusiness
example,mostproblemsthatatrstappeartobeindividual
mistakescanbetracedback
toproblemsintrainingortheoriginal
playbookforhowtheserviceistobedelivered.
LetmedemonstratehowusingtheFiveWhysalowedustobuild
theemployeetrainingsystemthatwasmentionedearlier.Imagine
thatatIMVUwesuddenlystartreceivingcomplaintsfrom
customersaboutanewversionoftheproductthatwehavejust
released.
1.Anewreleasedisabledafeatureforcustomers.Why?Because
aparticularserverfailed.
2.Whydidtheserverfail?Becauseanobscuresubsystemwas
usedinthewrongway.
usedinthewrongway.
3.Whywasitusedinthewrongway?Theengineerwhousedit
didn’tknowhowtouseitproperly.
4.Whydidn’theknow?Becausehewasnevertrained.
5.Whywasn’thetrained?Becausehismanagerdoesn’tbelievein
trainingnewengineersbecauseheandhisteamare“toobusy.”
Whatbeganasapurelytechnicalfaultisrevealedquicklytobea
veryhumanmanagerialissue.
MakeaProportionalInvestment
Here’showtouseFiveWhysanalysistobuildanadaptive
organization:consistently
makeaproportionalinvestmentateach
ofthevelevelsofthehierarchy.Inotherwords,theinvestment
shouldbesmalerwhenthesymptomisminorandlargerwhenthe
symptomismorepainful.Wedon’tmakelargeinvestmentsin
preventionunlesswe’recopingwithlargeproblems.
Intheexampleabove,theansweristoxtheserver,changethe
subsystemtomakeitlesserror-prone,educatetheengineer,and,
yes,haveaconversationwiththeengineer’smanager.
Thislaterpiece,theconversationwiththemanager,isalways
hard,especialyinastartup.WhenIwasastartupmanager,ifyou
toldmeIneededtoinvestintrainingmypeople,Iwouldhavetold
youitwasawasteoftime.Therewerealwaystoomany
other
thingstodo.I’dprobablyhavesaidsomethingsarcasticlike“Sure,
I’dbehappytodothat—ifyoucansparemytimefortheeight
weeksit’ltaketosetup.”That’smanager-speakfor“Nowayin
hel.”
That’swhytheproportionalinvestmentapproachisso
important.Iftheoutageisaminorglitch,it’sessentialthatwe
makeonlyaminorinvestmentinxingit.Let’sdothersthourof
theeight-weekplan.That
maynotsoundlikemuch,butit’sastart.
Iftheproblemrecurs,askingtheFiveWhyswilrequirethatwe
continuetomakeprogressonit.Iftheproblemdoesnotoccur
again,anhourisn’tabigloss.
again,anhourisn’tabigloss.
Iusedtheexampleofengineeringtrainingbecausethatwas
somethingIwasreluctanttoinvestinatIMVU.Attheoutsetofour
venture,Ithoughtweneededtofocusalofourenergieson
buildingandmarketingourproduct.Yetonceweenteredaperiod
ofrapidhiring,repeatedFiveWhyssessionsrevealedthatproblems
causedbylackoftrainingwereslowingdownproduct
development.Atnopointdidwedropeverythingtofocussolelyon
training.Instead,wemadeincrementalimprovementstothe
processconstantly,eachtimereapingincrementalbenets.Over
time,thosechangescompounded,freeinguptimeandenergythat
previouslyhadbeenlosttofirefightingandcrisis
management.
AutomaticSpeedRegulator
TheFiveWhysapproachactsasanaturalspeedregulator.The
moreproblemsyouhave,themoreyouinvestinsolutionstothose
problems.Astheinvestmentsininfrastructureorprocess
payo,
theseverityandnumberofcrisesarereducedandtheteamspeeds
upagain.Withstartupsinparticular,thereisadangerthatteams
wilworktoofast,tradingqualityfortimeinawaythatcauses
sloppymistakes.FiveWhyspreventsthat,alowingteamstond
theiroptimalpace.
TheFiveWhystiestherateofprogresstolearning,notjust
execution.StartupteamsshouldgothroughtheFiveWhys
whenevertheyencounteranykindoffailure,includingtechnical
faults,failurestoachievebusinessresults,orunexpectedchangesin
customerbehavior.
FiveWhysisapowerfulorganizationaltechnique.Someofthe
engineersIhavetrainedtouseitbelievethatyoucanderivealthe
otherLeanStartuptechniquesfromtheFiveWhys.Coupledwith
workinginsmalbatches,itprovidesthefoundationacompany
needstorespondquicklytoproblemsastheyappear,
without
overinvestingoroverengineering.
THECURSEOFTHEFIVEBLAMES
WhenteamsrstadoptFiveWhysasaproblem-solvingtool,they
encountersomecommonpitfals.Weneedsystems
likeFiveWhys
toovercomeourpsychologicallimitationsbecausewetendto
overreacttowhat’shappeninginthemoment.Wealsotendtoget
frustratedifthingshappenthatwedidnotanticipate.
WhentheFiveWhys
approachgoesawry,IcalittheFive
Blames.Insteadofaskingwhyrepeatedlyinanatemptto
understandwhatwentwrong,frustratedteammatesstartpointing
ngersateachother,tryingtodecidewhoisatfault.Insteadof
usingtheFiveWhystondandxproblems,managersand
employeescanfalintothetrapofusingtheFiveBlamesasa
meansforventingtheirfrustrationsandcalingoutcoleaguesfor
systemicfailures.Althoughit’shumannaturetoassume
thatwhen
weseeamistake,it’sduetodefectsinsomeoneelse’sdepartment,
knowledge,orcharacter,thegoaloftheFiveWhysistohelpussee
theobjectivetruththatchronicproblemsarecausedbybadprocess,
notbadpeople,andremedythemaccordingly.
IrecommendseveraltacticsforescapingtheFiveBlames.The
rstistomakesurethateveryoneaectedbytheproblemisinthe
roomduringtheanalysisoftherootcause.Themeetingshould
includeanyonewhodiscoveredordiagnosedtheproblem,
includingcustomerservicerepresentativeswhoeldedthecals,if
possible.Itshouldincludeanyonewhotriedtoxthesymptomas
welasanyonewhoworkedonthesubsystemsorfeatures
involved.
Iftheproblemwasescalatedtoseniormanagement,thedecision
makerswhowereinvolvedintheescalationshouldbepresentas
wel.
Thismaymakeforacrowdedroom,butit’sessential.In
my
experience,whoeverisleftoutofthediscussionendsupbeingthe
targetforblame.Thisisjustasdamagingwhetherthescapegoatisa
junioremployeeortheCEO.Whenit’sajunioremployee,it’sal
tooeasytobelievethatthatpersonisreplaceable.IftheCEOisnot
present,it’saltooeasytoassumethathisorherbehavioris
unchangeable.Neitherpresumptionisusualycorrect.
unchangeable.Neitherpresumptionisusualy
correct.
Whenblameinevitablyarises,themostseniorpeopleinthe
roomshouldrepeatthismantra:ifamistakehappens,shameonus
formakingitsoeasytomakethatmistake.InaFiveWhysanalysis,
wewanttohaveasystems-levelviewasmuchaspossible.
Here’sasituationinwhichthismantracameinhandy.Becauseof
thetrainingprocesswehaddevelopedatIMVUthroughtheFive
Whys,weroutinelyaskednewengineerstomakea
changetothe
productionenvironmentontheirrstday.Forengineerstrainedin
traditionaldevelopmentmethods,thiswasoftenfrightening.They
wouldask,“WhatwilhappentomeifIaccidentalydisruptor
stoptheproductionprocess?”Intheirpreviousjobs,thatwasa
mistakethatcouldgetthemred.AtIMVUwetoldnewhires,“If
ourproductionprocessissofragilethatyoucanbreakitonyour
veryrstdayofwork,shameonusformakingitsoeasyto
doso.”
Iftheydidmanagetobreakit,weimmediatelywouldhavethem
leadtheeorttoxtheproblemaswelastheeorttopreventthe
nextpersonfromrepeatingtheirmistake.
Fornewhireswhocame
fromcompanieswithaverydierent
culture,thiswasoftenastressfulinitiation,buteveryonecame
throughitwithavisceralunderstandingofourvalues.Bitbybit,
systembysystem,thosesmalinvestmentsaddeduptoarobust
productdevelopmentprocessthatalowedalouremployeesto
workmorecreatively,withgreatlyreducedfear.
GetingStarted
HereareafewtipsonhowtogetstartedwiththeFiveWhysthat
arebasedonmyexperience
introducingthistechniqueatmany
othercompanies.
FortheFiveWhystoworkproperly,therearerulesthatmustbe
folowed.Forexample,theFiveWhysrequiresanenvironmentof
mutualtrustandempowerment.Insituationsinwhichthisis
lacking,thecomplexityofFiveWhyscanbeoverwhelming.Insuch
situations,I’veoftenusedasimplifiedversionthatstilalowsteams
tofocusonanalyzingrootcauseswhiledevelopingthe
muscles
tofocusonanalyzingrootcauseswhiledevelopingthemuscles
they’lneedlatertotackletheful-blownmethod.
Iaskteamstoadoptthesesimplerules:
1.Betolerantofalmistakesthefirsttime.
2.Neveralowthesamemistaketobemadetwice.
Therstruleencouragespeopletogetusedtobeing
compassionateaboutmistakes,especialythemistakesofothers.
Remember,mostmistakesarecausedbyawedsystems,notbad
people.Thesecondrulegetstheteamstartedmakingproportional
investmentsinprevention.
Thissimpliedsystemworkswel.Infact,weuseditatIMVUin
thedaysbeforeIdiscoveredtheFiveWhysandtheToyota
ProductionSystem.However,suchasimpliedsystemdoesnot
workeectivelyoverthelongterm,asIfoundoutrsthand.Infact,
thatwasoneofthethingsthatdrovemetorstlearnaboutlean
production.
Thestrengthandweaknessofthesimpliedsystemisthatit
invitesquestionssuchasWhatcountsasthesameproblem?What
kindsofmistakesshouldwefocuson?andShouldwexthis
individualproblemortrytopreventawholecategoryofrelated
problems?Forateamthatisjustgetingstarted,thesequestionsare
thought-provokingandcanlaythegroundworkformoreelaborate
methodstocome.Ultimately,though,theydoneedanswering.
Theyneedacompleteadaptiveprocesssuchasthe
FiveWhys.
FacingUnpleasantTruths
YouwilneedtobepreparedforthefactthatFiveWhysisgoingto
turnupunpleasantfactsaboutyourorganization,especialyatthe
beginning.Itisgoingtocalforinvestmentsinprevention
that
comeattheexpenseoftimeandmoneythatcouldbeinvestedin
newproductsorfeatures.Underpressure,teamsmayfeelthatthey
don’thavetimetowasteonanalyzingrootcauseseventhoughit
wouldgivethemmoretimeinthelongterm.Theprocess
wouldgivethemmoretimeinthelongterm.Theprocess
sometimeswildevolveintotheFiveBlames.Atalthesejunctures,
itisessentialthatsomeonewithsucientauthoritybepresentto
insistthattheprocessbefolowed,thatitsrecommendationsbe
implemented,andtoactasarefereeifdisagreementsareup.
Buildinganadaptiveorganization,inotherwords,requires
executiveleadershiptosponsorandsupportthe
process.
Often,individualcontributorsatstartupscometomyworkshops,
eagertogetstartedwiththeFiveWhys.Icautionagainstatempting
todothatiftheydonothavethebuy-inofthemanagerorteam
leader.Proceedcautiouslyifyoundyourselfinthissituation.It
maynotbepossibletogettheentireteamtogetherforatrueFive
Whysinquiry,butyoucanalwaysfolowthesimpletwo-rule
versioninyourownwork.Wheneversomethinggoes
wrong,ask
yourself:HowcouldIpreventmyselffrombeinginthissituation
everagain?
StartSmal,BeSpecific
Onceyouarereadytobegin,Irecommendstartingwithanarrowly
targetedclassofsymptoms.Forexample,thersttimeIusedthe
FiveWhyssuccessfuly,Iusedittodiagnoseproblemswithoneof
ourinternaltestingtoolsthatdidnotaectcustomersdirectly.It
maybetemptingtostartwithsomethinglargeand
important
becausethatiswheremostofthetimeisbeingwastedasaresultof
aawedprocess,butitisalsowherethepressurewilbegreatest.
Whenthestakesarehigh,theFiveWhyscandevolveintotheFive
Blamesquickly.It’sbetertogivetheteamachancetolearnhowto
dotheprocessfirstandthenexpandintohigher-stakesareaslater.
Themorespecicthesymptomsare,theeasieritwilbefor
everyonetorecognizewhenit’stimetoscheduleaFive
Whys
meeting.SayyouwanttousetheFiveWhystoaddressbiling
complaintsfromcustomers.Inthatcase,pickadateafterwhichal
bilingcomplaintswiltriggeraFiveWhysmeetingautomaticaly.
Notethatthisrequiresthattherebeasmalenoughvolumeof
complaintsthathavingthismeetingeverytimeonecomesinis
complaintsthathavingthismeetingeverytimeonecomesinis
practical.Iftherearealreadytoomanycomplaints,picka
subseton
whichyouwanttofocus.Makesurethattherulethatdetermines
whichkindsofcomplaintstriggeraFiveWhysmeetingissimple
andironclad.Forexample,youmightdecidethateverycomplaint
involvingacreditcardtransactionwilbeinvestigated.That’san
easyruletofolow.Don’tpickarulethatisambiguous.
Atrst,thetemptationmaybetomakeradicalanddeepchanges
toeverybilingsystemandprocess.Don’t.Instead,keepthe
meetingsshortandpickrelativelysimplechangesateachofthe
velevelsoftheinquiry.Overtime,astheteamgetsmore
comfortablewiththeprocess,youcanexpandittoincludemore
andmoretypesofbilingcomplaintsandthentootherkindsof
problems.
AppointaFiveWhysMaster
Tofacilitatelearning,IhavefoundithelpfultoappointaFive
Whysmasterforeachareainwhichthemethodisbeingused.This
individualistaskedwithbeingthemoderatorforeach
FiveWhys
meeting,makingdecisionsaboutwhichpreventionstepstotake,
andassigningthefolow-upworkfromthatmeeting.Themaster
mustbeseniorenoughtohavetheauthoritytoensurethatthose
assignmentsgetdonebutshouldnotbesoseniorthatheorshewil
notbeabletobepresentatthemeetingsbecauseofconicting
responsibilities.TheFiveWhysmasteristhepointpersoninterms
ofaccountability;heorsheistheprimarychangeagent.
Peoplein
thispositioncanassesshowwelthemeetingsaregoingand
whetherthepreventioninvestmentsthatarebeingmadearepaying
of.
THEFIVEWHYSINACTION
IGNEntertainment,adivisionofNewsCorporation,isanonline
videogamesmediacompanywiththebiggestaudienceofvideo
videogamesmediacompanywiththebiggestaudienceofvideo
gameplayersintheworld.Morethan45miliongamers
frequent
itsportfolioofmediaproperties.IGNwasfoundedinthelate
1990s,andNewsCorporationacquireditin2005.IGNhasgrown
toemployseveralhundredpeople,includingalmostahundred
engineers.
Recently,Ihadtheopportunitytospeaktotheproduct
developmentteamatIGN.Theyhadbeensuccessfulinrecentyears,
butlikealtheestablishedcompanieswe’veseenthroughoutthis
book,theywerelookingtoacceleratenewproductdevelopment
andndwaystobemoreinnovative.Theybroughttogethertheir
engineering,product,anddesignteamstotalkthroughwaysthey
couldapplytheLeanStartupmodel.
ThischangeinitiativehadthesupportofIGN’ssenior
management,includingtheCEO,theheadofproductdevelopment,
thevicepresidentofengineering,thepublisher,andtheheadof
product.TheirpreviousefortsatFiveWhyshadnotgonesmoothly.
Theyhadatemptedtotacklealaundrylistofproblemareas
nominatedbytheproductteam.Theissuesvariedfrom
discrepanciesinwebanalyticstopartnerdatafeedsthatwerenot
working.TheirrstFiveWhysmeetingtookanhour,andalthough
theycameupwithsomeinterestingtakeaways,asfarastheFive
Whysgoes,itwasadisaster.Noneofthepeoplewhowere
connectedtoandknewthemostabouttheissueswereatthe
meeting,andbecausethiswasthersttimetheyweredoingthe
FiveWhystogether,theydidn’tsticktotheformatandwentoon
manytangents.Itwasn’tacompletewasteoftime,butitdidn’t
haveanyofthebenetsoftheadaptivestyleofmanagement
discussedinthischapter.
Don’tSendYourBaggagethroughtheFiveWhysProcess
IGNhadtheexperienceoftryingtosolvealofits“baggage”issues
thathadbeencausingwastedtimeformanyyears.Becausethisis
anoverwhelmingsetofproblems,ndingxesquickly
proves
overwhelming.
overwhelming.
IntheirzealtogetstartedwiththeFiveWhys,IGNneglected
threeimportantthings:
1.TointroduceFiveWhystoanorganization,itis
necessaryto
holdFiveWhyssessionsasnewproblemscomeup.Since
baggageissuesareendemic,theynaturalycomeupaspartof
theFiveWhysanalysisandyoucantakethatopportunitytox
themincrementaly.Iftheydon’tcomeuporganicaly,maybe
they’renotasbigastheyseem.
2.Everyonewhoisconnectedtoaproblemneedstobeatthe
FiveWhyssession.Manyorganizationsfacethetemptationto
savetimebysparingbusypeoplefromtherootcauseanalysis.
Thisisafalseeconomy,asIGNdiscoveredthehardway.
3.AtthebeginningofeachFiveWhyssession,takeafewminutes
toexplainwhattheprocessisforandhowitworksforthe
benetofthosewhoarenewtoit.Ifpossible,useanexample
ofasuccessfulFiveWhyssessionfromthepast.Ifyou’rebrand
new,youcanusemyearlierexampleaboutthemanagerwho
doesn’tbelieveintraining.IGNlearnedthat,whenever
possible,ithelpstousesomethingthathaspersonalmeaning
fortheteam.
Afterourmeeting,theIGNleadershipdecidedtogiveFiveWhys
anothertry.Folowingtheadvicelaidoutinthischapter,they
appointedaFiveWhysmasternamedTonyFord,adirectorof
engineering.TonywasanentrepreneurwhohadcometoIGN
throughanacquisition.HegothisstartwithInternettechnology,
buildingwebsitesaboutvideogamesinthelate
1990s.Eventualy
thatledtoanopportunityatastartup,TeamXbox,whereheserved
astheleadsoftwaredeveloper.TeamXboxwasacquiredbyIGN
Entertainmentin2003,andsincethattimeTonyhasbeena
technologist,leaderofinnovation,andproponentofagileandlean
practicesthere.
Unfortunately,Tonystartedwithoutpickinganarrowproblem
areaonwhichtofocus.Thisledtoearlysetbacksandfrustration.
areaonwhichtofocus.Thisledtoearlysetbacksandfrustration.
Tonyrelates,“AsthenewmasterIwasn’tverygoodattraversing
throughtheFiveWhyseectively,andtheproblemsweweretrying
tosolvewerenotgreatcandidatesintherstplace.
Asyoucan
imagine,theseearlysessionswereawkwardandintheendnotvery
useful.Iwasgetingquitediscouragedandfrustrated.”Thisisa
commonproblemwhenonetriestotackletoomuchatonce,butit
isalsoaconsequenceofthefactthattheseskilstaketimeto
master.Luckily,Tonypersevered:“HavingaFiveWhysmasteris
criticalinmyopinion.FiveWhysiseasyintheorybutdicultin
practice,soyouneedsomeonewhoknowsitwelto
shapethe
sessionsforthosewhodon’t.”
TheturnaroundcamewhenTonyledaFiveWhyssession
involvingaprojectthathadbeenmissingitsdeadlines.Thesession
wasfascinatingand
insightfulandproducedmeaningful
proportionalinvestments.Tonyexplains:“Thesuccesshadtodo
withamoreexperiencedmasterandmoreexperiencedatendees.
WealknewwhattheFiveWhyswas,andIdidarealygoodjob
keepingusontrackandawayfromtangents.Thiswasapivotal
moment.RightthenIknewtheFiveWhyswasanewtoolthatwas
goingtohavearealimpactonouroveralsuccessasateamandas
abusiness.”
Onthesurface,FiveWhysseemstobeabouttechnicalproblems
andpreventingmistakes,butasteamsdriveoutthesesupercial
wastes,theydevelopanewunderstandingofhowtoworktogether.
Tonyputitthisway:“IdaresaythatIdiscoveredthat
theFiveWhys
transcendsrootcauseanalysisbyrevealinginformationthatbrings
yourteamcloserthroughacommonunderstandingandperspective.
Alotoftimesaproblemcanpulpeopleapart;FiveWhysdoesthe
opposite.”
IaskedTonytoprovideanexampleofarecentsuccessfulFive
WhysanalysisfromIGN.Hisaccountofitislistedinthesidebar.
Whycouldn’tyouaddoreditpostsontheblogs?
Answer:Anypostrequest
(write)tothearticlecontentapiwas
Answer:Anypostrequest(write)tothearticlecontentapiwas
returninga500error.
Proportionalinvestment:Jim—We’lworkontheAPI,butlet’s
makeourCMSmore
forgivingfortheuser.Alowuserstoadd
andeditdraftswithouterrorsforabeteruserexperience.
WhywasthecontentAPIreturning500errors?
Answer:Thebson_extgemwasincompatiblewithothergems
itdependsupon.
Proportionalinvestment:King—Removethegem(alreadydone
toresolvetheoutage).
Whywasthegemincompatible?
Answer:Weaddedanewversionofthegeminadditiontothe
existingversionandtheapp
startedusingitunexpectedly.
Proportionalinvestment:Bennet—Convertourrailsapptouse
bundlerforgemmanagement.
Whydidweaddanewversionofageminproductionwithout
testing?
Answer:Wedidn’tthinkweneededatestinthesecases.
Proportionalinvestment:BennetandJim—Writeaunitor
functionaltestintheAPIandCMSthatwilcatchthisinthe
future.
Whydoweaddadditional
gemsthatwedon’tintendtouse
rightaway?
Answer:Inpreparationforacodepushwewantedtogetal
newgemsreadyintheproductionenvironment.Eventhough
ourcodedeploymentsareful
yautomated,gemsarenot.
Proportionalinvestment:Bennet—Automategemmanagement
Proportionalinvestment:Bennet—Automategemmanagement
andinstalationintoContinuousIntegrationandContinuous
Deploymentprocess.
Bonus—WhyarewedoingthingsinproductiononFriday
nights?
Answer:Becausenoonesayswecan’tanditwasaconvenient
timeforthedevelopertoprepareforadeployment
we’dbe
doingonMonday.
Proportionalinvestment:Tony—Makeanannouncementtothe
team.TherewilbenoproductionchangesonFriday,Saturday,
orSundayunlessanexceptionhasbeenmadeand
approvedby
David(VPEngineering).Wewilreevaluatethispolicywhen
wehaveafulyautomatedcontinuousdeploymentprocessin
place.
AsaresultofthisFiveWhyssessionandtheproportional
investmentswemade,ourdeploymentsareeasier,quicker,and
neveragainwilourprocessalowadevelopertoplacegems
intoproductionsystemswithunintendedconsequences.Indeed,
wehavenothadanotherissuelikethis.We
strengthenedour
“clusterimmunesystem”asyouwouldsay.
WithouttheFiveWhys,wewouldhaveneverdiscoveredal
oftheinformationwedidhere.Myguessisthatwewould
havetoldthatonedeveloper
tonotdostupidthingsonFriday
nightsandmovedon.ThisiswhatIemphasizedearlier,where
agoodFiveWhyssessionhastwooutputs,learninganddoing.
Theproportionalinvestmentsthatcameoutofthissessionare
obviouslyvaluable,butthelearningsaremuchmoresubtle,but
amazingforgrowingasdevelopersandasateam.
ADAPTINGTOSMALLERBATCHES
Beforeleavingthetopicofbuildinganadaptiveorganization,I
wanttointroduceonemorestory.Thisoneconcernsaproductthat
you’veprobablyusedifyou’veeverrunyourownbusiness.It’s
caledQuickBooks,anditisoneofIntuit’sflagshipproducts.
QuickBookshasbeentheleadingproductinits
categoryformany
years.Asaresult,ithasalargeanddedicatedcustomerbase,and
Intuitexpectsittocontributesignicantlytoitsbotomline.Like
mostpersonalcomputer(PC)softwareofthelasttwodecades,
QuickBookshasbeenlaunchedonanannualcycle,inonegiant
batch.Thiswashowitworkedthreeyearsago,whenGregWright,
thedirectorofproductmarketingforQuickBooks,joinedtheteam.
Asyoucanimagine,therewerelotsofexisting
processesinplaceto
ensureaconsistentproductandanon-timerelease.Thetypical
releaseapproachwastospendsignicantup-fronttimetoidentify
thecustomers’need:
Typicalytherstthreetofourmonthsofeachannualcycle
wasspentstrategizingandplanning,withoutbuildingnew
features.Onceaplanandmilestoneswereestablished,the
teamwouldspendthenextsixtoninemonthsbuilding.
Thiswouldculminateinabiglaunch,andthentheteam
wouldgetitsrstfeedbackonwhetherithadsuccessfuly
deliveredoncustomers’needsattheendoftheprocess.
Soherewasthetimeline:startprocessinSeptember,first
betareleaseisinJune,secondbetaisinJuly.Thebetais
essentialytestingtomakesureitdoesn’tcrashpeople’s
computersorcausethemtolosetheirdata—bythattimein
theprocess,onlymajorbugscanbexed.Thedesignofthe
productitselfislocked.
Thisisthestandard“waterfal”developmentmethodology
that
productdevelopmentteamshaveusedforyears.Itisalinear,large-
batchsystemthatreliesforsuccessonproperforecastingand
planning.Inotherwords,itiscompletelymaladaptedfortoday’s
planning.Inotherwords,itiscompletelymaladaptedfortoday’s
rapidlychangingbusinessenvironment.
YearOne:AchievingFailure
Gregwitnessedthisbreakdownin2009,hisrstyearonthe
QuickBooksteam.Thatyear,
thecompanyshippedanentirelynew
systeminQuickBooksforonlinebanking,oneofitsmostimportant
features.Theteamwentthroughroundsofusabilitytestingusing
mock-upsandnonfunctionalprototypes,folowedbysignicant
betatestingusingsamplecustomerdata.Atthemomentofthe
launch,everythinglookedgood.
TherstbetareleasewasinJune,andcustomerfeedbackstarted
cominginnegative.Althoughcustomerswerecomplaining,there
wasn’tsucientcausetostopthereleasebecauseitwastechnicaly
awless—itdidn’tcrashcomputers.Atthatpoint,Gregwasina
bind.Hehadnowayofknowinghowthefeedbackwouldtranslate
torealcustomerbehaviorinthemarket.Werethesejust
isolated
complaints,orpartofawidespreadproblem?Hedidknowone
thingforsure,though:thathisteamcouldnotaordtomissthe
deadline.
Whentheproductnalyshipped,theresultswere
terrible.It
tookcustomersfourtovetimeslongertoreconciletheirbanking
transactionsthanithadwiththeolderversion.Intheend,Greg’s
teamhadfailedtodeliveronthecustomerneedtheyweretryingto
address(despitebuildingtheproducttospecication),andbecause
thenextreleasehadtogothroughthesamewaterfalprocess,it
tooktheteamninemonthstox.Thisisaclassiccaseof“achieving
failure”—successfulyexecutingaflawedplan.
IntuitusesatrackingsurveycaledtheNetPromoterScore2to
evaluatecustomersatisfactionwithitsmanyproducts.Thisisa
greatsourceofactionablemetricsaboutwhatcustomersrealythink
aboutaproduct.Infact,IuseditatIMVU,too.One
thingthatis
niceaboutNPSisthatitisverystableovertime.Sinceitis
measuringcorecustomersatisfaction,itisnotsubjecttominor
uctuations;itregistersonlymajorchangesincustomersentiment.
uctuations;itregistersonlymajorchangesincustomersentiment.
Thatyear,theQuickBooksscoredropped20points,thersttime
thecompanyhadevermovedtheneedlewiththeNetPromoter
Score.That20-pointdropresultedinsignificantlosses
forIntuitand
wasembarrassingforthecompany—albecausecustomerfeedback
cametoolateintheprocess,alowingnotimetoiterate.
Intuit’sseniormanagement,includingthegeneralmanagerofthe
smalbusinessdivisionand
theheadofsmalbusinessaccounting,
recognizedtheneedforchange.Totheircredit,theytaskedGreg
withdrivingthatchange.Hismission:toachievestartupspeedfor
thedevelopmentanddeploymentofQuickBooks.
YearTwo:MuscleMemory
Thenextchapterofthisstoryilustrateshowharditistobuildan
adaptiveorganization.GregsetouttochangetheQuickBooks
developmentprocessbyusingfourprinciples:
1.Smalerteams.Shiftfrom
largeteamswithuniformfunctional
rolestosmaler,fulyengagedteamswhosememberstakeon
diferentroles.
2.Achieveshortercycletimes.
3.Fastercustomerfeedback,testingbothwhetherit
crashes
customers’computersandtheperformanceofnew
features/customerexperience.
4.Enableandempowerteamstomakefastandcourageous
decisions.
Onthesurface,thesegoalsseemtobealignedwiththemethods
andprinciplesdescribedinpreviouschapters,butGreg’ssecond
yearwithQuickBookswasnotamarkedsuccess.Forexample,he
decreedthattheteamwouldmovetoamidyearrelease
milestone,
eectivelycutingthecycletimeandbatchsizeinhalf.However,
thiswasnotsuccessful.Throughsheerdetermination,theteamtried
valiantlytogetanalphareleaseoutinJanuary.However,the
problemsthataictlarge-batchdevelopmentwerestilpresent,
problemsthataictlarge-batchdevelopmentwerestilpresent,
andtheteamstruggledtocompletethealphabyApril.That
representedanimprovementoverthepastsystembecause
issues
couldbebroughttothesurfacetwomonthsearlierthanunderthe
oldway,butitdidnotproducethedramaticalybeterresultsGreg
waslookingfor.
Infact,overthecourseoftheyear,theteam’sprocesskept
lookingmoreandmorelikeithadinprioryears.AsGregputit,
“Organizationshavemusclememory,”anditishardforpeopleto
unlearnoldhabits.Gregwasrunningupagainstasystem,and
makingindividualchangessuchasarbitrarilychanging
therelease
datewerenomatchforit.
YearThree:Explosion
Frustratedbythelimitedprogressinthepreviousyear,Greg
teamedupwiththeproductdevelopmentleaderHimanshuBaxi.
Togethertheytossedoutaltheoldprocesses.Theymadeapublic
declarationthattheircombinedteamswouldbecreatingnew
processesandthattheywerenotgoingtogobacktotheoldway.
Insteadoffocusingonnewdeadlines,Gregand
Himanshu
investedinprocess,product,andtechnologychangesthatenabled
workinginsmalerbatches.Thosetechnicalinnovationshelped
themgetthedesktopproducttocustomersfasterforfeedback.
Insteadofbuildingacomprehensiveroadmapatthebeginningof
theyear,Gregkickedotheyearwithwhattheycaled
idea/code/solutionjamsthatbroughtengineers,productmanagers,
andcustomerstogethertocreateapipelineofideas.Itwasscaryfor
Gregasaproductmanagertostarttheyearwithoutadenedlistof
whatwouldbeintheproductrelease,buthehadcondenceinhis
teamandthenewprocess.
Therewerethreediferencesinyearthree:
•Teamswereinvolvedin
creatingnewtechnologies,processes,
andsystems.
•Cross-functionalteamswereformedaroundnewgreatideas.
•Cross-functionalteamswereformedaroundnewgreatideas.
•Customerswereinvolvedfromtheinceptionofeachfeature
concept.
It’simportanttounderstandthattheoldapproachdidnotlack
customerfeedbackorcustomerinvolvementintheplanning
process.Inthetruespiritofgenchigembutsu,Intuitproduct
managers(PMs)woulddo“folow-me-homes”withcustomersto
identifyproblemstosolveinthenextrelease.However,thePMs
wereresponsibleforalthecustomerresearch.They
wouldbringit
backtotheteamandsay,“Thisistheproblemwewanttosolve,
andhereareideasforhowwecouldsolveit.”
Changingtoacross-functionalwayofworkingwasnotsmooth
sailing.Someteammembers
wereskeptical.Forexample,some
productmanagersfeltthatitwasawasteoftimeforengineersto
spendtimeinfrontofcustomers.ThePMsthoughtthattheirjob
wastogureoutthecustomerissueanddenewhatneededtobe
built.Thus,thereactionofsomePMstothechangewas:“What’s
myjob?WhatamIsupposedtobedoing?”Similarly,someonthe
engineeringsidejustwantedtobetoldwhattodo;theydidn’twant
totalktocustomers.Asistypicalythecaseinlarge-
batch
development,bothgroupshadbeenwilingtosacricetheteam’s
abilitytolearninordertoworkmore“eficiently.”
Communicationwascriticalforthischangeprocesstosucceed.
Altheteamleaderswere
openaboutthechangetheyweredriving
andwhytheyweredrivingit.Muchoftheskepticismtheyfaced
wasbasedonthefactthattheydidnothaveconcreteexamplesof
wherethishadworkedinthepast;itwasanentirelynewprocess
forIntuit.Theyhadtoexplainclearlywhytheoldprocessdidn’t
workandwhytheannualrelease“train”wasnotsetingthemup
forsuccess.Throughoutthechangetheycommunicatedtheprocess
outcomestheywereshootingfor:earliercustomer
feedbackanda
fasterdevelopmentcyclethatwasdecoupledfromtheannual
releasetimeline.Theyrepeatedlyemphasizedthatthenew
approachwashowstartupcompetitorswereworkinganditerating.
Theyhadtofolowsuitorriskbecomingirrelevant.
Historicaly,QuickBookshadbeenbuiltwithlargeteamsandlong
cycletimes.Forexample,inearlieryearstheil-fatedonline
bankingteamhadbeencomposedoffteenengineers,seven
qualityassurancespecialists,aproductmanager,andattimesmore
thanonedesigner.Nownoteamisbiggerthanvepeople.The
focusofeachteamisiteratingwithcustomersasrapidlyaspossible,
runningexperiments,andthenusingvalidatedlearning
tomake
real-timeinvestmentdecisionsaboutwhattoworkon.Asaresult,
whereastheyusedtohavevemajor“branches”ofQuickBooks
thatmergedfeaturesatthetimeofthelaunch,nowthereare
twentytotwenty-vebranches.Thisalowsforamuchlargersetof
experiments.Eachteamworksonanewfeatureforapproximately
sixweeksendtoend,testingitwithrealcustomersthroughoutthe
process.
Althoughtheprimarychangesthatarerequiredinanadaptive
organizationareinthemind-setofitsemployees,changingthe
cultureisnotsucient.AswesawinChapter9,leanmanagement
requirestreatingworkasasystemandthendealingwith
thebatch
sizeandcycletimeofthewholeprocess.Thus,toachievelasting
change,theQuickBooksteamhadtoinvestintoolsandplatform
changesthatwouldenablethenew,fasterwayofworking.
Forexample,oneofthemajorstresspointsintheatemptto
releaseanearlyalphaversionthepreviousyearwasthat
QuickBooksisamission-criticalproduct.Manysmalbusinessesuse
itastheirprimaryrepositoryforcriticalnancialdata.Theteam
wasextremelywaryofreleasingaminimumviableproductthat
hadanyriskofcorruptingcustomerdata.Therefore,evenifthey
workedinsmalerteamswithasmalerscope,theburdenofal
thatriskwouldhavemadeithardtoworkinsmaler
batches.
Togetthebatchsizedown,theQuickBooksteamhadtoinvestin
newtechnology.Theybuiltavirtualizationsystemthatalowed
themtorunmultipleversionsofQuickBooksonacustomer’s
computer.Thesecondversioncouldaccessalthecustomer’sdata
butcouldnotmakepermanentchangestoit.Thus,therewasno
riskofthenewversioncorruptingthecustomer’sdatabyaccident.
riskofthenewversion
corruptingthecustomer’sdatabyaccident.
Thisalowedthemtoisolatenewreleasestoalowselectedreal
customerstotestthemandprovidefeedback.
Theresultsinyearthreewerepromising.Theversionof
QuickBooksthatshippedthat
yearhadsignicantlyhigher
customersatisfactionratingsandsoldmoreunits.Ifyou’reusing
QuickBooksrightnow,oddsareyouareusingaversionthatwas
builtinsmalbatches.AsGregheadsintohisfourthyearwiththe
QuickBooksteam,theyareexploringevenmorewaystodrive
downbatchsizeandcycletime.Asusual,therearepossibilitiesthat
gobeyondtechnicalsolutions.Forexample,theannualsalescycle
ofboxeddesktopsoftwareisasignicantbarriertotruly
rapid
learning,andsotheteamhasbegunexperimentingwith
subscription-basedproductsforthemostactivecustomers.With
customersdownloadingupdatesonline,Intuitcanreleasesoftware
onamorefrequentbasis.
Soonthisprogramwilseethe
QuickBooksteamreleasingtocustomersquarterly.3
AsLeanStartupsgrow,theycanuseadaptivetechniquestodevelop
morecomplexprocesseswithoutgivinguptheircoreadvantage:
speedthroughtheBuild-
Measure-Learnfeedbackloop.Infact,one
oftheprimarybenetsofusingtechniquesthatarederivedfrom
leanmanufacturingisthatLeanStartups,whentheygrowup,are
welpositionedtodevelopoperationalexcelencebasedonlean
principles.Theyalreadyknowhowtooperatewithdiscipline,
developprocessesthataretailor-madetotheirsituation,anduse
leantechniquessuchastheFiveWhysandsmalbatches.Asa
successfulstartupmakesthetransitiontoanestablished
company,it
wilbewelpoisedtodevelopthekindofcultureofdisciplined
executionthatcharacterizestheworld’sbestfirms,suchasToyota.
However,successfulygrowingintoanestablishedcompanyisnot
theendofthestory.Astartup’sworkisneverdone,becauseaswas
discussedinChapter2,evenestablishedcompaniesmuststruggleto
ndnewsourcesofgrowththroughdisruptiveinnovation.This
imperativeiscomingearlierincompanies’lives.No
longercana
imperativeiscomingearlierincompanies’lives.Nolongercana
successfulstartupexpecttohaveyearsafteritsinitialpublic
oeringtobaskinmarket-leadingsuccess.Todaysuccessful
companiesfaceimmediatepressurefromnewcompetitors,fast
folowers,andscrappystartups.Asaresult,itnolongermakes
sensetothinkofstartupsasgoingthroughdiscretephaseslikethe
proverbialmetamorphosisofacaterpilartoabutery.
Both
successfulstartupsandestablishedcompaniesalikemustlearnto
jugglemultiplekindsofworkatthesametime,pursuing
operationalexcelenceanddisruptiveinnovation.Thisrequiresa
newkindofportfolio
thinking,whichisthesubjectofChapter12.
12
INNOVATE
Conventionalwisdomholdsthatwhencompaniesbecomelarger,
theyinevitablylosethecapacityforinnovation,creativity,and
growth.Ibelievethisiswrong.Asstartupsgrow,entrepreneurs
canbuildorganizationsthatlearnhowtobalancetheneedsof
existingcustomerswiththechalengesofndingnewcustomersto
serve,managingexistinglinesofbusiness,and
exploringnew
businessmodels—alatthesametime.And,iftheyarewilingto
changetheirmanagementphilosophy,Ibelieveevenlarge,
establishedcompaniescanmakethisshifttowhatIcalportfolio
thinking.
HOWTONURTUREDISRUPTIVEINNOVATION
Successfulinnovationteamsmustbestructuredcorrectlyinorderto
succeed.Venture-backedandbootstrappedstartupsnaturalyhave
someofthesestructuralatributesasaconsequenceofbeingsmal,
independentcompanies.Internalstartupteamsrequiresupport
fromseniormanagementtocreatethesestructures.Internalor
external,inmyexperiencestartupteamsrequirethree
structural
atributes:scarcebutsecureresources,independentauthorityto
developtheirbusiness,andapersonalstakeintheoutcome.Each
oftheserequirementsisdierentfromthoseofestablished
companydivisions.Keepinmindthatstructureismerelya
prerequisite—itdoesnotguaranteesuccess.Butgetingthestructure
wrongcanleadtoalmostcertainfailure.
ScarcebutSecureResources
Divisionleadersinlarge,
establishedorganizationsareadeptat
usingpoliticstoenlargetheirbudgetsbutknowthatthosebudgets
aresomewhatloose.Theyoftenacquireaslargeabudgetas
possibleandpreparetodefenditagainstincursionsfromother
departments.Politicsmeansthattheysometimeswinand
sometimeslose:ifacrisisemergeselsewhereintheorganization,
theirbudgetmightsuddenlybereducedby10percent.Thisisnota
catastrophe;teamswilhavetoworkharderanddomorewithless.
Mostlikely,thebudgethassomepaddinginanticipationofthis
kindofeventuality.
Startupsaredierent:toomuchbudgetisasharmfulastoolitle
—ascountlessdot-comfailurescanatest—andstartupsare
extremelysensitivetomidcoursebudgetarychanges.Itisextremely
rareforastand-alonestartupcompanytolose10percentofitscash
onhandsuddenly.Inalargenumberofcases,thiswouldbeafatal
blow,asindependentstartupsarerunwithlitlemarginfor
error.
Thus,startupsarebotheasierandmoredemandingtorunthan
traditionaldivisions:theyrequiremuchlesscapitaloveral,butthat
capitalmustbeabsolutelysecurefromtampering.
IndependentDevelopment
Authority
Startupteamsneedcompleteautonomytodevelopandmarketnew
productswithintheirlimitedmandate.Theyhavetobeableto
conceiveandexecuteexperimentswithouthavingtogainan
excessivenumberofapprovals.
Istronglyrecommendthatstartupteamsbecompletelycross-
functional,thatis,haveful-timerepresentationfromevery
functionaldepartmentinthecompanythatwilbeinvolvedinthe
creationorlaunchoftheirearlyproducts.Theyhavetobeableto
buildandshipactualfunctioningproductsandservices,notjust
prototypes.HandosandapprovalsslowdowntheBuild-Measure-
prototypes.Handosandapprovalsslowdownthe
Build-Measure-
Learnfeedbackloopandinhibitbothlearningandaccountability.
Startupsrequirethattheybekepttoanabsoluteminimum.
Ofcourse,thislevelofdevelopmentautonomyisliabletoraise
fearsinaparentorganization.Aleviatingthosefearsisamajor
goalofthemethodrecommendedbelow.
APersonalStakeintheOutcome
Third,entrepreneursneedapersonalstakeintheoutcomeoftheir
creations.Instand-alonenewventures,thisusualyisachieved
throughstockoptionsorotherformsofequityownership.Wherea
bonussystemmustbeusedinstead,thebestincentivesaretiedto
thelong-termperformanceofthenewinnovation.
However,Idonotbelievethatapersonalstakehastobe
nancial.Thisisespecialyimportantinorganizations,suchas
nonprotsandgovernment,inwhichtheinnovationisnottiedto
nancialobjectives.Inthesecases,itisstilpossiblefor
teamsto
haveapersonalstake.Theparentorganizationhastomakeitclear
whotheinnovatorisandmakesuretheinnovatorreceivescredit
forhavingbroughtthenewproducttolife—ifitissuccessful.As
oneentrepreneurwhoranherowndivisionatamajormedia
companytoldme,“Financialincentivesaside,Ialwaysfeltthat
becausemynamewasonthedoor,Ihadmoretoloseandmoreto
provethansomeoneelse.Thatsenseofownershipis
not
insignificant.”
Thisformulaiseectiveinfor-protcompaniesaswel.At
Toyota,themanagerinchargeofdevelopinganewvehiclefrom
starttofinishiscaledtheshusa,orchiefengineer:
Shusaareoftencaledheavy-weightprojectmanagersinthe
U.S.literature,butthisnameunderstatestheirrealrolesas
designleaders.Toyotaemployeestranslatethetermaschief
engineer,andtheyrefertothevehicleunderdevelopment
astheshusa’scar.Theyassuredusthattheshusahasnal,
absoluteauthorityovereveryaspectofvehicle
absoluteauthorityovereveryaspectofvehicle
development.1
Ontheipside,Iknowanextremelyhigh-prole
technology
companythathasareputationforhavinganinnovativeculture,yet
itstrackrecordofproducingnewproductsisdisappointing.The
companyboastsaninternalrewardsystemthatisbasedonlarge
nancialandstatusawardstoteamsthatdosomething
extraordinary,butthoseawardsarehandedoutbysenior
managementonthebasisof—nooneknowswhat.Thereareno
objectivecriteriabywhichateamcangaugewhetheritwilwin
thiscovetedlotery.Teamshavelitlecondencethattheywil
receiveanylong-termownershipoftheirinnovations.Thus,teams
rarelyaremotivatedtotakerealrisks,insteadfocusingtheir
energiesonprojectsthatareexpectedtowintheapproval
ofsenior
management.
CREATINGAPLATFORMFOREXPERIMENTATION
Next,itisimportanttofocusonestablishingthegroundrulesunder
whichautonomousstartupteamsoperate:howtoprotectthe
parentorganization,howtoholdentrepreneurialmanagers
accountable,andhowtoreintegrateaninnovationbackintothe
parentorganizationifitissuccessful.Recalthe“islandoffreedom”
thatenabledtheSnapTaxteam—inChapter2—to
successfuly
createastartupwithinIntuit.That’swhataplatformfor
experimentationcando.
ProtectingtheParentOrganization
Conventionaly,adviceaboutinternalinnovatorsfocuseson
protectingthestartupfromtheparentorganization.Ibelieveitis
necessarytoturnthismodelonitshead.
Letmebeginbydescribingafairlytypicalmeetingfromoneof
myconsultingclients,alargecompany.Seniormanagementhad
gatheredtomakedecisionsaboutwhattoincludeinthenext
gatheredtomakedecisionsaboutwhattoincludeinthenext
versionofitsproduct.Aspartofthecompany’scommitmentto
beingdata-driven,ithadtriedtoconductanexperimenton
pricing.
Therstpartofthemeetingwastakenupwithinterpretingthe
datafromtheexperiment.
Oneproblemwasthatnobodycouldagreeonwhatthedata
meant.Manycustomreportshadbeencreatedforthe
meeting;the
datawarehouseteamwasatthemeetingtoo.Themoretheywere
askedtoexplainthedetailsofeachrowonthespreadsheet,the
moreevidentitbecamethatnobodyunderstoodhowthose
numbershadbeenderived.
Whatwewereleftlookingatwasthe
numberofgrosssalesoftheproductatavarietyofdierentprice
points,brokendownbyquarterandbycustomersegment.Itwasa
lotofdatatotrytocomprehend.
Worse,nobodywassurewhichcustomershadbeenexposedto
theexperiment.Dierentteamshadbeenresponsiblefor
implementingit,andsodierentpartsoftheproducthadbeen
updatedatdierenttimes.Thewholeprocesshadtaken
many
months,andbythispoint,thepeoplewhohadconceivedthe
experimenthadbeenmovedtoadivisionseparatefromthatofthe
peoplewhohadexecutedit.
Youshouldbeabletospotthemanyproblemswiththis
situation:theuseofvanitymetricsinsteadofactionablemetrics,an
overlylongcycletime,theuseoflargebatchsizes,anunclear
growthhypothesis,aweakexperimentaldesign,alackofteam
ownership,andthereforeverylitlelearning.
Listeningin,Iassumedthiswouldbetheendofthemeeting.
Withnoagreed-onfactstohelpmakethedecision,Ithought
nobodywouldhaveanybasisformakingthecaseforaparticular
action.Iwaswrong.Eachdepartmentsimplytook
whatever
interpretationofthedatasupporteditspositionbestandstarted
advocatingonitsownbehalf.Otherdepartmentswouldchimein
withalternativeinterpretationsthatsupportedtheirpositions,and
soon.Intheend,decisionswerenotmadebasedondata.Instead,
theexecutiverunningthemeetingwasforcedtobasedecisionson
themostplausible-soundingarguments.
Itseemedwastefultomehowmuchofthemeetinghadbeen
Itseemedwastefultomehowmuchofthemeetinghadbeen
spentdebatingthedatabecause,intheend,theargumentsthat
carriedthedaycouldhavebeenmaderightatthestart.Itwasasif
eachadvocatesensedthatheorshewasabouttobe
ambushed;if
anotherteammanagedtobringclaritytothesituation,itmight
underminethatperson,andsotherationalresponsewasto
obfuscateasmuchaspossible.Whatawaste.
Ironicaly,meetingslikethis
hadgivendata-drivendecision
makingandexperimentationabadnameinsidethecompany,and
forgoodreason.Thedatawarehousingteamwasproducingreports
thatnobodyreadorunderstood.Theprojectteamsfeltthe
experimentswereawasteoftime,sincetheyinvolvedbuilding
featureshalfway,whichmeanttheywereneveranygood.“Running
anexperiment”seemedtothemtobecodeforpostponingahard
decision.Worstofal,theexecutiveteamexperienced
themeetings
aschronicheadaches.Theiroldproductprioritizationmeetings
mighthavebeenlitlemorethanabatleofopinions,butatleast
theexecutivesunderstoodwhatwasgoingon.Nowtheyhadtogo
througharitualthatinvolvedcomplexmathandreachedno
deniteoutcome,andthentheyendeduphavingabatleof
opinionsanyway.
RationalFears
However,attheheartofthisdepartmentalfeudwasavery
rational
fear.Thiscompanyservedtwocustomersegments:abusiness-to-
businessenterprisesegmentandaconsumersegment.IntheB2B
segment,thecompanyemployedsalesstatosellargevolumesof
theproducttoothercompanies,whereastheconsumersegment
wasdrivenmostlybyone-opurchasesmadebyindividuals.The
bulkofthecompany’scurrentrevenuecamefromB2Bsales,but
growthinthatsegmenthadbeenslowing.Everyone
agreedthere
wastremendouspotentialforgrowthintheconsumersegment,but
sofarlitlehadmaterialized.
Partofthecauseofthislackofgrowthwasthecurrentpricing
structure.Likemanycompaniesthatseltolarge
enterprises,this
structure.Likemanycompaniesthatseltolargeenterprises,this
onepublishedahighlistpriceandthenprovidedheavydiscounts
to“favored”corporateclientswhoboughtinbulk.Naturaly,every
salespersonwasencouragedtomakealofhisorherclientsfeel
favored.Unfortunately,thepublishedlistpricewasmuchtoohigh
fortheconsumersegment.
Theteaminchargeofgrowingtheconsumersegmentwantedto
runexperimentswithalowerpricestructure.Theteaminchargeof
theenterprisesegmentwasnervousthatthiswouldcannibalizeor
otherwisediminishitsexistingrelationshipswithitscustomers.
Whatifthosecustomersdiscoveredthatindividuals
weregetinga
lowerpricethantheywere?
Anyonewhohasbeeninamultisegmentbusinesswilrecognize
thattherearemanypossiblesolutionstothisproblem,suchas
creatingtieredfeaturesetssothatdierentcustomersare
ableto
purchasedierent“levels”oftheproduct(asinairlineseating)or
evensupportingdiferentproductsunderseparatebrandnames.Yet
thecompanywasstrugglingtoimplementanyofthosesolutions.
Why?Outoffearofendangeringthecurrentbusiness,each
proposedexperimentwouldbedelayed,sabotaged,andobfuscated.
It’simportanttoemphasizethatthisfeariswelfounded.
Sabotageisarationalresponsefrommanagerswhoseterritoryis
threatened.Thiscompanyisnotarandom,tinystartupwith
nothingtolose.Anestablishedcompanyhasalottolose.Ifthe
revenuefromthecorebusinessgoesdown,headswilrol.Thisis
notsomethingtobetakenlightly.
TheDangersofHidingInnovationinsidetheBlackBox
Theimperativetoinnovateisunrelenting.Withouttheabilityto
experimentinamoreagilemanner,thiscompanyeventualywould
suerthefatedescribedinTheInnovator’sDilemma:
ever-higher
protsandmarginsyearafteryearuntilthebusinesssuddenly
colapsed.
Weoftenframeinternalinnovationchalengesbyasking,How
canweprotecttheinternalstartupfromtheparent
organization?I
canweprotecttheinternalstartupfromtheparentorganization?I
wouldliketoreframeandreversethequestion:Howcanwe
protecttheparentorganizationfromthestartup?Inmyexperience,
peopledefendthemselveswhentheyfeelthreatened,andno
innovationcanourishifdefensivenessisgivenfreerein.Infact,
thisiswhythecommonsuggestiontohidetheinnovationteamis
misguided.Thereareexamplesofone-time
successesusingasecret
skunkworksoro-siteinnovationteam,suchasthebuildingofthe
originalIBMPCinBocaRaton,Florida,completelyseparatefrom
mainlineIBM.Buttheseexamplesshouldservemostlyas
cautionarytales,becausetheyhaverarelyledtosustainable
innovation.2Hidingfromtheparentorganizationcanhavelong-
termnegativeconsequences.
Consideritfromthepointofviewofthemanagerswhohavethe
innovationsprungonthem.Theyarelikelytofeelbetrayedand
morethanalitleparanoid.Afteral,ifsomethingofthis
magnitudecouldbehidden,whatelseiswaitingintheshadows?
Overtime,thisleadstomorepoliticsasmanagersareincentivized
toferretoutthreatstotheirpower,inuence,andcareers.Thefact
thattheinnovationwasasuccessisnojusticationforthis
dishonestbehavior.Fromthepointofviewofestablished
managers,themessageisclear:ifyouarenotontheinside,youare
liabletobeblindsidedbythistypeofsecret.
Itisunfairtocriticizethesemanagersfortheirresponse;the
criticismshouldbeaimedatseniorexecutiveswhofailedtodesign
asupportivesysteminwhichtooperateandinnovate.Ibelievethis
isonereasonwhycompaniessuchasIBMlosttheirleadership
positioninthenewmarketsthattheydevelopedusingablackbox
suchasthePCbusiness;theyareunabletore-createandsustainthe
culturethatledtotheinnovationinthefirstplace.
CreatinganInnovationSandbox
Thechalengehereistocreateamechanismforempowering
innovationteamsoutintheopen.Thisisthepathtowarda
sustainablecultureofinnovationovertimeascompaniesface
sustainablecultureofinnovationovertimeascompaniesface
repeatedexistentialthreats.Mysuggestedsolutionistocreatea
sandboxforinnovationthatwilcontaintheimpactofthenew
innovationbutnotconstrainthemethodsofthestartup
team.It
worksasfolows:
1.Anyteamcancreateatruesplit-testexperimentthataects
onlythesandboxedpartsoftheproductorservice(fora
multipartproduct)oronlycertaincustomersegmentsor
territories(foranewproduct).However:
2.Oneteammustseethewholeexperimentthroughfromendto
end.
3.Noexperimentcanrunlongerthanaspeciedamountoftime
(usualyafewweeksfor
simplefeatureexperiments,longerfor
moredisruptiveinnovations).
4.Noexperimentcanaectmorethanaspeciednumberof
customers(usualyexpressedasapercentageofthecompany’s
totalmainstreamcustomer
base).
5.Everyexperimenthastobeevaluatedonthebasisofasingle
standardreportoffivetoten(nomore)actionablemetrics.
6.Everyteamthatworksinsidethesandboxandeveryproduct
thatisbuiltmustusethe
samemetricstoevaluatesuccess.
7.Anyteamthatcreatesanexperimentmustmonitorthemetrics
andcustomerreactions(supportcals,socialmediareaction,
forumthreads,etc.)whiletheexperimentisinprogressand
abortitifsomethingcatastrophichappens.
Atthebeginning,thesandboxhastobequitesmal.Inthe
companyabove,thesandboxinitialycontainedonlythepricing
page.Dependingonthetypesofproductsthecompanymakes,the
sizeofthesandboxcanbedenedindierentways.Forexample,
anonlineservicemightrestrictittocertainpagesoruserows.A
retailoperationmightrestrictittocertainstoresorgeographic
areas.Companiestryingtobringanentirelynewproduct
tomarket
mightbuildtherestrictionaroundcustomersincertainsegments.
Unlikeinaconcepttestormarkettest,customersinthesandbox
Unlikeinaconcepttestormarkettest,customersinthesandbox
areconsideredrealandtheinnovationteamisalowedtoatempt
toestablishalong-termrelationshipwiththem.Afteral,theymay
beexperimentingwiththoseearlyadoptersforalongtimebefore
theirlearningmilestonesareaccomplished.
Wheneverpossible,theinnovationteamshouldbecross-
functionalandhaveaclearteamleader,liketheToyotashusa.It
shouldbeempoweredtobuild,market,anddeployproductsor
featuresinthesandboxwithoutpriorapproval.It
shouldbe
requiredtoreportonthesuccessorfailureofthoseeortsbyusing
standardactionablemetricsandinnovationaccounting.
Thisapproachcanworkevenforteamsthathaveneverbefore
workedcross-functionaly.
Therstfewchanges,suchasaprice
change,maynotrequiregreatengineeringeort,buttheyrequire
coordinationacrossdepartments:engineering,marketing,customer
service.Teamsthatworkthiswayaremoreproductiveaslongas
productivityismeasuredbytheirabilitytocreatecustomervalue
andnotjuststaybusy.
Trueexperimentsareeasytoclassifyassuccessesorfailures
becausetop-levelmetricseithermoveortheydon’t.Eitherway,the
teamlearnsimmediatelywhetheritsassumptionsabouthow
customerswilbehavearecorrect.Byusingthesamemetricseach
time,theteambuildsliteracyaboutthosemetricsacrossthe
company.Becausetheinnovationteamisreporting
onitsprogress
byusingthesystemofinnovationaccountingdescribedinPartTwo,
anyonewhoreadsthosereportsisgetinganimplicitlessoninthe
powerofactionablemetrics.Thisefectisextremelypowerful.Even
ifsomeonewantstosabotagetheinnovationteam,heorshewil
havetolearnalaboutactionablemetricsandlearningmilestones
todoit.
Thesandboxalsopromotesrapiditeration.Whenpeoplehavea
chancetoseeaprojectthroughfromendtoendandtheworkis
doneinsmalbatchesanddeliversaclearverdictquickly,they
benefitfromthepoweroffeedback.Eachtimetheyfailtomovethe
numbers,theyhavearealopportunitytoactontheir
ndings
immediately.Thus,theseteamstendtoconvergeonoptimal
solutionsrapidlyeveniftheystartoutwithrealybadideas.
solutionsrapidlyeveniftheystartoutwithrealybadideas.
Aswesawearlier,thisisamanifestationoftheprincipleofsmal
batches.Functionalspecialists,especialythosesteepedinwaterfal
orstage-gatedevelopment,havebeentrainedtoworkinextremely
largebatches.Thiscausesevengoodideastoget
boggeddownby
waste.Bymakingthebatchsizesmal,thesandboxmethodalows
teamstomakecheapmistakesquicklyandstartlearning.Aswe’l
seebelow,thesesmalinitialexperimentscandemonstratethata
teamhasaviablenewbusinessthatcanbeintegratedbackintothe
parentcompany.
HoldingInternalTeamsAccountable
WealreadydiscussedlearningmilestonesindetailinChapter7.
Withaninternalstartup
team,thesequenceofaccountabilityisthe
same:buildanidealmodelofthedesireddisruptionthatisbased
oncustomerarchetypes,launchaminimumviableproductto
establishabaseline,andthenatempttotunetheenginetogetit
closertotheideal.
Operatinginthisframework,internalteamsessentialyactas
startups.Astheydemonstratesuccess,theyneedtobecome
integratedintothecompany’soveralportfolioofproductsand
services.
CULTIVATINGTHEMANAGEMENTPORTFOLIO
Therearefourmajorkindsofworkthatcompaniesmustmanage.3
Asaninternalstartupgrows,theentrepreneurswhocreatedthe
originalconceptmusttacklethechalengeofscale.As
new
mainstreamcustomersareacquiredandnewmarketsare
conquered,theproductbecomespartofthepublicfaceofthe
company,withimportantimplicationsforPR,marketing,sales,and
businessdevelopment.In
mostcases,theproductwilatract
competitors:copycats,fastfolowers,andimitatorsofalstripes.
Oncethemarketforthenewproductiswelestablished,
Oncethemarketforthenewproductiswelestablished,
proceduresbecomemore
routine.Tocombattheinevitable
commoditizationoftheproductinitsmarket,lineextensions,
incrementalupgrades,andnewformsofmarketingareessential.In
thisphase,operationalexcelencetakesonagreaterrole,asan
importantwaytoincreasemarginsistolowercosts.Thismay
requireadierenttypeofmanager:onewhoexcelsin
optimization,delegation,control,andexecution.Companystock
pricesdependonthiskindofpredictablegrowth.
Thereisafourthphaseaswel,onedominatedbyoperatingcosts
andlegacyproducts.Thisisthedomainofoutsourcing,automation,
andcostreduction.Nonetheless,infrastructureisstilmission-
critical.Failureoffacilitiesorimportantinfrastructureor
the
abandonmentofloyalcustomerscouldderailthewholecompany.
However,unlikethegrowthandoptimizationphase,investmentsin
thisareawilnothelpthecompanyachievetop-linegrowth.
Managersofthiskindoforganizationsuerthefateofbasebal
umpires:criticizedwhensomethinggoeswrong,unappreciated
whenthingsaregoingwel.
Wetendtospeakofthesefourphasesofbusinessesfromthe
perspectiveoflargecompanies,inwhichtheymayrepresententire
divisionsandhundredsoreventhousandsofpeople.That’slogical,
astheevolutionofthebusinessinthesekindsofextremecasesis
theeasiesttoobserve.However,alcompanies
engageinalfour
phasesofworkalthetime.Assoonasaproducthitsthe
marketplace,teamsofpeopleworkhardtoadvanceittothenext
phase.Everysuccessfulproductorfeaturebeganlifeinresearchand
development(R&D),
eventualybecameapartofthecompany’s
strategy,wassubjecttooptimization,andintimebecameoldnews.
Theproblemforstartupsandlargecompaniesalikeisthat
employeesoftenfolowtheproductstheydevelopastheymove
fromphasetophase.Acommonpracticeisfortheinventorofa
newproductorfeaturetomanagethesubsequentresources,team,
ordivisionthatultimatelycommercializesit.Asaresult,strong
creativemanagerswindupgetingstuckworkingonthe
growth
andoptimizationofproductsratherthancreatingnewones.
Thistendencyisoneofthereasonsestablishedcompanies
Thistendencyisoneofthereasonsestablishedcompanies
struggletondcreativemanagerstofosterinnovationintherst
place.Everynewinnovationcompetesforresourceswith
establishedprojects,andoneofthescarcestresourcesistalent.
EntrepreneurIsaJobTitle
Thewayoutofthisdilemma
istomanagethefourkindsofwork
dierently,alowingstrongcross-functionalteamstodevelop
aroundeacharea.Whenproductsmovefromphasetophase,they
arehandedobetweenteams.Employeescanchoosetomovewith
theproductaspartofthehandoorstaybehindandbeginwork
onsomethingnew.Neitherchoiceisnecessarilyrightorwrong;it
dependsonthetemperamentandskilsofthepersoninquestion.
Somepeoplearenaturalinventorswhoprefertowork
without
thepressureandexpectationsofthelaterbusinessphases.Others
areambitiousandseeinnovationasapathtowardsenior
management.Stilothersareparticularlyskiledatthemanagement
ofrunninganestablishedbusiness,outsourcing,andbolstering
ecienciesandwringingoutcostreductions.Peopleshouldbe
alowedtofindthekindsofjobsthatsuitthembest.
Infact,entrepreneurshipshouldbeconsideredaviablecareer
pathforinnovatorsinsidelargeorganizations.Managerswhocan
leadteamsbyusingtheLeanStartupmethodologyshouldnothave
toleavethecompanytoreaptherewardsoftheirskilsorhaveto
pretendtotintotherigidhierarchiesofestablished
functional
departments.Instead,theyshouldhaveabusinesscardthatsays
simply“Entrepreneur”underthename.Theyshouldbeheld
accountableviathesystemofinnovationaccountingandpromoted
andrewardedaccordingly.
Afteranentrepreneurhasincubatedaproductintheinnovation
sandbox,ithastobereintegratedintotheparentorganization.A
largerteameventualywilbeneededtogrowit,commercializeit,
andscaleit.Atrst,thisteamwilrequirethecontinuedleadership
oftheinnovatorswhoworkedinthesandbox.Infact,thisisa
positivepartoftheprocessinthatitgivestheinnovatorsachance
positivepartoftheprocessinthatitgivestheinnovatorsa
chance
totrainnewteammembersinthenewstyleofworkingthatthey
masteredintheoriginalsandbox.
Idealy,thesandboxwilgrowovertime;thatis,ratherthan
movetheteamoutofthesandboxandintothe
company’sstandard
routines,theremaybeopportunitiestoenlargethescopeofthe
sandbox.Forexample,ifonlycertainaspectsoftheproductwere
subjecttoexperimentationinthesandbox,newfeaturescanbe
added.Intheonlineservicedescribedearlier,thiscouldbe
accomplishedbystartingwithasandboxthatencompassedthe
productpricingpage.Whenthoseexperimentssucceeded,the
companycouldaddthewebsite’shomepagetothe
sandbox.It
subsequentlymightaddthesearchfunctionalityortheoveralweb
design.Ifonlycertaincustomersorcertainnumbersofcustomers
weretargetedinitialy,theproduct’sreachcouldbeincreased.
Whensuchchangesarecontemplated,it’simportantthatsenior
managementconsiderwhethertheteamsworkinginthesandbox
canfendforthemselvespoliticalyintheparentorganization.The
sandboxwasdesignedtoprotectthemandtheparent
organization,
andanyexpansionneedstotakethisintoaccount.
Workingintheinnovationsandboxislikedevelopingstartup
muscles.Atrst,theteamwilbeabletotakeononlymodest
experiments.Theearliest
experimentsmayfailtoproducemuch
learningandmaynotleadtoscalablesuccess.Overtime,those
teamsarealmostguaranteedtoimproveaslongastheygetthe
constantfeedbackofsmal-batchdevelopmentandactionable
metricsandareheldaccountabletolearningmilestones.
Ofcourse,anyinnovationsystemeventualywilbecomethe
victimofitsownsuccess.Asthesandboxexpandsandthe
company’srevenuegrowsasaresultofthesandbox’sinnovations,
thecyclewilhavetobeginagain.Theformerinnovatorswil
becomeguardiansofthestatusquo.Whentheproductmakesup
thewholesandbox,itinevitablywilbecomeencumberedwiththe
additionalrulesandcontrolsneededformission-critical
operation.
Newinnovationteamswilneedanewsandboxwithinwhichto
play.
BecomingtheStatusQuo
Thislasttransitionisespecialyhardforinnovatorstoaccept:their
transformationfromradicaloutsiderstotheembodimentofthe
statusquo.Ihavefounditdisturbinginmycareer.Asyoucanguess
fromthetechniquesIadvocateaspartoftheLeanStartup,Ihave
alwaysbeenabitofatroublemakeratthe
companiesatwhichI
haveworked,pushingforrapiditeration,data-drivendecision
making,andearlycustomerinvolvement.Whentheseideaswere
notpartofthedominantculture,itwassimple(iffrustrating)tobe
anadvocate.AlIhadtodowaspushashardashumanlypossible
formyideas.Sincethedominantculturefoundthemheretical,they
wouldcompromisewithmea“reasonable”amount.Thankstothe
psychologicalphenomenonofanchoring,thisledtoa
perverse
incentive:themoreradicalmysuggestionwas,themorelikelyit
wasthatthereasonablecompromisewouldbeclosertomytrue
goal.
Fast-forwardseveralyearstowhenIwasrunningproduct
development.Whenwe’dhirenewpeople,theyhadtobe
indoctrinatedintotheLeanStartupculture.Splittesting,continuous
deployment,andcustomertestingwerealstandardpractice.I
neededtocontinuetobeastrongadvocateformyideas,
making
sureeachnewemployeewasreadytogivethematry.Butforthe
peoplewhohadbeenworkingthereawhile,thoseideashad
becomepartofthestatusquo.
Likemanyentrepreneurs,Iwascaughtbetweenconstant
evangelizingformyideasandconstantlyentertainingsuggestions
forwaystheycouldbeimproved.Myemployeesfacedthesame
incentiveIhadexploitedyearsbefore:themoreradicalthe
suggestionis,themorelikelyitisthatthecompromisewil
move
inthedirectiontheydesire.Iheardital:suggestionsthatwego
backtowaterfaldevelopment,usemorequalityassurance(QA),
uselessQA,havemoreorlesscustomerinvolvement,usemore
visionandlessdata,orinterpretdatainamorestatisticalyrigorous
way.
way.
Ittookaconstanteorttoconsiderthesesuggestionsseriously.
However,respondingdogmaticalyisunhelpful.
Compromisingby
automaticalysplitingthediferencedoesn’tworkeither.
I’vefoundthateverysuggestionshouldbesubjectedtothesame
rigorousscienticinquirythatledtothecreationoftheLean
Startupintherstplace.Can
weusethetheorytopredictthe
resultsoftheproposedchange?Canweincubatethechangeina
smalteamandseewhathappens?Canwemeasureitsimpact?
Whenevertheycouldbeimplemented,theseapproacheshave
alowedmetoincreasemyownlearningand,moreimportant,the
productivityofthecompaniesIhaveworkedwith.Manyofthe
LeanStartuptechniquesthatwepioneeredatIMVUarenotmy
originalcontributions.Rather,theywereconceived,
incubated,and
executedbyemployeeswhobroughttheirowncreativityandtalent
tothetask.
Aboveal,Ifacedthiscommonquestion:Howdoweknowthat
“yourway”ofbuildingacompanywilwork?What
other
companiesareusingit?Whohasbecomerichandfamousasa
result?Thesequestionsaresensible.Thetitansofourindustryare
alworkinginaslower,morelinearway.Whyarewedoing
somethingdiferent?
Itisthesequestionsthatrequiretheuseoftheorytoanswer.
ThosewholooktoadopttheLeanStartupasadenedsetofsteps
ortacticswilnotsucceed.Ihadtolearnthisthehardway.Ina
startupsituation,thingsconstantlygowrong.Whenthathappens,
wefacetheage-olddilemmasummarizedbyDeming:Howdowe
knowthattheproblemisduetoaspecialcauseversusasystemic
cause?Ifwe’reinthemiddleofadoptinganewwayof
working,
thetemptationwilalwaysbetoblamethenewsystemforthe
problemsthatarise.Sometimesthattendencyiscorrect,sometimes
not.Learningtotelthedierencerequirestheory.Youhavetobe
abletopredicttheoutcomeofthechangesyoumaketotelifthe
problemsthatresultarerealyproblems.
Forexample,changingthedenitionofproductivityforateam
fromfunctionalexcelence—excelenceinmarketing,sales,or
productdevelopment—tovalidatedlearningwilcauseproblems.
productdevelopment—tovalidatedlearningwilcauseproblems.
Aswasindicatedearlier,functionalspecialistsareaccustomedto
measuringtheireciencybylookingattheproportionof
timethey
arebusydoingtheirwork.Aprogrammerexpectstobecodingal
daylong,forexample.Thatiswhymanytraditionalwork
environmentsfrustratetheseexperts:theconstantinterruptionof
meetings,cross-functional
handos,andexplanationsforendless
numbersofbossesalactasadragoneciency.However,the
individualeciencyofthesespecialistsisnotthegoalinaLean
Startup.Instead,wewanttoforceteamstoworkcross-functionaly
toachievevalidatedlearning.Manyofthetechniquesfordoingthis
—actionablemetrics,continuousdeployment,andtheoveralBuild-
Measure-Learnfeedbackloop—necessarilycauseteamsto
suboptimizefortheirindividualfunctions.Itdoes
notmaterhow
fastwecanbuild.Itdoesnotmaterhowfastwecanmeasure.
Whatmatersishowfastwecangetthroughtheentireloop.
Inmyyearsteachingthissystem,Ihavenoticedthispaternevery
time:switchingtovalidatedlearningfeelsworsebeforeitfeels
beter.That’sthecasebecausetheproblemscausedbytheold
systemtendtobeintangible,whereastheproblemsofthenew
systemarealtootangible.Havingthebenetoftheoryis
the
antidotetothesechalenges.Ifitisknownthatthislossof
productivityisaninevitablepartofthetransition,itcanbe
managedactively.Expectationscanbesetupfront.Inmy
consultingpractice,for
example,Ihavelearnedtoraisetheseissues
fromdayone;otherwise,theyareliabletoderailthewholeeort
onceitisunderway.Asthechangeprogresses,wecanusetheroot
causeanalysisandfastresponsetechniquestogureoutwhich
problemsneedprevention.Ultimately,theLeanStartupisa
framework,notablueprintofstepstofolow.Itisdesignedtobe
adaptedtotheconditionsofeachspeciccompany.Ratherthan
copywhatothershavedone,techniquessuchastheFive
Whys
alowyoutobuildsomethingthatisperfectlysuitedtoyour
company.
Thebestwaytoachievemasteryofandexploretheseideasisto
embedoneselfinacommunityofpractice.There
isathriving
communityofLeanStartupmeetupsaroundtheworldaswelas
communityofLeanStartupmeetupsaroundtheworldaswelas
online,andsuggestionsforhowyoucantakeadvantageofthese
resourceslistedinthelastchapterofthisbook,“Jointhe
Movement.”
13
EPILOGUE:WASTENOT
ThisyearmarkstheonehundredthanniversaryofFrederick
WinslowTaylor’sThePrinciplesofScienticManagement,rst
publishedin1911.Themovementforscienticmanagement
changedthecourseofthetwentiethcenturybymakingpossiblethe
tremendousprosperitythatwetakeforgrantedtoday.
Taylor
eectivelyinventedwhatwenowconsidersimplymanagement:
improvingtheeciencyofindividualworkers,managementby
exception(focusingonlyonunexpectedlygoodorbadresults),
standardizingworkintotasks,thetask-plus-bonussystemof
compensation,and—aboveal—theideathatworkcanbestudied
andimprovedthroughconsciouseort.Taylorinventedmodern
white-colarworkthatseescompaniesassystemsthat
mustbe
managedatmorethantheleveloftheindividual.Thereisareason
alpastmanagementrevolutionshavebeenledbyengineers:
managementishumansystemsengineering.
In1911Taylorwrote:“Inthe
past,themanhasbeenrst;inthe
future,thesystemmustberst.”Taylor’spredictionhascometo
pass.Wearelivingintheworldheimagined.Andyet,the
revolutionthatheunleashedhasbeen—inmanyways—too
successful.WhereasTaylorpreachedscienceasawayofthinking,
manypeopleconfusedhismessagewiththerigidtechniqueshe
advocated:timeandmotionstudies,thedierentialpiece-rate
system,and—mostgalingofal—theideathatworkers
shouldbe
treatedaslitlemorethanautomatons.Manyoftheseideasproved
extremelyharmfulandrequiredtheeortsoflatertheoristsand
extremelyharmfulandrequiredtheeortsoflatertheoristsand
managerstoundo.Criticaly,leanmanufacturingrediscoveredthe
wisdomandinitiativehiddenineveryfactoryworkerand
redirectedTaylor’snotionofeciencyawayfromtheindividual
taskandtowardthecorporateorganismasawhole.Buteachof
thesesubsequentrevolutionshasembracedTaylor’scoreideathat
workcanbestudiedscienticalyandcanbeimprovedthrougha
rigorousexperimentalapproach.
Inthetwenty-rstcentury,wefaceanewsetofproblemsthat
Taylorcouldnothaveimagined.Ourproductivecapacitygreatly
exceedsourabilitytoknowwhattobuild.Althoughtherewasa
tremendousamountofinventionandinnovationintheearly
twentiethcentury,mostofitwasdevotedtoincreasingthe
productivityofworkersandmachinesinordertofeed,clothe,and
housetheworld’spopulation.Althoughthatprojectisstil
incomplete,asthemilionswholiveinpovertycanatest,the
solutiontothatproblemisnowstrictlyapoliticalone.Wehavethe
capacitytobuildalmostanythingwecanimagine.Thebigquestion
ofourtimeisnotCanitbebuilt?butShoulditbebuilt?This
placesusinanunusualhistoricalmoment:ourfutureprosperity
dependsonthequalityofourcolectiveimaginations.
In1911,Taylorwrote:
Wecanseeourforestsvanishing,ourwater-powersgoingto
waste,oursoilbeingcarriedbyoodsintothesea;andthe
endofourcoalandourironisinsight.Butourlarger
wastesofhumaneort,whichgooneverydaythroughsuch
ofouractsasareblundering,il-directed,or
inecient…arelessvisible,lesstangible,andarebut
vaguelyappreciated.
Wecanseeandfeelthewasteofmaterialthings.
Awkward,inecient,oril-directedmovementsofmen,
however,leavenothingvisibleortangiblebehindthem.
Theirappreciationcalsforanactofmemory,aneortof
theimagination.Andforthisreason,eventhoughourdaily
lossfromthissourceisgreaterthanfromourwasteof
materialthings,theonehasstirredusdeeply,whilethe
materialthings,theonehasstirredusdeeply,whilethe
otherhasmovedusbutlitle.1
Acenturyon,whatcanwesayaboutthosewords?Ontheone
hand,theyfeelarchaic.We
ofthetwenty-rstcenturyare
hyperawareoftheimportanceofeciencyandtheeconomicvalue
ofproductivitygains.Ourworkplacesare—atleastwhenitcomes
tothebuildingofmaterialobjects—incrediblywelorganized
comparedwiththoseofTaylor’sday.
Ontheotherhand,Taylor’swordsstrikemeascompletely
contemporary.Foralofourvauntedeciencyinthemakingof
things,oureconomyisstilincrediblywasteful.Thiswastecomes
notfromtheinecientorganizationofworkbutratherfrom
workingonthewrongthings—andonanindustrialscale.AsPeter
Druckersaid,“Thereissurelynothingquitesouselessasdoingwith
greateficiencywhatshouldnotbedoneatal.”2
Andyetwearedoingthewrongthingsecientlyalthetime.It
ishardtocomebyasolidestimateofjusthowwastefulmodern
workis,butthereisnoshortageofanecdotes.Inmyconsultingand
travelstalkingabouttheLeanStartup,Ihearthesame
message
consistentlyfromemployeesofcompaniesbigandsmal.Inevery
industryweseeendlessstoriesoffailedlaunches,il-conceived
projects,andlarge-batchdeathspirals.Iconsiderthismisuseof
people’stimeacriminalynegligentwasteofhumancreativityand
potential.
Whatpercentageofalthiswasteispreventable?Ithinkamuch
largerproportionthanwecurrentlyrealize.MostpeopleImeet
believethatintheirindustryatleast,projectsfailforgoodreasons:
projectsareinherentlyrisky,marketconditionsareunpredictable,
“bigcompanypeople”areintrinsicalyuncreative.Somebelieve
thatifwejustslowedeverythingdownanduseda
morecareful
process,wecouldreducethefailureratebydoingfewerprojectsof
higherquality.Othersbelievethatcertainpeoplehaveaninnate
giftofknowingtherightthingtobuild.Ifwecanndenoughof
thesevisionariesandvirtuosos,ourproblemswilbesolved.These
“solutions”wereonceconsideredstateoftheartinthenineteenth
century,too,beforepeopleknewaboutmodernmanagement.
century,too,beforepeopleknewaboutmodern
management.
Therequirementsofanever-fasterworldmaketheseantique
approachesunworkable,andsotheblameforfailedprojectsand
businessesoftenisheapedonseniormanagement,whichisaskedto
dotheimpossible.Alternatively,thengerofblameispointedat
nancialinvestorsorthepublicmarketsforoveremphasizingquick
xesandshort-termresults.Wehaveplentyofblametogoaround,
butfartoolitletheorytoguidetheactionsofleaders
andinvestors
alike.
TheLeanStartupmovementstandsincontrasttothishand-
wringing.Webelievethatmostformsofwasteininnovationare
preventableoncetheircausesareunderstood.Althatis
requiredis
thatwechangeourcolectivemind-setconcerninghowthisworkis
tobedone.
Itisinsucienttoexhortworkerstotryharder.Ourcurrent
problemsarecausedbytryingtoohard—atthewrong
things.By
focusingonfunctionaleciency,welosesightoftherealgoalof
innovation:tolearnthatwhichiscurrentlyunknown.AsDeming
taught,whatmatersisnotsetingquantitativegoalsbutxingthe
methodbywhichthosegoalsareatained.TheLeanStartup
movementstandsfortheprinciplethatthescienticmethodcanbe
broughttobeartoanswerthemostpressinginnovationquestion:
Howcanwebuildasustainableorganization
aroundanewsetof
productsorservices?
ORGANIZATIONALSUPERPOWERS
Aparticipantatoneofmyworkshopscameuptomeafewmonths
afterwardtorelatethefolowingstory,whichIamparaphrasing:
“KnowingLeanStartupprinciplesmakesmefeellikeIhave
superpowers.EventhoughI’mjustajunioremployee,whenImeet
withcorporateVPsandGMsinmylargecompany,Iaskthem
simplequestionsandveryquicklyhelpthemseehow
theirprojects
arebasedonfundamentalhypothesesthataretestable.Inminutes,I
canlayoutaplantheycouldfolowtoscienticalyvalidatetheir
plansbeforeit’stoolate.Theyconsistentlyrespondwith‘Wow,you
plansbeforeit’stoolate.Theyconsistentlyrespondwith‘Wow,you
arebriliant.We’veneverthoughttoapplythatlevelofrigortoour
thinkingaboutnewproductsbefore.’”
Asaresultoftheseinteractions,hehasdevelopedareputation
withinhislargecompanyasabriliantemployee.Thishasbeen
goodforhiscareerbutveryfrustratingforhimpersonaly.Why?
Becausealthoughheisquitebriliant,hisinsightsintoawed
productplansareduenottohisspecialinteligencebutto
havinga
theorythatalowshimtopredictwhatwilhappenandpropose
alternatives.Heisfrustratedbecausethemanagersheispitchinghis
ideastodonotseethesystem.Theywronglyconcludethatthekey
tosuccessisndingbriliantpeoplelikehimtoputontheirteams.
Theyarefailingtoseetheopportunityheisrealypresentingthem:
toachievebeterresultssystematicalybychangingtheirbeliefs
abouthowinnovationhappens.
PutingtheSystemFirst:SomeDangers
LikeTaylorbeforeus,ourchalengeistopersuadethemanagersof
moderncorporationstoputthesystemrst.However,Taylorism
shouldactasacautionarytale,anditisimportanttolearnthe
lessonsofhistoryaswebringthesenewideastoamore
mainstreamaudience.
Taylorisrememberedforhisfocusonsystematicpracticerather
thanindividualbriliance.HereisthefulquotefromThe
PrinciplesofScientic
Managementthatincludesthefamousline
aboutputingthesystemfirst:
Inthefutureitwilbeappreciatedthatourleadersmustbe
trainedrightaswelasbornright,andthatnogreatman
can(withtheoldsystemof
personalmanagement)hopeto
competewithanumberofordinarymenwhohavebeen
properlyorganizedsoaseficientlytocooperate.
Inthepastthemanhasbeenfirst;inthefuturethesystem
mustberst.Thisinnosense,however,impliesthatgreat
menarenotneeded.Onthecontrary,therstobjectofany
menarenotneeded.Onthecontrary,therstobjectofany
goodsystemmustbethatofdevelopingrst-classmen;and
undersystematicmanagementthebestmanrisestothetop
morecertainlyandmorerapidlythaneverbefore.3
Unfortunately,Taylor’sinsistencethatscientificmanagementdoes
notstandinoppositiontondingandpromotingthebest
individualswasquicklyforgoten.Infact,theproductivitygainsto
behadthroughtheearlyscienticmanagementtactics,suchastime
andmotionstudy,task-plus-bonus,andespecialyfunctional
foremanship(theforerunneroftoday’sfunctionaldepartments),
weresosignicantthatsubsequentgenerationsof
managerslost
sightoftheimportanceofthepeoplewhowereimplementing
them.
Thishasledtotwoproblems:(1)businesssystemsbecameoverly
rigidandtherebyfailedtotakeadvantageofthe
adaptability,
creativity,andwisdomofindividualworkers,and(2)therehas
beenanoveremphasisonplanning,prevention,andprocedure,
whichenableorganizationstoachieveconsistentresultsinamostly
staticworld.Onthefactoryoor,theseproblemshavebeentackled
headonbytheleanmanufacturingmovement,andthoselessons
havespreadthroughoutmanymoderncorporations.Andyetinnew
productdevelopment,entrepreneurship,and
innovationworkin
generalwearestilusinganoutdatedframework.
MyhopeisthattheLeanStartupmovementwilnotfalintothe
samereductionisttrap.Wearejustbeginningtouncovertherules
thatgovernentrepreneurship,
amethodthatcanimprovetheodds
ofstartupsuccess,andasystematicapproachtobuildingnewand
innovativeproducts.Thisinnowaydiminishesthetraditional
entrepreneurialvirtues:theprimacyofvision,thewilingnessto
takeboldrisks,andthecouragerequiredinthefaceof
overwhelmingodds.Oursocietyneedsthecreativityandvisionof
entrepreneursmorethanever.Infact,itispreciselybecausethese
aresuchpreciousresourcesthatwecannotafordto
wastethem.
ProductDevelopmentPseudoscience
IbelievethatifTaylorwerealivetoday,hewouldchuckleatwhat
constitutesthemanagementofentrepreneursandinnovators.
Althoughweharnessthe
laborofscientistsandengineerswho
wouldhavedazzledanyearly-twentieth-centurypersonwiththeir
featsoftechnicalwizardry,themanagementpracticesweuseto
organizethemaregeneralydevoidofscienticrigor.Infact,I
wouldgosofarastocalthempseudoscience.
Weroutinelygreen-lightnewprojectsmoreonthebasisof
intuitionthanfacts.Aswe’veseenthroughoutthisbook,thatisnot
therootcauseoftheproblem.Alinnovationbeginswithvision.It’s
whathappensnextthatiscritical.Aswe’veseen,toomany
innovationteamsengageinsuccesstheater,selectivelyndingdata
thatsupporttheirvisionratherthanexposingtheelementsofthe
visiontotrueexperiments,or,evenworse,stayingin
stealthmode
tocreateadata-freezoneforunlimited“experimentation”thatis
devoidofcustomerfeedbackorexternalaccountabilityofanykind.
Anytimeateamatemptstodemonstratecauseandeectby
placinghighlightsonagraphofgrossmetrics,itisengagingin
pseudoscience.Howdoweknowthattheproposedcauseand
eectistrue?Anytimeateamatemptstojustifyitsfailuresby
resortingtolearningasanexcuse,itisengagedin
pseudoscienceas
wel.
Iflearninghastakenplaceinoneiterationcycle,letus
demonstrateitbyturningitintovalidatedlearninginthenext
cycle.Onlybybuildingamodelofcustomerbehaviorandthen
showingourabilitytouseourproductorservicetochangeitover
timecanweestablishrealfactsaboutthevalidityofourvision.
ThroughoutourcelebrationofthesuccessoftheLeanStartup
movement,anoteofcautionisessential.Wecannotaord
tohave
oursuccessbreedanewpseudosciencearoundpivots,MVPs,and
thelike.Thiswasthefateofscienticmanagement,andintheend,
Ibelieve,thatsetbackitscausebydecades.Sciencecametostand
forthevictoryofroutineworkovercreativework,mechanization
overhumanity,andplansoveragility.Latermovementshadtobe
spawnedtocorrectthosedeficiencies.
spawnedtocorrectthosedeficiencies.
Taylorbelievedinmanythingsthathedubbedscienticbutthat
ourmoderneyesperceiveasmereprejudice.Hebelievedinthe
inherentsuperiorityinbothinteligenceandcharacterofaristocratic
menovertheworkingclassesandthesuperiorityofmen
over
women;healsothoughtthatlower-statuspeopleshouldbe
supervisedstrictlybytheirbeters.Thesebeliefsarepartandparcel
ofTaylor’stime,anditistemptingtoforgivehimforhavingbeen
blindtothem.
Yetwhenourtimeisviewedthroughthelensoffuturepractice,
whatprejudiceswilberevealed?Inwhatforcesdoweplace
unduefaith?Whatmightwerisklosingsightofwiththisinitial
successofourmovement?
ItiswiththesequestionsthatIwishtoclose.Asgratifyingasitis
formetoseetheLeanStartupmovementgainfameand
recognition,itisfarmoreimportantthatweberightinour
prescriptions.Whatisknownsofarisjustthetipofthe
iceberg.
Whatisneededisamassiveprojecttodiscoverhowtounlockthe
vaststoresofpotentialthatarehiddeninplainsightinourmodern
workforce.Ifwestoppedwastingpeople’stime,whatwouldthey
dowithit?Wehavenorealconceptofwhatispossible.
Startinginthelate1880s,Taylorbeganaprogramof
experimentationtodiscovertheoptimalwaytocutsteel.Inthe
courseofthatresearch,whichlastedmorethantwenty-veyears,
heandhiscoleaguesperformedmorethantwentythousand
individualexperiments.Whatisremarkableaboutthisprojectis
thatithadnoacademicbacking,nogovernmentR&Dbudget.Its
entirecostwaspaidbyindustryoutoftheimmediate
prots
generatedfromthehigherproductivitytheexperimentsenabled.
Thiswasonlyoneexperimentalprogramtouncoverthehidden
productivityinjustonekindofwork.Otherscienticmanagement
disciplesspentyearsinvestigatingbricklaying,farming,andeven
shoveling.Theywereobsessedwithlearningthetruthandwerenot
satisedwiththefolkwisdomofcraftspersonsortheparablesof
experts.
Cananyofusimagineamodernknowledge-workmanagerwith
thesamelevelofinterestinthemethodshisorheremployeesuse?
thesamelevelofinterestinthemethodshisorheremployeesuse?
Howmuchofourcurrentinnovationworkisguidedby
catchphrasesthatlackascientificfoundation?
ANewResearchProgram
Whatcomparableresearchprogramscouldwebeengagedinto
discoverhowtoworkmoreefectively?
Foronething,wehaveverylitleunderstandingofwhat
stimulatesproductivityunderconditionsofextremeuncertainty.
Luckily,withcycletimesfalingeverywhere,wehavemany
opportunitiestotestnewapproaches.Thus,Iproposethatwe
createstartuptestinglabsthatcouldputalmannerof
product
developmentmethodologiestothetest.
Howmightthosetestsbeconducted?Wecouldbringinsmal
cross-functionalteams,perhapsbeginningwithproductand
engineering,andhavethem
worktosolveproblemsbyusing
dierentdevelopmentmethodologies.Wecouldbeginwith
problemswithclearrightanswers,perhapsdrawnfromthemany
internationalprogrammingcompetitionsthathavedeveloped
databasesofwel-denedproblemswithclearsolutions.These
competitionsalsoprovideaclearbaselineofhowlongitshould
takeforvariousproblemstobesolvedsothatwecouldestablish
clearlytheindividualproblem-solvingprowessof
theexperimental
subjects.
Usingthiskindofsetupforcalibration,wecouldbegintovary
theconditionsoftheexperiments.Thechalengewilbetoincrease
thelevelofuncertaintyaboutwhattherightansweris
whilestil
beingabletomeasurethequalityoftheoutcomeobjectively.
Perhapswecouldusereal-worldcustomerproblemsandthenhave
realconsumerstesttheoutputoftheteams’work.Orperhapswe
couldgosofarastobuildminimumviableproductsforsolvingthe
samesetofproblemsoverandoveragaintoquantifywhich
producesthebestcustomerconversionrates.
Wealsocouldvarytheal-importantcycletimebychoosingmore
orlesscomplexdevelopmentplatformsanddistributionchannelsto
orlesscomplexdevelopmentplatformsanddistributionchannelsto
testtheimpactofthosefactorsonthetrueproductivityofthe
teams.
Mostofal,weneedtodevelopclearmethodsforholdingteams
accountableforvalidatedlearning.Ihaveproposedonemethodin
thisbook:innovationaccountingusingawel-denedfinancial
modelandengineofgrowth.However,itisnaiveto
assumethat
thisisthebestpossiblemethod.Asitisadoptedinmoreandmore
companies,undoubtedlynewtechniqueswilbesuggested,andwe
needtobeabletoevaluatethenewideasasrigorouslyaspossible.
Althesequestionsraisethepossibilitiesofpublic-private
partnershipsbetweenresearchuniversitiesandtheentrepreneurial
communitiestheyseektofoster.Italsosuggeststhatuniversities
maybeabletoaddvalueinmorewaysthanbybeingsimply
nancialinvestorsorcreatorsofstartupincubators,asisthecurrent
trend.Mypredictionisthatwhereverthisresearchisconductedwil
becomeanepicenterofnewentrepreneurialpractice,and
universitiesconductingthisresearchthereforemaybeableto
achieveamuchhigherlevelofcommercializationoftheirbasic
researchactivities.4
THELONG-TERMSTOCKEXCHANGE
Beyondsimpleresearch,Ibelieveourgoalshouldbetochangethe
entireecosystemof
entrepreneurship.Toomuchofourstartup
industryhasdevolvedintoafeedersystemforgiantmedia
companiesandinvestmentbanks.Partofthereasonestablished
companiesstruggletoinvestconsistentlyininnovationisintense
pressurefrompublicmarketstohitshort-termprotabilityand
growthtargets.Mostly,thisisaconsequenceoftheaccounting
methodswehavedevelopedforevaluatingmanagers,whichfocus
onthekindsofgross“vanity”metricsdiscussedin
Chapter7.What
isneededisanewkindofstockexchange,designedtotradeinthe
stocksofcompaniesthatareorganizedtosustainlong-term
thinking.IproposethatwecreateaLong-TermStockExchange
(LTSE).
(LTSE).
Inadditiontoquarterlyreportsonprotsandmargins,
companiesontheLTSEwouldreportusinginnovationaccounting
ontheirinternalentrepreneurshipeorts.Like
Intuit,theywould
reportontherevenuetheyweregeneratingfromproductsthatdid
notexistafewyearsearlier.ExecutivecompensationinLTSE
companieswouldbetiedtothecompany’slong-termperformance.
TradingontheLTSEwouldhavemuchhighertransactioncostsand
feestominimizedaytradingandmassivepriceswings.Inexchange,
LTSEcompanieswouldbealowedtostructuretheircorporate
governancetofacilitategreaterfreedomfor
managementtopursue
long-terminvestments.Inadditiontosupportforlong-term
thinking,thetransparencyoftheLTSEwilprovidevaluabledata
abouthowtonurtureinnovationintherealworld.Somethinglike
theLTSEwouldacceleratethecreationofthenextgenerationof
greatcompanies,builtfromthegroundupforcontinuous
innovation.
INCONCLUSION
Asamovement,theLeanStartupmustavoiddoctrinesandrigid
ideology.Wemustavoidthecaricaturethatsciencemeansformula
oralackofhumanityinwork.Infact,scienceisoneofhumanity’s
mostcreativepursuits.Ibelievethatapplyingitto
entrepreneurshipwilunlockavaststorehouseofhumanpotential.
Whatwouldanorganizationlooklikeifalofitsemployees
werearmedwithLeanStartuporganizationalsuperpowers?
Foronething,everyonewouldinsistthatassumptionsbestated
explicitlyandtestedrigorouslynotasastaling
tacticoraformof
make-workbutoutofagenuinedesiretodiscoverthetruththat
underlieseveryproject’svision.
Wewouldnotwastetimeonendlessargumentsbetweenthe
defendersofqualityandthe
cowboysofrecklessadvance;instead,
wewouldrecognizethatspeedandqualityarealiesinthepursuit
ofthecustomer’slong-termbenet.Wewouldracetotestour
visionbutnottoabandonit.Wewouldlooktoeliminatewastenot
visionbutnottoabandonit.Wewouldlooktoeliminatewastenot
tobuildqualitycastlesintheskybutintheserviceofagilityand
breakthroughbusinessresults.
Wewouldrespondtofailuresandsetbackswithhonestyand
learning,notwithrecriminationsandblame.Morethanthat,we
wouldshuntheimpulsetoslowdown,increasebatchsize,and
indulgeinthecurseofprevention.Instead,wewouldachievespeed
bybypassingtheexcessworkthatdoesnotleadtolearning.
We
woulddedicateourselvestothecreationofnewinstitutionswitha
long-termmissiontobuildsustainablevalueandchangetheworld
forthebeter.
Mostofal,wewouldstopwastingpeople’stime.
14
JOINTHEMOVEMENT
Inthepastfewyears,theLeanStartupmovementhasgoneglobal.
Thenumberofresourcesavailableforaspiringentrepreneursis
incredible.Here,I’ldomybesttolistjustafewofthe
best
events,books,andblogsforfurtherreadingandfurtherpractice.
Therestisuptoyou.Readingisgood,actionisbeter.
Themostimportantresourcesarelocal.Gonearethedayswhere
youhadtobeinSiliconValeytondotherentrepreneurstoshare
ideasandstruggleswith.However,beingembeddedinastartup
ecosystemisstilanimportantpartofentrepreneurship.What’s
changedisthattheseecosystemsarespringingup
inmoreandmore
startuphubsaroundtheworld.
ImaintainanocialwebsiteforTheLeanStartupat
htp://theleanstartup.com,whereyoucanndadditionalresources,
includingcasestudiesandlinkstofurtherreading.You
wilalso
ndlinkstheretomyblog,StartupLessonsLearned,aswelas
videos,slides,andaudiofrommypastpresentations.
LeanStartupMeetups
ChancesarethereisaLeanStartupmeetupgroupnearyou.Asof
thiswriting,thereareoverahundred,withthelargestinSan
Francisco,Boston,NewYork,Chicago,andLosAngeles.Youcan
ndareal-timemapofgroupshere:htp://lean-
startup.meetup.com/.Youcanalsondalistofcitieswherepeople
areinterestedinstartinganewgroup,andtoolstosetoneup
areinterestedinstartinganewgroup,andtoolstosetoneup
yourself.
TheLeanStartupWiki
NoteveryLeanStartupgroupusesMeetup.comto
organize,anda
comprehensivelistofeventsandotherresourcesismaintainedby
volunteers
on
the
Lean
Startup
Wiki:
htp://leanstartup.pbworks.com/
TheLeanStartupCircle
ThelargestcommunityofpracticearoundtheLeanStartupis
happeningonline,rightnow,
ontheLeanStartupCirclemailing
list.FoundedbyRichColins,thelisthasthousandsof
entrepreneurssharingtips,resources,andstorieseveryday.Ifyou
haveaquestionabouthowLeanStartupmightapplytoyour
businessorindustry,it’sagreatplacetostart:
htp://leanstartupcircle.com/
TheStartupLessonsLearnedConference
Forthepasttwoyears,IhaverunaconferencecaledStartup
LessonsLearned.Moredetailsareavailablehere:ht
p://slconf.com
REQUIREDREADING
SteveBlank’sbookTheFourStepstotheEpiphanyistheoriginal
bookaboutcustomerdevelopment.WhenIwasbuildingIMVU,a
dog-earedcopyofthisbookfolowedmeeverywhere.Itis
an
indispensableguide.Youcangetacopyhere:
htp://ericri.es/FourStepsorreadmyreviewofithere:
htp://www.startuplessonslearned.com/2008/11/what-is-customer-
development.html.Stevealsomaintainsanactiveandexcel
ent
development.html.Stevealsomaintainsanactiveandexcelent
blogathtp://steveblank.com/
BrantCooperandPatrickVlaskovitshavecreatedashortbut
excelentbookcaledThe
Entrepreneur’sGuidetoCustomer
Development,whichprovidesagentleintroductiontothetopic.
Youcanbuyithere:htp://custdev.comorreadmyreviewhere:
htp://www.startuplessonslearned.com/2010/07/entrepreneurs-
guide-to-customer.html
WhenIrstbeganbloggingaboutentrepreneurship,itwasnot
nearlyascommonanoccupationasitisnow.Veryfewbloggers
wereactivelyworkingonnewideasaboutentrepreneurship,and
togetherwedebatedandrefinedtheseideasonline.
DaveMcClure,founderoftheventurerm500Startups,writesa
blogathtp://500hats.typepad.com/.500Startupshasanexcelent
blogaswel:htp://blog.500startups.com/.Dave’s“StartupMetrics
forPirates”presentationlaidoutaframeworkforthinkingabout
andmeasuringonlineservicesthatgreatlyinuencedtheconceptof
“enginesofgrowth.”Youcanseetheoriginalpresentationhere:
htp://500hats.typepad.com/500blogs/2008/09/startup-
metri-
2.htmlaswelasmyoriginalreactionhere:
htp://www.startuplessonslearned.com/2008/09/three-drivers-of-
growth-for-your.html
SeanEliswritestheStartupMarketingBlog,whichhasbeen
inuentialinmythinkingabouthowtointegratemarketinginto
startups:htp://startup-marketing.com/
AndrewChen’sblogFuturisticPlayisoneofthebestsourcesfor
thoughtsonviralmarketing,startupmetrics,anddesign:
htp://andrewchenblog.com/
BabakNiviwritestheexcelentblogVentureHacksandwasan
earlyLeanStartupevangelist:htp://venturehacks.com/.He’ssince
goneontocreateAngelList,whichmatchesstartupsandinvestors
aroundtheworld:htp://angel.co/
OtherfantasticLeanStartupblogsinclude:
•AshMauryahasemergedasaleaderinhelpingbootstrapped
onlinebusinessesapplyLeanStartupideas.Hisblogiscaled
RunningLean,andhealsohasreleasedaneBookofthesame
name.
Both
can
be
found
here:
htp://www.runningleanhq.com/
•SeanMurphyonearly-stagesoftwarestartups:
htp://www.skmurphy.com/blog/
•BrantCooper’sMarketbyNumbers:htp://market-by-
numbers.com/
•PatrickVlaskovitsontechnology,customerdevelopment,and
pricing:htp://vlaskovits.com/
•TheKISSmetricsMarketingBlog:htp://blog.kissmetrics.com/
andHitenShah’sht
p://hitenism.com
FURTHERREADING
ClaytonM.Christensen’sTheInnovator’sDilemmaandThe
Innovator’sSolutionareclassics.Inaddition,Christensen’smore
recentworkisalsoextremelyhelpfulforseeingthetheory
of
disruptiveinnovationinpractice,includingTheInnovator’s
Prescription(aboutdisruptinghealthcare)andDisruptingClass
(abouteducation).
htp://ericri.es/ClaytonChristensen
GeoreyA.Moore’searlyworkisfamousamongalentrepreneurs,
especialyCrossingtheChasmandInsidetheTornado.Buthehas
continuedtorenehisthinking,andIhavefoundhislatestwork,
DealingwithDarwin:HowGreatCompaniesInnovateat
Every
PhaseofTheirEvolution,especialyuseful.
htp://ericri.es/DealingWithDarwin
ThePrinciplesofProductDevelopmentFlow:SecondGeneration
LeanProductDevelopmentbyDonaldG.Reinertsen.
LeanProductDevelopmentbyDonaldG.Reinertsen.
htp://ericri.es/pdflow
TheToyotaWaybyJefreyLiker.
htp://ericri.es/thetoyotaway
LeanThinking:BanishWasteandCreateWealthinYour
Corporation,Revisedand
UpdatedbyJamesP.WomackandDaniel
T.Jones.
htp://ericri.es/LeanThinking
ThePeople’sTycoon:HenryFordandtheAmericanCenturyby
StevenWats.
ht
p://ericri.es/ThePeoplesTycoon
TheOneBestWay:FrederickWinslowTaylorandtheEnigmaof
EficiencybyRobertKanigel.
htp://ericri.es/OneBestWay
ThePrinciplesofScienticManagementbyFrederickWinslow
Taylor.
htp://ericri.es/ScientificManagement
ExtremeProgrammingExplained:EmbraceChangebyKentBeck
andCynthiaAndres.
htp://ericri.es/EmbraceChange
ToyotaProductionSystem:BeyondLarge-ScaleProductionby
TaichiOhno.
htp://ericri.es/TaichiOhno
TheideaoftheBuild-Measure-Learnfeedbackloopowesalotto
ideasfrommaneuverwarfare,especialyJohn
Boyd’sOODA
(Observe-Orient-Decide-Act)Loop.Themostaccessibleintroduction
toBoyd’sideasisCertaintoWin:TheStrategyofJohnBoyd,
AppliedtoBusinessbyChetRichards.
htp://ericri.es/CertainToWin
OutoftheCrisisbyW.EdwardsDeming.
htp://ericri.es/OutOfTheCrisis
MyYearswithGeneralMotorsbyAlfredSloan.
htp://ericri.es/MyYears
Bily,Alfred,andGeneralMotors:TheStoryofTwoUniqueMen,a
LegendaryCompany,andaRemarkableTimeinAmericanHistory
byWiliamPelfrey.
htp://ericri.es/BilyAlfred
ThePracticeofManagementbyPeterF.Drucker.
htp://ericri.es/PracticeOfManagement
GetingtoPlanB:BreakingThroughtoaBeterBusinessModelby
JohnMulinsandRandyKomisar.
htp://ericri.es/GetingToPlanB
Endnotes
Introduction
1.Foranup-to-datelistingofLeanStartupmeetupsortondone
nearyou,seehtp://lean-startup.meetup.comortheLean
StartupWiki:htp://leanstartup.pbworks.com/Meetups.Seealso
Chapter14,“JointheMovement.”
Chapter1.Start
1.Manufacturingstatisticsandanalysisaredrawnfromtheblog
FiveThirtyEight:htp://www.fivethirtyeight.com/2010/02/us-
manufacturing-is-not-dead.html
Chapter2.Define
1.TheInnovator’sDilemmaisaclassictextbyClaytonChristensen
aboutthedicultyestablishedcompanieshavewithdisruptive
innovation.Alongwithitssequel,TheInnovator’sSolution,it
laysoutspecicsuggestionsforhowestablishedcompaniescan
createautonomousdivisionstopursuestartup-likeinnovation.
Thesespecicstructuralprerequisitesarediscussedindetailin
Chapter12.
2.FormoreaboutSnapTax,seehtp://blog.turbotax.intuit.com/
turbotax-press-releases/taxes-on-your-mobile-phone-
it%E2%80%99s-a-snap/01142011–4865andhtp://mobilized.
althingsd.com/20110204/exclusive-
intuit-sees-more-than-
althingsd.com/20110204/exclusive-intuit-sees-more-than-
350000-downloads-for-snaptax-its-smartphone-tax-filing-app/
3.MostinformationrelatingtoIntuitandSnapTaxcomesfrom
privateinterviewswithIntuitmanagementandemployees.
InformationaboutIntuit’sfoundingcomesfromSuzanneTaylor
andKathySchroeder’sInsideIntuit:HowtheMakersofQuicken
BeatMicrosoftandRevolutionizedanEntireIndustry
(Cambridge,Mass.:HarvardBusinessPress,2003).
Chapter3.Learn
1.TheoriginalvefoundersofIMVUwereWilHarvey,Marcus
Gosling,MatDanzig,MelGuymon,andmyself.
2.UsageintheUnitedStateswasevenmoreconcentrated;
see
htp://www.businessweek.com/technology/tech_stats/
im050307.htm
3.TohearmoreaboutIMVU’searlyconversationswithcustomers
thatledtoourpivotawayfromtheadd-onstrategy,see:htp://
mixergy.com/ries-lean/
4.Awordofcaution:demonstratingvalidatedlearningrequiresthe
rightkindofmetrics,caledactionablemetrics,whichare
discussedinChapter7.
5.ThiscasewaswritenbyBethanyCoatesunderthedirectionof
ProfessorAndyRachle.Youcangetacopyhere:htp://hbr.
org/product/imvu/an/E254-PDF-ENG
Chapter4.Experiment
1.Someentrepreneurshaveadoptedthissloganastheirstartup
philosophy,usingthe
acronymJFDI.Arecentexamplecanbe
seen
athtp://www.cloudave.com/1171/what-makes-an-
entrepreneur-four-leters-jfdi/
2.htp://techcrunch.com/2009/11/02/amazon-
closes-zappos-deal-
ends-up-paying-1–2-bilion/
ends-up-paying-1–2-bilion/
3.IwanttothankCarolineBarlerinandHPforalowingmeto
includemyexperimentalanalysisofthisnewproject.
4.InformationaboutKodak
Galerycomesfrominterviews
conductedbySaraLeslie.
5.TheVLSstorywasrecountedbyElnorRozenrot,formerlyof
InnosightVentures.AdditionaldetailwasprovidedbyAkshay
Mehra.FormoreontheVLS,
seethearticleinHarvardBusiness
Review:
htp://hbr.org/2011/01/new-business-models-in-
emerging-markets/ar/1orpresscoverageathtp://
economictimes.indiatimes.com/news/news-by-company/
corporate-trends/vilage-laundry-services-takes-on-the-dhobi/
articleshow/5325032.cms
6.FormoreontheearlyeortsoftheCFPB,seetheWalStreet
Journal’sApril13,2011,article“ForComplaints,Don’tCal
ConsumerBureauYet”;htp://online.wsj.com/article/SB100
01424052748703551304576260772357440148.html.
Many
dedicatedpublicservantsarecurrentlyworkinghardto
incorporatethisexperimentalapproachinthepublicsector
undertheleadershipof
PresidentObama.Iwouldliketothank
AneeshChopra,ChrisVein,ToddPark,andDavidForrestfor
introducingmetothesegroundbreakingeforts.
Chapter5.Leap
1.Forexample,CUCommunity,whichbeganat
Columbia
University,hadanearlyheadstart.Seehtp://www.slate.com/
id/2269131/.ThisaccountofFacebook’sfoundingisdrawn
fromDavidKirkpatrick’sTheFacebookEfect(NewYork:Simon
&Schuster,2011).
2.Actualengagementnumbersfrom2004arehardtond,butthis
paternhasbeenconsistentthroughoutFacebook’spublic
statements.Forexample,ChrisHughesreportedin2005that
“60%logindaily.About85%loginatleastonceaweek,and
93%loginatleastonceamonth.”htp://techcrunch.com/
93%loginatleastonceamonth.”htp://techcrunch.com/
2005/09/07/85-of-colege-students-use-facebook/
3.Irstheardthetermleapoffaithappliedtostartupassumptions
byRandyKomisar,aformercoleagueandcurrentpartneratthe
venturermKleinerPerkinsCaueld&Byers.Heexpandson
theconceptinhisbookGetingtoPlanB,coauthored
withJohn
Mulins.
4.
htp://www.forbes.com/2009/09/17/venture-capital-ipod-
inteligent-technology-komisar.html
5.“Acarefulyresearched
tablecompiledforMotormagazineby
CharlesE.Duryea,himselfapioneercarmaker,revealedthat
from1900to1908,501companieswereformedintheUnited
Statesforthepurposeofmanufacturingautomobiles.Sixty
percentofthemfoldedoutrightwithinacoupleofyears;
another6percentmovedintootherareasofproduction.”This
quoteisfromtheFordbiographyThePeople’sTycoon:Henry
FordandtheAmericanCenturybyStevenWats
(NewYork:
Vintage,2006).
6.JefreyK.Liker,TheToyotaWay.NewYork:McGraw-Hil,2003,
p.223.
7.htp://www.autofieldguide.com/articles/030302.html
8.Inthecustomer
developmentmodel,thisiscaledcustomer
discovery.
9.FormoreonthefoundingofIntuit,seeSuzanneTaylorandKathy
Schroeder,InsideIntuit.
10.FormoreontheLeanUXmovement,seehtp://www.cooper.
com/journal/2011/02/lean_ux_product_stewardship_an.htmland
htp://www.slideshare.net/jgothelflean-ux-geting-out-of-the-
deliverables-business
Chapter6.Test
1.htp://www.pluggd.in/groupon-story-297/
2.“Groupon’s$6BilionGambler,”WalStreetJournal;htp://
2.“Groupon’s$6BilionGambler,”WalStreetJournal;htp://
online.wsj.com/article_email/SB100014240527487048281045
76021481410635432-IMyQjAxMTAwMDEwODExNDgyWj.html
3.Thetermminimumviable
producthasbeeninusesinceatleast
2000aspartofvariousapproachestoproductdevelopment.For
anacademicexample,seehtp://www2.cs.uidaho.edu/
~biljunk/Publications/DynamicBalance.pdf
SeealsoFrankRobinsonof
PMDI,whoreferstoaversionof
theproductthatisthesmalestneededtoseltopotential
customers(htp://productdevelopment.com/howitworks/mvp.
html).ThisissimilartoSteveBlank’sconceptofthe“minimum
featureset”incustomer
development(htp://steveblank.com/
2010/03/04/perfection-by-subtraction-the-minimum-feature-
set/).Myuseofthetermherehasbeengeneralizedtoany
versionofaproductthatcanbegintheprocessoflearning,using
theBuild-Measure-Learnfeedbackloop.Formore,seehtp://
www.startuplessonslearned.com/2009/08/minimum-viable-
product-guide.html
4.Manypeoplehavewritenaboutthisphenomenon,usingvarying
terminology.Probablythe
mostwidelyreadisGeoreyMoore’s
CrossingtheChasm.Formore,seeEricVonHippel’sresearch
intowhathetermed“leadusers”;hisbookTheSourcesof
Innovationisagreatplacetostart.SteveBlankusestheterm
earlyvangelisttoemphasizetheevangelicalpowersofthese
earlycustomers.
5.“Tothecasualobserver,theDropboxdemovideolookedlikea
normalproductdemonstration,”Drewsays,“butweputin
aboutadozenEastereggsthatweretailoredfortheDigg
audience.ReferencestoTayZondayand‘ChocolateRain’and
alusionstoOceSpaceandXKCD.Itwasatongue-in-cheek
nodtothatcrowd,anditkickedoachainreaction.
Within24
hours,thevideohadmorethan10,000Diggs.”htp://answers.
oreily.com/topic/1372-marketing-lessons-from-dropbox-a-qa-
with-ceo-drew-houston/.Youcanseetheoriginalvideoaswel
asthereactionfromtheDiggcommunityathtp://digg.com/
software/Google_Drive_kiler_coming_from_MIT_Startup.
For
software/Google_Drive_kiler_coming_from_MIT_Startup.
For
moreonDropbox’ssuccess,see“Dropbox:TheHotestStartup
You’veNeverHeardOf”athtp://tech.fortune.cnn.com/2011/
03/16/cloud-computing-for-the-rest-of-us/
6.ThisdescriptioncourtesyofLifehacker:htp://lifehacker.com/
5586203/food-on-the-table-builds-menus-and-grocery-lists-
based-on-your-familys-preferences
7.Thislistwascompiledbymycoleague,ProfessorTom
EisenmannatHarvardBusinessSchool,LaunchingTechnology
VenturesforacasethatheauthoredonAardvarkforhisnew
class.Formore,seehtp://platformsandnetworks.blogspot.com/
2011/01/launching-tech-ventures-part-i-course.html
8.
htp://www.robgo.org/post/568227990/product-
leadership-
series-user-driven-design-at
9.
htp://venturebeat.com/2010/02/11/confirmed-google-buys-
social-search-engine-aardvark-for-50-milion/
10.Thisistheheartofthe
Innovator’sDilemmabyClayton
Christensen.
11.Formore,seehtp://bit.ly/DontLaunch
Chapter7.Measure
1.Bycontrast,Google’smaincompetitorOverture(eventualy
boughtbyYahoo)hadaminimumaccountsizeof$50,which
deterredusfromsigningup,asitwastooexpensive.
2.FormoredetailsaboutFarb’sentrepreneurialjourney,seethis
Mixergy
interview:ht
p://mixergy.com/farbood-nivi-grockit-
interview/
Chapter8.Pivot(orPersevere)
1.htp://www.slideshare.net/dbineti/lean-startup-at-sxsw-votizen-
pivot-case-study
2.FormoreonPath,seehtp://techcrunch.com/2011/02/02/
2.FormoreonPath,seehtp://techcrunch.com/2011/02/02/
google-tried-to-buy-path-for-100-milion-path-said-no/
and
htp://techcrunch.com/2011/02/01kleiner-perkins-leads-8–5-
milion-round-for-path/
3.Includesapproximately$30milionofassetsundermanagement
andapproximately$150milionofassetsunderadministration,
asofApril1,2011.
4.FormoreonWealthfront,seethecasestudywritenby
Sarah
Milstein
athtp://www.startuplessonslearned.com/2010/07/
case-study-kaching-anatomy-of-pivot.html.Formoreon
Wealthfront’srecentsuccess,seehtp://bits.blogs.nytimes.com/
2010/10/19/wealthfront-loses-the-sound-efects/
5.IMVU’sresultshavebeensharedpubliclyonafewoccasions.For
2008,
seehtp://www.worldsinmotion.biz/2008/06/imvu_
reaches_20_milion_regist.php;for2009see
htp://www.imvu.
com/about/press_releases/press_release_20091005_1.php,and
for2010seehtp://techcrunch.com/2010/04/24/imvu-revenue/
6.BusinessarchitectureisaconceptexploredindetailinMoore’s
DealingwithDarwin.
“Organizationalstructurebasedon
prioritizingoneoftwobusinessmodels(Complexsystems
modelandVolumeoperationsmodel).Innovationtypesare
understoodandexecutedincompletelydierentways
dependingonwhichmodelanenterpriseadopts.”Formore,see
htp://www.dealingwithdarwin.com/theBook/
darwinDictionary.php
Chapter9.Batch
1.
ht
p://lssacademy.com/2008/03/24/a-response-to-the-video-
skeptics/
2.Ifyou’rehavingtroubleacceptingthisfact,itrealyishelpfulto
watchitonvideo.Oneextremelydetail-orientedbloggertook
onevideoandbrokeitdown,
second-by-second,toseewhere
thetimewent:“Youlosebetween2and5secondseverytime
youmovethepilearoundbetweensteps.Also,youhaveto
youmovethepilearoundbetweensteps.Also,youhaveto
managethepileseveraltimesduringatask,somethingyou
don’thavetodonearlyasmuchwith[single-pieceow].This
alsohasafactorycorolary:storing,moving,retrieving,and
lookingforworkinprogressinventory.”Seetherestofthe
commentary
here:htp://lssacademy.com/2008/03/24/a-
response-to-the-video-skeptics/
3.TimothyFitz,anearlyIMVUengineer,deservescreditforhaving
coinedthetermcontinuousdeploymentinablogpost:ht
p://
timothyfitz.wordpress.com/2009/02/10/continuous-
deployment-at-imvu-doing-the-impossible-fifty-times-a-day/.The
actualdevelopmentofthecontinuousdeploymentsystemisthe
workoftoomanydierentengineersatIMVUformeto
give
adequatecredithere.Fordetailsonhowtogetstartedwith
continuousdeployment,seehtp://radar.oreily.com/2009/03/
continuous-deployment-5-eas.html
4.Fortechnicaldetailsof
Wealthfront’scontinuousdeployment
setup,seehtp://eng.wealthfront.com/2010/05/deployment-
infrastructure-for.htmlandhtp://eng.wealthfront.com/2011/
03/lean-startup-stage-at-sxsw.html
5.ThisdescriptionofSchoolofOnewasprovidedby
Jennifer
CarolanofNewSchoolsVentureFund.
6.Formoreonthelarge-batchdeathspiral,seeThePrinciplesof
ProductDevelopmentFlow:SecondGenerationLeanProduct
DevelopmentbyDonaldG.
Reinertsen:htp://bit.ly/pdflow
7.TheseleanhealthcareexamplesarecourtesyofMarkGraban,
authorofLeanHospitals(NewYork:ProductivityPress,2008).
8.ThisilustrativestoryaboutpulisdrawnfromLeanProduction
SimplifiedbyPascalDennis(NewYork:ProductivityPress,
2007).
9.Foranexampleofthismisunderstandingatwork,seehtp://
www.oreilygmt.eu/interview/fatboy-in-a-lean-world/
10.InformationaboutAlphabetEnergycomesfrominterviews
conductedbySaraLeslie.
11.FormoreonToyota’slearningorganization,seeTheToyota
11.FormoreonToyota’slearningorganization,seeTheToyota
WaybyJefreyLiker.
Chapter10.Grow
1.TheHotmailstory,alongwithmanyotherexamples,isrecounted
inAdamL.Penenberg’sViralLoop.FormoreonHotmail,also
seehtp://www.fastcompany.com/magazine/27/netef
ects.html
2.Formoreonthefourcustomercurrenciesoftime,money,skil,
andpassion,seehtp://www.startuplessonslearned.com/2009/
12/business-ecology-and-four-customer.html
3.
htp://pmarca-archive.posterous.com/the-pmarca-guide-to-
startups-part-4-the-only
4.ThisisthelessonofGeoreyMoore’sbestselingbookCrossing
theChasm(NewYork:HarperPaperbacks,2002).
Chapter11.Adapt
1.ToyotaProductionSystem:BeyondLarge-ScaleProductionby
TaichiOhno(NewYork:ProductivityPress,1988).
2.FormoreonNetPromoterScore,seehtp://www.
startuplessonslearned.com/2008/11/net-promoter-score-
operational-tool-to.htmland
TheUltimateQuestionbyFred
Reichheld(Cambridge,Mass.:HarvardBusinessPress,2006).
3.InformationaboutQuickBookscomesfrominterviewsconducted
byMarisaPorzig.
Chapter12.Innovate
1.JereyLiker,JohnE.Etlie,andJohnCreightonCampbel,
EngineeredinJapan:JapaneseTechnology-Management
Practices(NewYork:OxfordUniversityPress,1995),p.196.
2.Foroneaccount,seePCMagazine’s“LookingBack:
15Yearsof
PCMagazine”byMichaelMiler,htp://www.pcmag.com/
PCMagazine”byMichaelMiler,htp://www.pcmag.com/
article2/0,2817,35549,00.asp
3.ThefolowingdiscussionowesagreatdealtoGeorey
Moore’s
DealingwithDarwin(NewYork:PortfolioTrade,2008).Ihave
hadsuccessimplementingthisframeworkincompaniesofmany
diferentsizes.
Chapter13.Epilogue:WasteNot
1.htp://www.ibiblio.org/eldritch/fwt/ti.html
2.htp://www.goodreads.com/author/quotes/66490.Peter_Drucker
3.htp://www.ibiblio.org/eldritch/fwt/ti.html
4.Infact,somesuchresearchhasalreadybegun.FormoreonLean
Startupresearchprograms,
seeNathanFurr’sLeanStartup
ResearchProjectatBYU,htp://nathanfurr.com/2010/09/15/
the-lean-startup-research-project/,andTomEisenmannof
HarvardBusinessSchool’sLaunchingTechnologyVentures
project,htp://platformsandnetworks.blogspot.com/2011/01/
launching-tech-ventures-part-iv.html
Disclosures
Ihaveworkedwiththefolowingcompaniesnamedinthisbook
eitherasaconsultant,adviser,orinvestor.Ihavea
relationshipor
equityinterestineachofthem.
Aardvark
IMVU
Dropbox
Intuit
FoodontheTable
Votizen
Grockit
Wealthfront
Ihaveadditionalinterestsincompaniesthroughmyaliations
withventurecapitalrms.Ihaveinvestedinorworkedwiththe
folowingrmsaseitheraconsultantorasalimitedpartner.
Throughtheserms,Ihaveequityandrelationshipinterestsin
manymorecompaniesbeyondthoselistedabove.
500Startups
KleinerPerkinsCaufield&
Byers
Floodgate
GreylockPartners
SeraphGroup
Acknowledgments
Ioweatremendousdebtofgratitudetothemanypeoplewhohave
helpedmakeTheLeanStartupareality.Firstandforemostarethe
thousandsofentrepreneursaroundtheworldwhohavetestedthese
ideas,chalengedthem,renedthem,andimprovedthem.Without
theirrelentless—andmostlyunheralded—workeveryday,
noneof
thiswouldbepossible.Thankyou.
Realstartupsinvolvefailure,embarrassingmistakes,andconstant
chaos.Inmyresearchforthisbook,Idiscoveredthatmost
entrepreneursandmanagerswouldprefernottohavethe
realstory
oftheirdailyworktoldinpublic.Therefore,Iamindebtedtothe
courageousentrepreneurswhoconsentedtohavetheirstoriestold,
manyofwhomspenthoursintediousinterviewsandfact-checking
conversations.Thankyou.
Ihavebeengratefulthroughoutmycareertohavementorsand
colaboratorswhohavepushedmetoaccomplishmorethanIcould
haveonmyown.WilHarveyisresponsibleforbothrecruitingme
toSiliconValeyintherstplaceandfortrustingmewiththe
opportunitytotryoutmanyoftheseideasforthersttimeat
IMVU.IamgratefultomyotherIMVUcofoundersMarcusGosling,
MatDanzig,andMelGuymonaswelasthemany
IMVU
employeeswhodidsomuchoftheworkIdiscussed.Ofcourse,
noneofthatwouldhavebeenpossiblewithoutthesupportof
milionsofIMVUcustomersovertheyears.I’dalsoliketothank
DavidMilstone,KenDuda,FernandoPaiz,SteveWeinstein,Owen
Mahoney,RayOcampo,andJasonAltierifortheirhelpalongthe
way.
WealoweSteveBlankadebtfortheworkhediddeveloping
WealoweSteveBlankadebtfortheworkhediddeveloping
thetheoryofcustomerdevelopmentatatimewhenitwas
consideredhereticalinstartupandVCcircles.AsImentionedinthe
Introduction,Stevewasanearlyinvestorinandadviser
toIMVU.
Forthepastsevenyears,hehasbeenanadviser,mentor,and
colaboratortomepersonaly.Iwanttothankhimforhis
encouragement,support,andfriendship.
TheLeanStartupmovementismadeupofmanymore
thinkers,
practitioners,andwritersthanjustme.IwanttothankDave
McClure,AshMaurya,BrantCooper,PatrickVlaskovits,SeanElis,
AndrewChen,SeanMurphy,TrevorOwens,HitenShah,andKent
Beckfortheirideas,support,andevangelism.Severalinvestorsand
venturecapitalistswereearlysupportersandadopters.Iwouldlike
tothankMikeMaplesandAnnMiura-Ko(Floodgate),Steve
Anderson(Baseline),JoshKopelman(FirstRound
Capital),Ron
Conway(SVAngel),andJefClavier(SoftTechVC).
Asyoucanimagine,thisbookinvolvedatremendousamountof
feedback,iteration,andtesting.Ireceivedinvaluable,in-depthearly
feedbackfromLaura
Crescimano,LeeHoman,ProfessorTom
Eisenmann,andSachaJudd.ThanksalsotoMitchKapor,Scot
Cook,ShawnFanning,MarkGraban,JenniferCarolan,Manuel
Rosso,TimO’Reily,andReidHomanfortheirsuggestions,
feedback,andsupport.IoweaspecialnoteofthankstoRuth
KaplanandIraFayfortheirwisdomandfriendship.
Throughouttheprocessofwritingthebook,Ihadthebenetofa
custom-builttestingplatformtorunsplit-testexperimentson
everythingfromcoverdesigntosubtitlestoactualbitsofthebook
(youcanseetheresultsoftheseexperimentsathtp://lean.st).
PivotalLabsbuiltthissoftwareforme;theyarethepremier
practitionersofagiledevelopment.Specialthanks
toRobMee,Ian
McFarland,and—mostimportant—ParkerThompson,whoworked
tirelesslytobuild,experiment,andlearnwithme.
ThanksalsotoIMVUcofounderMarcusGosling,oneofthemost
talenteddesignersIknow,whodesignedthisbook’scover,after
countlessiterations.
Oneofthepremierwebanduserexperiencedesignrms,Digital
Telepathy,
designed
and
built
the
website
for
Telepathy,
designed
and
built
the
website
for
htp://theleanstartup.com,usingtheiruniqueIterativePerformance
Design
process.
It’s
awesome.
Learn
more
at
htp://www.dtelepathy.com/
Iwasextremelyfortunateto
havethesupportofthreelegendary
institutionsatvariouspointsinmyjourney.Muchoftheresearch
thatwentintothisbookwasgenerouslyunderwritenbythe
KaumanFoundation.AtKauman,IwanttoespecialythankBo
FishbackandNickSeguinfortheirsupport.Ispentthepastyearas
anentrepreneur-in-residenceatHarvardBusinessSchool,whereI
enjoyedtheopportunitytotestmyideasagainstsomeofthe
brightestmindsinbusiness.Iamespecialygratefulto
Professors
TomEisenmannandMikeRobertsfortheirsponsorshipand
support,aswelastothestudentsoftheHBSStartupTribe.Ialso
hadtheopportunitytospendabrieftimewithanoceatthe
premierventurecapitalrminSiliconValey,KleinerPerkins
Caueld&Byers,whereIreceivedanin-deptheducationintohow
entrepreneurshipisnurturedatthehighestlevels.ThankstoChi-
HuaChien,RandyKomisar,MatMurphy,BingGordon,
AileenLee,
andElenPao,andtomyoficemateandEIR,CyriacRoeding.
Myresearchteamhelpedmedocumentcasestudies,interview
hundredsofstartups,andlterthousandsofstories.Iwanttothank
MarisaPorzig,whologgedcountlesshoursdocumenting,cross-
referencing,andinvestigating.Additionalcasestudieswere
developedbySaraGaviserLeslieandSarahMilstein.
Traditionalpublishingisacomplicatedandinsularbusiness.I
benetedfromadviceandconnectionsfrommanypeople.Tim
FerrissandRamitSethisetmestraightearlyon.Iamalsograteful
toPeterSims,PaulMichelman,MaryTreseler,Joshua-MichéleRoss,
ClaraShih,SarahMilstein,AdamPenenberg,Gretchen
Rubin,Kate
Lee,HolisHeimbouch,BobSuton,FrankieJones,RandyKomisar,
andJefRosenthal.
AtCrown,theherculeantaskofturningthisideaintothebook
youarereadingfeltoahugeteamofpeople.Myeditor,
Roger
Schol,sawthevisionofthisbookfromtheverybeginningand
shepherdeditthroughtheentireprocess.IwanttoalsothankTina
Constable,TaraGilbride,andMeredithMcGinnisandeveryoneelse
Constable,TaraGilbride,andMeredithMcGinnisandeveryoneelse
whoworkedonmakingthisbookareality.
Thosewhohadthemisfortuneofreadinganearlydraftknowjust
howmuchgratitudeIowetoLaureenRowland,whoprovided
essentialeditorialhelponanunbelievablytightschedule.Ifyou
enjoyedanypartofthisbook,shedeservesyourthanks.
Myadviser,partner,andconsiglierethroughoutthepublishing
processhasbeenmyphenomenalagent,ChristyFletcher.Shehas
theuncannyabilitytopredictthefuture,makethingshappen,and
keepeverystakeholderhappy—alatthesametime.Shetruly
understandsthemodernmedialandscapeandhashelpedme
navigateitscrazywatersateveryturn.AtFletcherand
Company,I
alsowanttothankAlyssaWol,whohasbeenatirelessadvocate
andgatekeeper,andMelissaChinchilo,whoisworkingtobring
thisbooktonewregionsandlanguages.
Iknowitisaclichétosay,
“Noneofthiswouldhavebeen
possiblewithouttheconstantsupportofmylovingfamily.”Butin
thiscase,itissimplythetruth.Myparents,VivianReznikand
AndrewRies,havealwayssupportedmyloveoftechnologywhile
stilinsistingontheimportanceofaliberalartseducation.Without
theirconstantloveandsupport,Iwouldneverhavehadthe
couragetostepintothevoidofentrepreneurshiporhavefoundmy
ownvoiceasawriter.Iknowmygrandparentshavebeen
withme
everystepofthisjourney—theybelieveddeeplyinthepowerof
writingandtooksupremejoyinmysisters’andmyevery
accomplishment.TomysistersNicoleandAmandaandmybrother-
in-lawDov,Icanonlysay:
thankyouforsupportingmealthese
years.
Mywife,TaraSophiaMohr,hasbeenaconstantsourceofjoy
andcomforteverystepoftheway.Shehasexperiencedeverystress,
everyhigh,andeverylow
throughthisverylengthyprocess.Tara,
youareanincrediblybriliant,strong,andcompassionatewoman.
WordscannotexpresshowmuchIappreciateyoursteadfast
support,youroverwhelminglove,andthedailyadventurethatis
ourlifetogether.Thankyou.
AbouttheAuthor
ERICRIESisanentrepreneurandauthorofthepopularblog
StartupLessonsLearned.HecofoundedandservedasCTOof
IMVU,histhirdstartup.Heisafrequentspeakerat
businessevents,
hasadvisedanumberofstartups,largecompanies,andventure
capitalrmsonbusinessandproductstrategy,andisan
entrepreneur-in-residenceatHarvardBusinessSchool.HisLean
Startupmethodologyhas
beenwritenaboutintheNewYork
Times,theWalStreetJournal,theHarvardBusinessReview,the
HufingtonPost,andmanyblogs.HelivesinSanFrancisco.