AlsobyMichaelE.Raynor

TheStrategyParadox

TheInnovator’sSolution(coauthoredwithClaytonM.Christensen)

Disclaimer:Thisworkcontainsgeneralinformationonlyandisnotintendedtobeconstruedasrenderingaccounting,business,financialinvestment,legal,tax,orotherprofessionaladviceorservices.Thisworkisnotasubstituteforsuchprofessionaladviceorservices,norshoulditbeusedasabasisforanydecisionoractionthatmayaffectyourbusiness.Theauthorandpublisher

disclaimanyliability,loss,orriskthatisincurredasaconsequenceoftheuseandapplicationofanyofthecontentsofthiswork.

Copyright©2011byMichaelE.RaynorForewordcopyright©2011byClaytonM.Christensen

Allrightsreserved.

PublishedintheUnitedStatesbyCrownBusiness,animprintoftheCrownPublishingGroup,adivisionofRandomHouse,Inc.,NewYork.

www.crownpublishing.com

CROWNBUSINESSisatrademarkandCROWNandtheRisingSuncolophon

areregisteredtrademarksofRandomHouse,Inc.

LibraryofCongressCataloging-in-PublicationDataRaynor,MichaelE.

Theinnovator’smanifesto:deliberatedisruptionfortransformationalgrowth/MichaelE.Raynor.—1sted.

p.cm.1.Disruptivetechnologies.2.Creativeabilityinbusiness.

3.Successinbusiness.I.Title.HD45.R2962011

658.4063—dc222010052634

eISBN:978-0-385-53167-2

JacketdesignbyJeanTraina

v3.1

CONTENTS

CoverOtherBooksbyThisAuthorTitlePageCopyright

FOREWORD:FromArttoSciencePROLOGUE:TheFive-Percentage-PointSolution

PARTI:PredictionCHAPTERONE:AProblemofPredictionCHAPTERTWO:BetterbyHalf

PARTII:ExplanationCHAPTERTHREE:HowCHAPTERFOUR:IfCHAPTERFIVE:WhenandHowLong

PARTIII:ApplicationCHAPTERSIX:DeliberatelyDisruptiveCHAPTERSEVEN:InvestingforInnovationCHAPTEREIGHT:OrganizingforInnovation

EPILOGUE:CreativeCreation

ACKNOWLEDGMENTS

APPENDIX A: Delimitations and Limitations of theExperimentsAPPENDIXB:ImageIllusionsAPPENDIXC:ExperimentalMethodandStatisticalAnalysisAPPENDIXD:DivisionalRevenuesGeneratedbyDisruptionatJohnson&Johnson

NOTES

FOREWORD

FROMARTTOSCIENCE

byClaytonM.Christensen

Inthisfascinatingbook,MichaelRaynortellsusthattheworldofinvestingtocreatesuccessfulbusinessesisabouttochange.Justastheoriesintheworldofbiology or physics have allowed us to predictably create desired outcomes inmedicine or engineering, Raynor shows here that Disruption promises muchgreaterpredictabilityintherealmofcreatingsuccessfulnewbusinesses.Raynorshows us that there are certain technologies and strategies that succeedmuchmoreoftenthanothers.Heshowsuswhattheyare,whytheywork,andhowtoapplythem.Science—atleastinthisoneinstance—trulyismakingadifferenceinthepracticeofmanagement.TheultimatesignificanceofTheInnovator’sManifestowillberevealedonly

over time. I, however, have high hopes for its longevity and impact becauseRaynor’s work falls very neatly into a well-established pattern for thetransformation of tacit, intuitive knowledge—art, if you will—into codified,well-understood,explicitrules—inotherwords,science.IbelievethatRaynorisplayingacentralroleintransformingthemanagementofinnovationfromanarttoascience.Thiswilltrulybealandmarkwork.Toseethesignificanceofthiscontribution,considerthatintheearlystagesof

any field, our collective knowledge is little more than an assortment ofobservationscollectedovermanygenerations.Therearemanyunknowns,andsotheworkiscomplexandintuitive,andtheoutcomesarerelativelyunpredictable.Only skilled experts are able to cobble together adequate solutions, and theirwork proceeds through intuitive trial-and-error experimentation. This type ofproblem-solving process can be costly and time-consuming, but there is littlealternativewhenourknowledgeisstillinitsinfancy.Creating new, successful innovations still looks very much like this today.

Investment decisions and strategic choices are typically based on intuition;learning, if it happens at all, is a very expensiveby-product of trial and error.

Entrepreneurs and new venture investors alike live a perpetual contradiction,convincedonacase-by-casebasisthattheventuretheyhavejustlaunchedwillsucceed,evenastheycannotescapethefactthat90percentofallnewventures—including theirs—ultimately fail. In such a world, we can make no clearconnectionamongtheattributesofthenewbusiness,theoversightprovidedbythe investors, the management methods of the leadership team, and finaloutcomes.Thatmakesitveryhardtolearnhowtosucceedatinnovation.In the face of this uncertainty, some widely accepted rules of thumb have

emerged.Forexample,amantraformostventurecapitalistsisthatitisfollytomakeinvestmentdecisionsbaseduponthestart-up’stechnologyorstrategy.TheVCshaveconcludedfromtheirtrialsanderrorsthateventhey—thebestintheworld—cannotpredictinadvancewhetherthetechnologyorstrategydescribedinastart-up’sbusinessplanwillactuallywork.Asaresult,theytypicallyassess—intuitively—whether the management team has the intuition to succeed. Ifmembersoftheteamareexperiencedandperceptive,theVCsreason,theycandeveloptherighttechnologyandtherightstrategy—becausetheyandonlytheywill have the instinct to change direction when needed. As far as affectingoutcomesinameaningfulandpredictableway,however,thisapproachranksuptherewith“feedacold,starveafever.”Itislittlemorethananaphorismbasedonselectivememory,theforceofrepetition,andthehopethatatleastitdoesnoharm.Getting beyondmyth requires thatwe first carefully document patterns that

repeat over time. This does not provide any guarantee of success, but it doesprovide at least some confidence that there is a correlation among factors ofinterest. Ultimately these patterns of correlation are supplemented with anunderstandingofcausality,whichmakestheresultsofgivenactionsmuchmorepredictable. Work that was once intuitive and complex becomes routine, andspecific rules are eventually developed to handle the steps in the process.Abilities that previously resided in the intuition of a select group of expertsultimately become so explicitly teachable that rules-based work can beperformedbypeoplewithmuchlessexperienceandtraining.To illustrate, consider the evolution of medical science. At its core, the

problem in medicine historically is that the human body has a very limitedvocabulary from which it can draw when it needs to declare the presence ofdisease. Fever, for example, is one of the “words” through which the bodydeclares that something inside isn’t quite right. The fever isn’t the disease, ofcourse. It is a symptomatic manifestation of a variety of possible underlyingdiseases, which could range from an ear infection to Hodgkin’s lymphoma.Medicationsthatamelioratethefeverdon’tcurethedisease.Andatherapythat

addressesoneofthediseasesthathasfeverasasymptom(asampicillincancureanear infection)maynotadequatelycuremanyof theotherdiseases thatalsohappentodeclaretheirpresenceswithafever.As scientists work to decipher the body’s limited vocabulary, they are

teachingus thatmanyof the thingswe thoughtwerediseasesactuallyarenot.They’re symptoms. For example, we have learned that hypertension is like afever—it is a symptomatic expression of a number of distinctly differentdiseases.Therearemanymorediseasesthanthenumberofphysicalsymptomsthatareavailable,sothediseasesenduphavingtosharesymptoms.Onereasonwhy a therapy that effectively reduces the blood pressure of one patient isineffective in another may be that they have different diseases that share thesame symptom. When we cannot properly diagnose the underlying disease,effective care generally can be provided only through the intuition andexperienceofhighlytrained(andexpensive)caregivers—medicine’sequivalentofWarrenBuffett.At the other end of the spectrum, we define precision medicine as the

provisionofcarefordiseasesthatcanbepreciselydiagnosedandforwhichtheunderlying causes are understood. This makes it possible to develop apredictablyeffective therapy. In thesecircumstances,caregiverssuchasnursesandtechnicianscangiveeffectivecareandatlowercostthanispossibletodayby the best clinicians.Most infectious diseases live here: we have dispositivetests for theirpresenceandwell-understoodandhighlyeffective treatmentsfortheircure.Wecanallbutguaranteeanoutcomeforanindividual;exceptionsarerareandnoteworthy.Notallofmedicinefallsintothe“intuitive”or“precision”category,however.

Thereisabroaddomaininthemiddlecalledempiricalmedicine.Thediagnosisand treatment of a pathology falls into this third categorywhen a fieldhas anincomplete but still very valuable set of causalmodels and validated patterns.The connections between actions and outcomes are consistent enough thatresultscanbeusefully,ifimperfectly,predicted.Whenwereadstatementslike“98 percent of patients whose hernias were repaired with this procedureexperienced no recurrence within five years, compared to 90 percent for theothermethod,” we’re in the realm of empiricalmedicine. Empiricalmedicineenables caregivers to follow the odds. They can generally guarantee theprobabilisticoutcomeonlyforapopulation.WhatmakesTheInnovator’sManifestososignificantisthatitisperhapsthe

firstand inmyviewthemostsignificantandsuccessfuleffortyet tomove thefield of innovation from the intuitive stage into the world of empiricalmanagement. Building upon groundbreaking research at Intel Corporation,

Raynor has quantified the improvements in predictive accuracy and survivalratesthatarepossiblethroughthecarefulapplicationofDisruptiontoearly-stagebusinesses. He has elaborated upon particular elements ofDisruption inwaysthatmakeclearwhenandhowthetheorycanbeapplied.Andhehasprovidedframeworks for its application that will enable most any business to reap thebenefitsthatDisruptionmakespossible.Achieving suchanoutcomemeans that this isnotyour typicalmanagement

book. There are no “just-so” stories attributing the success of the latest bottlerocket to a newbuzzword. Instead, youwill find the careful collectionof realdata,consideredandcircumspectanalysisthatrecognizesshortcomingswithoutbeing paralyzed by them, a rigorous and reflective treatment of some of thechestnuts of popularmanagement thinking, and a genuine appreciation for thechallengesofapplyingreal theoryintherealworld.Youwillhavetoreadthisbookcarefullyandreflectuponitdeeply.Butitwillbeworthit.AsIhavesaidelsewhere,myadmirationforMichaelRaynorhasnoend.The

integrity of Disruption theory has improved substantially sinceMichael and IcoauthoredTheInnovator’sSolution,andmuchofthatimprovementIattributetomy continued collaborationwith him. I love just to sit in his presence andlisten as his magnificent mind goes to work on the complicated puzzles ofmanagement.ThoughIhaveabusylife,forMichaelRaynorIalwayshavetime.Ihopethatyouwillenjoybeingwithhimasyoureadthisbook.

ClaytonM.ChristensenistheRobertandJaneCizikProfessorofBusinessAdministrationattheHarvardBusinessSchoolinBoston,Massachusetts.

PROLOGUE

THEFIVE-PERCENTAGE-

POINTSOLUTION

“Disruption,” used in a technical sense, is a theory of innovation—of howparticular types of newproducts and services, or “solutions,” come to achievesuccessordominance inmarkets,oftenat theexpenseof incumbentproviders.Disruptionwas discovered byClaytonChristensen, a professor at theHarvardBusiness School, in 1992when hewas a doctoral student there. (When using“disruption” or its cognates in a technical sense I will use an uppercaseD.)Christensen’s 1997 best-selling book,The Innovator’s Dilemma, was the firstpopular expression of his ideas. Christensen and I collaborated on TheInnovator’s Solution, published in 2003. At least seven more books andhundreds of articles have been published since then exploring the theory’simplications in different contexts.1 It is in widespread use as an organizingprincipleforinnovationatorganizationsaroundtheworld.Manywhohaveusedithavecrediteditwithasignificantroleincreatingsuccessfulnewbusinesses.And yet, thanks to the confusing world of applied management research,

Disruption is still seen bymany as “just another theory.”One newbook afteranothercascades into themarketplaceof ideas,attempting toexplain the latestsuccessstoryorallegedlyrevolutionaryphenomenonwithanewlycoinedtermand a fresh set of case studies as supporting evidence. How are practicingmanagers to decide which frameworks, theories, approaches, or 2×2s areapplicabletotheircircumstancesandtrulyusefultothem?HowisonetoknowwhethertouseDisruptionorsomethingelsetonavigatethroughthechallengesassociatedwithinnovatingsuccessfully?

EXPLANATIONANDPREDICTION

Onewaytosortoutwhatisusefulandaccuratefromthenoiseistotakeapagefromthephilosophyofscience.Inhis2010bookNonsenseonStilts,MassimoPigliucci points out that the type of evidence one adduces in support of apositiondependsinlargepartonthesortofargumentonehopestomake.2If,forexample, a theory is intended merely to be useful—that is, instrumental inachieving a desired outcome—then one needs to demonstrate predictiveaccuracy. In other words, theories are useful if they tell us what will happennext,andthemostusefultheoriesaresimplytheonesthatdothatbest.Assessingpredictiveaccuracyrequiresverycarefullycontrolledandrepeated

experiments and at times a remarkably high tolerance for experimental error.Physics, the queen of the hard sciences, has risen to this challenge time andagain, and as a result that discipline’s long-term project has made enormousprogress.Wehaveabandonedtheoriesofphlogistonandtheetherforquantummechanics and the standardmodel of elementary particles thanks to a carefulaccumulationofdataunder increasinglywell-controlledconditions.It isa longandcomplexchainfromformulatinga theorytocontrolledexperiments testingthe theory’s propositions to usefulness in the everydayworld ofmiddle-sized,middle-distance objects. But every link holds (well enough) for the predictivepowerofphysicstomanifestitselfinmanyandrepeatedsuccessfulapplicationsinfieldssuchasengineering.Predictivepowerestablishesthatatheoryisuseful,butitdoesnotprovethata

theory is true; a true theory explains reality. Galileo, for example, would notlikelyhavebeen in suchhotwaterwith theCatholicChurchauthoritiesofhisday ifhehad saidmerely that theheliocentricviewof the solar systemwasausefulmethodforpredicting thefuturelocationsoftheplanets.Hegothimselfin trouble by claiming that it explained why the planets moved as they did,namely,becausetheplanetsreallydoorbitthesunandnottheearth.Predictionandexplanation requireverydifferent sortsofevidenceand rules

of inference. Experiments to establish predictive power admit of sometimessignificantmeasurementandothersortsoferror.Evenunderthemostcarefullycontrolled conditions there remains a great deal that is, well, uncontrolled;indeed, experiments that come out too close to perfect are often suspected ofhaving been fudged.We insist that the theory be specified in advance of theexperiments, rather than creating our theory after the fact: our unconscious

biasesmight leadus to create a theory that fitsourdataperfectly, and sinceadatasetisusuallyonlyasample,thiskindofinterpolationunderminesatheory’sbroaderapplication.Theories“win”basedonthestatisticalsignificanceoftheirresultsoveranumberoftrialsandtheirparsimony—theirabilitytoexplainthebroadestrangeofoutcomeswiththefewestandsimplesttheoreticalconstructs.Incontrast,explanatoryframeworksaddressafixedandunchangingpast.We

cannottestaproposedexplanationofwhathasalreadyhappenedbyturningbacktheclockandseeingifhistoryplaysoutthesamewayagain.Wemustthereforedecidewhatwinsbasedon the completenessof the explanation, theweightofcircumstantialevidence,andwhereverpossiblewhatPigliuccicallsa“smokinggun”: one or two critical facts that no other competing theory can plausiblyaccountfor.So, for example, how do we know that an asteroid impact explains the

extinctionofthedinosaurssixty-fivemillionyearsago?Wecanreasonablyinferfrom what we know about asteroid impacts in general that an asteroid ofsufficient size could trigger amass extinction.What we need to show is thattherewasanimpactbyanasteroidofsufficientsizeatabouttherighttimeandthatthepatternofextinctionsisconsistentwiththeexpectedconsequences.Overthe years enough circumstantial evidence has accumulated to convince mostinformedobservers that thiswas thecase.Forexample, there isacraterof theright size in the floor of the Gulf ofMexico (which was also an ocean backthen),alongwithevidenceofdevastatingtsunamisalongancientcoastlines.Wealsohaveatelltalelayerofiridiumoreofjusttherightconcentrationlaiddownatjusttherighttimeinrockstrataaroundtheworld.Finally,competingtheories—suchastheriseofegg-eatingmammalsorclimatechangeduetoeccentricitiesin the earth’s orbit—cannot account for the fact that the dinosaurs wereextirpatedsimultaneouslywithagreatmanyplantandmammalspeciesaswell,norfortherapiditywithwhichthemassextinctionsoccurred.Due to these differences in purpose and hence evidence, establishing

explanatory power says nothing about a theory’s predictive power. That thedinosaurswerewipedoutbyanasteroidimplieslittleaboutwhatwillcausethenextmassextinction.Itjustturnsoutthatanasteroidstrikecausedthatone.Considernowthelastmanagementbookyouread.Whatkindofevidencedid

itprovideinsupportofitscentralclaims?Itverylikelyreliedforevidenceonananalysis of case studies, and out of that analysis emerged a frameworkpurportingtoexplainwhyeventsturnedoutastheydid—whyagivencompanysucceededorfailedorwhyagivenproductwasahitoraflop.Veryoften, however, the explicit claim is that theprinciples that havebeen

extractedfromananalysisof thepastcanbeused toshapefutureoutcomes in

desiredways.Typically,authorsseemtobelievethatcase-studyevidencealonesupportsprescriptiveclaims.Inotherwords,mosteverymanagementbookIamfamiliarwith—andcertainlymostof thebest sellers—makespredictive claimsbasedonexplanatorypower.Whetherdeliberateornot,itisamostunfortunateandpotentiallydamagingformofconceptualbaitandswitch.Is there any way to avoid this, though? After all, the subject matter of

managementresearch—actualorganizationsfunctioningintherealworld—doesnot lend itself to thekindsofcarefullycontrolledexperiments thatallowus totest predictive accuracy in the usual ways. Perhaps we can do no better thansimplytoinferpredictivepoweronthebasisofexplanatorypersuasiveness.

THREEOBJECTIVES

Idisagree.ThefirstobjectiveofthisbookistodemonstratethatDisruptionhastruepredictive power. I hope to show this usingwhat is formany people themost persuasive evidence there is when it comes to prediction: controlledexperiments.MyhopeisthatyouwillfindthesedatasufficientlycompellingtoconcludethatDisruptionisuniqueinhavingevidencetosupporttheclaimthatitisgenuinelyuseful.Second,IwillmakethecaseforDisruption’suniqueandsuperiorexplanatory

power. Iwill layout adefinitionofDisruptionprecise enough thatDisruptiveinnovations can be accurately identified in advance of knowing how theyultimately fare and their results in the marketplace explained more fully andparsimoniously than by any other theory.To the extent I succeed, I hope youwillconcludethatDisruptionisfarmorethanmerelyausefulperspectivebutisinfacttrue.Finally, Iwilloffer some thoughtsonhowonecangoaboutapplying these

conceptstogreatesteffectattheleastexpense.Totheextentthisthirdobjectiveisachieved,IhopeyouwillconcludethatDisruptionispractical.AndifIcanconvinceyouthatDisruptionisuseful,true,andpractical,Iwill

gofurtherandhopethatyouwillwantandbeabletouseit insupportofyourinnovationefforts.

PREDICTION:CHAPTERS1AND2

Chapters 1 and 2 describe the design and results of carefully controlledexperimentstestingthepredictivepowerofDisruption’scentralclaims:thataninnovation has the best chance of success when it has a very differentperformance profile and appeals to customers of relatively little interest todominantincumbents,andtheorganizationcommercializingitenjoyssubstantialstrategic and operational autonomy. In contrast, attempts to introduce better-performingsolutionstargetedatcustomersvaluedbysuccessfulincumbentswillfail. To test these propositions I use a portfolio of forty-eight new businessproposalsthatreceivedseedfinancingfromIntelCorporation.Tosummarizetheresults,testsubjectsimprovedtheirpredictiveaccuracyby

as much as 50 percent when they applied Disruption theory to make their

choices. Specifically, in the actual portfolio of funded businesses just over 10percent survived. The portfolio chosen by MBA students who did not useDisruption theory had a similar survival rate, while students usingDisruptiontheorytopickwinnersbuiltaportfoliowithasurvivalrateofuptoslightlymorethan 15 percent.That five-percentage-point gain is a 50 percent improvement.(More recently, Intel reports that the survival rateof its fundedbusinesseshasincreased,inpartduetotheapplicationofDisruptiontheory.)Ofcourse,neither thedatanor theexperimentaldesignisperfect(andIwill

havemoretosayabouttheprecisenatureoftheimperfectionsofthisworklateron),butperfection is thewrongbenchmark. In themortal realm,all success isrelative, and the most important question is not “What are the flaws of thisdesignandthesedata?”but“Arethisdesignandthesedatabetterthanwhatyouhaveseenelsewhere?”Notealso thatIamnotclaimingthatIhaveshownthatDisruptiontheoryis

betterthansomeothertheory.Rather,IamclaimingthattheevidenceinsupportofDisruption theory’s predictive power is better than the evidence supportinganyotherrelevanttheory’spredictivepower.Toseethedifferencebetweenthesetwoclaims,considertestsfortheefficacy

ofnewpharmaceuticaldrugs.ImaginethatDisruptionisadrugthatpurportstotreat a given condition, and some other theory is a different drugmaking thesame claim. The evidence in these first two chapters supports the claim thatDisruption actually “treats the condition”: it improves predictive accuracy. IhavenotshownthatDisruptionworksbetter thananyotherdrug; that requirescomparingtherelativeeffectivenessoftwodrugs.Atthesametime,however,asfarasIknownoonehasshownthatanyotherdrugactuallytreatstheconditionatall.What I hope to convince you of at the outset, then, is that Disruption can

claimmore legitimately thananyother theory tomakeyoubetter thanyouarewithrespecttoonecriticallyimportantdecision:assessingwhichbusinesseswillliveordie.

EXPLANATION:CHAPTERS3TO5

A common challenge in research of any kind, and certainly in the field ofapplied management, is determining the extent to which one can “generalizebeyond the sample.” For example, if someone does a study on large publiccompanies,dothefindingsapplytosmall,privatelyheld,family-runbusinesses?Toextendourpharmaceuticaldrugtestinganalogy,considerclinicaltrialson

adrugthattreatshighbloodpressure.Suchtrialstypicallyincludethousandsofpeopleandyearsofobservationinordertodeterminewhetheranewdrugissafe(doesnoharm)andeffective(actuallyhelpsinthedesiredway).Assumeforthesakeofargument that thedrugprovedsafeandeffective,but it turnedout thattherewerenosubjectsnamedPhil.AdministeringthedrugtopeoplenamedPhilwith theexpectationofsafeandeffectiveoutcomes isgeneralizingbeyond thesample. One is therefore open to the possibility that the drug could have adifferenteffectonpeoplenamedPhilthanitdidonthoseobservedinthestudy.Thankfully, we can claim a credible understanding of what will happen in

circumstanceswehavenottesteddirectlyifwehaveacorrectunderstandingofwhyresultsturnoutastheydo.Inthepharmaceuticalexample,ifweunderstandthemechanismsofactionforaparticulardrugandweknowwithahighdegreeofcertaintythatbeingnamedPhilhasnomaterialimpactonadrug’seffect,thenwearejustifiedingeneralizingbeyondthesample.If,however,thereareotherattributes thatwebelievemightaffect thedrug’sefficacy—say,apatient’ssexorageorbeingdiabetic—inwaysthatwedonotfullyunderstand,thenwearenotjustifiedingeneralizingbeyondthesample.Inreality,asisoftenthecase,suchjudgmentsarenotbinary:oneismoreor

lessjustifiedingeneralizingbeyondthesampledependingonthesample,whatonehopestogeneralize,andhowfarbeyondthesampleonewishestogo.Inthelarge public/small private company example, we might ask what therelationships are between behaviors and outcomes being investigated andwhethertherearemeaningfuldifferencesbetweenthesetypesofcompaniesthatmightaffecttherelationshipsweobserveinoursample.Astudyaboutprocessesfor implementing a quality-management system might generalize across suchdiverse companies much better than a study on governance processes, forexample,sincethepublicorprivatestructureofacompanyhasadirectbearingontherelevantlegalandregulatorygovernancerequirements.Withthisinmind,theextenttowhichwecanreasonablyexpectthepredictive

powerofDisruptiontobeevidentincontextsthatwerenotdirectlytestedintheexperimentsturnsonwhetherDisruptioncanaccountforitspredictivepowerbyspecifying when it should be applied and providing sufficiently powerful andcompellingexplanationsforwhyitworks.Inotherwords,thegeneralizabilityofdemonstratedpredictivepowerisafunctionofexplanatorypower.The experiments in chapters 1 and 2 test whether Disruption improved the

ability of MBA students to predict the survival of very early-stage businessplans.Chapters3through5exploretheextenttowhichothertypesofpeopleindifferentcircumstancescandoanythingwiththesefindingsbymakingthecasefor Disruption’s explanatory power. Unlike the tests of predictive power, this

entails a direct comparison of the explanatory power of Disruption with theexplanatory power of competing theories when accounting for specificoutcomes.The test case, explored in chapter 3, is Southwest Airlines, for although

Southwest has been analyzed seemingly ad nauseam, the signal feature ofSouthwest’s performance—its nearly twenty-year run of slow growth anddecliningprofitabilityfromtheearlyseventiestotheearlynineties,withasharpturnaround and a decade of record-setting growth, increasing profitability, andshare-price appreciation—has had no parsimonious explanation. Disruption,however, explains notmerelywhy Southwestwas successful but alsowhy itsgrowthoccurredpreciselywhenitdid.IwillarguethatDisruptionexplainsthesalientfeaturesofSouthwest’sperformanceinawaythatnoothertheorydoes,and inaway thatwouldhavemade itpossible topredictSouthwest’ssuccess.This is the sort of “smoking gun” required to establish that Disruption is therightexplanation,ratherthanmerelyaplausibleone.Now, proving that Southwestwas aDisruptor says nothing about any other

company. Nor am I claiming that every successful innovation is a Disruptiveone.Sochapter4describeshowtodeterminewhetherornotagivenopportunityhaseven thepotential tobeDisruptive.For example, I explainhowso far thehotel industry, strategy consulting, and the discovery of new patentablepharmaceuticals have been immune to Disruptive innovation, not (to use aphraseyouwillseerepeatedly)asamatteroftheoreticalnecessitybutmerelyasa matter of empirical fact. The key message here is that an integral part ofDisruptiontheoryisthecriteriafordeterminingwhenitisapplicable.HavingdefinedthecircumstancesunderwhichDisruptionispossible,chapter

5addresseshowtoassessthetimingandextentofDisruption.Forexample,whydid Disruption take so long in the automotive sector (Toyota’s rise to globalleadership took almost seventy years) and so quickly in telecommunicationsequipment (Cisco was an industry leader less than fifteen years after goingpublic).Chapter5explainswhytheseDisruptionsplayedoutastheydid.Thissecondsectionmakesthecaseforgeneralizingbeyondtheexperimental

sampleandsuggeststhatDisruptioncanbeusedtodomorethanmerely“pickawinner.”For example, thanks to its combinationofpredictive andexplanatorypower,Disruptioncanbeapplied:

If you are an investor: to pick with greater accuracy which businesseshavethebestchanceofsurvival.Thisisthemostdirectapplicationofthe

experimentalresults.Ifyouareanentrepreneur:toshapeyourideasandyourstrategysothatyour newbusinesses have a better chanceof surviving, getting additionalfunding,andultimately thriving.Since lookingatanewventure from theperspectiveoftheentrepreneurisjusttheotherendofthesituationfacedbythe investor, this is perhaps the most direct extension of Disruption’sapplicability.Inshort,ifyouunderstandwhatmakesacompanysuccessfulfrom an investor’s point of view, you have a better shot at building abusinesswiththosecharacteristics.If you are amanager trying to growan existingbusiness: to improvematerially your ability to identify or create opportunities to innovatesuccessfully. What makes Disruptive innovations successful is theirtrajectory of performance improvement: the ways and rate at which aproductorservicegetsbetter.ItisbecauseDisruptionallowsyoutoassessanddeterminethesevariablesthatitmakesforbetterinvestmentdecisions.Consequently, if you want to improve your chances of success in anexistingbusiness,DisruptionprescribesthatyouguideyourowninnovationeffortsinwaysthatmakeyouDisruptivetootherswheneverpossible.IfyouareincorporateM&A:toidentifyviabletargetsandmanagetheminways that are likelier to create value. Althoughmaterially different inimportantwaysfromlaunchinganewbusinessfromwithinanestablishedcompanyorpilotingagoingconcern,acquiringanexisting firmdemandsthat you think carefully about the strategy you hope to advancewith theacquisition.Disruptiontheoryprovidesawaytothinkabout thisproblem,with important implications for how tomanage the integration process inparticular.

At thesametime,Disruptionisnota theoryofeverything.Thereare lotsofotherquestionsyouwillhavetoanswernomatterwhichoftheserolesyoufill.For example, as an investor, you likely have to worry about the risk/returnstructureofyouroverallportfolio.Ifyouareanentrepreneuryoulikelyhavetoworryabouthowtoraisecapital.Ifyouaremanaginganexistingbusiness,youprobablyhavetoworryaboutorganizationalpoliticsandthechallengesofhead-to-head competition in your coremarkets. If you are in corporateM&A, youlikely have to worry about how best to finance the deal and realize costsynergies.Theseare importantquestions,butDisruptiondoesnotbeardirectlyon them.WhatDisruption can do ismaterially and significantly contribute toyouroveralllikelihoodofsuccess.

APPLICATION:CHAPTERS6TO8

Whatever scientific rigor and theoretical elegance might characterizeDisruption, the proof of the pudding is in the eating. And so how to applyDisruptionsuccessfullyisaddressednext.Chapter6isanexplorationofhowDisruptioncanbeusedtoshapeaspecific

product innovation. We follow the evolution of what is now Johnson &Johnson’sSEDASYS™automated sedation system froman early-stagepartialequity stake in a small start-up to a commercialized product aimed atrevolutionizing a wide and increasing range of surgical procedures the worldover.Itisafactthatnon-Disruptiveinnovationscansucceedandthatbreakthroughs

by new entrants sometimes revolutionize an industry—something thatDisruption theory cannot account for. Consequently, chapter 7 explores theimplications of deliberately pursuing this sort of unexpected (to Disruptiontheory,atleast)successforspecificmanagementprocesses.Highlightingthekeysuccess factors, probability of success, magnitude of initial investment, timehorizon, requisite autonomy, and connections to the established business foreach type of success should be helpfulwhen deciding howmuch to invest indifferent typesof innovation. Inotherwords,wherechapter6exploreshowtouseDisruptiontoshapeasingleproject,chapter7looksathowDisruptionmightfitintoabroaderportfolioofinnovations.Finally,chapter8takesaprocessperspectiveontheapplicationofDisruption.

IsDisruptionatheorythatcanbepluggedintoexistingwaysofthinkingaboutandfosteringinnovation,orisafundamentalshiftinmind-setrequiredtomakethe most of what Disruption implies? The claim here is that the existingparadigm of innovation is evolutionary (variation, selection, retention) and,despitetheexhortationto“failfast,”isunavoidablyprofligate.Disruptionadmitsofadifferenttack:beginwithaclearfocusonareasripeforDisruption;shapeideassothattheyareconsistentwiththeprescriptionsofthetheory;andpersistinthepursuitofaDisruptivestrategy,learningandadaptingalongtheway.Theexamplesandtools in thesechaptersareintendedtostartyou—whether

youareaninvestor,anentrepreneur,amanager,oracorporateM&Astrategist—downtheroadtousingDisruptioneffectively.

HOWMUCHISENOUGH?

TheMBAstudentsintheexperimentsimprovedtheirpopulation-levelpredictiveaccuracybyuptofivepercentagepoints.Thatdoesnotmeanyoucanexpecttodothesame.3Whatdotheseresultsmeanforyou,then?There are at least two questions worth asking yourself as you answer this

question.First, istheevidenceIprovidesufficienttosupportmyconclusion?Ihave attempted to make my case for Disruption’s predictive and explanatorypowerwith as full an accounting of its shortcomings as I am able to provide.Youmightfindstillotherflaws.Iwouldencourageyou,however,toassessthesignificance of these shortcomings in light of the evidence supporting claimsmade by other investigators or, for that matter, your current views aboutinnovation.Without some sort of critical parity there is a danger that onewillendupholdingon toexistingbeliefsnotbecause theyarebettersupportedbutonly because they are existing beliefs. Consequently, whether you personallyshouldaccepttheclaimsmadehereandaddDisruptiontoyourarsenalofideasdependsnoton theobjectivemerits ofmycasebutonhowwellmyevidenceandmyargumentcomparewiththefoundationsofcompetingviews.Second, even if you believemy findings, are theymeaningful?After all, a

bump in the survival rate of a portfolio from 10 percent to up to 15 percentacrossapopulation isnoguaranteeof riches foryou,personally,onyournextendeavor. If I could credibly make such a promise, I would not sell you theknowledge. But five percentage points is still a 50 percent increase over thebaselinesurvivalrateof10percent.Putting those five percentage points in a broader context, it is worth

rememberingthatevenphysics—soimpressiveinitspredictiveandexplanatorypower—is a long way from having everything figured out. In addition to thelong-standing difficulties of reconciling quantum mechanics and generalrelativity, current thinking is that we actually do not understand what theuniverseismadeof.Galaxies are rotating so fast that the gravitational force of the stars within

themisinsufficienttokeepthosegalaxiesfromflyingapart.Toaccountfortheircoherence,physicistshaveinvokedthenotionof“darkmatter,”whichisreallyjustalabelforwhateveritisthatisgeneratingtheadditionalgravitationalforceunaccounted for by the mass of the stars. At the same time, the universe isexpanding,notcontracting,which iswhat it shouldbedoing thanks toall thatdarkmatterthatissupposedlyoutthere.Sotocounteracttheeffectsofthedarkmatter, cosmologists have ginned up “dark energy,” which is whatever is

overcomingthedarkmatterandpushingtheuniverseoutward.Whenyouput it all together, according to current estimates, theuniverse is

made up of 24 percent darkmatter (whatever that is), 72 percent dark energy(whatever that is), and only 4 percent matter—the bit we actually think weunderstand, putting aside the schism between quantum theory and generalrelativity,ofcourse.4Andyet,withourarmsbarelyaroundbarely4percentoftheuniverse,lookwhatwehavebeenabletoaccomplish.Maybefivepercentagepointsisprettygood,afterall.

PARTIPREDICTION

CHAPTERONE

APROBLEMOFPREDICTION

Ifthepurposeofatheoryistoinformourchoicestoday,wemustdemandmore than compelling explanations of the past. For a theory to have alegitimateclaimonourallegiancetheremustbeevidencethatitimprovesourabilitytopredictfutureoutcomes.

Creatingandbackingwinningbusinesses isbyallaccountsa low-probabilityendeavor. Far more new businesses fail, or at least do little better than limpalongmired inmediocrity, than actually break away from thepack and createrealwealth.Thereismoretothisstatementthansimplythenecessarytruththatonly10percentofallbusinessescanbeinthetop10percent:thebestbusinessestendtodofabulouslywell,whilemostoftherest,iftheysurviveatall,generatereturns that are embarrassingly small in comparison.5 We have becomecollectively resigned, it seems, to the notion that successful innovation isunavoidablyunpredictable.Despite the challenges and the longodds, there is no shortageof players in

this great game.Hedge funds and venture capital partnerships channel capitalinto thebusinesses theyfeelwillsucceed.Manycorporationsmaintain internalventure functions for strategic purposes, some seeking to create ecosystemsaroundacorebusinessortostakeaclaimtopossiblenewgrowthopportunitiesinadjacentmarketsortoestablishalineofdefenseagainstpossibleusurpersofavaluableentrenchedposition,tonameonlythreepossibleobjectives.Take,forexample,IntelCorporation,bestknownforitssignificantroleover

thelastthirtyyearsintheglobalmicroprocessorindustry.In1998Intellaunchedthe New Business Group (NBG) in order to coordinate and more effectivelymanage the company’s attempts to diversify beyond the microprocessorindustry.6WithinNBG,approximately$20millionwasearmarkedfor theNewBusiness Initiatives (NBI) group, which had the remit to identify, fund, anddevelop new businesses thatwere especially far afield, such as Internet-basedbusinesses and consumer products. NBI’s mandate included exploring newtechnologies,newproducts,newmarkets,andnewdistributionchannelsandhadaninvestmenthorizonoffivetotenyears.NBIoperatedasalargelyautonomousunitwithinNBG.Unliketherelatively

formal and structured annual planning and budgeting processes that drovesustainedsuccessinthemicroprocessorsegment,NBItypicallycommittedonlyseed capital to new business ventures, ramping up its level of commitment asvarious strategic and financialmilepostswere reached. In addition, leadershipexplicitly accepted the inherent unpredictability of incubating new businessesalongwithanunavoidableimplicationofthatuncertainty:thatsomeandperhapsmanyoftheventuresthatwerelaunchedcouldfail.IntelOpticalLinks (IOL)wasoneofNBI’s investments.ThomasThurston,

thenanattorneyinhismidtwentieswithanMBAandlawdegree,joinedIOLin2005, excited at the prospect of helping launch a new venture inside anestablished company. Although successfully incubated, IOL was sold offfollowingIntel’sbroaderdivestitureofopticalcomponentandcommunicationsbusinesses.However,Thurston’scuriositywaspiquedbythisinitialexposuretotheinternalventuringprocess:hewantedtounderstandbetterhowInteldecidedwhichinitiativestosupportandwhy.Something in excess of seventy business proposals are explored by NBI’s

investmentdirectorseachyear.Theyworkwitharangeofpeopleandsources,both inside and outside Intel, to determine the potential of a given idea. Theconstant challenge is to find the “diamonds in the rough”—the concepts thathavewithinthemtheseedsofsustainablesuccessandperhapsgreatness.Itisaninherentlyriskyundertaking,andtheonlywaytoavoidfailureentirelyistodonothing,whichofcoursereducesone’schanceofsuccesstozeroaswell.It is this unavoidable uncertainty that leadsmany observers to prescribe an

investmentstrategybasedon“rapidfailure”:thewillingnesstoattemptasmanydifferentinitiativesaspossiblewithaneyetolearningwhatdoesnotworkastheinevitableprerequisitetodiscoveringwhatdoes.InIntel’sworld,however,bonefideinitiatives—thekindsofeffortsthatactuallyteachyousomethinguseful—can get very expensive very quickly. NBI executives are therefore forced tomake difficult tradeoffs between the need to husband their investment capitalandtheriskofoverlookingthenextblockbusterproductorservice.For present purposes, the salient features ofNBI’s investment processwere

the Seed Approval Meeting (SAM) and Business Approval Meeting (BAM).Proposals thatwereapprovedat theSAMreceived fundingofseveralhundredthousanddollarstotypicallylessthan$1million,withanupperrangethatrarelyexceeded$2million.Thisallowedateamtogetbeyondtheideastageandfleshout a business plan, perhaps by developing a prototype, collaborating withpotential customers, doing market research, and so on. BAM funding wascontingent on having demonstrated an increased level of viability and broughtwith it investment capital that ranged from severalmillion dollars to in some

cases as much as $20 million. Ultimately, NBI’s goal was to transition orgraduateonenewbusinessopportunityperyear toanexistingornewbusinessunitwithinIntel.(Noteveryventurehadtopassthroughbothstagesofapproval:some ventureswere graduated directly from SAM to an operating division inlightoftheirstrongperformance.)Inteltakesaveryrigorousapproachtounderstandingcompetitors,technology,

customers, market structure, and a host of other variables when analyzingopportunitiesforgrowth.UnfortunatelyforIntel,andeveryoneelsewhoseekstoinnovateinordertogrow,therearenodataaboutthefuture,andsothereoftenremained many important but unanswered questions. Consequently, well-informed,experiencedexecutivescouldlookatthesameopportunityandcometodifferentconclusionsaboutthatventure’schallenges,financialpotential,andsoon.Worse,onlywhenaventurewasfundedcouldthemeritsofthedecision-makingprocessemployedbeassessed, since if somethingwas turneddown, itrarelygotfundedviaotherchannels,andsotheopportunitycostofpassingonwhatwouldhavebeenawinnerwasalmostalwaysincalculable.Thurstonundertookaforcedmarchthroughthepopularmanagementresearch

intoinnovationinsearchofamorenearlyrules-basedapproachinthebeliefthat,given the importance of the subject and the wealth at stake, any frameworkholdingevenascintillaofadvantageovertheotherswouldbereadilyidentified.YetThurstondiscoveredthatinsteadofavibrantmarketplaceofideaspopulatedby challengers seeking to unseat the reigning champion, the agora wheretheoreticaldominanceisestablishedischaracterizedbygeneraldisarray.Therewereagreatmanyframeworkssupportedbycompellingevidence,yetwhentheyconflictedandcounseleddifferentcoursesofaction,therewaslittlebasisintheevidence to guide someone in choosing one approach over the others. Whendifferentapproachesdidnotconflict,itwasdifficulttotreatthemascumulativeand attempt to follow the sum total of their collective advice, since doing soresultedinaparalyzinglylongto-dolist.7In lightof this theoreticalcacophony, inall likelihoodNBIexecutivesmade

their choices in largely the same way most early-stage investors make theirchoices:dothebestyoucanwiththedatayouhaveavailable,whilenecessarilyrelying on your experience and your wits to fill in the sometimes significantgaps.Theverybestpractitioners typicallydoall theycantocreateasolidfactbase, but personal judgment generally figures prominently inmaking the finalchoice.8 It is simply the nature of the beast that evaluation criteria differ frompersontopersonandprojecttoproject.ThurstonrecountsthatatNBI,thismeantthat sometimes the emphasis was on technology, sometimes on managementexpertise, sometimes on the promise of themarket opportunity, sometimes on

thestrengthof linkageswithIntel’scorebusiness.It isaprocess thatseemstohaveservedIntelwell, for there isnoreasonto thinkthat itsachievementsareanythingotherthanrepresentativeoftheverybesteffortsinthisspace.Theprevalenceof thissortofapproachisanunderstandableconsequenceof

therelianceofpopularmanagementresearch into innovationonposthoccase-study evidence to support its claims. What Thurston was looking for wasevidencesupportingpredictiveaccuracyinadditiontotherequisiteexplanatorypower.Andnotheoryhecouldfindprovidedboth.

CLOSE,BUTNOCIGAR

Christensen’sfirstbook,TheInnovator’sDilemma, introducedtheworldtothenotionof“disruptive technology.”Christensendescribedhow large, successfulincumbentorganizationsinalltypesofindustriesweretoppledbymuchsmallerstart-ups. Entrants typically succeeded by developing solutions for relativelysmallandunattractivemarketsthatwereofessentiallynointeresttosuccessfulincumbents. These constituted the entrants’ “foothold” markets. Sometimescustomers in these footholdmarketswere quite happywith inferior butmuchless expensive solutions; sometimes they required solutions with a vastlydifferent performance profile. Either way, entrenched players, focused on theneeds of their established customers, proved systemically unable to devoteinvestmentfundstothosemarkets.Incontrast,drivenbytheirdesiretogrow,theentrantswerestronglymotivated to improve their initialofferings inways thatwould allow them to compete effectively for the larger, more lucrativemainstreammarkets.Thiswastheentrants’“upmarketmarch,”andentrantsthatmarchedupmarketsuccessfullyeventuallycapturedthecustomersthathadbeentheincumbents’lifeblood.Christensen observed that when entrants attacked successful incumbents by

adoptingtheincumbents’modelsandtechnologicalsolutions,theytendedtofail.Theytendedtosucceedbycombiningabusinessmodelsuitableforarelativelylessattractivemarket—theentrants’ foothold—withanability to improve theiroriginalsolutionsinwaysthatallowedthemtoprovidesuperiorperformanceina manner incumbents were unable to replicate—the upmarket march.Christensencalledtheunionofthesetwoelementsadisruptivestrategy.Thearchetypal illustrationof thisphenomenon isChristensen’s all-inclusive

studyofinnovationandcompetitionintheU.S.diskdriveindustryfrom1976to1994. In themidseventies, companies such as Storage Tech andControlDataweremaking fourteen-inchdiskdrives formainframecomputermakers.Thesecompanies,amongthemAmdahlandUnisys,wantedStorageTechandControlDatatoinnovate:greaterstoragecapacity,fasterdata-retrievaltimes,andlowercostspermegabyte.Whenminicomputerswere first brought tomarket by start-ups such as Sun

Microsystems and Hewlett-Packard, they required very different disk drives:smaller, more modular, and less expensive. To achieve these outcomes, disk-drivemakers found theywouldhave to reducestoragecapacity, increasedata-

retrievaltimes,andaccepthighercostspermegabyte.Theresult,theeight-inchdisk drive,was close to the antithesis ofwhatStorageTech andControlDatawould countenance as an innovation; it was, if anything, a technological stepbackward in the interest of serving a small and highly uncertain newmarket.Thatopenedthedoorforstart-updrivemakerssuchasMicropolisandMaxtortodevelopsomethingthatwastechnologicallytrivialtoStorageTechandControlDatabutstrategicallyimpossibleforthemtolaunch.In the short run, no harm done: Storage Tech and Control Data went on

printing money in the fourteen-inch disk-drive market while Micropolis andMaxtor eked out a living selling technically inferior eight-inch disk drives tosmallminicomputermakers.But thenKryder’s law—thedisk-drive equivalentofMoore’s law inmicroprocessors—asserted itself: the areal density of disk-drivestoragespacewasdoublingannuallythankstoimprovementsinrecordingmedia,softwarecorrectioncodes,andotherkeytechnologies.Inaddition,otherdimensions of minicomputer performance were improving rapidly, fueled inlarge part by advances inmicroprocessor technology and software design. Asminicomputers began to encroach on the mainframe market, and ultimatelypushedmainframesintodecline,thefourteen-inchdiskdrivemakerscastaboutfor new markets but found only the minicomputer makers buying, and theywanted eight-inch drives. Thanks to their relative unfamiliarity with theinnovations first commercialized by the eight-inch disk drive makers (e.g.,greatermodularityand smaller size), thecompaniesmaking fourteen-inchdiskdriveswereataninsuperabledisadvantage.Mostwentoutofbusiness,andnonewasabletomaintainitsmarketdominanceinthedisk-driveindustry.Thestart-upeight-inchdiskdrivemakersfoundafootholdbyfirstexploiting

tradeoffsamongdifferentdimensionsofperformanceandappealingtotheneedsof an economicallyunattractivemarket.TheyDisrupted the fourteen-inchdiskdrivemakersbyultimatelybreaking those tradeoffsandremaining theprimarydisk drive suppliers to the newly dominantminicomputer companies. In otherwords,as themost lucrativeand largestendcustomersforcomputersswitchedfrommainframestominis, thefourteen-inchdiskdrivemakersendedupgoingdownwiththeirchip.(Sorry.)Accept for themoment thatDisruption is a good explanation for a specific

phenomenon:theseeminglyunlikelyabilityofentrantstotopplewell-resourcedand well-managed incumbents on their home turf. Still more remarkably,however, Christensen observed that over the eighteen years of competition indisk drives that he documented, Disruptive strategies had a much higherfrequency of success, and when successful were muchmore successful thansustainingstrategies.

On the strength of this, Thurston felt that Disruption was among the mostpromisingoftheframeworkshehadstudied.Hewasparticularlyencouragedbythe fact thatDisruption lent itself to fairly straightforward predictions ofwhatwouldworkandwhatwouldnot.

FIGURE1:THEFREQUENCYOFSUCCESSOFDISRUPTIVEANDSUSTAININGSTRATEGIES

And thenThurston ran intoabrickwall.Therewerenodata tosupportanyclaims of predictive accuracy for Disruption. Christensen and others haddeveloped a robust library of literally hundreds of cases across dozens ofindustries thatwere explained byDisruption—but the samewas true ofmanyother theories out there.Worse, for just about every case study explained byDisruptiontherewerecompetingexplanationsthatdrewonentirelydifferentsetsof concepts. (Academic journals continue to debate whether Disruption is thebest explanation of the disk-drive industry’s evolution.) And even if it werepossibletowinthebattleforexplanatory-powerbraggingrights,untiltherewassomeevidenceinsupportofDisruption’spredictivepoweritcouldnotclaimtobetherighttheorytouseformakingdecisionsaboutthefuture.Thurstoncould

havenomoreconfidenceintheprescriptionsofDisruptionthanhecouldinanyothertheory.

EVERYONECOMPLAINSABOUTTHEWEATHER

Intel hasworkedwithChristensen for some years, and the company has usedDisruption theory in its own strategic planning processes. In fact, Christensenand former Intel CEOAndyGrove appeared together on the cover ofForbesmagazine in January 1999 under the headline “Andy Grove’s Big Thinker.”Consequently, when Thurston approached NBI’s leadership about exploringwhetherornotDisruptionmighthavepredictivepowerwhenapplied toNBI’sportfolioofinvestments,divisionalleadershipprovidedThurstonthelatitudeandsupportnecessarytoconductsomepreliminaryinvestigations.Thurston began by stating Disruption’s predictions. Specifically, Disruptive

innovationsaredefinedasproductsorservicesthatappealtomarketsormarketsegmentsthatareeconomicallyunattractivetoincumbents,typicallybecausethesolution is “worse” from the perspective ofmainstream, profitablemarkets ormarket segments. Disruption predicts that leading incumbents with so-calledsustaining innovations—innovations targetedat theirmost importantcustomers—typicallysucceed.Newentrantswithsustaininginnovationstypicallyfail.Disruptions typically succeed, whether launched by incumbents or entrants,

butonlywhen theventures launching themarehighlyautonomousandable todesign strategic planning processes and control systems and financialmetrics,among other characteristics, independently of systems built for incumbentorganizations. This element is important and hardly unique to Disruption:established, successful businesses can and should be held to very differentmeasuresofperformanceandexpectationsforfutureperformancethanstart-uporganizations, and for at least two reasons. First, a start-up typically has atrajectory of growth and profitability that is very different from that of anestablished business. Second, start-ups typically must change, sometimesdramatically, material elements of their strategy as they grapple with theunpredictable nature of customer reaction, competitive response, and theperformance of key technologies. Consequently, start-upsmust find their ownway,andthatispossibleonlywhentheyenjoytherequisiteautonomytodoso.In short, Thurston inferred that Disruption predicts that success awaits

sustaining initiatives launched by successful incumbent organizations andDisruptiveinitiativeslaunchedbyautonomousorganizations.Everythingelseispredictedtofail.(Seefigure2forasummaryofThurston’shypotheses.)NowThurstonneededdatawithwhichtotest thosepredictions.Fortunately,

NBI had retained a robust archive of the materials supporting many of itsprevious efforts. This allowed Thurston to compile a portfolio of forty-eightventures thathadreceivedat leastSAM-level fundingover the ten-yearperiodending in2007.SAMfunding, recall,wasvery early-stage support, analogousperhapsto“angel”investing.Usingthe“pitchdecks”thatwereusedtoexplaineachbusinesstoNBIexecutivesaspartofitsfundingprocess,Thurstonassessedthese SAM-approved businesses for “incumbent” or “entrant” status based onthe degree of Intel’s participation in the market targeted by the start-up andassessed the start-up’s product or service as sustaining orDisruptive based onhowitcomparedtoexistingsolutionsinthattargetedmarket.These decks were typically exemplars of business planning and

communication.Theybeganwithasummaryofthetechnologyinvolvedandthebenefits to Intel of commercializing it. The most optimistic projections wereusually for devices or services that were demonstrably superior to existingsolutionsofferedbycompetitors.ThegrowthopportunitywasoftenarguedtobegreatestwhenInteldidnotalreadycompeteinthatmarket.A review of the management team’s expertise then followed. It was not

uncommon for ventures to be run by an impressive cross section of Intelveterans,newhireswithexperience in the targetmarket, andotherswithdeepexpertiseinfunctionssuchasmarketingordesign,dependingonwhatwasseenascriticaltolong-termsuccess.

FIGURE2:THURSTON’SHYPOTHESES

In framing the predictions implied by Disruption in this way, Thurston wasemphasizing two elements of Disruptors: they start out targeting markets ormarket segments that incumbents do not value, and they have significantautonomy.Butheignoredoneotherelementthatwillprovecrucial:Disruptorsmustimproveinwaysthatallowthemtocompeteformainstreammarketsfromapositionofstructuraladvantage.Thatis,itisnotenoughsimplytoappealtoamarket or market segment that is unattractive to incumbents; that is a nichestrategy.Wewilltieoffthislooseendattheconclusionofchapter4.Fornow,focusonwhatThurstonwastryingtogetdone:hewaslookingfor

actionableadvicethatwouldhelphimpredictwhetherastart-upwouldsucceedor fail, andDisruption—ashe interpreted it—provided the kindsofpredictive,falsifiablestatementsthathecouldtest.Thencameadetaileddescriptionofthevalueproposition.Thiswastheteam

“making good” on its claims of superiority, often including endorsements ofprototypes by customers the team was targeting as early adopters. This wasfollowedbyanimplementationplan:whichmarketsegmentswouldbetargetedinwhatsequence,withspecificdescriptionsofhowIntelwouldbesuccessfulineach, often accompanied by a multigenerational product road map. Finally,financial projections, complete with sensitivity analysis, described theanticipated economic value of the business to Intel, usually over three to fiveyears.Tokeep thingsassimpleaspossible,hedefined“success”assurvival—that

is,theventurewasstillfunctioningasagoing-concernventure,whetherornotitwas still controlled by Intel—and “failure” as “dead”—that is, no longer acommercial going concern. Without knowing the actual outcomes for theseventures,ifThurstoncouldassesstherelevantcharacteristicsoftheNBI-backedventures and predict subsequent “success” and “failure”more accurately thanchancealone,hewouldhavesolidevidencesupportingDisruption’spredictivepower.HereishowitworkedwithImageIllusions,adisguisedNBI-backedventure.

Image-processingtechnologies,suchasprintersorphotocopiers,typicallyusealarge number of application-specific integrated circuits (ASICs) to handledifferent elements of image manipulation, such as shrinking or rotating animage,priortoprinting.ASICsareveryefficient,butthisefficiencybringswithittwodrawbacks.First,becauseeachASICishighlycustomized,manufacturingeconomies of scale are limited,which keeps costs up. Second,ASICs are notprogrammable,sochangingthefeaturesofaproducttypicallyrequiresdesigningandsourcinganentirelynewchip,whichiscostlyandslowsdowndevelopmenttimes.

Alternatives to ASICs, such as media processors, digital signal processors,andcentralprocessingunits,providedvastly increasedeconomiesof scaleandprogrammability but sacrificed performance to such an extent that they wererarely viable. In other words, there was a sharp tradeoff among performance,flexibility, and cost. Manufacturers of image-processing technology—forexample, the folks who make printers and photocopiers—would find it veryvaluabletobreakthattradeoff,forthentheycouldintroduceagreaterrangeofmorepowerfulnewproductsfasterandatlowercost.Intelisanincumbentinoneofthesethreealternativetechnologiesmentioned

above.ImageIllusionssoughttoleveragethispositiontocreateanewsolutionthatprovidedbothefficiencyandflexibility.BycompetingwithASICs, ImageIllusionswouldbeleveragingoneofIntel’scorecompetenciestoexpandintoa“whitespace”opportunitytogeneratenew,innovation-drivengrowth.In collaboration with a key potential customer—a large, successful

manufacturer of digital imaging technology—the Image Illusions teamdeveloped a highly sophisticated and demonstrably superior solution based onproprietaryintellectualcapital.ItcostalmosttwiceasmuchperunitasASICs,buttheteamfelt(andthecustomercorroborated)thatthehigherpricewasmorethanoffsetbytheincreasedperformanceandflexibility.Inotherwords,theteamhadbrokenthecriticaltradeoffthatwaslimitingtheperformance,cost,andpaceofinnovationinimage-processingtechnology.There were, of course, challenges. The largest companies that made image

processors—includingtheonethatImageIllusionshadcollaboratedwithandallof the targeted early adopters—had their own in-house ASICs design staffs.Many of these peoplewere also on the internal committees that assessed newtechnologies.Toadoptanon-ASICssolutionwaseffectivelytoput themselvesoutofajob.ThatmeantImageIllusionswouldlikelyhavetobevastlysuperiorbefore customers would switch in volume, since the in-house ASICs designteamswouldbe stronglymotivated to show that theycouldup theirgameandmatchthenewtechnology.The Image Illusions team had reason for optimism. The image-processing

marketwasfiercelycompetitive,andthevastperformanceimprovementsImageIllusionscouldprovidemeantthatalltheteamneededwasonemajorplayertoadoptitssolutionandtherestwouldfollowsuit.TheabilitytoleverageIntel’sstrongbrandandcustomeraccessmade theoddsofgettingonedomino to fallseemveryfavorable.Thecash-flowprojectionsforImageIllusionsestimatedanetpresentvalue(NPV)between$9millionand$100millionoverfiveyears,arangethatreflectedboththeteam’sconfidenceandtheunavoidableuncertaintythatcomeswithlaunchinganewbusiness.

Assessingtheprospectsofsuchaventureisreasonablyseenasacomplexandchallengingtask.Isthetechnologyreallythatmuchbetter?Isit“betterenough”to overcome the entrenched interests of the customers’ in-house designfunctions? Is the management team at Image Illusions up to the challenge ofovercoming the inevitable and unforeseeable twists and turns on the road tosuccess?IsIntelsufficientlycommittedtothisventuretosupportitfortheone,two,orthreeyearsneededtomakeittopositivecashflow?Itwouldappearthattopredictwithanyconfidencewhatwillhappenonemusthavedeepexperienceand expertise in the relevant technologies andmarkets, strong familiaritywiththe management processes at Intel, and an intuitive but accurate take on theabilitiesoftheleadershipteam.Not if you are Thomas Thurston trying to test the predictive accuracy of

Disruption.Forhim,theonlyquestionsthatmatteredwerethefollowing:

1.IsIntelanincumbentinthismarket;thatis,doesIntelalreadysellthissortofproducttothissortofcustomer?2.IsIntel’sinnovationsustainingorDisruptiveinnature?ADisruptivesolutionmakesmateriallydifferenttradeoffsthantheexistingsolutionspurchasedbymainstreamcustomers;asustainingsolutionisstraightforwardlybetter.3.IftheinnovationisDisruptive,doesthenewbusinesslaunchingitenjoyoperationalandstrategicautonomyfromIntel’sestablishedprocesses?

In the Image Illusions case the answers were pretty clear. Intel was a newentrant: it did not sell image processors. The Image Illusions technology wassustaining:itpromisedbetterperformancethanASICs,asdefinedbythelargestand most profitable customers. According to Disruption, an entrant with asustaininginnovationcanexpecttofail.SothatiswhatThurstonpredicted.

CHAPTERTWO

BETTERBYHALF

Experimental data show that large populations can use Disruption toimprove their ability to predict accurately which new businesses willsurvive and which will fail. Specifically, Disruption theory increases thepredictiveaccuracyoflargegroupsofMBAstudentsbyanaverageofuptofivepercentagepoints,or50percentabovethebaselinesurvivalrate.

You are probablywondering how accurateThurston’s predictionswere—notjust for Image Illusions, but for the entire portfolio of forty-eight businesses.Beforerevealingtheanswer,itisworthconsideringexplicitlyjusthowdifferenttheapproachheusedwasfromthatofmanyseasonedinvestmentprofessionals.First, it isworthunderliningthatThurstonwasnottestingwhetherornothe

wasbetter at predictingoutcomes than the collective efforts of thosewhohadassembled the NBI portfolio. Rather, he was testing whether the transparentapplicationof a set of specific ruleswould result inmore accuratepredictionsthancouldbeexpectedbychancealone.Second,theapplicationoftheserulesdoesnotrequiremuchinsightintomany

dimensionsofcompetitiveanalysisthatareoftencentraltothiskindofdecision.HerearejustafewoftheelementsoftraditionalbusinessplanningthatarenotpartofDisruptionasThurstontestedit:Marketattractiveness.Youwouldthinkwhetherornotamarkethasgrowth

potentialisworthatleastlookingat.Insomeofthemostpopularandtime-testedbusiness planning frameworks,market attractiveness iscentral, andwith goodreason:ifyoudonothaveanycustomers,whatisthepoint?Organizationalcapability.Agreatstrategyisfine,butyouhavetobeableto

implement that strategy, right? And so you should consider—no, you shouldobsessover—whetherornottheorganizationinquestioncandowhatitplanstodo,right?Organizationalfit.Especiallywhenlookingatnewventureslaunchedbyan

established company, some concept of “fit” almost always enters into thepicture. It isnotuncommonto thinkthatnewventureshaveabetterchanceofsuccess if they are leveraging an established core competency or can takeadvantage of other resources or capabilities, such as marketing and sales,

distributionchannels,andsoon.Leadership. Surely there is little thatmattersmore to a business’s success

than leadership.Manyventure-capital investorswill tell you theybetmore ontheteamthanonthetechnologyorthestrategy—largelybecause,intheirview,the strategy and technology are all highly uncertain and difficult to assess.Consequently,whatmattersmost is the ability of the team to copewith thoseuncertainties.LookingbackattheImageIllusionscase,wecanseethatitpassedwithflying

colors these four criteria, and one could continue at length extolling theventure’s many other virtues. It was targeting a large and growing market,measured in tens of millions of units per year, with significant interest frommajor would-be customers. Image Illusions had the capability and thecompetencies to prosecute the proposed strategy: from product design tosoftwaredevelopment tomanufacturingtomarketingandsales, therewas littlethedivisioncouldnotdoordidnotwant to leverage inorder togeneratenewgrowth.And the Image Illusions teamwas an archetype for any start-up,withjust the right mix of grizzled veterans, fresh-faced entrepreneurs, technicalexperts,andmarketingmavens.Butitsoriginalplanfailed.JustasDisruptionpredicted.To see why, look at Image Illusions now through a Disruptive lens. First,

Image Illusions was proposing a sustaining technology, targeting the largest,most importantmanufacturers of image-processing technologywith a solutionpositionedasbetter than the alternative, theASICsprovidedby its customers’in-house designers. The competitive response one would expect from theincumbents in this space (the in-house design shops) was vigorous andimmediate:facedwiththelossoftheirmostimportant—andinmanycasesonly—customer, the in-house designers were not only likely to have a less-than-objective (nevermind favorable) take on themerits of an alternative solution;theywerestronglymotivatedtocomeupwithsomethingbetterthemselves.Second, Intel was an entrant into this space. Compared to the depth of

knowledgeofincumbentsinthatspace,theIntelmanagerstowhomtheleadersof Image Illusions reportedwouldbe likely tosee themarketasuncertainandImage Illusions’ initial sales as small, even if the venture was a promisinggrowth opportunity. In contrast, Intel’s primary markets were large andrelatively well understood. Consequently, any tradeoff between providingresources for Image Illusions and protecting the success of Intel’s corebusinesseswouldclearlyandreasonablyfavorthelatter.Incontrast,incumbentsin the digital image-processing industry would defend their home markets atalmostanycost.Thisasymmetryinmotivationleavestheentrantfarmorelikely

thantheincumbenttoabandonthefield.Nevertheless, thebusinessplan receivedSAMfunding.Theventuremade it

through prototyping and even secured low-volume orders as major customersaroundtheworldbeganevaluatingtheproductinearnest.Itwasthenthatcracksbegan to show. As a new product, there was a dearth of software tools andapplications that enabled customers to take full advantage of the chip’sprogrammingfeatures.Asastart-up,ImageIllusionshaddifficultyfundingthedevelopment of those tools on its own. Lacking scale, the per-unit cost wassignificantly higher than that ofASICs-based solutions,while in-house designteams, spurred by the competitive threat, began responding with increasedperformanceandreducedcost.Worse,evenifsomecustomerswereenthusiasticabout the future of the technology, they would not commit to significantvolumesquickly:sincethiswasamajoradvanceonestablishedtechnologies,itwouldtaketimetoconfirmthatthenewproductnotonlyworkedbutcouldbeintegrated intomajorproducts,whichmeant testingeverything frombackwardcompatibilitywithsoftwaredriverstotheabilitytoprovideafter-salesservice.ItisimportanttonotethatDisruptiondidnotpredictthesespecificchallenges,

nor can it predict the specific causes of failure for any particular“sustaining/entrant”initiative.Rather,thetheorypositssimplythattheforcesatworkwillresultinsomesetofeventsresultinginfailureforsuchventures.So,althoughwecannotknowpreciselywhatwillgowrongwhenacompanyentersanewmarketwithasustaininginnovation,wecanexpectthatsomethingwill.For Image Illusions, even as the venture moved out from under NBI’s

umbrellaandbecamepartofanoperatingdivisioninIntel, itslong-termfuturewasbeingquestioned.Assoonas ImageIllusionshad tocompetewitha largeand established business for funding and engineering talent, the uncertaintiesthat still surrounded its future became debilitating. However promising thefinancialprojectionsforImageIllusionswere,themoneytobemadeandyear-on-year growth were much less in absolute terms than those of its parentdivision. Small is a problem because large corporations would rather employ$100millionattheircostofcapitalwithnearcertaintythan$1millionatall-but-ridiculous multiples of their cost of capital, and for entirely rational reasons:eventhoughImageIllusionswasaverysmallinvestmentforIntel,managementtimeandattentionarescarceresourcesthatcanrarelybeallocatedinproportiontothemagnitudeofaninvestment.Abusinessthatconsumes0.1percentofthecapital budget still consumes an hour at the quarterly reviewmeeting—alongwiththedozensofe-mailsandphoneconversationsandthewatercoolerchitchatthatprecedethequarterlyreviewmeeting.Thatistimenotavailabletoexistingbusinessesthatkeepthelightson.

Uncertaintyisaproblembecausethekindofgrowthrequiredtojustifysmallinvestments often seems utterly fantastic. It matters little that many—maybeevenmost—of thebiggest successes typicallystartoutverysmall.There isanintuitive and entirely accurate sense that new businesses fail more often thantheysucceed,andsincewecannot tell thedifferencebetween thewinnersandthelosersinadvance,thebiasshouldbetowardsayingno.DoesthatmeanitwasimpossibleforImageIllusionstohavesucceededasa

sustainingentrant?That ishard to saybut forpresentpurposes irrelevant.ThequestioniswhetherDisruptionaccuratelypredictedwhathappenedbasedsolelyoninformationthatwasavailableatthetimeofinitialfunding.Thurstongotthatquestion right, and he got it right by applying Disruption only to what wascontained in a pitch deck developed fully five years before Image Illusionsultimatelyfolded.

ITCANNOTBETHATEASY

Perhapsoneof themostcounterintuitiveelementsofThurston’sapplicationofDisruption theory topredictingsurvivaloutcomeswashowrelatively fewdatawere required tomake a prediction. It often took littlemore than reading theexecutive summary of what was an otherwise lengthy and carefully crafteddiscussionofawiderangeofissues—everythingfromtechnologytrendstotheevolution of potential customers’ needs. A proposal claiming to have a bettermousetrap (sustaining innovation) and targeting a new growth opportunity forIntel (new entrant) was predicted to fail. The rest of the deck, from theperspectiveofmakingtheprediction,becameinstantlyirrelevant.Usingonlythesetwolitmustests,Thurstonmadehislive/diepredictionsfor

the forty-eight businesses in his test portfolio—in complete ignorance ofwhatactuallyhappened.9Assessinghisaccuracy,however,wasnotasstraightforwardasonemightthink,fordeterminingwhatcountedas“alive”or“dead”wasbothconceptuallyandpracticallychallenging.Forexample,someventureshadbeenspunofforsoldtootherbusinesses.In

suchcasesThurstonhadtotrackdowntheacquirertodeterminethefateoftheNBI-backedventure.Inothercases,assessingaventure’sultimatefatebecameanearlyphilosophical question.Oneventurehaddeliberatelybeen turned into anonprofitenterprise,managedand fundedby Intel’sphilanthropicarm.Oneortwo had been mothballed but not completely shut down: someone might bebabysitting the patent portfolio simply to ensure that Intel got appropriatecompensation from other companies incorporating its innovations into theirproducts.Stillothershadbeenrolled into thecorporateoverheadfunctionasaform of research and development, with no further expectation of directcommercial application.Whichof these should be considered alive andwhichdead?Thurstonapplieda setof criteria thatmade thebar for survival lowbutnot

trivial. Any venture that was functioning in any form as a commercial entityselling products or services counted as having survived. Sold off but stilloperating?Alive.RolledupintocorporateR&Dortransformedintoanonprofit?Dead.Collectingannuitychecksonpatents?Dead.(Thislastoneisagrayarea;the deciding feature for Thurston was that such a business was not a goingconcern.10)Here is how Thurston’s predictions, made using Disruption, stacked up

againstNBI’sactualresults.Oftheforty-eightSAM-approvedbusinessesintheoriginal portfolio, five succeeded (survived) and forty-three failed (werediscontinued), foranoverall success rateof10.4percent.11Thurstonpredictedsixsuccessesandforty-twofailures.Thurstoncorrectlyidentifiedfourofthefiveactualsuccessesforan80percentaccuracyrate;hecorrectlyidentifiedforty-oneoftheforty-threeactualfailuresforanaccuracyrateof95percent.Overall,hecorrectlypredictedtheoutcomesforforty-fiveoftheforty-eightbusinesses,foranaccuracyrateof94percent.Interpretingtheseresultsisnotstraightforward.Sincetheoverallsuccessrate

oftheactualportfoliowasjustover10percent,simplypredictingfailureallthetimewouldresultinanearly90percentoverallaccuracyrate.However,suchanapproach is hardly of much use, since it would have a 0 percent chance ofsuccess.Inaddition,statisticaltestscanidentifythelevelsofaccuracyrequiredwhenpredictingbothsuccessesandfailurestoruleoutluckyguesses.Predictingall failures might have a high overall accuracy rate, but it would not bestatistically different from random guessing due to the poor accuracy inidentifying successes (0 percent). In contrast, Thurston’s outcome would beexpectedbychancealoneonlyoncein2,500trials.

FIGURE3:THURSTON’SRESULTS

TheshadedcellsarethekeyentriestoconsiderwhenassessingThurston’saccuracy:theycapturethefrequencywithwhichhepredictedsuccessesorfailurescorrectly.

Togetamoreintuitivesenseof theremarkablenatureofThurston’sresults,imagine for amoment thatwe had one of those “neuralyzer”memory-erasing

gadgetsfromthemovieMeninBlackandwezappedthegoodfolksatNBIandpresented themwith the same set of forty-eight businesses and asked them topredictwhichwouldliveandwhichwoulddie.Theywouldalmostcertainlynotpredict survival for themall: suchaprocess rarelyhasa100percentpass rateandindeedisfarmorelikelytorejectatleastasmanyproposalsasitaccepts.Supposenowthatthisinvestmentmulliganresultsinonlyhalfoftheoriginal

portfolio, or twenty-four proposals, getting approved.12 To demonstrate astatistically significant improvement over chance alone, the neuralyzedexecutiveswouldhavetocaptureallfiveactualwinnersintheirnewportfoliooftwenty-four approved proposals. Even then, this result would be expected tooccurbymerechanceonceinonlytwentytrials.Inotherwords,theymighthavecaughtallthewinners,buttheywerenotgoodenoughatweedingouttheloserstoshowasignificantimprovement.Thurston,incontrast,pickedfourofthefiveactualwinnerswhilepickingas

survivorsonlysixoftheforty-eightproposals,oronly12.5percentoftheactualportfolio. In other words, although Thurston’s application of Disruptionpredicted failure far more often than in the actual portfolio, this increasedpessimismdidnotcomeattheexpenseofadebilitatingrateoffalsenegatives:DisruptionallowedThurston toseealmostall thesuccesseswhileweedingoutalmostallofthefailures.Tobring intosharperrelief thestartlingsimplicityofThurston’sevaluations

usingDisruption,recallthatonlythreeattributesofeachventurewereassessed:incumbentorentrant;sustainingorDisruptive;andifDisruptive,autonomousornot.Figure4 summarizeshowThurston’spredictions stackedupagainst thosecriteriaandtheaccuracyofeachtypeofprediction.Certainly this research design, like every other, has its limitations (see

appendix A for a discussion). But the bottom line is this: Thurston predictedfailure forty-two times—the forty sustaining/entrant ventures and the twoDisruptive/nonautonomous ventures.Of these, forty-one failedwhile only onesustaining entrantmanaged to survive. He predicted success six times; all sixsustaining/incumbent ventures were expected to survive, of which four did.TherewerenoDisruptive/autonomousventures.Theoverallaccuracyrateof94percentgrabsourattention,buttherealnewsisthatThurstonhadasignificantlyhigher accuracy rate for successes and failures as individual categories thanexpectedbychancealone.

INGODWETRUST.ALLOTHERS,BRINGDATA.

Oneexperiment ishardlyconclusive,whatever theoutcome.That iswhydrugcompanies must demonstrate the safety and efficacy of new compounds withmultiple studies: unless a result can be repeated, it is difficult to have muchconfidence in it. Consequently, however suggestive Thurston’s initialexperiment might be, few would counsel abandoning existing practices forsomethingcompletelydifferentjustyet.

FIGURE4:THURSTON’SRESULTSBYINNOVATIONTYPE

Whatotherkindsof testsmightoneapply todetermine ifDisruptioncanbeusedtopredictcertainelementsofthefuture?Formanyobservers,oneacidtestis theability topickstocks: ifonecandemonstratesuperiorreturnsbyusingagivenmethod to identifywinners, then thatmethod is as right as anymethod

needstobe.TherearesomedatatosuggestthatDisruptioncanbeusedsuccessfullyinthis

way. InDecember2002a fundexplicitlydedicated to taking longpositions incompanies identified as Disruptors was launched by Rose Park Advisors, itsDisruptiveInnovationFund(DIF).MattChristensen,ClaytonChristensen’sson,plays a leading role in determining the fund’s investment strategy and stockpicks.Due toSECrestrictions, thefundcannotpublish its returns,but through2010 it has handily outperformed the S&P 500, with much of thatoutperformancecomingfrom2002to2007,aperiodofgeneralbuoyancyfortheS&P. From 2008 to 2010 the fund outperformed the S&P 500, albeit by aslimmermargin.Theproblemwiththissortofevidenceisthatthemagnitudeanddurationof

outperformanceneededtodemonstratethatsomethingspecialisgoingonisfarbeyondwhat isavailableforalmostanyinvestmentfund,nevermindthemereeight years available for the DIF and its antecedents. Over any given timehorizon therewill always be some number of fundswith remarkable, perhapseven astonishing, performance. But there is not necessarily anything specialabout any of them.The track record required to stand above the noise of thissystemtakessolongtoestablishthatverytypicallyotherfactors impinge—forexample,fundsmergeoraredisbandedduringarunofpoorreturns—makingiteffectively impossible for any fund to achieve it. Consequently, the key toincreasing our confidence in Disruption’s predictive power is replicatingThurston’sresults inanexperimentalsetting.Only thencanweget thesamplesizesrequiredtoseparatesignalfromnoise.Partlyinsupportofthisgoal,IntelfundedayearlongsabbaticalattheHarvard

BusinessSchool forThurston to continuehis investigations into thepredictivepowerofDisruptionincollaborationwithChristensen.Aspartofthatresearch,Thurston conducted a second predictive experiment at the MassachusettsInstituteofTechnology’sSloanSchoolofManagement(MIT)andtheHarvardBusinessSchool(HBS) in2009.Itwas then thatChristensen introducedmetoThurston’s work, and the three of us collaborated on a second experiment atHBSin2009.IconductedathirdexperimentattheIveyBusinessSchoolattheUniversityofWesternOntario(Ivey)inLondon,Canada,in2010.Each experiment followed the same basic design. Thurstonwrote disguised

two-pagesynopsesoftheforty-eightbusinessplansthathehadevaluated.Thesecases had the same general structure and tended to follow the structure andcontentof thepitchdecks supportingeachactual fundedventure’sproposal totheSAMcommittee.Thecasesdescribedthefollowing:Marketneed.What product or servicemarketwas being targeted,who the

customerswere,andwhatsortofperformanceprofilethosecustomersseemedtoprefer.Businessunit.ThestructureoftheorganizationalunitwithinIntelthatwould

pursuethisventure.Wasitastand-alone?Wasitpartofanexistingunit?Whowerethecompetitors?Whatcapabilitieswouldtheunitcontrol,andwhatwoulditsrelationshipwithotherInteloperatingcompaniesbe?Managementteam.Whowouldbedoingwhatforthenewventure,andwhat

theirpedigreesandcredentialswere.Financials.Thismightincludeunitvolumes,prices(andhowtheycompared

with competitors’), growth projections, and net present value (NPV) estimatesandrangesforsuchprojections,whereavailable.AppendixBisthecaseusedforImageIllusions.MBAstudentswithlittletonofamiliaritywithDisruptionwerethengivena

randomly selected set of six cases and told that each described a real butdisguised business that had been funded and launched by a major U.S.corporation.Theywerethenaskedtopredictwhetherthebusinessdescribedineachcasehad“survived”or“died.”Theyweretomaketheirassessmentsusingwhateverframeworks,logic,orintuitiontheywished.Thiswasthe“control”inthe experiments—an assessment of the students’ predictive accuracy prior tolearningaboutDisruption.Onewouldexpectthestudentsonaveragetodoaboutthe same as random chance, that is, collectively to build a portfolio with thesamesurvivalrateastheactualportfolio.The students were then given varying degrees of instruction in Disruption.

This constituted the “treatment.” Finally, each student was asked to makepredictions on a second set of six cases. No student evaluated the same casetwice.This design provides insight into three aspects of Disruption. First, it tests

predictive power. Disruption “works” if, after having received training, thestudentswereabletopredictoutcomesmoreaccuratelythanchancealonewouldprovide for. The extent of its predictive power lies in the magnitude andstatistical significance of the difference between the control and treatmentgroups.Second,wecanseefromtheseexperimentswhetherDisruptionis teachable.

Oneof themajordrawbacksformanymanagementtheories is thedifficultyofdefiningconcepts inwaysthatcanbeappliedsuccessfullybyothers.Take,forexample,thenotionof“corecompetence.”Thishasbeenanenormouslypopularconceptforovertwentyyears.Yetifyourexperiencewiththetermisanythinglikemine,althoughyoumightfeelyouknowwhatitmeansandhowtouseit,itisofteneye-openingtorealizehowdifferentothers’interpretationscanbe.

Whether or not something isDisruptive has often been subject to the samesort of ambiguity.ClaytonChristensen is rarely confused about theDisruptivenature of an innovation. When management teams attempt to employ theconcept,chaosoftenensuesaspeopledebatethemeaningoftheterm.BystatingDisruptionanditsimplicationsinstark,unadornedtermsandsubjectingittothisform of experimental test, we can assess whether applying Disruption issomething“anyone”cando.This is potentially important, for not only is most of the support for most

managementtheoriesmerelytheposthocinterpretationofselectedcasestudies,thisposthocinterpretationofselectedcasestudiesisdonebythecreatorofthetheory. Larger-scale experiments on a population that closely mirrors themanagerialranksofmanycorporationsprovideinsightintowhetherDisruptionhasthepotentialtobewidelyandsuccessfullyapplied.Third, these three experiments test for differences in the impact of different

types of instruction in Disruption on its predictive power. In the Iveyexperiment,studentshadonlyaone-hourlectureonDisruptionfromme.Thereis only so much ground one can cover in sixty minutes, so students wereequippedwithonlyabasicunderstandingofthesalientfeaturesofDisruption.IntheMIT/HBSexperiment,studentshadasix-sessionmoduleonDisruption

delivered by Clayton Christensen, Willy Shih, also a professor at HBS, orRaymondGilmartin,aprofessorofmanagementatHBSanda formerCEOofMerck & Co., the pharmaceutical company. This consisted of short (twenty-minute) lectures from the instructor followed by class discussion. The HBSexperimentconsistedofafullsemestercourse(twenty-sixsessions)acrossthreesectionsofMBAstudents,oneofwhichwastaughtbyChristensenandtwoofwhich were taught by Gilmartin. The content of both of these treatmentsfollowedverycloselythecontentofTheInnovator’sSolution.Thestudents’overallaccuracyrateisnotespeciallyrelevant;whatmattersis

whether our “treatment” groups had an accuracy rate—for both successes andfailures—thatwasstatisticallysignificantlygreaterthanthatoftheportfoliotheystartedwith.

DRUMROLL,PLEASE

Adetailedexplanationof theexperimentalmethodanda statisticalanalysisofthe results are provided in appendix C. Here are the two most intriguingfindings. First, Disruption improved predictive accuracy significantly andmaterially.DisruptioncanbetaughttolargepopulationsofMBAstudentswhothenareabletomakebettercallsaboutthesurvivalorfailureofnewbusinessesthan they could before they knew anything about Disruption. In other words,Disruptionworks.Second, the more nuanced and seemingly complete the instruction the

students received, the less significant and the less material was theirimprovement.ItwastheIveyMBAs,whoreceivedonlytheone-hourlectureinDisruption, who improved the most. TheMIT/HBS seminar students showedmaterialimprovementbutcamesecond,andtheHBSstudentswhogotthe“fullMonty”—a three-month, second-year HBS elective, with one-third of thestudents taught by Christensen and two-thirds by a seasoned and successfulformerCEO—werethird,showingessentiallynoimprovementatall.

FIGURE5:IMPROVEMENTBYTREATMENTTYPE

Theexperimentaldesignisnotperfectandtheresultsarenotoverwhelming,butasfarasIknow,noonehaseversubjectedanyothertheoryofinnovationtoanykindofpredictivetest.Disruptionisperhapstheonlytheoryof innovationthat has evidence to support the claim that it can improve predictive power.WhileconceptuallydistinctfromtheclaimthatDisruptionisbetterthanallothertheoriesinthisregard,inpracticaltermsitamountstothesamething.Thesignificanceofthisishardtooverstate.Theresultsoftheseexperiments

demonstratethatDisruptionworksandcanbeappliedsuccessfullyinanexplicit,rules-basedway—rather thanrelyingon the intuitivepatternrecognitionof the

appropriatehighpriesthood.Thereisnolengthyapprenticeship,nosecretsauce,noblackart.Justthestraightforwardapplicationofthreesimpleyes/noquestionswith explicitways of determining the relevant answers.Whatever ambiguitiesmightremaininmakingtherelevantcategorizations(e.g.,incumbentvs.entrant;sustaining vs. disruptive), the data reveal that overall Disruptionmakes betterpredictionpossible.Theseexperimentsdonotconstituteahead-to-headcomparisonbetween the

processusedbyNBItofundbusinessesandtheprescriptionsofDisruption.Thatwould require using each approach independently to evaluate and fundbusinesses from a commonpopulation of proposals and then seeing how eachportfolioperformed.Andwewouldhavetodoitmultipletimesinordertohaveconfidence in the findings. One cannot conclude, therefore, that a portfoliochosenusingDisruptionfromthepopulationofproposalsthatNBIevaluatedforSAM funding would have fared better than NBI’s actual portfolio. We can,however, get some idea of the potential impact of using Disruption to pickwinnersinatleasttwoways.First,considerthe“improvementoverrandom”observedwitheachtreatment.

What if we increased NBI’s observed success rate by the same incrementobserved in each treatment?Assume that therewerewinners in the proposalsthatwererejected,sothatitwaspossibletohavemorewinnersoutoftheforty-eightthanwereactuallyobserved.Whatwouldthathavemeantforthenumberof winners and losers in the total portfolio if we hold the number of fundedproposalsconstant?

FIGURE6:IMPACTOFPREDICTIVEIMPROVEMENTONADENOVOPORTFOLIO

Bythismeasure,Thurstonwouldappeartobeinaleagueofhisown.Withatwo-to-one hit rate, we should expect him to be taking calls from Warren

Buffett.Whetherhecanreplicate thatkindofaccuracy issomethingonly timewill tell: he took what he learned back to NBI, which is in the process ofcodifyingandimplementingthefruitsofThurston’syearatHBS.Fortheothertreatments, the results are less eye-popping but nevertheless impressive.Specifically, thewin/loss ratio reflects the increasedaccuracyratesof the IveyandMIT/HBSexperiments.Withabetterlikelihoodofpredictingsuccesses,theIvey treatment wins when picking a portfolio of investments from a broaderpopulationofproposals.AsecondandmoreconservativewaytoestimatetheimpactofDisruptionon

the value of the portfolio is to apply Disruption as a screen for NBI-fundedprojects.That is, ifwe took theproposalsNBIchose tofundandpushed themthroughtheevaluationprocesscapturedineachexperiment,whatwouldwebeleftwith?NowtheMIT/HBStreatment(theDisruptionmodule)offersthebestresults:

thestudents’abilitytopredictmoreaccuratelywhatwouldfailoutweighshavingloweraccuracy than theIvey treatment incalling thesuccesses. It isonlywiththelargerpercentageofsuccessesinareselectedportfoliothatthesinglelecturegiventotheIveystudentsoutdoesthefullmodule.

FIGURE7:IMPACTOFPREDICTIVEIMPROVEMENTONTHEORIGINALPORTFOLIO

Translatingtheseestimatesofimprovementinsurvivalratesintodollarvaluesisimpossible:wedonotknowthereturnsgeneratedbyeachofthesuccessesorthe losses incurred as a result of funding the failures. As a result, it isconceivablethatthesuccessesDisruptionmissedwereofsuchenormousvalue,and the failures that would have been avoided by using Disruption of suchmarginal impact, that excluding them from the portfolio might well havedecreasedtheoverallvalue.Wesimplycannotknow.

WHENIGNORANCEISBLISS

We also have the question of the relative impact of our three differenttreatments.ThurstonandIexpectedthelecturetohaveamildeffect,themoduleto have a moderate effect, and the full course to have the biggest and mostsignificanteffectofall. It cameout inprecisely theoppositeorder: thecoursehad almost no impact, themodule had a small but significant impact, and thelecture had the largest and most significant impact of all. That requires anexplanation.Typically,managementframeworksareusedtoexplainoutcomes.Theoretical

parsimony is a valuable attribute: theories that can explain complex outcomesusingonlyafewkeyvariablesaretypicallythemostpowerful.Themarketplaceofideas,however,hasbeenconditionedtocravecompleteexplanations.Toseethismostclearly,examinethenatureofthecriticismtypicallyleveledatagiventheory:itisalmostalwaysintheformof“butwhataboutX,”whereXissomeexample that does not seem to fit themodel. These counterexamples are thentypically used to justify a different approach that allegedly provides a betterexplanation. Defenders of the theory under attack typically respond byelaboratingontheirtheoryinordertofittheallegedcounterexamplesintotheiroriginalstructure.Theresultisgreaterexplanatorypower,butattheexpenseofsimplicity.Disruption’s defenders have sought to explain seeming counterexamples by

developinganever-expanding toolkitof ideas thathavebeengraftedonto thebasic concepts. What began as a theory of customer dependence (targetcustomers incumbents do not want) and technological constraints (you likelycannotmakesomethingbetter,somakesomethingworseandsell it inmarketsthat will settle for less) now includes constructs designed to address, amongother strategic questions, the optimal scope of the firm, the role of seniormanagement in strategy formulation, and methods for assessing underlyingcustomerneeds.SinceIhavebeenpartofthedevelopmentofsomeofthisadditionalfiligree,

itwillnotcomeasasurprisethatIthinkmuchofthisworkisveryhelpful.ButIhave toaccept theverdictof thedata:notonlydo theseadditionaldimensionsnotincreasepredictiveaccuracy,buttheydemonstrablygetintheway,at leastwithintheparametersoftheseexperiments.Thereasonbeginstoemergewithananalysisofthestudents’explanationsof

theirpredictions.CommentsfromstudentsintheHBStreatment(thetwenty-six-session course) covered just about every dimension of Disruption. For ImageIllusions, for example, some studentsmade the correct prediction (failure) fortherightreasons:itwasasustaininginnovationbyanentrantorganization.ButpartofChristensen’scourseexploresthecharacteristicsofsuccessfulmanagers,andsomestudentsargued that themanagementof ImageIllusionshadhad theright sorts of experiences to overcome any deficiencies in the strategic plan,leading them to predict success. Others tapped the unit’s value-chainconfiguration as reason for optimism, while still others felt that the strategy-formulationprocesswassufficientlyemergentthattheventurewouldultimatelysucceed. For each business plan, students were much more likely to findsomething the plan did “right” inDisruption terms because they had somuchmoretheorytoworkwith.Yetitturnsoutthattheseadditionaldimensionsaddnopredictivepower.ThepredictivepowerofDisruptionseemsinsteadtolieinits focus on the underlying strategy and technology—not the “implementationissues”surroundingplanning,organizationaldesign,andsoon.ThishelpsexplainwhythestudentsintheMIT/HBStreatment(theDisruption

module) did so much better: they had much less to work with, namely, onlythose elements of Christensen’s course that addressed the core elements ofDisruption.Theydidnotget “distracted”byall these additional—even if verypowerful—explanatory dimensions. Similarly, the Ivey treatment (the lecturealone)consistedof themostbare-bones instructionofall,and thestudentshadnotimetogetdistractedorforgetwhattheyhadlearned.When assessing predictive power in the context of these experiments,

counterexamples do not really count for much, since each counterexample issimply a prediction that the theory got wrong, and no one ever claimed 100percentpredictiveaccuracyformuchofanything,nevermindthissortofsocialscience. We need to calibrate our evaluation of Disruption’s merits with ourcurrent stateofunderstanding.After all, an investorwith a20percent successratewouldbelionizedforprescience,notridiculedforan80percentfailurerate.Those failuresmightwellbeexplainedusingcriteriaandframeworksdifferentfromthoseusedtomaketheoriginalprediction,butthereisnoreasontobelieve—simply on the basis of explanatory completeness—that the alternative orexpandedexplanatoryframeworkswillprovidebetterpredictions thenext timearound.So,isDisruptiontrue?Inmyview,yes,butinaverypreciselydefinedsense

ofthatword.Disruptionisnottrueinvirtueoflogicalnecessity,asistheclaim2+2=4.Rather,theavailableevidencesuggeststomethatwecansaymorethan“maybe” while still accepting the possibility—even the likelihood—that

something better will come along. The merits of Disruption will need to bereevaluatedaswelearnmore.Butifyouareevaluatingabusinessstrategyandwanttoknowwhetheritwillsucceedorfail,DisruptionistheonlytheoryIamawareof thathasdemonstratedutilityandthatyou,personally,haveafightingchanceofapplyingcorrectly.This type of claim, made this forcefully, is highly unusual. The standard

approachistomakesweepinggeneralizationswithcircumspection,anudgeanda wink. You can “cross the chasm,” sail the “blue oceans,” “innovate to thecore,” you name it. But you must embark on these endeavors accepting theenormousuncertaintythat,despitetheinsightsoffered,continuestosurroundthechallengesofinnovatingsuccessfully.In contrast, I make a more limited and modest claim: that Disruption can

deliver statistically significant and practically material improvement in yourability to innovate successfully. It is not a guarantee of success. Rather,Disruptionoffersnothingmore—orless—thananevidence-basedhopeofbeingbetteratcreatingorpickingwinnersthanyouwouldbeotherwise.AndImakethis limited claimwith vigor thanks to the nature ofmy supporting evidence:proofofimprovedpredictiveaccuracy.If youwantmoredata, say, in the formof repeating this experimentwith a

differentportfoliofromanothercompany,youarenotalone;I’dwelcomethataswell.Butfornow,torepeatachallengemadeintheprologue, inassessingthesoundness and validity of these data and this evidence, compare it not to aPlatonic ideal of perfection but to the evidence that supports the beliefs youcurrentlyhold.Few of us should hope to replicate Thurston’s result; it might well be an

outlier. The differing treatments among the three experiments revealed whichelements of Disruption have true predictive power. And it is in the Iveytreatment,withits50percentimprovementoverrandom,thatwefindour“five-percentage-pointsolution.”Andthechanceofrealizingthosefivepercentagepointsislikelyworthalot.

PARTIIEXPLANATION

CHAPTERTHREE

HOW

Ifwearegoingtogeneralizebeyondtheparametersoftheexperiments,weneedtoknowwhentouseDisruption.ThischapterexplainshowDisruptorscome to compete for mainstream markets and, as a consequence, whyincumbentscannotrespondeffectively.

Is it enough that Disruption is the “five-percentage-point solution”?Will theexperimentalevidencereportedinthepreviouschapterscarrythedayandleaveDisruptiontheunquestionedchampionofinnovationtheory?Probablynot,andforatleasttworeasons.First,somewillnotfindthedataconvincing.Theexperimentaldesignhasits

limitations: among others, what would Disruption have done with all thebusinessesthatNBIrejected?Wesimplycannotknow.Istheeffectstatisticallysignificant enough? And is the effect size—five percentage points, or a 50percent improvement—enough to really matter? The four experimentaltreatments (Thurston’s initial assessment and the MIT/HBS, HBS, and Iveyexperiments) were all done with the same portfolio of investments; we couldhavemoreconfidenceinDisruptionifwehadanotherportfoliofromadifferentcompanytoworkwith.Andsoon.BetteranswersthanwhatIcanofferherewillonly be possible with more experiments. However, I continue to believe thatdespite these shortcomings, the findings so far place Disruption on aqualitativelydifferentandfirmerfootingthanothertheoriesofinnovation.Second, there is a big hole in the data: the Intel portfolio did not include a

singleDisruptive/entrantbusiness.

FIGURE8:EXPERIMENTALRESULTSFROMTHEINTELPORTFOLIO

PREDICT+EXPLAIN=TRUE

Thelackofdatainthetwoofthethree“PredictedSuccess”sectorsofthefigureabove is both ironic and vexing. It is ironic because Christensen discoveredDisruption throughhis investigationsof thedriversof successofentrantsoverwell-resourced, well-managed incumbents. The case library of this type ofDisruptive success is substantial and constantly expanding; there is a lot ofevidence that the Disruptive/entrant quadrant is well populated. It was theimplications ofDisruption for the sustaining/entrant andDisruptive/incumbentsectorsthatgaverisetothepredictivetestsdescribedintheprecedingchapters.Itisnot,then,thattherearenoexamplesofsuccessesintheDisruptive/entrantsectors.Rather,thedataavailableforthepredictivetestssimplyhappenednottoinclude any.To add such cases to the samples used in the experimentswouldhavemeantincludingcasesthathadnotbeenfundedbyNBI,whichwouldhavecompromisedthecontrolsforimplementationexpertise,timeframe,andalltheotherunobservedorganizationalconstraintsthataffectoutcomes.ThelackofexamplesfromtheNBIportfoliothatfallintheDisruptive/entrant

sector is vexing because without those data there is no way directly to testwhether Disruptive/entrants actually succeed more frequently than thoseexamples falling in the other sectors. Neither canwe test the proposition thatbeing trained in Disruption allows people to spot Disruptive successes withgreateraccuracy.Asaresult,thereisnoexperimentaltestofthecentralclaimofthis book: that success lies with launching either (a) Disruptive attacks onmarketsyoucurrentlydonotcompeteinor(b)Disruptiveattacksonyourhomemarketsviawhollyautonomousunits.Infact,acriticalobservermightarguethatthedatapresentedsofarcouldjust

as well be interpreted as support for the long-standing view that incumbentcompaniesshould“sticktotheirknitting.”Afterall,asshowninfigure8,Inteldid finewith innovations that fell in the sustaining/incumbent sector andwithoneexceptionfailedwithinnovationsthatfellinanyothersector.Theonlyway, fornow, thatwecanbuildacase forDisruption’spredictive

powerbeyondtheexperimentaldata is tocombine thedemonstratedpredictivepower in thepopulated sectorswithexplanatorypowerusingcases thatwouldhavefalleninthosesectorsthatare,inthecaseofIntel’sportfolio,empty.Thisrequiresmeetingtwotests.First,thesamecausalmechanismsthatexplainthosecircumstanceswhereDisruptionhasdemonstratedpredictivepowermustimply

successforDisruptorsintheemptysectors.Second,Disruptionmustbetherightexplanation (not just a plausible one) for observed successes by DisruptiveentrantsorDisruptivebutautonomousunitswithinincumbentorganizations.This second criterion is more demanding than it might at first seem. It is

typicallythecasethatforanygivensuccessorfailurethereareanynumberofcompeting explanations, and it is often difficult to demonstrate definitivelywhich explanation is right. Why did the Sony Walkman succeed but theMiniDiscfail?WhydidtheAppleNewtonfailbuttheiPodsucceed?Therearelots of opinions, yet strangely which explanation any of us finds convincingoftenseemstoboildowntoamatteroftaste.Consequently, toestablishDisruption’sexplanatorypower Imustbeable to

show that Disruption explains the observed patterns of success and failure aswellasorbetterthanthecurrentleadingcandidates.Thiscannotbedonebyanexhaustive analysisof everypossible case. Instead, consider the following testcase, an instance of an entrant’s success that has been extensively studied butthat, I will argue, is best explained as an archetypal case of Disruption:SouthwestAirlines.

WHATNEEDSEXPLAINING?

Somereaderswillberollingtheireyesat thispoint.“NotSouthwestagain?!?”The company has been the subject of hundreds of books and thousands ofarticlesovertheyears.Howcantherepossiblybeanythingnewtosay?The standard explanation for Southwest’s performance has been its

development and continuing commitment to what has become known as the“low-cost carrier” (LCC) model. Thanks to a great deal of sound theoreticalworkand rigorous field research, thedefiningattributesof thismodelarewellunderstood.Theyinclude,amongotherthings,flyingonemodelofplane(ratherthanseveral,eachchosenforitssuitabilityforadifferenttypeofroute),apoint-to-pointroutestructure(ratherthanhub-and-spoke),oneclassofservice(ratherthanseveral),nomeals,noassignedseating,andsoon.ButwhataspectsofSouthwest’sperformancedothesefeaturesexplain?The

company’s profitability started out strong in 1980 but declined steadily from1980through1991,onlytoturnaroundandimprovefortenyears,untilitbegana long-term decline in 2000. Revenue grew relatively slowly from the firm’s

inceptionuntil1991whenafifteen-yearrunofstrongrevenuegrowthbegan.Asforitsstockprice,that,too,showedmaterialbutmoderateappreciationuntiltheearly1990s,when,foralmosttenyears,thecompanydeliveredamongthebestratesofreturninthehistoryofAmericanbusiness.

FIGURE9:SOUTHWEST’SPERFORMANCE,1980–2008

Any credible explanation of Southwest needs to explain these changes inperformance. Yet the LCCmodel, which is claimed bymany to be what hasdriventhecompany’sperformance,didnotdemonstrablychange.Thecompanycertainlygotbetteratwhatitdidovertime,butthemagnitudeanddirectionofitschangesinperformancecryoutforafarmoredynamicexplanation.Afterall,SouthwesthadthesameLCCmodelin1981thatitdidin1987,yetitsreturnonassets(ROA)hadfallenfrom12percentto2percent.Itsfundamentaloperatingmodelin1993wasthesameineverymaterialrespectasithadbeenin1990,yet

itsROAhadclimbedfrom3percentto6percentwhileitsstockpricehadrisen650 percent—even as the industry was stagnant or declining. The challenge,then,istoshowhowDisruptionprovidesageneralyetaccurateexplanationforthesesignalfeaturesofSouthwest’sperformance.Disruption as it is has been developed and applied to date offers an

explanation forSouthwest’s success.This standard formulation, however, failstospecifyinsufficientdetailpreciselyhowthecompanywasableeventuallytocompete for a sizable chunk of the U.S. commercial aviation market. Recallfromchapter1thatcentraltoasuccessfulDisruptionistheabilitytoimproveinways that enable growth, typically by competing successfully for moremainstreammarkets that incumbents value,without sacrificing the advantagesthat made success possible in the foothold markets that incumbents wererelativelywilling to cede.There are still othergaps in the standardDisruptionmodel, which collectively have left Disruption merely “one explanation” forSouthwest’ssuccess,not“theexplanation.”Inordertoclosethesegaps,wewillhavetoreviewhowDisruptionisusually

described and apply that description to Southwest. That will allow us to seeprecisely where it falls short and so reformulate the theory in terms that willultimatelynarrowitsfocusbutincreaseitsexplanatorypower.According to theDisruption canon, a newentrant designs a businessmodel

tailoredtotheneedsofcustomersthatareeconomicallyunattractivetopowerfulincumbents.Success in this relativelycompetitively isolatedmarketconstitutesthe entrant’s foothold.As the entrant refines its offering and then improves itover time, it is eventually able to compete for established markets in waysincumbentscannotmatch—theupmarketmarch.TheprocessofDisruptionhastypicallybeencapturedinwhathasbecomeknownasthe“SAM”(surface-to-airmissile)chartofDisruption.

FIGURE10:THESAMCHARTOFDISRUPTION

At , incumbentsarefocusingontheneedsofthosecustomersthatprovidethecombinationofvolumeandmarginthatmaximizestheirreturns.Therationalpursuit of profitmaximization leads them to “overshoot” theneedsof the lessdemanding and less economically attractive lower end of the market—or,alternatively, to ignore other smaller, newmarkets that offer lower returns inspiteofhighermargins.This overshoot of low-end markets, or ignorance of new markets, is what

allowsnewentrantstofindtheirfootholdat .Successinthefootholdmarketdemands a business model that is different from that employed by theincumbents,onethatisdesigndtomeetprimarilytheneedsofthesemarketsormarket segments. Incumbentsdonot respondbecause theyhavebigger fish tofryintheirmainstreammarketsandaremotivatedtopursuestillmorelucrativesegments farther upmarket. The “inability” of incumbents to defend againstentrants’ incursions and their establishment of a foothold is not a function oforganizationalincompetencebutratherofsoundeconomicreasoning:whygiveupadollartodefendadime?Thetraditionalstrategyemployedbyincumbentairlineswasbasedonnational

and international hub-and-spoke (H&S) route structures designed to serve as

efficiently as possible as many cities as possible. Smaller aircraft flew lessheavily traveled routes and fed larger aircraft serving the more populardestinations.Variousclassesofservicecatered to thedifferent requirementsofdifferentmarketsegments,andsoon.When it comes to servingspecificpairsofcities,however,dedicatedplanes

flyingpoint topointare farmoreefficient.Southwest,whichoriginallyservedonlytheDallas—Houston—SanAntoniotriangle,wasabletooffermuchlower-cost service by dedicating its fleet and its operations for just those routes.Incumbentairlineswerenotmotivatedtorespondespeciallyvigorouslybecausedoingsodistractedthemfromthefaster-growing,moreprofitableopportunitiesin their nationwidenetwork. Inotherwords,Southwesthad found its footholdwith theLCCmodel,andinsodoinghadcreatedrelatively littleheartburnforthemuchlargerincumbentairlines.WhatSouthwestwasabletodo,however,wasexecutethe“upmarketmarch”

( in the diagram). In general terms, this takes the form of the entrantimprovingitsofferingsinwaysthatmattertothemainstreamcustomerswithoutsacrificingtheadvantagesthatmadeitprofitableinitsfoothold.Thispartofthedefinition is crucial: if the entrant competed for the mainstream markets byadopting a business model similar to the incumbents’, all we would get isanother combatant in a messy bar fight. In such battles, incumbents typicallyprevail.But Disruptive entrants are fundamentally different. The solutions that

establish their footholds prove able to improve and bring to market betterofferings, yet preserve the cost or other advantages that drove success in therelatively unattractive (to incumbents) footholdmarkets. This leads to : theentrant has improved enough to compete for mainstream markets, yet haspreservedenoughofitsfootholdadvantagestodisplacetheincumbentfirms.For Southwest thismeant the ability to fly tomore andmore cities and to

appeal to more and more customers over time without compromising thedefiningelementsoftheLCCmodel.Allthewhile,incumbentairlinesremainedhandcuffed by their H&Smodel, with all its complexity and implications forcost.Totheskepticalreader,claimingthatSouthwesthada“differentmodel”ina

“foothold” and then executed an “upmarket march” might seem no differentfrom the more traditional explanation based on the improvement of the LCCmodelandtheconsequentgrowthofcarriersemployingit.Inotherwords,howis aDisruption-based explanation different from and better than one based onconventional notions of strategic differentiation and market penetration? Inparticular, nothing so far explains precisely how entrants that begin with

manifestly worse solutions targeted at customers incumbents do not want areable eventually to offer manifestly better solutions targeted at customersincumbentsdowant,withoutcopyingtheincumbents’strategies.Neither does this mapping of Southwest’s success onto Disruption’s SAM

chart explain why Southwest in particular and Disruptors in general prevailwhentheycomeintodirectcompetitionwithincumbents.Whycanincumbentsnot respond effectively once the dire nature of a Disruptive threat becomesclear? One can accept easily enough why incumbents have no interest indefending a would-be Disruptor’s foothold, but successful incumbents haveenormousincentivetofightbackonceDisruptorsarecompetingfortheirbreadandbutter.Whyshouldwebelievethatincumbentsthatcaninnovateeffectivelyinpursuitofprofitandrepelnearlyanysustainingattackaresuddenlycaughtinthe headlights when it comes to innovating in the service of survival whenthreatenedbyaDisruptoronanupmarketmarch?We are left, then, with three questions: (a) what is the difference between

establishinga footholdmarketandstrategicdifferentiation; (b)whatdrives theupmarket march; and (c) why are incumbents incapable of countering trulyDisruptive attacks? Answers will provide a Disruptive explanation ofSouthwest’s specific patterns of profitability, growth, and share-priceappreciation.

STRATEGY,INNOVATION,ANDDISRUPTION

ItwillhelptobeginbyrelatingDisruptiontoanestablishedconceptofstrategyandthebroaderconceptofinnovation.

STRATEGY:EXPLOITINGTRADEOFFS

In his 1996 article “What Is Strategy?” Harvard Business School professorMichael E. Porter synthesizes over twenty years of writing, research, andreflection on the implications of microeconomic theory on businesscompetition.13Heconcludesthatdifferentstrategiesaredefinedbythetradeoffsin theperformanceof theactivities thatdefine thevaluecreatedbyabusinessmodel.14 Porter illustrates this framework using two dimensions of customervalue:priceandnonprice.Thepricevalueacompanycanprovideisafunctionofitsrelativecostpositionversusthecompetition,sincethepriceitcanchargeover time cannot fall below its total costs (which include a capital charge).Nonprice value is a vector of all the different dimensions of performancecustomerswant. In the caseof automobiles, for instance, thismightbe safety,acceleration,styling,roominess,andsoon.Deliveringanygivenbundleofnonpricebenefitsalways incursacost—it is

tough, after all, to get something for nothing. The minimum cost required toachieve a specified nonprice value is not some fixed Platonic ideal: it iswhatever cost is incurredby the lowest-cost provider in themarket.Similarly,thelevelofanynonpricevaluethatcanbeprovidedatanycosthasamaximum:itdoesnotmatterwhatyouarewillingtopay,youcannotgetfromNewYorktoLosAngeles in fiveminutes.The limits ofwhat canbeprovided atwhat costdescribethe“productivityfrontier”forabusinessmodelatapointintime.

FIGURE11:THEPRODUCTIVITYFRONTIER

At ,afirmcanappeartobreaktradeoffsanddelivergreaternonpricevaluewithout an increase in cost; that is, it canmove “right” to (anincreaseinnonpricevalue)withoutmoving“down”(anincreaseincost).This is because a firm is merely wringing out inefficiencies that othersalreadyknowhowtoavoid.In other words, at a firm really can get something for nothing by

workingsmarterratherthanharder.Firmsthathavereachedthefrontierofwhat a given business model can do are “operationally excellent,” inPorter’sterms.Once a firm gets to , however, that is as smart as it can work: the

frontierdefinesthelimitsofwhatispossibleatthatmoment.Ofcourse,onecould exploit different types of tradeoffs to reach a different point on thefrontier,competinginsteadat bymoving“up”(areductionincost)fromwithoutmoving“left”(areductioninnonpricevalue).Oncefirmsareat

thefrontier,however,changesincostandnonpricevalueareinextricablylinked:moreofonenecessarilymeanslessoftheother.Thatmakes and

qualitativelydifferentstrategiesbecause theyareatdifferentpointsonthesamefrontier.

Porterapplies this thinking toSouthwestAirlines,explaining thecompany’ssuccessintermsofitsuniquestrategicposition.ItsLCCstrategyplaceditaton the frontier, whereas most of the major H&S air carriers were clusteredaround with largely similar combinations of price and nonprice attributes(e.g., meals, frequent-flyer points, in-flight entertainment, on-time departures,connectingflights,andsoon).AccordingtoPorter,Southwest’sLCCmodelisbothprofitableandrelatively

immunetoeffectivecompetitiveresponsesbyH&Sairlinesnotmerelybecauseit is different in the ways already specified but because those differencesmanifestverydifferenttradeoffsthatH&Scarrierscannotbreak.Forexample,apoint-to-point route structure obviates expensive hubs.ButH&S carriers needhubs tokeep theirplanes full.TheLCCmodel typically routes flights throughsecondary airports, which keeps landing fees low and reduces congestion,enabling fast turnarounds, frequent, reliable departures, and more nearlyoptimizingtheuseofitsaircraft.ButH&Scarrierscatertodemandingcustomersegmentswhowill not tolerate the significantly longer drives from secondaryairportstomajorcitycenters.BecauseLCCcarrierstypicallyuseonlyonetypeof plane, very often the Boeing 737, flight crews are more nearlyinterchangeable and everything frommaintenance to grooming the equipmentbetween flights is consistently faster and more efficient than at competingairlines. But H&S carriers need smaller planes for smaller routes and largerplanes toaccommodatemore in-flightamenitiesforbusiness-class travelers. Inshort, the LCC and H&S models occupy very different and fundamentallyincompatiblepositionsontheproductivityfrontier.What aspects of Southwest’s performance does this positioning model

explain?Debutingin1971,thecompanyquicklydominatedasmallnicheoftheairline market. But for the ensuing twenty years there was little that wasremarkable about Southwest’s performance. After a period of potentiallymisleadingdouble-digitgrowth(thankstohavingbegunfromasmallbase),thecompanyseemedhemmedin toasmallandtightlydefinedniche.Throughthe1980s, Southwest ranked fourth in share-price appreciation behind the twolargestairlines,UnitedandAmerican,andanothernicheplayer,AlaskaAirlines.Itwassecondinrevenuegrowthoverthedecade,butstillfromasmallbase:intotal revenue it was sixth of the seven listed below, ahead only of AlaskaAirlinesby about $140million in revenue, or 13percent ofAlaska’s top line.What did set the company apartwas its profitability, ranking a strongnumber

oneinreturnonassets(ROA).

FIGURE12:PERFORMANCEDATAONSELECTEDU.S.-BASEDAIRLINES:1980–1990

This performance record seems to have strongly influenced how analystsviewed thecompanyat that time.Theydescribed its strategy inessentially thesame way Porter and many others have since, leading many investors toconclude that the company had staked a claim to a profitable position withlimited growth potential on the shared frontier of the airline industry. Thecompanywasgenerallyrateda“hold,”andinthewordsofoneanalystwritinginNovember1989,itwasa“buy”onlyfor“longterminvestorsfocusingonqualitycompaniesoperatinginwell-establishednichemarkets.”15Famouslastwords.TheverynextyearSouthwestbeganadecade-longrunof

growth, increasing profitability, and astonishing share-price appreciation.Secondonly toContinentalover thedecade in share-priceappreciation (whichwasreboundingfromadismalpriortenyears),thecompanyledthepackinbothrevenuegrowth—despitenowbuildinguponamuchlargerbase—andROA.

FIGURE13:PERFORMANCEDATAONSELECTEDU.S.-BASEDAIRLINES:1990–2000

We cannot explain this shift in terms of positioning because Southwest’spositionontheindustry’sproductivityfrontierneverchanged.EveryattributeofSouthwest’sLCCbusinessmodel thatsomanycommentatorshaveadducedasan explanation of its success was in place pretty much from the start, andcertainlyunderwentnodemonstrable,material,relevantshiftbetween,say,1990and1993,aperiodduringwhichthecompany’sstockpriceincreasedmorethansixfold.Sowhatdidhappen?

INNOVATION:BREAKINGTRADEOFFS

An answer requires that we take seriously an objection that has often beenleveled against the notion of “strategy as positioning” but that has remainedsomewhat nebulous: the dynamics of strategy. The durability of strategicpositionshasbeendecliningforsometime:manycommentatorshavenotedthatthe tenure of companies in theFortune 500 has been falling for decades.Theallegedcauseofthiscompetitiveturmoilistheabilityofthebestcompaniestochange their strategic positions quickly or of new entrants to define new andsuperiorpositionsandimplementthemrapidly.This does not make strategy dynamic, however; it merely casts strategic

change in termsof a successionofdifferent strategicpositions—similar to theillusionofmovementcreatedbyaseriesofstaticimages.Tomakestrategy—theexploitationoftradeoffs—trulydynamic,weneedtointroducesomemechanismbywhichcompaniesbreaktradeoffsovertime.Tradeoffsdefinethelimitsofwhatispossibleatapointintime,notwhatis

possible forall time.We learn.We improve.We innovate. Inotherwords,wefigure out how to getmore for less. From the pace of technological advance,capturedinsomethinglikeMoore’slaw(whichdescribesthedecliningcostand

rising power of microprocessors), to process improvements captured in the“learningcurve”(whichquantifiestherateofcostreductionsforeachdoublingofcumulativevolume),organizationshavebeenwildlysuccessfulineventuallybreakingthetradeoffsthatdefinethefrontierofanygivenbusinessmodel.Companies are motivated to innovate (that is, break tradeoffs) because

innovation holds the promise of enormous growth. By breaking tradeoffs, acompany is able to reach a point in “strategic space” that competitorsfundamentally cannot.The first firm to incorporate a given innovation into itsbusiness model can deliver, depending on the nature of the innovation,performance or price that competitors simply cannot match. For example,automakers might differentiate themselves based on cost and power becausethereisatradeoffbetweenthesetwoattributes.Innovationsinenginetechnologyallow lower-cost engines to deliver ever greater power. The company thatintroducesthisinnovationfirstenjoysaninnovation-basedadvantage,deliveringhigher-powered automobiles at prices competitors cannot match. The mostpowerful of such innovations do not even require material change in anorganization’s business model or even its strategy; rather, a technology thatdefined part of the productivity frontier for that business model broke animportanttradeoff.

FIGURE14:INNOVATIONEXPANDSTHEFRONTIER

An innovation thatexpands the frontierofabusinessmodel issustaining tothatbusinessmodel.Thankstothemarketplacefornewtechnologyandprocessimprovements, sustaining innovations typically disseminate across an industryquite quickly, regardless of the strategies being pursued by the differentcompanies in that industry, and somost competitors typically catch up, oftenquite quickly. In our automotive example, other car companies would beexpected to reverseengineer thenewpowerplant and incorporate the relevanttechnologies into their engines.When this happens, competition is once againbasedonstrategicpositioningalongagivenfrontier.Andso,despiteitsseemingattractiveness,thissortofinnovation-basedadvantagerarelylastslong.Itisforthis reason that we are often told that a given innovation is not a source ofcompetitiveadvantage.Instead,somearguethattheabilitytoinnovate—tokeeppushingoutthefrontierandtohaveyourcompetitionconstantlyplayingcatch-up—isthekeytosustainedoutperformance.

DISRUPTION:DIFFERENTTRADEOFFS,FASTEREXPANSION

Consistent and industry-leading sustaining innovation can be a formula forlong-term success. But sustaining innovation is of little use when seeking toenternewmarkets.Afterall,thecentralphenomenonChristensenexplainswithDisruption is the toppling of successful, well-managed organizations byseeminglyunderresourcedupstarts.Howwasit that thesecompanieswereableto enter the disk-drive industry and ultimately overtake well-managedincumbents, even as incumbents vigorously and continuously expanded theirownproductivityfrontier?Disruptorsdonotmerelypick adifferent spoton the frontierof an existing

business model. Instead, they create a new business model with an entirelydifferentfrontier.Takethepersonalcomputer(PC)asanexample.Beginningasatoytargetedathobbyists,improvementsincomponenttechnologyallowedPCstogetgoodenoughtoperformtasksthathadpreviouslybeenthesoledomainofminicomputers, yet do so at costs that were orders of magnitude less.Consequently, when PC makers got their start, they were not choosing adifferent spot on the frontier defined by the dominant business model in theminicomputer industry. In other words, they did not merely have a differentstrategy but had instead created an entirely new businessmodelwith its own,verydifferentfrontier.Consequently, thoseearlyPCswerenot justworse thanminicomputers, theywereworse thanminicomputerscouldpossiblybe,andatpricesthatminicomputerscouldnevermatch.

FIGURE15:DISRUPTIONISAFUNCTIONOFTHERELATIVENATUREANDPACEOFFRONTIEREXPANSION

PCandminicomputermakerseachpushedtheirfrontiersoutwardthroughtheirownsustaininginnovations.ThePCdisruptedtheminicomputerwhenitsfrontierexpandedinwaysthatallowedPCstodeliverasuperiorbundleofpriceandnonpricevaluetoenoughcustomersthatitsignificantlyerodedthegrowthandprofitabilityoftheminicomputermakers.

Sustaining innovation was taking place along both frontiers. Minicomputermakerswere lockedin theirownsustainingbattlesamongthemselves, thebestof themworking feverishly, incessantly, andwith great creativity to push outtheir frontier by improving performance and cutting costs. Those innovationsbroke real tradeoffs and gave minicomputer customers more for less whileaffordingthemostinnovativeminicomputermakersabriefopportunitytoenjoysuperiorpricingpowerandgrowth.Persistentoutperformancecomparedtootherminicomputermakerswasafunctionofconsistentsustaininginnovation.However,thefrontierofthePCmakers’businessmodelwasalsoexpanding,

thankstotheirownsuccessionofsustaininginnovations.Therateandnatureofthatexpansionwasdrivenlargelybyinnovationsintheircomponents,andthose

componentsgotsogoodsofastthatPCseventuallycaughtuptoandsurpassedthe frontier of theminicomputermakers’ businessmodel for all but themostdemandingcustomersegments.InTheInnovator’sSolution,ClaytonChristensenandIidentifiedtwotypesof

Disruption,basedonthedifferenttrajectoriesofimprovementtheyfollow.16InthePCexampleabovewesee“lowend”Disruption,socalledbecausetheinitialmarket segment targetedby theeventualDisruptor settles for seeminglyworseperformanceinexchangeformateriallylowercosts.“Newmarket”Disruptionisamirror imageof this:aneventualDisruptormaywellbechargingsignificantsums for a solution, but it offers a very different bundle of nonprice valuecomparedtoincumbentsolutions.Overtime,thisfrontierisexpandedthankstocostreductionsthatdonotunderminetheperformancefeatures.Afavoriteexampleofanew-marketdisruptionistheriseofmobiletelephony.

Whenfirstlaunched,mobiletelephonyservicesweremuchmoreexpensiveandlessreliablethanlandlinetelephony,butmobilephoneshadonefeaturelandlinephones could not match: they were mobile. Over time, technologicalimprovements made it possible to reduce the cost and price of the service toconsumers even as the quality of the services improved. This resulted in theDisruptionoflandlinetelephonybymobileservices.Now,smartphonessuchasthe RIM BlackBerry and Apple iPhone are raising the ante, making mobile“phones” more expensive, but this time in the service of Disrupting personalcomputers:thenewdevicesallowconsumerstoaccesstheInternetandperformanynumberofothertasksthatwereoncethesoleprovinceofthePC.

FIGURE16:NEW-MARKETDISRUPTION

Mobiletelephonyfounditsfootholdbyprovidingalevelofnonpricevaluethat landline telephony could not match at any price. Then, as mobiletelephony’s underlying technologies made both phones and networkconnections less expensive and more capable, mobile services Disruptedthelandlineinfrastructure.

In short, all innovation is about breaking tradeoffs. Sustaining innovationsbreak the tradeoffs that define a particular frontier by pushing that frontieroutward.Disruptiveinnovationsarethosethatpropeladifferentcurveoutwardinwaysthatallowiteventuallytoovertakethefrontieroccupiedbyincumbentplayerscompetingforadifferentmarketormarketsegment.

CHANGINGISHARDERTHANNOTCHANGING

Competitors employing essentially the same businessmodel are subject to thesameconstraintsandtradeoffs.Whatisdifferentistheirstrategies—thatis,theirrespectivepositionsonthefrontierdefinedbytheirsharedbusinessmodel.Asaconsequence,one’scompetitorscan typicallycopysustaining innovationratherquickly.ADisruptiveentrantandtheincumbentaredifferentinanumberofways,but

fromtheperspectiveofDisruption,theonethatmattersmostiseachcompany’srelative ability tomakemoney in the entrant’s footholdmarket. Entrants thatultimatelyproveDisruptivehaveabusinessmodel that ismoreprofitable thanthoseofincumbentsinthosefootholdmarketsyetistypicallylessprofitablethanthe incumbents’ model applied in the markets incumbents are motivated tocompetefor.ThefirstPCmakersfoundearlysuccesswithabusinessmodelsuitedtothe

needsofhobbyistswhovalued theearlymachinesdespite their limitations.Tobeprofitable,PCmanufacturershadtorelyonmuchcheaperandconsequentlyinferior technology and modular designs. As a result, their computers were,unsurprisingly, worse than the minicomputers and mainframes purchased bymainstream computer users, which were at the time largely corporations,universities, andgovernments. In exchange, though,PCmakers suchasAppleComputer were able to make the machines cheaply enough and in sufficientvolume to serve a market niche that was of no interest—at first—to othercomputermanufacturers.Inotherwords,theycreatedanewproductivityfrontierforanewproducttargetedatadifferentmarketsegment.Minicomputermakershad no interest in exploring that frontier because they had bigger and moreprofitable opportunities in the markets they already served using a businessmodeltheyhadalreadymastered.SuccessforthePCmakersinthisfootholdwastransformedintoanultimately

Disruptiveupmarketmarchthankstoimprovementsintechnologiesthatdefinedthe cost and performance profile of the personal computer: disk drives,microprocessors, software, communications networks, monitors, sound cards,videodrivers,andsoon.Aseachof theseimproved, thePCgotbetterandPCmakerswere able to compete for larger,more demanding, andmore lucrativemarketswithouthavingtochangetheirbusinessmodel.In contrast, for incumbent minicomputer makers to exploit the improving

capabilitiesofthecomponentsdrivingthePC’sascendancytheywouldhavehadtochangejustabouteverythingabouttheirbusinessmodel,includinghowtheydesigned,manufactured,sold,andsupported theirproducts.Worse, respondingto the rapid substitution of the minicomputer by a PC-based IT architecturewouldhaverequiredthatminicomputermakerschangejustabouteverythingjustaboutimmediately.This last bit is critical. Changing a business model, thanks largely to the

organizational change this typically implies, is crushingly difficult. Disruptivegrowth does not require much change for the Disruptor. Disruptors build abusinessmodelthatissuccessfulinthefootholdmarketbutthenareabletorideupmarketonimprovementsinthekeytechnologiesthatdeterminetheirproductor service’s performance and cost profile.Whatever change is required for aDisruptor to march upmarket is nothing compared to what incumbents arerequiredtotakeonwhenattemptingtoturnandfaceaDisruptivechallenge.Certainlythesortofchangerequiredforincumbentstofightbacksuccessfully

is possible. There are many well-known cases of corporate reinvention, fromIBM’s shift from computer hardware to various IT-related services andconsulting to Nokia’s transformation from a forestry company to a mobilecommunicationspowerhouse toSamsung’s successful transition fromawidelydiversified chaebol with close ties to the Korean government to a globalcompetitorwithawell-knownbrandandmanymarket-leadingproducts.ButallI need for my argument to go through is for not changing to be easier thanchanging.Andthatdoesnotseemtooheroicanassumption.Itisthisdifferenceinthelevelandtempoofchangerequiredthatleavesthe

Disruptoratadistinctandstructuraladvantageovertheincumbent.Thisiswhyincumbents, however well managed or successful, typically cannot fend offgenuinely Disruptive attackers. The Disruptor has fundamentally changed thegame, and by the time the incumbent realizes this, it no longer has the timeneededtolearnthenewrulesormasterthenewskills.

DISRUPTIONTAKESFLIGHT

ForSouthwest’s success tobea functionofDisruption, rather thanstrategyorevengenericinnovation,threethingshavetobetrue:

1.Southwesthadtocreateaproductivityfrontierthatallowedittoreachapointincost/valuespacethatincumbentscouldnotreachwiththeirexistingbusinessmodelandhadnointerestinreachingwithanewbusinessmodel.2.Southwest’sfrontierhadtoexpandfasterthanthatofincumbentswithoutSouthwesthavingtomakematerialchangestoitsbusinessmodel.3.Forincumbentstorespondeffectively,theywouldhavehadtomakematerialchangestotheirbusinessmodels.

ANEWBUSINESSMODEL

Some very carefulwork has demonstrated that Southwest enjoys amaterialcostadvantageoveritscompetitors.17Oneestimateattributesfully70percentofitsadvantagetothecompany’sbusinessmodel.

FIGURE17:DRIVERSOFSOUTHWEST’SCOSTADVANTAGEOVERINCUMBENTAIRLINES

Itisonethingtohaveadifferentbusinessmodelwithlowercosts.TheclaimthatSouthwestwasatrueDisruptor,however,requiresthatthisbusinessmodeldescribeanewproductivity frontier; that is, incumbentH&ScarriersmustnothavebeenabletoreachSouthwest’scostswithintheconstraintsoftheirbusinessmodel. Only then can it be shown that Southwest was reaching a point incost/performance space that H&S carriers could not, as opposed to the moreconservativeclaimmadebyPorterandmanyothers thatSouthwesthadstakedoutthelow-costendoftheindustry’sfrontier.Althoughwe do not have data fromSouthwest’s start-up days,more recent

analysis suggests that today Southwest and LCC airlines indeed are on adifferentcostcurve.Thespecificshapesinfigure18capturethescaleeconomiesof theindustry,for ingeneral the“averageseatmile”(ASM)costdropsas theaverage stage length (route distance) increases. (Note that the curves are,correctly, not extrapolatedmuch beyond the observed data: the LCCs operateacrossamuchnarrowerrangeofroutelengthsthandotheH&Scarriers,andforreasons explored below, their costs would likely increase substantially if they

were to fly longer routes.) In light of the consistency of Southwest’s businessmodel over the years and its sustained profitability advantage over incumbentcarriers, it does not seem too much of a leap to believe that these data arereflectiveofSouthwest’srelativecostpositionformuchofitsexistence.

FIGURE18:RELATIVECOSTCURVESFORLCCANDH&SCARRIERS

Thanks to this fundamentally different business model, incumbent airlineswouldbeunlikely tomatchSouthwest’s costwith theirbusinessmodel at any

levelofoperationalexcellence.Thatwouldhavebeentryingtoeatsoupwithaknife. For instance, Southwest has only one class of service, but few H&Sairlines would think of sacrificing the richmargins earned on the last-minutepurchasesoffirst-classtickets.Conversely, Southwest’s differentiation left it quite likely incapable of

deliveringthekindofservicethatwasthecoreoftheH&Saircarriers’business;thatwouldhavebeencuttingsteakwithaspoon.Forexample,Southwestcouldnotofferdifferentclassesofserviceandassignedseatswithoutcompromisingakeyelementofitsprofitability:quickgroomingofaircraft,noneedforcomplexprovisioning(e.g.,nomeals),andfastboarding(noassignedseats),allofwhichreduce turnaround time.Those attributes, however, came at a price: nomatterhowfast the turnaround,Southwestwasatadisadvantagewhencompetingforhigh-maintenancebusinesstravelers.

ANEXPANDINGFRONTIER

Sofar, there isno realdifferencebetweena“strategy-based”explanationofSouthwest’s performance and a “Disruption-based” explanation: both merelyaccount for the company’s profitability. The challenge is to explain whySouthwestdidnotgrowforsolongandthengrewsoquicklyandsosuddenly.Incumbentairlines, spurredprimarilyby theircompetitionwithoneanother,

expandedtheirfrontierthroughtheirownsustaininginnovations.Someoftheseinnovationswere their doing, such as the establishment of feeder networks ofregionalairlines inorder toboost load factorsand increase their serviceareas;lounges; frequent-flyer programs; and so on. Others came courtesy of theirsuppliersintheformoflargerorsmallerplanesbettersuitedtothedemandsofspecific routes. These innovations broke the tradeoffs that had defined thefrontier,allowingthebestincumbentairlinestoofferbetterandbroaderserviceatlowerpriceswhilestillgrowingandremainingprofitable.Southwest, meanwhile, was hard at work pushing its own frontier outward

withitsownsustaininginnovations.Thecompanyimproveditsuniqueapproachto motivating and empowering employees; it wrestled with ground-crewefficienciesandflight-crewscheduling;itevenadoptedsomeoftheinnovationsoftheincumbents,introducingitsownfrequent-flyerprogramin1987.Despiteall these improvements, however, even after almost twenty years of relentlesseffort, the company had yet to appeal, relative to the incumbent airlines, tomaterialnumbersofcustomers.Whatwasholdingitback?Southwest’sgrowthprospectswerehemmed inonboth the lowendand the

highendbythenatureofitsbusinessmodel.ThereisalimittohowshortarouteSouthwestcanserve,determinedinlargepartbythefixedcostoftakingaplane:getting to theairport,parking, clearing security, and so forth. In somecities itcan take ninetyminutes to get from your driveway to the boarding gate for athirty-five-minuteflight,afterwhichcomesanother thirty tosixtyminutes inacar to get to your destination—especially if you are flying into a secondaryairport.Whenthealternativeisathree-hourdrive,thecarwins.Inotherwords,there are tradeoffs defining Southwest’s lower limit that Southwest cannot domuchaboutandthatstructuralconsiderations—suchaswhereyoucansafelyputanairport—makeitdifficultforanyonetobreak,ever.Similarly,thereweretradeoffsdefiningSouthwest’supperlimit.Itcouldonly

openslotsatairportssoquickly;itwasconstrainedbytheWrightAmendment,whichprohibitedanyoneflyingoutofDallas’sLoveFieldfromflyinganywherebutwithinTexas or the four states adjoining it; and itwas constrained by theeconomics of flying the 737-200 (the only plane Southwest flew until 1984),which cost approximately one cent per revenue passenger kilometer more tooperate than the 747,DC-10, or L-1011s being flown byH&S carriers at thetime.18This last element is especially critical. Many commentators have observed

thatSouthwest foragoodpartof itshistory flewonlyshorter routes, typicallylessthanfivehundredmiles(eighthundredkilometers).Thisisoftenportrayedassomesortofkeenstrategicinsight.Farmorelikelyisthatsincethe737costaboutapennymoreforeverypassengerkilometerflown,carryingonehundredpassengersonethousandmilescostSouthwest$1,600morethanitscompetitors—very likelya sizablechunkofSouthwest’son-the-groundcostadvantage. Inotherwords,Southwesthad to flyshorter routesbecause itcouldnotafford toflylongerones.Looking at the big picture, we see that from 1975 to 1991 Southwest was

walkingatightrope.Togrow,thecompanyhadtoincreasethenumberofroutesit served.Thanks to its industry-leadingprofitability, thisgrowthcreatedvalueanddrove thecompany’s stockup.However, the routes thecompanychose toadd tended to be of gradually increasing length, and thanks to the relativeinefficiencyofthe737at thetime,theselongerroutesweremoreexpensivetoserve. The effect was a steadily declining ROA, which served to limit thecompany’srevenuegrowthandstockprice.The introduction of the more efficient 737-500 was just what Southwest

neededtoturneverythingaround.Witharange53percentgreaterthanthe737-200—at4,449kilometers,itcouldgoanywhereinthecontinentalUnitedStates—Southwestcouldnowflyjustaboutanywhereitwantedto.Andwithadirect

operatingcostper revenuepassengerkilometer (DOC/RPK)within0.5¢of theother 7-series Boeing jets, Airbus 320s, and MD-11s flown by competitors,Southwest was no longer giving back its other cost efficiencies when it flewlongerroutes.19Southwestquicklyincorporatedthenewplaneintoitsfleetandexploiteditsgreaterrangeandlowercoststobeginaddingmoreroutesofgreateraveragelength.Inadditiontosimplyservingmorecities—whichdroverevenuegrowth—asSouthwestbuiltoutitscoverage,manybusinesstravelerswhomightwellhavestayedwithincumbentprovidersforshort-haultripsinordertobuildtheir frequent-flyer account balances could now switch enough of their totalbusinesstoSouthwesttoenjoysimilarbenefits.Inotherwords,amoreextensiveroutestructuremayhavehadasecond-ordereffectonrevenuegrowth.Giventhesensitivityofairlineeconomicstoloadfactors,evenmarginalincreasescanhaveasignificantimpactonoverallprofitability.

FIGURE19:EVOLUTIONOFROUTESTRUCTUREANDFLEETCOMPOSITION:1971–2009

With both growth and profitability increasing rapidly, the stock price rosequickly and significantly. All three performance measures continued on anunbrokenupwardtrajectoryuntil2000,bywhichtimeSouthwesthadbecomeamajorplayerintheU.S.commercialaviationmarket.Inshort,the737-500wasto Southwest what better microprocessors were to PC makers: the keytechnologythatmadeitpossibleforacompanytoimproveitsofferingswithouthavingtochangeitsbusinessmodel.We can translate this evolution of Southwest into the graphical language of

productivityfrontiers.

FIGURE20:THEPACEANDNATUREOFSOUTHWEST’SRELATIVEFRONTIEREXPANSION

Southwest’sandincumbentairlines’respectivefrontiersexpandedovertheintervalindicatedby from1971to1990.Eachpushedoutbyrelativelythesameamount,leavingSouthwest’srelativecompetitivepositionlargelyunchanged.Butfrom1990to2000,thetwofrontiersexpandedtheintervalindicatedby .ThiswasenoughtogiveSouthwestadistinctadvantageincompetingforarelativelylargemarket.

Finally, the company’s relative decline since 2000 isworth examining.ThesharpdropinROAin2001wasalmostcertainlylargely(andperhapsentirely)aresultoftheaftermathofthe9/11attacks.ROAhasstayedlow,quitelikelydueto a confluence of a number of other factors. First, there has been littleimprovement in the relative cost position of the 737 compared to whatSouthwest’scompetitorsareflying.Infact,otherairlinesareenjoyingthefruitsof economical regional jets fromBombardier andEmbraer that can seatmorethanonehundredpassengers.Southwestcannotrespondinkind:itiscommittedtoonetypeofaircraft.Second,increasesintherangeofsubsequentgenerationsof the 737 are of little competitive use to Southwest. With the 737-500 the

companycouldflyeconomically(enough)coast tocoast.Thatleavesitwith—almostliterally—nowhereelsetogo:transoceanicflightswouldmeancorruptingits “U.S.-only” strategy, and seven hours on a planewith nothing but peanutsand jokes to sustain you is not likely to passmusterwith toomany travelers.Increasesinfuelpricessince2000havebeensignificant,depressingROAforallairlines.And finally,otherLCCcompanies, suchasAirTran (founded in1992and acquired by Southwest in 2010), Frontier Airlines (1994), Allegiant Air(1997),JetBlue(1998),andVirginAmerica(2004),haveenteredthemarketbycopyingSouthwest’smodel,puttinganewlevelofcompetitivepressureonthecompany’sprices.Some of these pressures—better planes for its competitors, fewer growth

opportunities for Southwest, and new LCC competitors—can be expected tolowerSouthwest’s relativeperformance.Others, suchas fuelpricesor reducedair travel,arestructural innatureandshould reduceonlySouthwest’sabsoluteROA.Whenwelookatthecompany’sperformancesince2000comparedtoallthepublicly tradedairlineswithcomparabledataover thesameperiod,weseemore evidence of absolute than of relative declines. The company’s revenuegrowthrate remainsstrong:amongmajorairlines, it issecondonly toUSAir,whichpostedstronggrowth largelydue to itsacquisitionofAmericaWest; itsROAissecondonlytoSkyWest.Perhapsmostsurprising,however,isthesharpdecline in share price: capital markets have a tendency to overshoot on bothgood news and bad, and lacking any clear sense of when Southwest’s highlyprofitable growth opportunities would run out, the company’s growth for thepriordecadehadbeenextrapolatedtoofarintothefuture.When,asmustalwaysbethecase,therealizationsetinthattreesdonotgrowtothesky,thestockpricefell—notaspunishmentforpoorperformancebutinreactiontofinallyacceptingtheimpossibilityofuninterruptedexponentialgrowth.

FIGURE21:PERFORMANCEDATAONSELECTEDU.S.AIRLINES:2000–2008

INEFFECTIVERESPONSEBYINCUMBENTS

Southwest’s low-costmodel isnotespeciallydifficult tounderstandorevenimplement effectively: from WestJet in Canada to Ryanair and easyJet inEurope,alongwithahostofothers,manyhave,tovaryingdegrees,adoptedandadapted this strategy. Yet established airlines have found it very difficult toimitate Southwest’s business model—although not for lack of trying.Continental was the first to give it a go with Continental Lite (1993–95),followed by United’s Shuttle by United (1994–2001), which overlapped withDelta’s Delta Express (1996–2003). US Airways took a kick at the can withMetroJet (1998–2001).Delta’s Song (2003–2006)was a second at-bat for theAtlanta-basedcarrier,andUnitedtriedagainwithTed(2004–2009).With six attempts by four airlines over thirteen years—during which time

Southwest’s revenue almost tripled—one cannot help but think that theconsistent failure of the “airline within an airline” approach speaks to an

underlyingweaknessinthestrategymorethananyputativelackofoperationalcompetence.Whatkeptgettingintheway?Theanswerappearstolieinthedeeplydifferentbusinessmodelsthatdefine

the twoapproaches.Shifting fromonepositionona frontier to anotheron thesame frontier—essentially changing strategy—is difficult but not impossible.Thesubsidiaryairlineslaunchedbyincumbentswereabletoachievelowercostpositions than their parent organizations. However, changing frontiers is anentirelydifferentandvastlymorechallengingtask.Inbroadstrokes,whenH&SairlinesattemptedtosetupanLCCsubsidiary,theywereforcedintocomparingthe marginal cost of leveraging existing assets such as planes, airport gates,reservationsystems,loyaltyprograms,andstaffwiththetotalcostofsettingupsomethingfromscratch.Inaddition,theseLCCsubsidiarieswerecreatingmorecapacity inwhatwasthenanoverservedmarket,whichthreatenedtobleedoffvaluablebusinessfromtheirestablishedoperations.Theneteffectwastomakeiteffectively impossible for thenewLCCdivisions tocompeteonequal termswithSouthwestanditscohortof“true”LCCairlines.20

THESIGNIFICANCEOFTHESOUTHWESTCASE

The many and unique facets of Southwest’s business model are undeniablycriticalelementsof thecompany’ssuccess.From1971 to1991Southwestwasable to carve out and defend a niche in the airline market. But the definingfeatures of the company’s strategy—its LCC model—cannot have been anexplanationforthecompany’ssustainedexceptionalperformanceinthe1990s:itsuniquebusinessmodelhadbeeninplaceforalmosttwentyyearspriortothe1990sanddidnotchangemateriallyevenasthefirm’sgrowthandprofitabilityimproveddramaticallyduringthatdecade.What did change was a critical underlying technology that allowed

Southwest’sproductivityfrontier toexpandatarateandinawaythatenabledSouthwest toDisrupt incumbentH&Sair carriers.The connections among theadoption of the 737-500 by Southwest, the change in the company’s routestructure,andsustainedimprovementsinperformancearepreciselythekindof“smokinggun”evidencerequiredtoestablishDisruption’ssuperiorexplanatorypower.And so I assert that is it not strategy that explainsSouthwest’s success, nor

merelyitsbusinessmodel.ItisthefactthatSouthwestwasaDisruptor:ithadadifferentbusinessmodelthatdefinedanewfrontierthatwaspropelledoutwardbyakeyenabling technology inaway that left incumbentsunable to respond.Disruption does not merely provide a helpful or valuable perspective onSouthwest’s success. Rather, Disruption explains what happened and when ithappenedandcanaccountforhowlongitlasted.Thatis,Disruptionistherightexplanation.Disruptionthereforehastherequisiteexplanatorypowerwhenaccountingfor

particular types of entrant success against dominant incumbents. The samemechanisms have demonstrated predictive power when applied to incumbentsuccessorfailurewithsustainingorDisruptiveinnovations.ItisonthestrengthofthiscombinedpredictiveandexplanatorypowerthatDisruptionisdefensiblyseentobetrue.But wait—how do we know when Disruption applies? What are the

“particular types of entrant success” that Disruption explains better than thealternatives?JustbecauseSouthwestwasaDisruptordoesnotmeanthateverycompany with a lousy product ends up a world beater. If we want to knowwhethersomeotherhistoricalexampleisexplainedbyDisruption,weneedtodo

thesamekindofcarefuldatacollectionandanalysisthatallowedustodrawthisconclusionforSouthwest.Andweneedtobeconfidentthatthiscanbedoneinadvanceevenwhendealingwithcomplexsituations.Worse,whenitcomestoapplyingthisinsightintheserviceofpredictions,we

have the additional complication that Disruptors often follow very differentpathsand takeverydifferent lengthsof time tobreak through.Howarewe toknowifanentrantwithareasonablefootholdwillbelimitedtofeedingoffthetablescrapsofasuccessfulincumbentorwilleventuallybecomeamortalthreat?And how can we tell when that upmarket march will begin and, just asimportantly,howlongitwilllast?

CHAPTERFOUR

IF

Perhaps the best way to understand if Disruption applies to a specificcompetitivebattleistoillustrateindetailcircumstanceswhenitdoesnot.

Translating Disruption into the “tradeoff” language of microeconomics andstrategyprovidesamorepreciseunderstandingofthekeycausalmechanismsofDisruption,specifically,the“how”oftheupmarketmarch.Companiescaptureafoothold by defining a new productivity frontier that reaches far intomarkets(new-market Disruption) or market segments (low-end Disruption) that arebeyond the reach of the incumbents’ frontier and so are economicallyunattractive to them. Since the new frontiers defined by Disruptors breaktradeoffs that define the incumbents’ frontiers,Disruptorsmust create entirelynewbusinessmodelsthatareenabledbydifferent,orverydifferentlyemployed,technology. (For example, the 737 was certainly not new technology to theairlineindustry,butSouthwest’smodelusedtheplaneinauniqueway.)Every productivity frontier expands over time thanks to its own sustaining

innovations, that is, innovations that break tradeoffs among the behaviors thatconstitute that frontier’s business model. The relative pace and shape of theexpansion of a would-be Disruptor’s frontier determines whether there is anupmarketmarch,andhenceaDisruption,tobehad.Ifthenewfrontierdefinedby an entrant (like Southwest) expands in ways that break the tradeoffs thatdefinetheincumbents’frontier,thentheentrantcancompetefortheincumbents’mainstreammarketsfromapositionofstructuraladvantage.A critical implication of this insight—one that I believe answers a long-

standingcriticismofDisruption—is thatsuccess ina footholdmarketdoesnotlead inevitably to the sort of upmarket march that culminates in marketdominance.PCmakerswereabletoDisruptminicomputermakersonlybecausemicroprocessors, disk drives, and so on got better in the right sorts of ways.Without those innovations, any attempts by PCmakers to compete for large,lucrativecorporatemarketswouldhaverequiredthemtochangefundamentallytheirstrategies.Mostlikely,thatwouldhavemeantcreatingproductsverymuchlike those offered by minicomputer makers and adopting their organizationalforms as well. In this alternate history, PC makers that tried to transformthemselves into minicomputer makers would almost certainly have failed.Similarly for Southwest: the only way it could have competed effectively for

longer-haulmarketsbeforethe737-500cameonthescenewouldhavebeentoadoptothertypesofaircraftthatweremoreefficientthanthe737-200and737-300.Thatwouldhavecorruptedakey featureof itsbusinessmodeland likelyhavemadeitallbutimpossibleforittosucceed.The general principle at stake is whether or not looking at Disruptive

opportunitiesinthiswayallowsustopredictwhetherornotDisruptionisevenapossibility for a given entrant’s business model. Unless we can predict if aparticularfrontierwillexpandintherightway,weareleftwonderingwhetheragivensuccessfulbusinesswill foreverbe relegated to itsnicheorcanmountabonafideDisruptivechallengefromits initial footholdmarket.Companiesthathavefoundavaluablenichebasedontraditionalstrategicdifferentiationcanbetremendously successful. Disruptors, on the other hand, have an entirelydifferentgrowthtrajectoryaheadofthem.

DISRUPTION-PROOFINDUSTRIES

Relatively few competitive battles are between aDisruptor and an incumbent.Themajority of the struggles for corporate survival are Tennyson-like red-in-tooth-and-clawbarfightsamongcompaniesplayingbythesameruleswherethecombatantwith the best right crosswins. In these sorts of contests creativity,tenacity,andnotinfrequentlyluckareamongthedecidingfactors.However,because“disruption”isacommonlyusedwordmeaning“tothrow

into confusion or disorder,” the technical sense of the term, as Christensendefined it, isoften lost.Anycircumstance inwhich incumbents stumbleandanew order takes shape is chalked up to Disruptive innovation. This isunfortunate,forwhenatermisusedtodescribeeverythingitquicklycomestodescribenothing.SoitisworthwhiletoillustratewhatDisruptionisbyshowingwhat it is not.Agoodplace to start iswith industries that are, so far at least,seeminglyimmunetoDisruption.

HOTELS

Inthesummerof1951KemmonsWilson,hiswife,andtheirfivechildrensetoutfromtheirhomeinMemphisforthefamily’ssummervacation:aroadtriptoWashington,D.C. Surprisingly, itwas not his kids’ backseat antics that droveWilson to distraction; it was the lack of suitable, moderately pricedaccommodations. Downtown hotels were beyond his budget, andaccommodations he could afford were typically mom-and-pop motels withunpredictable standards of cleanliness, facilities, and pricing that ranged fromexcellenttoawful.Wilson saw an opportunity to provide a clean, friendly, moderately priced

motel to serve the growing segment of postwar families who were travelingtogetherbyautomobile.Theresult, launchedin1952,wasHolidayInn,anditssuccess is attributable to classic,well-executed strategic differentiation. Priceswerekeptlowbykeepingthefacilitiescleanandcomfortablebutbasic,withnoindoor parking, no bellman, no room service, no concierge, and no specialrequests. He positioned his motels near on-ramps to the interstate highways,whichwasbothcheapforhimandconvenientforhiscustomers.Heprovidednoamenities save one: a swimming pool so the kids could burn off energy after

spendinghoursinacar.Itwasaperfectexampleofhowtoexploitthetradeoffsinherent inprovidingdeeplydifferent typesofaccommodations. Incumbents—theotherhotelsormotelchains—couldofcoursechoosetocompetewithhim,buttodosotheywouldhavetoreplicatehismodelandabandontheirexistingone.At theother extremeof thehospitality industry’sproductivity frontier stood

Isadore Sharp. In 1961, asWilsonwas hitting his stridewith theHoliday Innfranchise, Sharp launched the Four Seasons hotel chain with the first FourSeasonsMotorHotel inToronto.Wanting to establish himself immediately atthe higher-end market, he innovated by supplying guests with personal-sizeshampoo (a first for the industry); thicker, all-cotton towels; and, over time,dressing gowns, Belgian chocolates placed on the high-thread-count cottonpillowcases at night, fine dining, and gymnasiums and spas as his clientelebecamemorehealthconscious.Perhapsmostcritically,Sharpkeptthenumberofroomsinhispropertieslow

—muchlowerthanatotherhigh-endhotels,andmuchlowerthanhisinvestorstypicallywanted.Inhisbiographyofthebusiness,Sharprecounts(inachapteraptlyentitled“StartingattheTop”)negotiationswithBritishinvestorsinwhichhesecuredtheirbackingonlybyagreeingtopayrentona320-roomproject—eventhoughhewaspreparedtobuildonly230rooms.21Inconversationin2004,Sharp said that he still insisted on building nothing bigger than four hundredrooms, even though other luxury properties built much larger facilities andequityanalystssuggestedstronglythatSharpwasleavingmoneyonthetablebystaying small.ButSharp’sviewwas thatonlybykeeping theproperties smallwas itpossible toprovide the levelof service required todominatehischosenniche.Holiday Inn and Four Seasons were successful because each identified

specific customer groups with specific needs and were willing to accept thetradeoffs required to serve them well. At Holiday Inn, it took tremendouscouragetostickwith“just thebasics”whenthesirencallofhighermarginsathigher price points beckoned. Similarly, it can be just as challenging for FourSeasons to stick with its “pillow menu” (buckwheat filled, hypoallergenic,naturaldown,andahostofshapesandsizes)whenadip inbusinessseems todemand cutting such seeming extravagances. But sticking to your guns evenwhen you seem to be running out of bullets is the key to successfuldifferentiation: typically, competitors were unwilling or unable to make thesetradeoffs and so were not able to compete as effectively for the business ofcustomersinthosesegments.Unwaveringandconsistentfocusiswhatkepteachofthesecompaniesontopoftheirrespective—andverydifferent—games.

WilsonretiredaschairmanofHolidayInnin1979,whenthechainwasstillstrugglingtorecoverfromtheslumpinlong-haulautotravelinthewakeoftheoil embargo of the early 1970s. Those who were traveling were increasinglyprice sensitive, and new competitors cropped up with an even lower coststructure.DaysInnandMotel6offeredstillmorespartanservicesforstilllowerprices—aclassicsegmentationattackonHolidayInn’sniche.The newmanagement atHoliday Inn looked to reignite growth by entering

higher-margin segments. They seemed to feel that they could compete byleveragingtheir“corecompetencies”inhoteloperationsandattackinghigh-endhotelsfrombelow,justastheywerebeingattackedfrombelowthemselves.Itmight seem that this is textbookDisruption: a low-cost competitor taking

what it has learned in a low-end segment and competing for high-marginbusiness. It did not turn out that way, however. Holiday Inn launched theCrowne Plaza division in 1983. Today the chain is a perfectly crediblecompetitorinthehigher-endhotelsegment,withoverfourhundredhotels,andisgrowing rapidly. (The Holiday Inn family of hotels, including Crowne Plaza,waspurchasedbytheInterContinentalHotelsGroupin1990.)ButithasnotDisruptedanyone,althoughnotasaresultofanyshortcomings

onitspart.Rather,whatHolidayInndiscoveredwasthattheluxurysegmenthadneeds that could only be met with a higher-cost business model: superiorfurnishingsand linens,moreandbetter-trainedstaff, round-the-clock fine foodservice, and so forth. In other words, the hotel business resists Disruptionbecause there isnoenabling technology thatallowsabusinessmodel intendedforsegmentsunattractivetopowerfulincumbentstoimproveinwaysthatbreakthetradeoffsthatdefinethefrontieroftheincumbents’businessmodel.Theonlyway to compete for a different niche is to accept a different set of tradeoffs,adopt the relevant business model, and try to outdo the competitors throughtraditional,sustaininginnovation.

FIGURE22:NONEWFRONTIER,JUSTSUSTAININGINNOVATION

Everyfrontierexpandsovertime;competitiveforcesmakesureofthat.Butin the hotel business there has not been any true Disruption. Instead,competition has largely taken the form of jockeying for position along asteadilybutslowlyexpandingfrontier.Some“extreme”solutionsmightbeseenasdefininganewcurve; forexample, the“capsule”hotels inTokyoareverylowcostandprovidelittlemorethan2.5cubicmetersofspaceandcommunal toilets. Critically, however, there remains no “enablingtechnology” that might allow a competitor to improve such an offeringwithintheconstraintsofthatmodel:sofar,atleast,theonlywaytoprovidemorespaceistoprovidemorespace.

Toputthatinplainerlanguage,theonlywaytohaveaconciergeistohireaconcierge, and the only way to have a better concierge is to hire a betterconcierge. Typically, incumbents do that better than entrants. Consequently,Crowne Plaza is a solid competitor in a challenging industry, but it is not aDisruptor.The performance implications of this kind of competition can be sobering.

FourSeasonsenjoyedatwenty-five-yearrunofannual increasesinstockpriceof11.5percent—butonlya2percentannualgrowthrateinrevenueandwithanROA that fluctuated wildly. Holiday Inn fared little better, with a revenuegrowthrateovertwentyyearsof13.9percent,anannualstock-priceappreciationof4.6percent,andanaverageROAof4.9percentthatfluctuatedwithinarangeofabout2.5percent.Onecancertainlyexplaintheseoutcomesinhindsight.Forpresentpurposes,

however, the key insight is that given the absence of a new business modeldefining a new productivity frontier and an enabling technology for breakingtradeoffs,we can see thatDisruptionwould have predicted the absence of thekind of consistent, breakthrough growth that tends to characterize successfulDisruptors.Ofcourse,greatperformancedoesnotrequireDisruption:bothFourSeasons

and Holiday Inn are great success stories. But for investors neither the FourSeasonsnorHolidayInnwereSouthwest.

FIGURE23:SELECTEDPERFORMANCEDATAONFOURSEASONS(1982–2006)

STRATEGYCONSULTING

The business of strategy consulting also seems to be resistant toDisruptiveforces.Thehistoryofcompetitionintheindustryreadsverymuchlikethehotelbusiness:asuccessionofsmallplayersfocusingonaspecificnichegrowingintomainstream competitors largely by replicating the business models of thesuccessfulincumbents.McKinsey,forexample,gotitsstartexploitinganorganizationalinnovation—

theM-form(ormultidivisional)structure.Asthefirmgrew,ittookonanever-widerarrayofgeneralmanagementproblemsandcreatedabusinessmodelthatallowed it to become a large and successful general management consultingprovider. Its competitive success is a function of its differentiation along thefrontierofthatbusinessmodel.

FIGURE24:SELECTEDPERFORMANCEDATAONHOLIDAYINN(1966–1988)

Otherfirmshavetendedtogettheirstarts,andthengrow,inmuchthesameway.TheBostonConsultingGroup(BCG)beganbydifferentiatingitselfbasedon its unique expertise with the experience curve concept. This allowed it totackle efficiently and profitably smaller engagements than were attractive toMcKinsey—aformoflow-endstart.YetasBCGgrewandbegantotakeonanever-wider array of general management problems, it ended up essentiallyreplicating McKinsey’s business model. As a result, BCG became a crediblecompetitorbutwasnotaDisruptor.MonitorGroup’sstoryisverysimilar:froma beginning built around the application of Michael Porter’s Five Forcesframework, Monitor’s subsequent success has led it to build the capabilitiesrequiredtotackleanever-widerarrayofgeneralmanagementproblems—anditdoes so in fundamentally the same way as just about every other strategy

consultingfirm.Thehotelbusinesshasbeen immune toDisruptionbecause theonlyway to

havemarblecountertopsandacity-centerlocationistohavemarblecountertopsandacitycenterlocation.StrategyconsultingseemstobeimmunetoDisruptionbecause no one has figured out how to create structured problem solving thatenablesinexperienced(andhencelower-cost)consultantstoprovidethelevelofservice that more experienced consultants can. The significant investmentsconsulting firms make in recruiting, knowledge-sharing infrastructure, thedevelopmentofintellectualcapital,project-managementmethods,andsoonareall sustaining innovations—attempts to push out the frontier of the existingbusinessmodel.Noonehasbeenabletostartwith,say,astructurallylower-costbusinessmodel that delivers consulting services to small companies, then takethat capability upmarket and break the tradeoffs that define the existingindustry’s frontier. Instead, consultants that serve smaller companies typicallyjust make less money, and as they are successful and grow they get moreprofitable—butenduplookingjustliketheincumbents.Inshort,newentrantsdonotcreateadifferentbusinessmodelthatdescribesa

newproductivityfrontier.Instead,theyembraceadifferentsetoftradeoffs—atleastatfirst—inordertofindanicheofthemarketthattheycanserveprofitablyandthatincumbentsarerelativelyuninterestedin.However,asthatinitialnicheisexhaustedandthelureofprofitsinmoreattractivenichesbecomestoostrongto resist, entrantsmigrate to where themoney is but are forced to accept thesametradeoffsastheincumbentstheycompetewith.Fundamentally, then, Disruption is only possible when there is an enabling

technology or process that allows one businessmodelwith a given frontier tobreak the tradeoffs that define the frontier of a different business model thatservesmoreattractivemarketsormarket segments.For thePCmakers, itwasthe evolution of the microprocessor, disk drives, and software, among othercomponents,thatpropelledthemupmarket.ForSouthwest,itwastheincreasedefficiency of the Boeing 737-500 that enabled its Disruption. Still otherDisruptors, such asToyota (see below), have reliedonprocess improvements.But in hotels and strategy consulting, to nameonly two examples, there is nosuchenabling technology thatallowsafundamentallydifferentbusinessmodeltocompeteformainstreamsegmentsfromapositionofstructuraladvantage.

PSEUDO-DISRUPTORS

Itistemptingtoconcludethatnewcompaniesthatappeartoreplicatetheearly-stagestrategyofasuccessfulDisruptorwillbesuccessfulDisruptorsthemselves.Applyingthis“enabling technology” litmus testallowsus to tell thedifferencebetweenthosethatfoundadefensiblenicheandthosethatcanlegitimatelyhopetomovefromafootholdtofull-bloodedDisruption.

RETAIL

Retailinghas beenwell and trulyDisruptedby the discount retail format ingeneral. The dominant discount retailers have hundreds of billions in revenueandhaveoperationsspanningtheglobe.Commentatorsonthediscountretailingphenomenon quite often point to the vast scale of such operations as a keycontributor to the format’s success: purchasing in huge volumes allows largediscounters to secure lower prices from many suppliers than its smallercompetitorstypicallycan.However,scaleisveryrarelyasourceofadvantageintheearlydays;afterall,

withvery fewexceptions,nothingbig startsbig.Mostoften,discount retailersand their close cousins thewarehouse clubs usually start out relatively small,focusingonimproving“worse”retailingsolutionstargetedatcustomersegmentsthat are relatively unattractive to incumbent retailers. In other words, it is atextbook low-end foothold, created through a relentless and typicallyidiosyncraticfocusonmeetingaspecificsetofcustomerneeds.Thediscountretailersthathaverisentomarketdominanceareoftenidentified

with charismatic, highly visible, and influential founders early in theirascendancy,buttheseinspiringleadersarerarelythefuelofDisruptivegrowth.Rather, the successful discount retailers that grewDisruptively did so becausethey used their early success as a platform for investing in, among othercapabilities, highly differentiated, hard-to-replicate logistics-managementprocesses.Thissortofinfrastructurewasforthemostsuccessfuldiscountersthe“enabling technology” of their Disruption of established department storeretailers:acompetitiveadvantagethatfueledtheircontinuedupmarketmarchinboththebreadthofproductofferingsandtherangeofretailformats.22AmongthemostsuccessfuldiscountershasbeenWalmart,whichoverthelast

fourdecadeshasturnedinacompoundedannualgrowthrate(CAGR)inrevenueofover27percentandanadjustedstock-priceCAGRofalmost25percent,allthewhiledeliveringanalmostpreternaturallystable returnonassetsaveraging9.6 percent and in thirty-seven of thirty-eight years never straying more thanthreepercentagepointsawayfromit.Understandably, other companies have attempted to replicate these results,

and Dollar General has been among the more credible efforts.23 By the late2000s, Dollar General had seemingly burst onto the scene as a leader in“extreme-valueretail.”Likemostovernightsuccesses,though,ithadbeenalongroad.FoundedbyJ.L.TurnerandhissonCal in1939, thecompany’scurrentform emerged when it morphed into a “closeout” store in 1955, sellingoverstock, discontinued, surplus, and distressed merchandise. This allowedDollar General to keep prices very low, but at the expense of carrying aninconsistentassortmentofitemsofinconsistentquality.Nevertheless,themodelproved successful, and thanks to consistently savvy purchasing, the companygrewtoover1,300locationsintwenty-threestatesby1990.

FIGURE25:SELECTEDPERFORMANCEDATAONWALMART(1970–2008)

Sofar,sogood:seenthroughthelensofDisruption,DollarGeneralin1955isboth “worse” and enjoys the necessary autonomy to warrant a prediction ofsurvival. Making it through not just to 1990 but to 2010 certainly counts ashaving survived, and the company has continued to grow profitably, so thetheory’s predictive accuracy is intact. But has the company been a genuineDisruptor?Through the early2000s,DollarGeneral’sproductivity frontier seems tobe

quite different from the discount-retail model. Most discount retailers havemarchedupmarket thanks to thepowerof their logisticscapabilities.Themostsuccessful of them have used this capability to launch ever-larger stores(100,000 square feet and more) and carry an ever-broadening array ofmerchandise(insomecasesapproaching100,000stock-keepingunits—SKUs).Mostnotably, somearebreaking thedefining tradeoff in retail: remainingcostandpriceleadersevenastheyappealtomoreaffluentsegmentsofthemarket.In contrast, Dollar General remained strongly focused on low-income

households:over40percentofthestores’customerscamefromhouseholdswith

less than$30,000peryear in total income.Makingmoneyonthe thinmarginsthis segment offers hasmeant embracing a very different set of tradeoffs thanhave other successful retailers. Where Walmart’s stores were typically verylarge,withawideselection,DollarGeneralkept itsstoressmall (6,900squarefeetonaverage)anditsrangeofproductsnarrow(4,900SKUs),withalimitednumber of high-turnover product categories, such as household cleaners,batteries,anddrygroceries.Inaddition,formuchofitshistorythecompany’sprocesseshavebeenvastly

different from those of the large, successful discount retailers as well. DollarGeneral had historically had no sophisticated cross-docking system, noproprietary satellite-based communications infrastructure, and no nationwidesharing of sales data. Instead, it had typically pressed everyone in its leanlystaffedstores intoserviceon theweekly“TruckDay” tounloaddeliveriesandrestocktheshelves.ThisapproachleftDollarGeneralablenotonlytoserveverylow-income households and to operate profitably but also to operate in verysmall towns: a community of twenty thousand would be ample for a DollarGeneral location,whereas large, successfuldiscountershavecome to requireasmallcityofatleastfiftythousandbeforeconsideringopeninganewlocation.

FIGURE26:SELECTEDPERFORMANCEDATAONDOLLARGENERAL(1967–2007)

The results speak for themselves: Dollar General has had a thirty-year runwith a revenue CAGR of over 15 percent, stock-price appreciation over 16percent,andanaverageROAofover9percent.Iwouldagreewithyouifyousaidthissoundslikeadifferentbusinessmodel

abletoreachsegmentsthatwereunattractiveorevenunprofitablefortraditionalretailers.Unlikeourhotelsorstrategyconsultancies,DollarGeneralhaddefinedadifferentfrontier.Itwasoutcompetingtheother“extreme-valueretailers”andpushing out the frontier of that business model in ways that its most directcompetitorshaddifficultymatching.Butdidithaveanenablingtechnologythatallowed it to pursue a truly Disruptive upmarket march? Could it break thetradeoffsthatdefinedthefrontierfordiscountretailersordepartmentstores?Muchofthecompany’ssuccesscanbeattributedtothealignmentofitsmodel

with twomajormarketplace trends. First, the “bargain hunter/treasure hunter”mentalityamongU.S.consumersmadeit“okay”forpeoplewhocouldaffordtoshop more traditional retail channels to economize on everyday purchases,whichinturnfreedupincomeformoreaspirationalpurchasesfromhigher-end

channels. Second, the rising percentage of U.S. households in lower incomebracketsmeantthatthereweremanymorehouseholdsthathadlittlechoicebuttoshopthelowest-priceretailavailable.Combined,thesetwotrendsmeantthatthepercentageofU.S.householdsthathadshoppedatsometypeofdollarstorerose from55 percent in 2000 to 67 percent in 2005. In otherwords,much ofDollar General’s growth can more reasonably be attributed to shifts in themarket than to a significant expansion in the company’s productivity frontier,thatis,toinnovation.(Thisisbynomeansacriticism—youstillgettokeepthemoney!ButIamtryingtodistinguishbetweengrowthfueledbyDisruptionandgrowthdrivenbyotherfactors.)The more compelling reason to question the company’s Disruptive nature,

however, emerges from an analysis of the company’s changes in its businessmodeloverthelastfiveyearsorso.Since2005,thecompanyhasresumedlocalcircular advertising for the first time in more than a decade. It has investedheavily in inventory management and point-of-sale data systems, which havebroughtthecompanyuptodatewithcommonindustrypractices.Ithasinstalledcard readers, moving away from the cash-and-carry limitations of the past,whichhasservedtoincreasebasketsizeand,byimplication,increasethelengthof the average customer’s visit. It has added coolers so it can sell perishablessuch as milk and eggs, and it has had to adapt its store stocking proceduresaccordingly:“TruckDay”hasbeensupplantedbycontinuousreplenishmentand“direct store delivery” suppliers. And the company has rationalized its stores,closingmanyofthelessprofitable,andtypicallysmaller,locations.Thesearenot innovationsthatexpandDollarGeneral’sexistingproductivity

frontier; rather, they amount to a fundamental shift in business model. Thischangeverylikelymakesperfectfinancialandstrategicsense.Butitdoessignalpretty strongly that Dollar General’s original model did not have Disruptivepotential.Rather, thealternativeproductivityfrontier that thecompanydefinedas a leading extreme-value retailer, which manifested a very different set oftradeoffs and constraints than other low-cost retailers, has exhausted its niche.That model was less likely to continue growing at historical rates, and tomaintain itsgrowthDollarGeneralappears increasingly tohaveclimbedon totheproductivity frontierof established low-cost anddiscount retailers, evenasthe most successful discounters are experimenting with smaller-scale retailformats. Dollar General and the leading discounters are well-managed,strategicallyinsightfulplayers,andtheyallbringtheirownsourcesofpotentialadvantagetothisbattle.Butthisisshapinguptobea“clashofthetitans,”notaDisruptive battle. Disruption, therefore, has nothing to say about what theoutcomewillbe.

Whether one thinks Dollar General is a Disruptor or not says a great dealaboutwhat kind of growth strategy is likeliest to succeed: canDollarGeneralexpand its SKUs, its store size, its volume, or its margins from within theconfinesof itsexistingmodel?Ordoes itneedtoprepare itselffor thekindoforganizational transformation that comeswith a fundamental shift in strategy?Andfromtheperspectiveofleadingdiscountretailers,whetherDollarGeneralisa Disruptor or not determines what kind of defense is appropriate: do thediscount retailers need to create autonomous and entirely different divisionsfocusedonexploringthepotentialandlimitsofanewretailformat?Orshouldthey remain focused on honing their existing model through a vigorous andnever-ending quest for incremental or breakthrough—but fundamentallysustaining—innovations?

FIGURE27:DEFININGANEWFRONTIER,THENJUMPINGTOANESTABLISHEDONE

The discount-retail format pioneered a new business model that wasfundamentallydifferentfromthatoftraditionaldepartment-storeretail.Asthebestdiscount retailersgrew, theywereable toexploitnewsourcesofprofitability, such as economies of scale, which had been unavailable tothem as start-ups. This concatenation of competitive advantages allowedthe successful discounters to expand their business model’s frontier andDisrupt department-store retailers. Dollar General also defined a newfrontier with the “closeout” retail format, but the segment that modelappealed to was not large enough to satisfy the company’s growthambitions. As a matter of fact (rather than theoretical necessity), thecloseoutretailfrontierhadnoenablingtechnologythatwouldhaveallowedDollarGeneral to growbeyond this originalmarket segment, and so thecompanyhashadto“jump”toanewfrontierandnowemploysastrategythatmorenearlyreflectsthetraditionaldiscountretailmodel.

PHARMACEUTICALS

Asimilarsortof“frontierhopping”hasbeenmistakenforDisruption in thepharmaceutical industry.Genericdrugmakershave longhadaviablebusinessmanufacturingoff-patentdrugs.Nothavingtohavemadethesamekindsofveryexpensive investments indrugdevelopmentand testing that theoriginalpatentholders did, generic manufacturers typically need recover only manufacturinganddistributioncostsandsocanaffordtopricetheirdrugsfarlower.Recently, some generic manufacturers, such as Dr. Reddy’s in India, have

enteredthedrugdiscoverybusiness.Thishasbeentakenbysomeasevidenceofan upmarket march, perhaps fueled by lower-cost research talent in India orelsewhere.Certainlyitistruethatgenericdrugmakershaveanentirelydifferentbusiness

modelthatdefinesaverydifferentfrontier,whichpermitsgenericmanufacturerstoreachsegmentsofthemarketthatareunattractivetomoreresearch-intensivepharmaceuticalcompanies.However,as thegenericdrugcompaniesmoveintoresearch, they are finding that they have no structural cost advantagewhen itcomes to drug discovery. Any labor arbitrage they hope to exploit in thedevelopmentortestingofdrugsiseasilyreplicatedbyincumbents.Otheraspectsof the industry that promise to change the game, such as so-called openinnovation,arenotantitheticaltosuccessfulincumbentfirmsandinfactwouldappear to magnify the advantages enjoyed by “big pharma” rather than

undermine them. In contrast, a trueDisruptive advantagewould have to stemfromafundamentallydifferentapproachtodrugdiscoverythatincumbentfirmswould be relatively unable to adopt. Such advantages may well exist, but nogenericdrugmakersIamawareofpossessthem.Consequently, justasDollarGeneralhasdriven itsgrowthbyhoppingfrom

itsdistinctivefrontier tothatoftheincumbents,genericdrugmakersthatenterthedrugdiscoverybusiness are notDisruptors butmerely tradingone frontierforanother—withalltheattendantrisksthatsortofstrategicchangeentails.

THEREALTHING

Thisreviewofnon-Disruptorsandpseudo-Disruptorswill,Ihope,bringintosharper relief the three defining characteristics of a true Disruption. First, thealleged Disruptor must have a new business model that defines a differentfrontier.Simplystakingoutadifferentpositionalongtheexistingfrontierisnota viable starting point for Disruption. The hotel and strategy-consultingexamplesareillustrationsofthis.WhetheritisHolidayInnversusFourSeasonsorBCGversusMcKinsey,whennewentrantsgettheirstartmerelybyacceptingdifferent tradeoffs along the same frontier, they have found a niche. This istraditionalstrategicdifferentiation,and itcanbeapath tocompetitivesuccess.Butanicheisnotafoothold—andsostrategicdifferentiationisnotDisruptiveinnovation.Second,evenwhenanewfrontierhasbeendefined,wecannotnecessarily,or

evenmostof the time,expecta trueDisruption to follow.Disruptiondemandsthat the businessmodel that defines this new frontier be pushedoutwardby atechnologyorsetofprocessesthatincumbentsareatadisadvantageinadopting.In the case of Southwest, itwas the 737-500 that pushed its frontier outward,allowing the company to compete for more and longer routes in ways thatincumbent airlines couldnotmatch.Similarly,PCsDisruptedother computingarchitectures thanks to improvements in microprocessor technology andpackaged software, advances thatminicomputermakerswere ill positioned toincorporateintotheirstrategiestosimilareffect.In contrast, as in the Dollar General and generic pharmaceutical examples,

companiesthatstartwithadifferentbusinessmodelandadifferentfrontiercanfind themselvesunable tocompete for themoreattractivesegments theycovetfromwithintheconstraintsoftheirmodel:theylackanenablingtechnologythatdrives their frontier outward quickly enough and in the right sorts of ways.Consequently, achieving the performance required to be competitive in moreattractive markets can turn out to require taking on the business model ofsuccessfulincumbents.Whenentrantsstartmimickingtheincumbentstheyhopetodefeat,theyarenolongermountingaDisruptiveattack.Third and finally, successful Disruptions culminate with the new frontier

expandingenoughthatitallowsDisruptorstodeliverlevelsofnonpricevalueata cost that incumbents simply cannot match. That is, not only does the newfrontierhave toexpand, ithas toexpandat sucha rateand insuchaway that

eventuallyourerstwhileentrantsareable tocompetesuccessfully for lucrativeestablished markets or market segments—and end up dominant incumbentsthemselves.

FIGURE28:THEDEFININGCHARACTERISTICSOFDISRUPTIVEINNOVATIONS

It is time to tie off that loose end from chapter 1. Thurston’s criteria fordeterminingwhetherabusinessisaDisruptionwere“Isthesolutionworsethanthat of incumbents?” and “Does the business have the necessary autonomy?”These criteria identify viable niches: developing “worse” solutions thansuccessful incumbentsaimsaventureatsegments thatare likelypoorlyservedandunlikely tobevigorouslydefended;havingautonomyprovides the latitudeneeded for the new venture to develop a viable solution for that segment. Bythese lights, both Dollar General and Holiday Inn would have qualified asDisruptors,sincetheypassbothlitmustests.Certainly,thesetwoplayerswouldhave been predicted to survive—which they have certainly done—but not, Imustclaim,becausetheywereDisruptors.Inaway,thisisheartening.Workingwithonly“half”ofDisruption—howto

identify a foothold—Thurston was able to improve materially the predictiveaccuracyofinvestmentdecisions.Inanotherway,itisdisturbingthatDisruptionmightbegettingtherightanswersforthewrongreasons.HowdowesquaretheincompletenatureofThurston’sworkingdefinitionofDisruptionforpredictivepurposeswiththeunderlyingcausalmechanisms?As it turns out, Thurston’s incomplete predictive criteria did not undermine

his predictive power for two very important, but entirely contingent, reasons.First,asexploredinappendixA,veryfewofthebusinessesconsideredbyNBI

were in fact “worse”—that is, they failed the foothold test and instead were“sustaining”toanindustryinwhichIntelwasanentrant.SincethedatasetusedintheexperimentscontainednoDisruptive/entrants,abigpartofthepredictiveaccuracyofDisruptionintheexperimentsreportedherestemsfromidentifyingventures that would ultimately fail. Distinguishing between a niche and afootholdmattersmorewhenpredictingsuccess.Second, most of Intel’s new businesses tended to be built around

technological platforms that improved in precisely the manner required forsuccessfulDisruption. In thenext chapter Iwill havemuchmore to sayaboutthis, but businesses built on, for example, nanotechnology, computing, orcommunications technology, and perhaps increasingly biotechnology, arehitchedtointrinsicallyDisruptivewagons.Findawaytobuildanichesolutionthat is powered by technologies that improve this rapidly, and on so manydimensions, and you have a much better shot at Disruptive success.Consequently, the twonewbusinesses thatNBI funded thatwere “worse” butfaileddid in fact fail because they lacked the requisite autonomy,not becausetheylackedtherightsortofenablingtechnology.Noteverycompanyis Intel,ofcourse,sonoteveryonecancountonhaving

the good fortune of systematically building new businesses on a potentiallyDisruptive platform.As a result, ifwe are to generalize Thurston’s predictiveframework, we must capture all of the defining elements of Disruption.Thurston’scriteriagiveusawaytoidentifyaniche;whatweneedisawaytodeterminewhetherthatnicheisatruefoothold,thekindthatcantranslateintoanupmarketmarchthatleadstoDisruption.In light of this distinction, consider nowan expandedversionofThurston’s

originaldecisiontree.Notethattheentireright“branch”aftera“yes”responseto the“autonomous”question results inapredictionof survival—but for threeverydifferentreasons.Companiesthatdonotdefineanewbusinessmodelcanbeexpectedtosucceedthankstoclassicstrategicdifferentiation,thatis,findingunoccupied but valuable real estate on an industry’s existing productivityfrontier.This is theHolidayInnstory.Acompanythatdefinesanewbusinessmodel but lacks an enabling technology will likely do better because it isreachingintoasegment that is fundamentallyoutofreachfor incumbents—itsdegree of strategic differentiation is much greater because it has defined anentirely new frontier. It is a strategic innovator: it has found an entirely newstrategy,whichservestoinsulateitthatmuchmoreeffectivelyfromcompetitivepressures.ThisistheDollarGeneralstory.Finally, we have bona fide Disruption when a new frontier is propelled

outward by a key enabling technology, one that eventually permits our new

entrant to compete for mainstream markets from a position of structuraladvantagewithouthavingtochangeitsbusinessmodel.ThatistheWalmartandSouthweststory.And then there is FourSeasons. Frankly,Disruption theorywould not have

likedthiscompany.Itwasenteringthehighendof thehotelbusinesswith theexplicitintentofofferingbetterservicethanestablishedincumbents—theRoyalYork in Toronto, the Dorchester in London, and the Ritz-Carlton just abouteverywhereelse.Andyetitsucceeded.ThisfurtherdemonstrateswhatIobservedattheoutset:Disruption,evenifa

betterway to predict outcomes, is not perfect.But, in one of those strokes ofgoodfortune thatare toogood topassup, reflect foramomenton therelativeperformanceoftheexamplesdiscussedsofar.

FIGURE29:EXPANDEDDECISIONTREEFORDISRUPTIVEINNOVATION

TrueDisruptorsareDisruptorsbyvirtueofhavingdefinedanewfrontierthathas an enabling technology that drives the upmarket march. Dollar Generaldefinedanewfrontier forextremevalue retailingwith itsnewbusinessmodelbut lacked the fuel that powers true Disruption; it was a “strategicdifferentiator”: it createdanewproductivity frontier thanks to itsmasteryofanewbusinessmodel,butit isnoDisruptor.HolidayInnshouldbeevaluatedintwo phases. During its first two decades (it went public seven years after itsfounding in 1951) the company enjoyed strong growth and returns to capitalwith solid profitability thanks to its strong strategic differentiation—eventhrough the oil shocks of the early 1970s, not a great time to be a motelappealingtofamiliesoncartrips.WhenHolidayInnbeganexperimentingwithnewformats,suchastheCrownePlaza,whereitwasanentrantseekingtoofferbetter solutions, its growth and returns to capital sagged appreciably. FourSeasons is an anomaly: a survivor and a success despite entering establishedmarketswiththeintentofofferingasuperiorsolution.Itsomehowmanagedtomuscle its way into an attractive niche and build a successful and enduringbusiness.Butnotethatitsperformanceistheworstofthelot.

FIGURE30:SELECTEDPERFORMANCEDATAONSELECTEDSTRATEGYTYPES

Asever,theseperformancecomparisonsareentirelyrelative.Allfiveofthese

companies are justly heralded in their industries and in the business pressgenerally. There is much to learn from each about how to be a greatorganization.IsadoreSharp’smemoironFourSeasonsseemsjustascompellingandenlighteningareadasSamWalton’swordsofwisdomonWalmart.Botharewell-runcompanieswithstrongleadersandgreatorganizations.TheabsenceofDisruption at Dollar General or in the hotel industry is not an indictment ofanyone’s managerial acumen. It might simply mean that the nature of thebusinessmodelsrequiredtodeliverthoseservices,atleastfornow,isresistanttothekindsoftechnologiesthatdriveDisruptivegrowth.Whatever might drive the susceptibility of an industry to Disruptive attack

(moreon this in thenext twochapters), fornowwhat is fascinating is that theDisruptorshavethestrongestgrowth,thehighestreturnstocapital,andthemostprofitablebusinesses.Ourstrategicinnovatorcomessecond,whileourstrategicdifferentiatoristhirdandouranomalyfourth.Withsuchasmallsample,andanonrandomoneatthat,anykindofgeneralizationisunwarranted.ButIcannothelp but feel some satisfaction in observing that the performance of thesecompanieslinesupwiththeviewthatthemorenearlyacompanycandefineforitself a Disruptive trajectory the more durable and rewarding its performancewillbe.

CHAPTERFIVE

WHENANDHOWLONG

Because we now understand the mechanics of Disruption and thecircumstancesunderwhich it applies,wecanalsoestimatewhenagivenDisruptionwilloccurandhowlongitwilllast.

Understanding the how of Disruption allows us to predict whether or not agivenopportunityhas trueDisruptivepotential—the if.This is certainlya stepforward,butitwouldhelpalottoknow—oratleasttohavesomeideaof—whenDisruptivegrowthcanbeexpectedtokickinandhowlongitwilllast.Afterall,there seems to be enormous variability along both of these dimensions. InChristensen’s signature disk drive example we witness four generations ofDisruptioninfifteenyears,eachlastingonlyafewyears.YetSouthwestbumpedalongaslittlemorethanacompellingnicheplayerforovertwodecadesbeforebeginningitsdecade-longrunofDisruptivegrowth.Walmartwasatleastfifteenand more like twenty-five years in the relative wilderness before it becameapparent—to the capital markets, at least—that the company was on tosomething big. Whether one is a manager or an investor, knowing when anenabling technology is about to transform a foothold into the explosive andprofitable growth of an upmarket march, and how long that growth will last,wouldseemtobeinvaluable.

THEPRODUCTLIFECYCLE

Perhaps one of themost durable and powerful frameworks for thinking aboutwhengrowthwillbeginandhowlongitwilllastisthe“productlifecycle”:theideathatproductsprogressinorderthroughthestagesofbirth,growth,maturity,anddecline.Sopervasivehasthismodelbecomethatitisalmostsecondnaturetomanymanagers,asortoflinguafrancafortalkingaboutthecurrentandfutureprospectsofanybusiness.Theproductlifecycleframeworkisbasedoninvestigationsintothediffusion

ofinnovationsfromtheearlypartofthelastcentury.24Bythe1960s,overfivehundredstudieshadbeendoneonhowvariousinnovationshadbeenadoptedbyspecific market segments, and a recurring pattern had emerged: successfulinnovationsfollowedthearchetypalsigmoidal,orS-shaped,growthcurve.When first introduced,an innovation sells in relatively lowvolume.Growth

accelerates fora time,only toslowagain.Thegeneralphenomenon, then,wasthat growthwas not a slow, steadyprogression but one subject to exponentialincrease—asteadygrowthpercentageonanever-increasingbase—followedbydramatic slowdowns, since even under the best-case scenario, market sharecannotexceed100percent.To explain this fact, researchers invoked the notions of “customer

innovativeness” and “market saturation,” arguing that more innovativecustomerstendtoadoptnewproductssoonerthanlessinnovativecustomersandthat eventually all available customers will have adopted. Since the observedpattern of product adoption had a sigmoidal curve, it followed that thedistribution of innovativeness among customers must similarly be normallydistributed, since a normal curve captured as a cumulative distribution yieldspreciselytheS-shapedgrowthtrajectorythatwastobeexplained.It turned out that instruments designed tomeasure customer innovativeness

using such attributes as exposure to media, level of education, and so forthmapped roughly to just such a normal distribution. The result was a formalexpression of the old commercial for Faberge Organic Shampoo: an earlyadoptertellstwofriends,andtheytelltwofriends,andsoon,andsoon…andassuming enough time passes and the friends have the requisite level ofinnovativeness,onewillobservepreciselythekindofsigmoidalgrowthonehadhopedtoexplain.

FIGURE31:DISTRIBUTIONOFCUSTOMERSBYINNOVATIVENESS

Itisthemovementofaproductthroughthesedifferentsegmentsofamarketthat drives its passage through the various stages of its life cycle. When aproductisfirstintroduced,asmallnumberofhighlyinnovativeconsumersbuyit. This is the birth phase. These innovative consumers are not infrequentlytrendsetters,sotheproduct“catcheson”andinvadessuccessivesegmentsofthemarket.AsthemorepopulousEarlyMajorityandLateMajoritysegmentsenterthemarket,theproductpassesthroughthegrowthphase,withlargenumbersofcustomers jumping on the bandwagon. When the Laggards finally adopt theproduct,themarketfortheinnovationissaturatedandtheproductismature.Atbest, one can hope that the product enters a steady state of low to no growth.More frequently, a product then shifts to its decline phase, as consumers loseinterestoranewproductwithsuperiorperformancedisplacesit.Theproductlifecycleisapowerfuldescriptiveframework.Andwhendealing

with relatively static customer preferences and product attributes, it can haveimpressive predictive accuracy. For example, the seminal studies in diffusionlooked at the adoption of new farming practices (seed varieties, tools andimplements,cultivation techniques).Elaborationof theseearlystudies, suchasthe 1943 investigations into the adoption of new seed types by Iowa cornfarmers, focused on generalizing those findings to other locales, such asColombia andBangladesh. These investigations demonstrated that even in theabsenceofmassmediaandhighlyliterateconsumers,theunderlyingsigmoidaladoption pattern held. Further, it proved possible to predict which customers(farmers)wouldadoptthenewproduct(cornseed)basedontheirrelativelevelofinnovativeness.There is, however, a critical limitation to the product life cycle’s predictive

power.Theobjective, recall, is to specify (within reason)whenaproductwillpassfromthepromiseportendedbytheembraceoftheEarlyAdoptersegmentto the path to glory that acceptance by the Early Majority represents. Foralthoughthesigmoidalgrowthcurvehasacharacteristicshape,itcantakeaverynearlyinfinitenumberofdifferentspecificformsdependingonwhetherandhowquickly the product makes the leap from one customer segment to the next.There is a big difference between spending a year in the birth phase andspendingadecade.Itmattersagreatdealwhetherthegrowthphasewilllasttwoyears or ten and whether the maturity phase offers a warm glow of gradualsenescenceorthemarketplaceequivalentofslammingintoabrickwallatsixtymilesperhour.Andthereislittleintheproductlifecycleframeworktohelpusunderstandthoseparameters.Muchoftheproblemstemsfromanissuethatwasraisedintheprologue:the

groundsforgeneralizingfindingsbeyondthesample.Themarketsandproductsuponwhich diffusion-basedmodels are built tend to be essentially static. Forexample, the corn seed in the early studies did not change during the fifteenyears thatmarket data were collected, and neither did the farmers’ functionalrequirements.Consequently,asadvertised,theproductlifecyclemodelcapturesthe diffusion of an innovation through a population. It can tell you who willadoptnext,butitcannottellyouwhentheywilladopt.25Other products and markets, such as personal computers, are far more

dynamic than corn seed in Iowa in the 1940s, yet they frequently follow thisclassic S-curve as well. Consequently, it seems reasonable to explain theevolutionofthesemarketsintermsofthesameunderlyingconstructofcustomerinnovativeness: we all carry in our minds the stereotype of the whiz kidstinkeringintheirgarageswiththeearliestAltairs,onlytobefollowedmuchlaterby their far less innovativeparentswhen theyadoptedPCs into theireveryday

worklives.Theproblemwith thiskindofnarrativeaboutdynamicmarketssuchas that

for PCs is that it was not the Altair that was adopted much later by lessinnovative consumers; it was a very different machine—and this subsequentgeneration of PC was adopted by both the whiz kids and the less innovativeamong us. Indeed, it is only by the most generous of interpretations that theAltairfromtheearly1970sandtheIBMPCthatbrokeintothecorporatemarketin the1980s canbe considered in the sameclassof product.ThePCchangedrapidlyandextensivelyoverthattimeandhascontinuedtochange.Productlifecycletheory,ifinterpretedstrictlyinkeepingwithitsunderlying

data on diffusion rates, becomes very difficult to apply as a predictive tool insuchcircumstances.EarlyAdoptersareeasyenoughtoimagine:theyadopteachnew generation of computer first, in accordance with their underlyinginnovativeness.Buthowareweto thinkabout theLaggards?Bythetimetheywere ready to enter themarket for Altairs, much-improved technologywouldhave been available—the Apple or the Apple II. Would less innovative PCconsumers (Laggards) buy a demonstrably worse, more difficult to use, lesspractically useful computer (the Altair)? Or would they enter the market bypurchasingthemostup-to-datetechnology(theAppleII)?The first possibility seems absurd conceptually and is all but impossible

practically.Conventionalnotionsofutilitysuggestthatconsumerswillbuywhatbest meets their requirements, and subsequent generations of computers didmore,moreeasily.Andgenerationsofcomputersappearedanddisappearedsoquickly that by the time theLaggardswere “ready” tobuyAltairs, therewerenoneavailable.The problem with the second possibility is that it is inconsistent with the

underlying construct of customer innovativeness. After all, Early Adopters ofonetechnologyarelikelyearlyadoptersofallsimilartechnologies.Itisthesameunderlyinginnovativeness thatmotivatesbehavior,so thosewhowerequicktoexperiment with Altairs were likely to be the first to purchase subsequentmodels.However, if theLaggards in theputativepopulationofAltairadoptersenter themarket by purchasing theApple II during the birth phase of its lifecycle,weareleftwiththeratherunusualconclusionthattheEarlyAdoptersandthe Laggards—populations with very different underlying levels ofinnovativeness—are entering the market for the same technology at the sametime.

WHEN:CATCHINGTHECOMPETITION

It is important to remember that the product life cycle is a general theory ofgrowth,notatheoryofDisruptivegrowth.AndsinceDisruptivegrowthisaveryparticular type of growth, it should come as no surprise that the product lifecycle cannot cope effectively with it. A Disruption penetrates a market notmerely as a function of less innovative customers finally “getting it.” Rather,Disruptive innovations grow because they have improved in ways that allowthem to appeal to the needs of more demanding customers. That is, theproductivityfrontierofaDisruptionexpands,drivenbyitsenablingtechnology.As its performance profile improves, it surpasses incumbent solutions andtriggers adoption by segments of the market that have more stringentrequirements. According to Disruption, then, the later adopters are not lessinnovativebutmoredemanding.Consider the shift from centralized to decentralized computer architectures.

PCs might have begun with hobbyists in garages and eventually displacedmainframes—and consumers that shifted from mainframes to minicomputersandeventuallytoPCswereakeycomponentofindustrygrowth.Itisdifficulttoclaim,however,thattheselateradoptersofPCswerelessinnovativethanearlieradopters,fortheyadoptedmainframeswhenearlyPCusersneverdid.Itismucheasier to countenance the view that mainframe and minicomputer users weremore demanding—and that as the PC caught up with their more exactingrequirementswhileofferingotherbenefits,theymadetheshift.Similarly,asthesoftware that drove PC performance got better—easier to use, more reliable,withawiderrangeofapplications—large-scaleconsumermarketsforPCswerecreated,andthePCbegantodisplaceotherformsofrecreation.ThishasreachedfullflowerinthelastdecadeastheInternethasonceagaintransformedthePC,servingasakeyenablingtechnology,muchasthemicroprocessordidduringthe1980sand1990s.The implication is that we can predict in a way that the product life cycle

framework cannot, when a Disruptor is going to begin its run of Disruption-fueledgrowthby tracking therateofexpansionof itsproductivity frontierandmappingthisagainsttheperformanceoftheincumbentsitisontracktodisrupt.InthecaseofSouthwest,forexample,thecompany’sabilitytocompeteonanygivenroutewasafunctionofitstotalcostadvantage,whichwasinturnlargelyafunction of its on-the-ground efficiency. The various defining elements of its

businessmodel limned an entirely new productivity frontier. As the companypushedthatfrontieroutwardthroughvariousinnovations,itwasabletoexpandits routestructure,growingbeyondTexasand intoCaliforniaandanumberofotherlocationsthroughthe1980s.Thanks to the relative inefficiencyof the737-200and737-300over longer-

haulroutes,Southwestwaslimitedtoflyingrelativelyshortroutes.Onanythingoverapproximatelyfivehundredmiles,thecompany’son-the-groundefficiencyadvantage was overcome by the higher costs of its plane. Consequently,Southwestdidnotenjoythecostadvantagerequiredforittoofferlowerpriceson longer routes. However, large-scale success in the general commercialaviation market demanded just this kind of route coverage. This allowedSouthwesttopreserveitsbusinessmodelandhaveasufficientlylowertotalcoston longer routes that, even with its price leadership, the company was moreprofitable.Asamatteroffact(ratherthanlogicalnecessity),IarguethatitwasimpossibleforSouthwesttoclosethatgapexclusivelythroughimprovementsinon-the-groundefficiency:by1985thecompanyhadbeenat it forfifteenyearsandwasstillaregional,short-haulcarrier.Thekeyadvancewouldhavetocomefromairframetechnologyintheformofamoreefficient737.Manytechnologieshaveatrajectoryofimprovementthatcanbemappedand

extrapolatedwithoftenimpressiveaccuracy.Perhapsthemostfamousexampleofthisisthemicroprocessor,theimprovementofwhichiscapturedbyMoore’slaw,developedbyGordonMoore, thecofounderand formerchairmanof IntelCorporation.Moore’s law states that the number of processors per area on anintegrated circuit doubles every two years, a trend first observed byMoore in1965thathascontinueduninterruptedforoverfourdecades.ThesekindsoftrendsmakepredictingthetimingofaDisruptionconceptually

straightforward, if at times technically challenging.When an innovator has abusinessmodelenabledbyatechnologythathasawell-understoodtrajectoryofimprovement,itbecomespossibletomakeplausibleestimatesofwhenagivenDisruptor will turn the corner from a profitable foothold into full-scaleDisruptivegrowth.InthecaseofSouthwest,onemighthavedonethisintheearly-to-mid-1980s

byextrapolatinghistoricaladvances inairframeefficiency toestimatewhenanaircraftwiththenecessaryefficiencywouldarrive.Butthisisasomewhatdodgyexercisewhendealingwithposthocdata: it isverydifficult toensure that thetechnical projections one creates are not informed by knowledge of whatactually happened.With so many design parameters to model (length, width,wheelbase,cruisingaltitude,cargoversuspassengerload,etc.),onecanpickandchoose plausible candidates that give the “right” answer, as well as perfectly

reasonableparametersthatgivevery,verywronganswers.Thegoodnews is that froman investmentperspective,at least, this levelof

precisiondoesnotappeartobeneeded.Theperformanceparametersofthe737-500 were known by the late 1980s. The plane was launched in 1989, andSouthwest incorporated it into its fleet in 1990. A backward look at thecompany’sstockprice,however,doesnotsuggestanythingisafootuntil1993,andisonlyunambiguousby1998,bywhichpointthestockhadrealizedhalfofits total appreciation for the decade. In otherwords, it would appear that oneneed not predict the improvement trajectory of the enabling technologyespecially accurately in order to get sufficient precision on the outcome ofinterest.Rather,oneneedmerelyunderstandtherelationshipamongthefrontierofagivenbusinessmodel,itsenablingtechnology,andthelevelofperformanceprovided by successful incumbents. When the necessary technologicalimprovementsarerealized,thatisthetimetobuy.Thismight lookdangerouslysimilar to the rather simplisticobservation that

an entrant will be successful when it is better than the competition. Do notoverlook, however, the importanceof theprocess bywhich theDisruptor getsbetter than the incumbents and is able to compete for lucrative mainstreammarkets. An ultimately Disruptive entrant creates a new business model thatallows it to reach combinations of price and nonprice value that incumbentscannotmatchandthatmostcustomersdonotvalue.Thefrontierofthatbusinessmodel is driven outward thanks to an enabling technology that allows theDisruptor to surpass the incumbents’ ability to serve increasingly demandingtiersof themarket.WhentheDisruptor’sfrontierovertakes the incumbents’ inthemainstreammarkets,Disruptionisimminent.Tosummarizeinjargon-heavybutmore precise language, the key is to understandwhat sort of performanceprofile is demanded by mainstream markets, and then determine when thefrontier of the Disruptive business model will overtake that portion of theincumbents’ frontier that meets the needs of the largest and most lucrativemarketsegments.ViewingDisruptivegrowthinthiswaygoesalongwaytowardexplainingthe

widevarianceoneseesinthepaceofDisruptionindifferentindustries.Toyota’sDisruptionoftheglobalautomotiveindustry,andGeneralMotorsinparticular,tookdecades,withlongperiodsoftimerequiredfortheJapaneseautomakertoreachsignificantmilestonesinitsrelativesize.Thismakessenseinlightofthenature of the improvements Toyota was counting on to fuel its growth: thesteadyaccretionofsmalladvancesinthemyriadprocessstepsrequiredtodesignand assemble an automobile. Nucor, the minimill steelmaker, and Southwesttookwhatwouldappear tobe shorterpaths toDisrupting incumbents, in large

part because they were fueled not only by their own process insights (e.g.,Nucor’s focus on safety andSouthwest’s on aircraft turnaround) but also, andperhaps primarily, by key advances in mechanical technologies such ascontinuouscastingortherangeandefficiencyoftheBoeing737.

FIGURE32:YEARSREQUIREDFORSELECTEDDISRUPTORSTOREACHSPECIFIEDRELATIVESIZE

Totherightoffigure32(above),Walmartmovesstillmorequicklythankstoits ability to tradeon advances in information technology to fuel its inventorymanagementsystems,butthepaceisslowedthankstothedramaticchangesinthelargeandcomplexprocessesrequiredtotakefulladvantageoftheseinsights.OnlywhenWalmart ismore thanhalf thesizeofSearsdoes itappear tohaveestablished sufficient scale to accelerate its growth, passing Sears in totalrevenueonly threeyears later.Cisco andCompaqmove themost quickly andconsistently of all, in keepingwith the fact that theirDisruptionswere fueledalmost entirely by advances in the electronic technologies at the heart of theirrespectiveproducts.This analysis suggests a general principle: businesses with productivity

frontiersfueledbyrapidlyevolvingtechnology—forexample,electronics—canbe expected to realize their Disruptive potential far more quickly than thoserelying on less rapidly improving technologies—for example, mechanicaltechnologies—while those relying primarily on process improvements willdisruptmostslowlyofall.26

HOWLONG:OVERSHOOTINGTHEMARKET

An entrant crosses the threshold from its foothold market to fully fledgedDisruptionwhenit is finallygoodenough—asaconsequenceof theexpansionoftheproductivityfrontierofitsuniquebusinessmodel—todoabetterjobformainstream customers than solutions provided by incumbents. In otherwords,Disruptionbeginswhentheentrantcatchesupwiththecompetition.Sadly,nothinglastsforever.NotrajectoryofDisruptivegrowthisaticketto

infinitelylongperiodsofdouble-digitgrowth.Thisprincipleisbothapracticalreality and a mathematical necessity. Success engenders imitation, andeventuallysomeoneelsefiguresouthowtocompetewithandbesteventhemostentrenchedmarketleader;andanycompanythatgrowsfasterthantheeconomyfor longenougheventuallybecomes theeconomy,whichbrings itsownsetofproblems.Knowingthatsomethingwillendisverydifferentfromknowinghowlongit

will lastbefore itdoes.Andseeing theendcomingcanbeenormouslyhelpfulfor both investors and managers. Investors want to know when to shift theirexpectationsforacompanyfrom“growthstock”to“valuestock”(thatis,fromwealthcreatedbysharepriceappreciation towealthgarnered fromdividends),andmanagersaretryingtodecidewhentoshifttheiremphasisfromexploitingtheexistinggrowthenginetofindingthenextone.Theconventionalproduct lifecycleviewholds thatgrowthwill slowas the

marketbecomessaturatedandonlytheLaggards(theleastinnovativecustomers,andthelasttoadopt)remaintoenterthemarket.Sincethesecustomersarethemostdifficulttoconvinceandtherearerelativelyfewofthem,asyouenterthematurityphasewithoneproductyouwanttobesureyouarewellonyourwaytocrossing into the growth phase with something new. As before, however, theproductlifecyclebeginstofraywhenproductschangeovertime.LaterversionsofexistingproductsareadoptedbybothEarlyAdopters,whotradeup,andbythelessinnovativeyetfirst-timebuyers.Sincetheybothadoptthesameproductat the same time, it is difficult to see them as having different levels ofinnovativeness.This iswhy I introduced thenotionof theDisruptor “catchingthe competition,” rather than simply adoption by customers, in order tounderstandwhenDisruptivegrowthwillbegin.Productsarenot theonly things thataredynamic,however.Customerneeds

changeaswell,ofteninresponsetoadeeperunderstandingofhowaparticular

solutionfitsinto,andperhapsevenchanges,theirlives.Theautomobilestartedoutappealingtoanarrowrangeofinterestsforasmallnumberofpeopleandhascometoshapepracticallyeveryaspectofoursociety;thesamecanbesaidofthemobilephone,thecomputer,andtoalesserextentahostofothertechnologiesandservicesthatmanyofususeeveryday.Westartoutquitesatisfiedwiththeperformanceof agivendevice, andaswecome todependon it and findnewwaystouseit,ourexpectationsofitinvariablyrise:whatwasonceamarvelandboonbecomesaconstraintandasourceoffrustration,notbecauseithasgottenworsebutbecausewehaveraisedthebar.Consequently, product improvements can be rewarded even after one has

bested the competition. In a very nearly inscrutable positive feedback loop,betterperformanceenablesadditionaluses,whichcreatesaneedforstillbetterperformance—avariationonSay’s lawof“supplycreates itsowndemand.” Itwould appear, then, that there is an endlessmarket for ever-better-performingproducts.JustasSay’slawhasagrainoftruthbutcannotbeuniversallygeneralized,it

turnsout—again,asamatteroffact,nottheoreticalnecessity—thatthepaceoftechnological improvement almost always outstrips the pace at which we canincorporatethoseimprovementsintoourlives.Betterperformanceatsomepointsimplyconfusesus:wecannolongeradaptourlivestotheusefulexploitationofmore-better-faster, so we become fundamentally indifferent to what used toexcite us, accepting “better” onlywhen that implies “cheaper.”This is, in theargot of established Disruption theory, “overshooting” the market. That is,providingmoreperformancethanmainstreamcustomersarewillingtopayfor.ItisherethatDisruptiontheorytakesonasurfacesimilaritywiththeproduct

life cycle model. For just as growth explodes for a Disruptor as it bests theincumbents, growth slows when the Disruptor exceeds the performancerequirementsofmainstreamcustomers.Butnowthesigmoidalgrowthcurvethatissooftenobservedinproductmarketsisexplainedbyanormaldistributionofcustomer “demanding-ness”: small numbers of undemanding customers in thefootholdmarket,largenumbersofmoredemandingcustomersinthemainstreamresultinginexplosivegrowth,andsmallnumbersofverydemandingcustomersatthehighend.Whatdrivesthisdeclineingrowthrate,however,isnotmarketsaturation as much as a mismatch between the rate of product performanceimprovementandtheabilityofcustomerstomakeuseofthoseimprovements.We can see this phenomenon atwork in the last twenty years of growth at

Intel.Intelbeganlifeasamakerofmemorychips(DRAMs),butsincethemid-1980sthecompanyhasbeendefinedbyitsmicroprocessorbusiness.Anengineof Disruption for the PC, it provides as clear an example as there is of how

overshootplaysout.The first microprocessors Intel offered to the market were, by today’s

standards, very low power. They were not good enough for the mainstreammarketsoftheday:themainframeandminicomputermakers.Asaresult,Intelhad tosettle fora foothold in the relativelyundemandingmarketpopulatedbycheaphandheldcalculators.Intel’sfirstvolumecustomerforitsprocessorswasBusicom,amanufacturerofsimplefour-functioncalculators.ForInteltogrowitsbusiness,itneededtofindwaystoappealtoever-more-

demanding customers. Thismeant improving the processor speed (megahertz)and power (number of transistors) of its chips. Such improvements wereenormouslydifficult todeliver,andtheengineeringchallengeswereoftenseenasnearlyinsuperable.Buttheincentivetofindawaywasextremelypowerful:moreprofitableandmorerapidgrowth.Eventually, Intel was able to bring to market a series of previously

unimaginableimprovements.Eachnewgenerationofchipdidnotsufferthrougha “birth” phase that bore any resemblance to what diffusion research mightpredict. Rather, each new generation propelled the underlying growth of theindustry for at least three reasons. First, Early Adopters had integrated thetechnology into their livesandfound theyneededstillmorecomputingpower.So they traded up.New adopters found that the new generations of computerwerefinallypowerfulenoughtomeettheirneedsforeasy-to-usesoftwarewithintuitive graphical interfaces.And improvements in processor speeds, togetherwithotherimprovementsinassociatedtechnologiessuchasdisk-drivecapacity,resulted in the displacement of mainframes and minicomputers by a moredistributedcomputingarchitecture.Andforaslongasthemarketwasrewardingthese improvements, Intelcouldconfidentlyconclude that itwas in thegrowthphaseofthelifecycleofthepowerandspeedtrajectory.Thechallenge,ofcourse,istopredictwhenthistrajectorywillovershootthe

willingness of consumers to pay for still more improvements. This can bedifficult tododirectly,butwecanobservedeclines incustomerwillingness topay for each generation of improvement and from that infer that overshoot issetting in. Early products tend to command high price premiums for a longperiodoftime,sincetheyappealtolargesegmentsthatvaluetheperformanceonoffer. Consequently, when new generations of chips begin to suffer pricedeclinessoonaftertheirintroduction,wecanconcludethatagiventrajectoryhas“overshot”thedemandsofmainstreammarkets.By 1985, the Intel-fueled Disruption of minicomputers by PCs was in full

flight. The x386 chip, the brains of the IBM PC, stayed on the market forseventy-five months, suffering a 67 percent price decline from its price at

introductionbeforebeingremovedfromthemarket.Subsequentgenerationsofchip, such as the x486 andPentium I, commanded significant price premiumsover earlier chips,which suggests that customerswere stillwilling to pay forincreases in performance. However, each subsequent generation enjoyed ashorterlifespan,andhenceitspricefellfartherandfaster.Inotherwords,therewere fewer customers willing to pay for this increase in performance—if notfewercustomerswillingtoacceptthisperformanceatasufficientlylowprice.Partof the reason for the shortened life spanof subsequentgenerationswas

competitivepressure.Intelhaslongbeenaleaderinthemicroprocessorindustry.Bytheearly1990s,maintainingthisleaddemandedanever-acceleratingrateofnew product introduction, with each generation of microprocessor deliveringgreater advances in processing speed and power. In short, to stay ahead, Intelwasbeingforcedtomovealongitssustainingtrajectoryatanever-fasterrate.The Herculean efforts protected Intel’s performance lead, but at a cost: its

ever-better products had ever-shorter life spans. This signals that by the early2000s,Intel’strajectoryofever-greaterspeedandpowerhadovershottheneedsofthemainstreammarket.Highinitialpricesfornewgenerationsofchipswerepaid only by a very few, very demanding customers at the high end of themarket.Theirnumberswerequicklyexhausted,andsocomparativelysoonafterintroduction,themarketfortheseproductshadbeensaturated.Lessdemandingcustomers—by now themajority of total customers—were quite happy to usefaster chips; they justwere notwilling to paymuchmore for them than theywerepayingforthechipstheywerealreadyusing.Maturityhadarrived.Orsoitseemed.Asearlyasthemid-1990stherewasevidencethatovershoot

wassettingin.AMD,alongtimecompetitor,wasenjoyingmaterialmarket-sharegains:in1999,forthefirsttime,AMDhadhighermarketsharethanIntelintheU.S.retaildesktopsegment,with43.9percent,thankslargelytoitsgainsinthesub-$1,000-systemsegment.27AMDhadgainedthisleadbybeginningearly—inthemid-1990s—tofocusonlessdemandingtiersofthemarketwherechipsthatwerelesspowerfulthanthebestthatIntelhadtoofferwerewelcomedwithopenarms, especially since theywere being sold atmuch lower prices than Intel’shighest-performingproducts.Inotherwords,thedimensionsofperformancethatconstituted theobjective function and thatwere theconstraints had essentiallyflippedinwhathadbecomethemainstreammarket.ForIntel’sentirehistoryasadesignerandmanufacturerofmicroprocessors,theobjectivefunctionhadbeenpower and speed,while the constraint had been price. That is, customers hadwanted to maximize the performance of their computer systems subject to abudget.AMD provided “good enough” performance for these less demandingsegments (the constraint) while seeking to keep price down (the objective

function).

FIGURE33:PRICESOFINTELMICROPROCESSORS,1985–2005

Inshort,Intel’smicroprocessorbusiness—whichhadfortwentyyearsbeenanengineofDisruptioninthecomputerbusiness—wasnowsubject toDisruptionitself. AMD had drawn a new curve, reaching a new point in nonprice/pricespace, andwas appealing to the low end of themicroprocessormarket. Therewas also good reason to think that AMD would be able to systematicallyimprove the performance of those relatively low-end chips inways that couldultimatelyDisruptIntel.CertainlyAMD’searlysuccesswasworryingtoseniorIntelmanagement,especiallyAndyGrove,thenCEO.Intel’sresponsewastoestablishanewunitinIsrael,farawayfromthecore

operations inSantaClara,California, tofocusonbuildingwhatwouldbecometheCeleron processor. Launched in 1998, Celeron quickly became the largestlineofprocessorsbyrevenueinIntel’shistory.BasedonthePentium“chassis,”

the Celeron was a deliberate attempt to fight back with a lower-cost, lower-priced,lower-performancemicroprocessorinordertodefendIntelfromnascentDisruptive attacks. Consequently, the Celeron was launched not as a better-performing,higher-pricedchipbutasa“goodenough”chipwithanintroductoryprice onlymarginally higher than the prices at which previous generations ofchipshadbeenremovedfromthemarket.Onlytoday,morethantenyearsafterits launch, is Celeron beginning to be phased out of the low end of themicroprocessormarket, being replaced byAtom, Intel’s new line of low-pricemicroprocessors.Grove explicitly credits the application ofDisruptionwith enabling Intel to

seethelong-termsignificanceofAMD’sinroadsinthelowendofthemarket.Inaddition,Grovemaintainsthat,althoughDisruptiondidnotprovidetheanswerdirectly,itdidgiveIntel’speoplethetoolsrequiredtodiscoverwhatneededtobedone,enabling themtodevelopcompetitiveresponses thathebelievesIntelwouldotherwisevery likelyhavemissed,orat leastnot seenuntilmuch later,therebymakinganeffectivecounterattackthatmuchmoredifficult.28Launchinganeffective “counter-Disruption”onyourownbusiness is a rare

andremarkablefeat.Itnotonlyrequiresstrategicinsightandapowerfulabilitytoimplementthoroughgoingorganizationalchange,italsohastobestructurallypossible. In some industries, once theDisruptive trajectory has run its course,there is little else one can do to grow in a Disruptive manner. Now thatSouthwest can fly competitively on just about every route in the continentalUnitedStates—including,asof2009,NewYorktoLosAngeles—thereislittleelseforittodobutcontinuetosqueezetheincumbentsonflightswithintheU.S.market. For although the Boeing 737-900 has a range of up to ten thousandkilometers,therestofSouthwest’sbusinessmodelisunlikelytosupportmaterialvolumeonverylongflights.29Consequently,forSouthwesttocompetefortheseroutes,itwouldlikelybecompelledbytherequirementsofthemarkettoadoptbehaviorsthatwouldcorruptitsstrategicdifferentiation:meals,seating,classesofservice,etc.,whichwouldinturnundermineitscostadvantage.In sum, then, a given Disruptive trajectory lasts until one has overshot the

performancerequirementsofthemarketoneisDisrupting.Thiscanbepredictedtypicallyinoneoftwoways.First,watchforstrongdownwardpricingpressuredue to customer resistance rather thancompetitive incursions;whencustomerswantforfreewhattheyusedtopayforgladly,chancesareyouareprovidingtoomuch of a good thing. Second, look for structural limits to the level ofperformance a business model can provide: the enabling technology mightpermitstillgreaterlevelsofperformance,butthisfuelsDisruptivegrowthonly

solongastherestofthebusinessmodelneednotchangeinordertoexploitnewopportunities. When either of these constraints becomes binding, Disruptivegrowthisatanend.

PARTIIIAPPLICATION

CHAPTERSIX

DELIBERATELYDISRUPTIVE

Chapters 1 and 2 made the case for Disruption’s predictive power.Chapters3–5madethecaseforthetheory’sexplanatorypower.Chapters6–8 illustrate practical applications of the theory. This chapter exploreshowJohnson&JohnsonhasusedDisruptiontoshapetheevolutionof itsSEDASYS™automatedsedationdeliverysystem.

One questionmany people have about examples of Disruptive companies is“Was Southwest [or Nucor or Maxtor or Walmart or …] Disruptive onpurpose?”Itisagoodquestion,andtheansweristypicallyno.Mostcompaniesthat end up Disrupting powerful incumbents start out focused primarily, andoftenexclusively,onconnectingwithaspecificsegmentofthemarket—onethatispoorly servedby incumbents.Oftenmuch smaller than the incumbents theyhopetounseat,strappedforresourcesofalltypes,andlackingacapitalbasetocoveroperatinglosses,would-beDisruptorsaretypicallyhungryforrevenueanddesperate for profits. Consequently, they are looking only for some relativelyuncontestedmarketwhere their disadvantages compared to incumbents do notmattermuchandareperhapsevenabenefitofsomesort.ItistheseconstraintsthathavehistoricallyfosteredDisruption,notstrategicinsight:mostcompaniesthatendupDisruptingstartoutlookingnotforapotofgoldbutratherforsomeniche where they have a much better chance of success than if they were tocontestmainstreammarketswheresuccessfulincumbentsdominate.In many cases, eventual Disruption—the migration from a niche to

mainstreammarkets—hasalsobeenafunctionofluckasmuchasanythingelse,and for two reasons. Disruption is driven by improvements in key enablingtechnologies, and very often those improvements are provided by suppliers ofthosetechnologies,notbytheDisruptoritself.Consequently, theDisruptorhasto have had the good luck first to have built its business model around atechnologythatcanimproveintherightsortsofways.Second,theDisruptorhasto have suppliers that actually do improve their products in the right sorts ofways,sincetherequisiteimprovementsarebynomeansguaranteed.WehavestuckwithSouthwest this long,sowemightaswellfinishthejob.

KeytoSouthwest’sbusinessmodel,andthenewproductivityfrontieritwasable

to create, was standardizing on the 737 aircraft. This standardization allowsSouthwest’s crews to be much more flexible in the routes they fly; it makesmaintenance more efficient; and it provides a host of other benefits thatcontributemateriallytoSouthwest’sprofitability.CentraltoSouthwest’sgrowthand eventual Disruption of established airlines, however, was the increasedrange and efficiency of subsequent generations of the 737. So how didSouthwest come tomake suchanastute strategicdecision?DidHerbKelleherforesee the 737’s future improvements in range and operating efficiency andunderstandtheirimplicationsforSouthwest?Perhaps … but unlikely. In fact, it might even be a stretch to argue that

Southwest management understood the impact of standardization onprofitability,nevermindonthegrowththatwastwentyyearsinthefuturewhenthe737reducedtheoperating-efficiencygapwithotheraircraft.AnexplanationthatIfindmuchmoreplausibleisthatLamarMuse,CEOofSouthwestin1971andtheninchargeofbuyingSouthwest’sfirstaircraft,securedadealforthreesurplus737-200sthatBoeingwaswillingtounloadata20percentdiscountandwith90percentvendorfinancing.30Havingexhaustedmostof itscashinlegalbattles to get permission to operate in themore heavily regulated commercialairlinemarketsoftheearly1970s,thatkindofdealwaslikelytoogoodtopassup.Overtime,thebenefitsofstickingtoasingletypeofaircraftintheserviceofSouthwest’sfootholdmarketbecameclear,andsoasthecompanygrewwithinthatnicheitpreservedthiselementofitsbusinessmodel.By the1980sSouthwestwasan important enoughcustomer forBoeing that

the airline could begin influencing the development path of the aircraft itpurchasedandsocouldplayamoreactiverole inshapingitsownfuture.Thisallowed the company to push for, say, better efficiency and range rather thanincreased seatingcapacityand flexible cabinconfigurations.But these sortsofadvancements only made sense in light of a commitment made more than adecadeprior,andverylikelywithnoparticularlyprescientinsightintohowthecompany’scompetitivefuturewouldbedeterminedbytheevolutionofthe737.Many other Disruption stories include similarly generous helpings of good

fortune.Hondawasdetermined topenetrate theU.S.motorcyclemarket in the1950susingwhatwewouldrecognizeasasustainingattack:sellingwhatHondathoughtwerebetterbikes to themostattractivesegmentsof themarket. Itwasfailing miserably and turned things around only when, desperate to sellinganything to anyone, it stooped to peddling its 50 cc Super Cub minibikes insporting-goods stores. From that foothold, the company was able to rideimprovementsinsmall-enginetechnologyanditsownproductionprocessestoa

position of leadership in themotorcycle and eventually automotive industry.31AndhowcouldWalmarthaveknown that advances in information technologywould be central to itsDisruptive growth in retailwhen it began focusing oninventorymanagementandexcellenceinlogisticsinthe1970s?Doesgoodluckmatter?Absolutely.Thelargerquestion,ofcourse,iswhatthe

restofusaresupposedtodoifwewanttoDisruptonpurposeratherthanrelyingonthewhimsoffate.The answer implied by what has been said so far is conceptually

straightforward:

Identifyalarge,lucrativemainstreammarketyouwanttoDisrupt.Identifywhattradeoffsyouneedtobreakinordertostealthatmarketfromthesuccessfulincumbentsthatcurrentlydominateit.Identify a market segment of relatively little interest to any dominantincumbents that values a different set of tradeoffs among the samedimensionsofperformance.Buildabusinessmodelthatservesthosecustomersprofitablyandwell.Besurethatbusinessmodelisenabledbyarapidlyimprovingtechnologythatwillultimatelyallowyoutocompetefor themainstreammarketsyouwant to disrupt from a position of structural advantage by breaking thetradeoffsyouoriginallyidentified.Map the rate of change in the incumbents’ productivity frontier, yourproductivityfrontier,andchangingcustomerrequirementsso thatyoucandeterminetheappropriatetimetableforinvestment.

Ifyouarerollingyoureyesindisbelief,thatistherightresponse.Asexploredinthepreviouschapter,Disruptiontakesplaceinanenvironmentofconstantlyshifting customer requirements and technological performance profiles. Yetpredicting either customer preferences or the paths of technologicalimprovement with any meaningful accuracy is essentially impossible, andgetting both right simultaneously—while folding in the uncertainties ofexecutinganystrategy—isimpossiblesquared.Weneednot,however,abandonallhope.AlthoughDisruptionoftenturnson

luck,itisaluckthatplaysoutwithinlaws,andforthosewhounderstandthoselaws, it is possible to influence materially when and how luck will strike. Inotherwords,whenitcomestoDisruption,aswithsomuchelse,itispossibleto

makeyourownluck.

DISRUPTIONATJOHNSON&JOHNSON

To see this in sharpest relief, consider the role Disruption has played in thegrowth of Johnson& Johnson (J&J) over the last fifteen years.Thanks to thecompany’s diversified portfolio of businesses and decentralized managementphilosophy,J&Jillustratesmanyoftheprinciplesthathavebeenexploredsofar.J&J has the three main operating divisions: Consumer Products (Band-Aid

brand bandages, Tylenol pain reliever, etc.; revenue: $15 billion),Pharmaceuticals(Remicade,ananti-inflammatory;Procrit,whichstimulatesred-blood-cell production, etc.; revenue: $22 billion), and Medical Devices andDiagnostics (MD&D—Ethicon sutures, Cordis coronary stents, etc.; revenue:$23 billion). A high-level analysis of each division’s main products revealsdifferences in the importance ofDisruption to each for reasons that appear tolineupnicelywithwhatthetheorypredicts.(IdescribehowandwhydifferentlinesofbusinessareallegedtobeDisruptiveorsustaininginappendixD.)

FIGURE34:PERCENTAGEOFDIVISIONALREVENUESGENERATEDBYDISRUPTIVEPRODUCTSATJ&J,1996–2009

Specifically,thePharmaceuticaldivisionappearstohaveverylittleinthewayofDisruptiveinnovationdrivingitsrevenue,butnotbecauseofanyfailingsonitspart.Rather,asexploredinchapter4,thepharmaceuticalindustryseemstoberelatively immune to Disruption. (Once again, this appears to be a fact, nottheoreticalnecessity.)Foralthoughtherearealternativebusinessmodelsthatcanreach different points in price/nonprice value space (e.g., generic drugmanufacturersversusdiscovery-drivendrugcompanies),therehasbeensofarnoenabling technology that can propel the frontier defined by that, or any other,alternativebusinessmodeloutwardinawaythatleadstotrueDisruptivegrowth.J&J’s Consumer Products division has had one product group, Women’s

Health(createdin1999),thathasDisruptivehome-basedpregnancytestingandotherproducts.Thegrowthofotherproductgroupshasgenerallybeengreaterthan that of Women’s Health, however, and the sharp drop in 2007 was afunction of the acquisition of Pfizer Consumer Health by J&J—an astute butclassicallysustainingacquisition.TheMD&DdivisionhaslonghadthemostsignificantDisruptivecomponent

toitsrevenue.Thedivisionhasshownagraduallong-termincreaseinrevenue

fromDisruptions, climbing from25percent to justover40percent.This is inpartafunctionofthetypesofindustry-levelvariablesdiscussedsofar.Medicaldevices and diagnostics can turn to advances in materials science,miniaturization, computer technology, and other electronics to expand theirproductivity frontiers, and many of these disciplines enjoy the kinds ofexponential increases in performance that have driven so many other high-profileDisruptions(diskdrives,computers,etc.).AgoodexampleofthisisJ&J’sCordisdivision.Cordis’sprimaryproductis

thearterialstent,adevicethatpropsopen(typicallycoronary)arteriesthathavebecome dangerously occluded with plaque. Historically, severe arterialblockageswere treatedwithcoronaryarterybypassgrafts(CABG,pronounced“cabbage”).Thissophisticated,expensive,andriskysurgicalprocedureinvolvesharvestingalengthofveinfromthepatient,usuallyapieceofthesaphenouslegvein, and bypassing blocked segments of the coronary arteries. It is a veryeffectivetreatmentforalife-threateningcondition.Onemightthinkthatinlightoftheskyrocketingincidenceofheartdiseasein

theUnitedStatestherewouldbeasimilarincreaseinCABGprocedures,butthishasnotbeenthecase;instead,thenumberofCABGprocedureshasbeenfallingsteadily.Howarealltheseblockedcoronaryarteriesbeingtreated?Theanswer isangioplastywithcoronarystenting,aprocedureperformedby

the relatively new specialty of interventional cardiology.Angioplasty involvestheinsertionofaballoon-tippedcatheterwithametalstent—essentiallyawirecagecollapsedaroundtheballoon—intothefemoralarteryandthreadedthroughthe blocked segment of a coronary artery. The balloon is then inflated,expandingthewirecageandreestablishingbloodflow.Angioplasty has been Disruptive because of the way it evolved in the

marketplace. The very first angioplasty procedures were performed onextraordinarilysickpatients,peoplesoillandfragilethat thetraumaofCABGsurgeryposedunacceptable risks.From thesevery early-stage applications theprocedurewasimprovedtothepointthatitwasgoodenoughformoregeneralapplications.Atfirst,however,comparedtoCABGtheprocedurecouldaddressonlyrelativelylowlevelsofblockageinrelativelyeasy-to-accessarteries.Ithadthe advantage in some of these cases, however, of being less risky and lessexpensive—essentially reaching a point in price/nonprice value space thatCABG could not and that cardiac surgeons, who performed the CABGprocedures, had little interest in addressing: they were busy enough with thepatientstheyhad.The frontier of angioplasty expanded as a consequence of a number of

innovations.Somewereprocessbased,bornsimplyasaconsequenceofanever-

increasing number of procedures. Specialists learned which types of patientsweregoodcandidates, andas theprocedureprovedsafeacrossamorediversepatient population, angioplasty began to obviate future surgeries for patientswith relatively early-stage heart disease and eventually substituted for someCABG procedures. Other innovations were more substantive, including thedrug-elutingstent,whichpermittedstents tostay inplace longerbycombatingrestenosis,orthereblockageofthestentedarteryduetoanimmuneresponsetothe presence of the stent in the body.The result has been an explosion in thenumber of angioplasties and a precipitous fall in the number of CABGprocedures.CordisrodethatwaveofDisruption.Revenuesfromstentsin1996were$790

million and peaked in 2006 at just over $4 billion. Since then, competitivepressure in the stent business has depressed revenue to $2.7 billion in 2009.Nothing lasts forever, but it was quite a ride, and the division remains asignificantandprofitableelementoftheMD&Dportfolioofbusinesses.AlthoughthestrategyandgrowthplansatCordisweredeliberate,itisfairto

saythatCordiswasnotdeliberatelyDisruptive,atleastinitsearliestdays.Sincethen,however,manybusinessdevelopmentexecutivesandseniorleadersatJ&Jhavebecomeexpert in thetheoryandpracticeofDisruption.Convincedof thepower of Disruption theory, MD&D has begun developing new businessopportunitieswiththeexplicitintentofrealizingDisruptivegrowth.Anexampleplayingoutnowisthedivision’sSEDASYS™automatedsedationsystem.HowSEDASYS™has been developed and shepherded through its early stages andinto themarketplace isanarchetypalexampleofhowDisruptiongetsgoing inthe realworld—warts andall.Topredict the successofSEDASYS™,wecanevaluate how it has been developed against the expanded decision tree ofsuccessfulDisruption.Specifically:

Itissustaining?Does the division commercializing SEDASYS™ have the requisiteautonomyfromestablishedbusinesses?Doesthesolutionbeingdevelopeddescribeanewproductivityfrontier?Aretherekeyenablingtechnologiesthatwillallowthecurrentsolutiontoimproveinwaysthatwillallowittocompeteformainstreammarkets?

AVOIDSUSTAININGINNOVATION

What is now SEDASYS™ began life as a partial equity stake by J&J’sventure capital arm, Johnson & Johnson Development Corporation (JJDC) inScott Labs (Scott) in 1998. An anesthesiologist, Dr. Randy Hickle, foundedScott to develop a technology that could automate the delivery of moderatesedationwithnitrousoxideandpropofol,twohighlyeffectiveanestheticdrugs.JJDC’s investment funded Scott to take the idea from concept to “first inhuman” trials, a milestone reached in 2000 when the device was used undercarefully monitored and controlled clinical experiments at the University ofCaliforniaatSanFranciscoandtheUniversityofLouisville.AsaninvestmentbythecorporatearmofJ&J,ratherthantheMD&Dgroup

ofoperatingcompanies,thisearly-stagestakecanbeseenasaformofstrategicoption.32 Once the technical uncertainties and clinical utility had been largelyresolved, however, SEDASYS™ became the responsibility of MD&D,specifically J&J’sEthiconEndo-Surgery (EES) division, led byKarenLicitra,Worldwide Company Group Chair, who leads the EES and AdvancedSterilizationProductsbusinesses, togetheroneof the largest groupswithin theMD&Dsector.As an early-stagemedical device in the hands of a growing and profitable

operating division, SEDASYS™ faced all the usual challenges. Although theinitial capital investment in Scott Labs had been covered by JJDC, MD&Dwouldhavetofundseveralmoreyearsofpotentiallyexpensivedevelopmentandtesting: proof of concept is a long way from Food and Drug Administration(FDA)approvalandmarketplaceacceptance.Consequently,astereotypical—butbynomeanscaricatured—applicationofconventionalbusinessplanningtothiskindofproductwouldtargetdevelopmenteffortsatlarge,lucrativemarketswiththeintentofminimizingboththelikelihoodoffailureandthepaybackperiod.A solid conventional business plan would begin by identifying who used

propofol and finding away to appeal to thosemarkets.Thiswasnotdifficult:propofol in particular is very popular with physicians because it allows for arapidonsetandquickoffset.Thisenablestheusertoquicklyreachthetargeteddrug effect for the patient. Also, because it is so fast acting and its effectdissipatessoquickly,itsdosagecanbeadjustedinrealtimetoinducedeeperorlighterlevelsofsedationdependingontheprogressofaprocedure.However, this flexibility comes at a cost: the FDA-approved label (list of

approved uses and conditions of use) for propofol specifies that it is to beadministered only by anesthesia professionals (APs), specifically,anesthesiologists, who are MDs, or Certified Registered Nurse Anesthetists

(CRNAs),who are notMDsbutwhohave three years of specialty training inanesthesia.Consequently,asbusinessplanningforSEDASYS™developmentandlaunch

beganwithinMD&Din2001, the targetmarketconsistedofprocedureswhereAPsusedpropofol.Given thecostofhavinganAPin thesurgicalsuite, thesetended to be relatively complex and expensive surgical procedures: colonsurgery,cholecystectomy,hysterectomy,lungsurgery,herniarepair,andsoon.SecuringadoptionoftheSEDASYS™systemsfortheseprocedures,andinthissetting,wouldmeanoneoftwothings:displacingtheanesthesiaprofessionalsorincreasing their productivity. Each of these alternatives promised significantchallenges.In the first instance, displacingAPs in theoperating roomwouldmean that

SEDASYS™wouldhavetobebetterthantheywerealongallthedimensionsofvaluetheyprovidedtothepatientandsurgicalteam.Yetthefirstgenerationofthedevicewouldbeunablecrediblytomakesuchclaims,andforatleastthreereasons. First, in order to ensure patient safety, the algorithms that govern theSEDASYS™systememployconservativeparameters.Forexample,shouldthepatient’sbloodoxygenlevelsorrespiratoryratedeviatemore thanavery littlefrom acceptable levels, the device automatically reduces or stops theadministration of propofol until the patient’s condition has been resolved,effectively putting the procedure on hold. In contrast, an AP canmuchmorereadily adapt tominor deviations and allow theprocedure to continuewithoutcompromisingpatientsafety.Second,becauseoftherelativelynarrowparameterswithinwhichautomated

systems can function, higher-risk patients, for example themorbidly obese orthose with a severe systemic disease, would not be good candidates forSEDASYS™: these patients require the presence of an AP to manage anypotential complications. Third, and also a consequence of the limitations ofautomated systems, SEDASYS™ was designed only for procedures whereminimaltomoderatelevelsofsedationaretargeted,whereasanAPistrainedtothefullcontinuum,fromminimalsedationtogeneralanesthesia.These hurdles might not have been technically insurmountable. The irony,

however, is that building a business plan around targeting the largest then-existing markets for propofol—something intended to minimize risk andmaximizereturn—wouldhaveimpliedmuchlonger,moreexpensive,andmoretechnically uncertain development in order to meet the performancerequirementsofthatmarket.Inotherwords,targetingthe“best”marketactuallyincreasedrisk.Theotherconventionalalternative—increasingtheproductivityofanesthesia

professionals—would not have made much sense, either. PositioningSEDASYS™as an “APmultiplier,” inwhich oneAPmightmonitormultipleSEDASYS™ devices deployed in concurrent procedures, would lead toscheduling,billing,andlogisticaldifficulties,makingthefullrealizationofsuchefficienciesquitedifficult.Worse, by targeting these seemingly lucrativemarkets, SEDASYS™would

be forced to focus on a benefit—cost reduction—that was of relatively littleinterest to keydecisionmakers.Theprocedureswhere propofolwas generallyused had long includedAPs, and they are integralmembers of the operating-roomteam.AnAPisahighlytrainedmedicalprofessionalwhobringsskillsandinsight that go beyond administering pain-management medication; noautomated replacement could ever match the full range of contributions APsmakeinthesurgicalsuite.Inotherwords,thecostofhavingAPsinthesurgicalsuite reflected the value they provided. SEDASYS™ would not be a simple“same for less” productivity proposition. Instead, to displace APs theSEDASYS™teamwouldhave toconvinceall the relevantdecisionmakers toacceptadifferenttradeoff:getpainmanagementatalowercost,buttheneitherdowithouttheadditionalvaluethatAPsprovidedorfindotherwaystoreplicateit.Finally, and perhaps worst of all, the FDA, which regulates new medical

devices in theU.S.,must take into consideration a greatmanymore variableswhenadevice’s label includesabroadrangeofapplications.Byattempting togo“wherethemoneyis,”SEDASYS™wouldbetargetingallprocedureswhereAPs administered propofol. This was a long list, which would make it time-consumingandexpensive todemonstrate thatSEDASYS™could clear all therelevanttechnicalhurdles.Had the SEDASYS™ team gone after the market of “AP-administered

propofol,” Disruption theory would have predicted failure. As an entrantpitching a sustaining innovation, you rarely have much of a chance; if youroffering is a sustaining innovation that is onlymarginally better in ways thatcustomersdonotparticularlyvalueand thatcouldbeperceived tocomeat thecostofwhattheyvaluemost,youhaveeffectivelynochanceatall.Andifthatvalue proposition implies a lengthy, expensive development and approvalprocesstoboot,well…’nuffsaid.Instead,theSEDASYS™teamexplicitlydevelopedaDisruptivestrategyand

targetedprocedureswhereAPs,andhencepropofol,hadessentiallynopresence:gastrointestinal(GI)proceduressuchascolonoscopies,orlower-GIendoscopy,and esophagogastroduodenoscopy (EGD), or upper-GI endoscopy. In 2001 thevast majority of these procedures were done without the support of an AP

becausetheyweretypicallyperformedonhealthyadultpatientsandsorequiredonly minimal to moderate sedation in the interest of patient comfort andcooperationduringtheexams.GIphysiciansrequestedAPsupportforthatsmallsegment of patients who, due to health conditions or comorbidities, wereconsidered at risk for complications during even simpleGI procedures; but in2002,lessthan2percentofGIproceduresinvolvedanAP.At the same time, theeffectsofpropofol,with the right safetymechanisms,

can be customized to each patient, and it is an excellent sedative for GIprocedures.However,asdiscussedearlier,propofolmustbeadministeredbyanAP, and this limitation had led most GI physicians to choose from a set ofalternativesedatives thataregenerallyconsideredinferior topropofolfor theserelativelyshortoutpatientprocedures.Itwas a cumbersome tradeoff: use anAPandpropofol and reduce the time

required for the procedure and improve patient outcomes, but at materiallyincreased cost, or settle for suboptimal solutions that were less expensive buttooklongertoadministerandrecoverfrom.ForGIphysicians—thenonconsumersofpropofol—SEDASYS™offeredtwo

benefits: (1) the ability to automate one element of the procedure, therebyminimizing procedure-to-procedure variability with the potential to improveproductivity,and(2)anopportunitytousepropofol,which,withSEDASYS™,promisesmoreprecise sedationspecific to thepatient’s individualneedswhileminimizing the risk of oversedation. This last benefit typically resulted in asignificantreductioninrecoverytimeandsoaquickerreturntonormalactivity.SEDASYS™ could therefore offer benefits to the GI physician, the overallperformance of the practice, and the patientwithout collidingwith the vestedinterests of the APs or removing a skilled medical professional from theprocedureroom.SEDASYS™hadfoundits“rebar,”its“SanAntonio—Houston—Dallas,”its“ruralretail”—inotherwords,itsfoothold.Pursuingthatfootholdrequiredastrategiccommitment.Productdevelopment,

FDAfilings,advisoryboards,andtheengagementofAPsontheSEDASYS™AnesthesiaAdvisoryPanel—allwerenotmerely coloredor influencedby thisdecisiontotargetGI,butrevolvedaroundthischoice.Between2000and2004SEDASYS™hadtheGIsuiteinitssightsandneverwavered.TheSEDASYS™team wanted GI physicians not because they were big users of propofol butbecause they used it so very little. Their strategic objective was to make itpossibletodosomethingGIphysicianswantedtobutcouldnotdobecausetheylackedtherequisitecredentials,skills,andmoney.Unfortunately forSEDASYS™,during thisdevelopmentperiod thebenefits

of propofol use in theGI suite had become increasingly clear tomuch of the

relevantmedicalcommunityaswell.Despitetheincreasedcost,by2004severalregions in the United States had experienced exponential growth in propofolsedationbyAPsinGIprocedures.InhighlypopulatedregionsliketheNortheastandSoutheast thepenetrationofAPs, andhencepropofol, in theGI suitewasnear100percent.Asreportedbylargecommercialpayers,theincrementalcostof this trend ran from $400 to $1,400 per procedure. Today, according to GIprofessional societies, billings for gastrointestinal anesthesia (which arepredominantly, but not exclusively, for propofol) have added somewherebetween $5 billion and $6 billion per year in costs to the U.S. healthcaresystem.33Nowwhat?TheSEDASYS™ teamhad committed to developing its device

forGIproceduresandhadpursuedapprovalfromtheFDAonthatbasis.IthadchosenthispathpreciselybecauseGIprocedureswereofrelativelylittleinteresttoAPs.NowthatAPshadcolonizedupwardsof50percentofallGIproceduresbeforeSEDASYS™wasreadytogotomarket,whatwastheSEDASYS™teamtodo?ThereweresomestrongindicatorsthatstayingfocusedontheGIsuitecould

be the right answer. Insurance companies, for instance, were willing to offerreimbursement for SEDASYS™ as an alternative to AP-delivered propofolsedation,assoonasFDAapprovalwassecured.Ifnotafirstforanewmedicaldevice, this kind of enthusiasm is certainly very rare. It suggested that justmaybeSEDASYS™mightbegoodenoughforakeyconstituency—thepayer—thatitcouldwininahead-to-headsustainingbattleagainstAPs.Ontheotherhand,theAPswerenowentrenchedandsuccessfulincumbents

in theGIsuite.TheSEDASYS™customer—theGIphysician—wasno longerassuredtobeanonconsumerofpropofolandAPservices.ForthosewhowereusingAPs,theexplicitperformancetargetwouldnowbe“asgoodastheAPbutmoreaffordable.”Inotherwords,forthosesegmentsSEDASYS™wouldhavetobreaktheprice/performancetradeoffallatonce.ThiswouldbelikeSouthwesttryingtocompetefortheNewYork—LosAngelesroutewitha737-200.In addition to the technical challenges, SEDASYS™would face resistance

fromtheAPs,whonowhadalottolose,namely,theirnew,verylarge,andfast-growingmarket.ThiswasverysimilartothechallengeImageIllusionsatIntelhadfacedwhenbeingevaluatedby thedesignersof thechips itwashoping toreplace.Inshort,shiftsinthemarketplace,ratherthanthestrategicchoicesmadebytheSEDASYS™team,hadtransformedanattempttoestablishaDisruptivefootholdintoasustainingattack.

ENSURESUFFICIENTAUTONOMY

TreatingtheGIsuiteasanundifferentiatedmarketwouldviolatethetenetsofDisruption.Buthowcouldtheteamchangetack?ItwasonethingtoconvinceMD&D management to pursue a large, even if nonconsuming, market withmaterialupside.Whatdidyoudowhenhalfofthatmarketdisappeared?Strategically, the answer was clear: you simply took a much finer-grained

approachtotheGImarket,targetingthosephysicianswhohadnotadoptedAP-administeredpropofol.DisruptionprescribedthatSEDASYS™bedevelopedforand targetedat thoseGIpractices thatwere still notusingpropofoldue to thecost of adding an AP but would like to use propofol due to its superiorefficiency.Andindeed,thisispreciselywhattheSEDASYS™teamhaschosentodo.Itisunderjustsuchcircumstancesthattheneedfororganizationalautonomy

becomes clear. J&J and EES leadership set up the SEDASYS™ group as astand-alone entity inside EES, led by General Manager Mike Gustafson. ForMikeandhis team, thestrategic imperativewas topursueaDisruptivecourse.Andalthoughitmighthaveseemedclearin2001whatthatDisruptivestrategywouldlooklike,by2008conditionshadchangedradically:thefootholdmarketwasnow50percentsmallerthanoriginallyenvisioned.HadSEDASYS™been a subunitwithin one ofEES’smanybusinesses, its

needtoadaptitsstrategytothesenewconditionscouldhavebeendebilitating.Aunitwitha two-or three-year revenueplan that includedmaterial contributionsfromSEDASYS™wouldnowbefacedwithapainfultradeoff:pushaheadwithtargeting the full population of GI physicians and hope that the merits ofSEDASYS™would carry the day, or take on difficult discussions withmoreseniormanagementtochangethoseprojections—whichwould,ofcourse,haverippleeffectsasbudgetswererolledup.The right choice—changing the strategy—only seems easy from our

perspective. Data are never conclusive in these sorts of situations: theenthusiasm of health insurers, for example, and the strong support of thegastroenterologysocietiesfortheSEDASYS™systemwouldhavemadeiteasytoconcludethattakingonAPsintheGIsuiteswasaviablestrategy.Andforallanyoneknows,itmighthaveworked:wecannotgobackintimeandtryagain.However, because of J&J’s bias against decentralization and autonomy, and ageneralawareness in theseniormanagementranksof thepowerofDisruption,KarenLicitrawasablequicklytosecurebroad-basedandhigh-levelsupportforthechangeincourse.SEDASYS™wasrelievedofthetraditionalfinancialandstrategicplanningconstraints,andGustafsonandhisteamwerefreetoadaptin

ways that kept them on a Disruptive path. And so SEDASYS™ is going tomarket with a focus on nonconsumption (GI physicians not using AP-administeredpropofol)asitsfoothold.

DEFINEANEWPRODUCTIVITYFRONTIER

For a solution to have even a hope of a Disruptive future it must have afrontier that reaches a point in price/nonprice space that incumbent solutionscannot.Ifitisstronglydifferentiatedfromothersolutionsbutmerelyoccupiesadifferentpositionon the same frontier,Disruptionwillbe impossible.There isnothing wrong with competitive advantage based on strategic differentiation;that was Holiday Inn’s genius. But the upside of a strategy based ondifferentiationaloneismoreconstrainedthanforDisruptions:itwasitsinabilitytoDisruptthatultimatelylimitedHolidayInn’sgrowthprospects.SEDASYS™, likeessentiallyall formsofautomation,substitutescapital for

labor.Because thedevicesubstitutesfor relativelyexpensive labor—ultimatelyforananesthesiaprofessional—thedevicerathereasilyreachesacostpointthatincumbentsolutionscannot:aslongasahighlytrainedprofessionalneedstobemonitoring the administration of propofol, SEDASYS™ will be the lessexpensivesolution.Yet, for reasons already explored, SEDASYS™ is not as good as an AP

because itwasnotdesigned for the administrationofdeep sedationorgeneralanesthesia.Norshould itbeusedfor theat-riskpatientwhereanAPisclearlyindicated. In other words, SEDASYS™ can reach a lower cost point forrelatively lower levels of performance but cannot today at any cost reachperformancelevelsthatincumbentsolutionscan.Inshort,SEDASYS™definesa new frontier, rather than simply staking out a different point on the samefrontier.

EXPLOITENABLINGTECHNOLOGIES

AlthoughthepresenceofAPsinGIproceduresissignificantandgrowing,theaccepted standard of care for GI sedation is still a drug delivered by a GIphysician and nurse care team without anesthesia support. By targeting GIphysicianswhodo not use propofol, SEDASYS™passes a critical test for aneventualDisruptor:itisfocusedonafootholdmarketwherethecompetitionisstillpredominantlynonconsumption.Fortheseclinicians,SEDASYS™offersaclear step up in performance at typically lower cost. Thanks to the new

productivityfrontiertheSEDASYS™technologyhasdefined,APsareunlikelytobeabletocompeteagainstSEDASYS™inthissegmentoftheGImarket.ThishasgottenustoDollarGeneralterritory:astrategicinnovator.Thisisa

strongpositiontocreate,andtypicallyahighlysuccessfulone.Thekeyquestionnow, however, is whether SEDASYS™ will be a true Disruptor: will itsproductivity frontier expand in a way and at a rate that allows it to replaceanesthesiaprofessionalsinanincreasingnumberofprocedures?Answering thisquestion requires anunderstandingof the four elements that

determine the range of procedures for which SEDASYS™ could one daybecome an effective substitute for APs: (1) the computer that is the system’s“brain”;(2)theaccuracyandcostofmonitoringthepatient’sphysiologywhileunder sedation; (3) thedrugs thatSEDASYS™ is approved todeliver and thesuiteofcomplementarydrugsavailabletobeusedinconcertwithSEDASYS™;(4)long-termtrendsinthedevelopmentoflessinvasivesurgicalproceduresthatare reducing the degree of pain management required for many surgicalinterventions.

CHEAPERELECTRONICS

The core of SEDASYS™ is essentially a personal computer. Programs thatrunonthecomputergatherdatafromthepatient-monitoringunitandadjusttheadministrationofthepropofolviaapumpconnectedtoanintravenous(IV)feed.Notsurprisingly,theSEDASYS™teammadeitapointtodesignitssystemsothat it could use what are essentially off-the-shelf hardware components. Notonlyisthislessexpensivethanacustom-madesystemwouldbe,butitissubjecttodramatic andpredictable cost reductionsover time.Asmicroprocessors andotherkeycomponentsofPCsystemsofagivenperformance levelcontinue tofall in cost and price, and as computer assemblers are able to provide PCs ofsufficient performance levels for SEDASYS™ at ever-falling prices,SEDASYS™will be able to pass those cost reductions along to its customerswithnoimpactonitsprofitability.This decline in cost will serve to make SEDASYS™ more attractive to

smallermedicalpracticesinNorthAmericaandEurope.Suchgrowth,however,is likely to be marginal: independent estimates place the total cost ofSEDASYS™-deliveredpainmanagementat less than60percentof thecostofananesthesiaprofessional.Sincepropofolisconsiderablysuperiortomostotherpain-managementdrugsforGIprocedures,thiscostadvantageoverAPsislikelyenoughtoconvincemostpracticesthatarenotusingpropofolnowtomakethe

switch at existing cost levels. While welcome and valuable, future costreductionsduetocheaperhardwarewillnotlikelybeahugedriverofadoptionindevelopedmarkets.Therealbenefittothissortofcostreductionisinthecreationofentirelynew

consumptioninmarketssuchasChinaandIndia.There, thecompetitionisnotthe AP, and for two reasons. First, labor costs are much lower, so the costadvantageofSEDASYS™overanAP ismuch less, if itexistsatall.Second,the regulatory regimes in other countries often have different labels, so theremay not be the same monitoring requirements for the use of propofol.Consequently,thebindingconstraintontheuseofsedationisnotthecostoftherelevantprofessionalsasmuchas it is their availability,whatever the requisitetraining.Therefore,SEDASYS™neednotbe thatmuchcheaper thanAPs (ortheir equivalent) in these countries.Rather, it needmerely fall enough in costand price that it becomes affordable, since J&J can manufacture moreSEDASYS™machinesfasterthanmedicalschoolscangraduatequalifiedAPs.Seeninthis light, thecostreductionsSEDASYS™islikelytoenjoyarenot

drivers of Disruption as much as they are drivers of growth. Disruption,remember,isdefinedbyimprovementsthatallowaninnovationtocompeteformainstreammarkets inways that incumbents cannot effectively respond to. Inthe case of SEDASYS™, that requires the system to be able tomanage paineffectively enough to cope with a wider range of circumstances—in terms ofbothpatientriskfactorsandtypesofprocedures.Extendingthecostadvantageof theSEDASYS™systemiscertainlyworthwhile,but itdoesnotexpand thesystem’sfrontieringenuinelyDisruptiveways.However,thethreeotherfactorsthatarepushingouttheproductivityfrontierforSEDASYS™areallpushinginpreciselythatdirection.

PATIENTPHYSIOLOGYMONITORING

Theheartofeffectivepersonalizedpainmanagementisadministeringaslittleofthedrugaspossiblewhilestillachievingthedesiredresult.Whenapatientissedated,keepinganaccurateandtimelywatchonkeyphysiologicalindicatorsiscrucial to adjusting the level of drug being administered. Currently, theseindicators are primarily respiratory rate, blood oxygen levels, blood pressure,andheartrate.ThesoftwarealgorithmsthatruntheSEDASYS™systemcontrolsecondbysecondtheamountofpropofolbeingadministeredbasedontheseandotherindicators.Theaccuracyandprecisionpossible from thecurrent setof indicatorsallow

SEDASYS™ to serve as an effective automated pain-management system formany common GI procedures. But improvements are on the horizon. Forexample,itwilllikelysoonbepossibletomeasurepreciselyhowmuchpropofolis being respired by the patient with each exhaled breath, providing a moretimely and accurate indication of the amount of drug beingmetabolized. Thiscould allow the SEDASYS™ system to adjust the amount of propofol beingdelivered through the IV on a continuous basis. This fine-grained and verynearly real-timeadjustment indrugdeliverywouldallowSEDASYS™todealsafely and effectivelywith higher-risk patients, since the chance of deliveringtoomuchofthesedativewouldbereduced.Anothernewmonitoringtechnologyisreal-timeelectroencephalogram(EEG)

wave form (i.e., “brain wave”) analysis, which could enable SEDASYS™ tomonitorandmanageanypatientdiscomfortthatisnotapparentthroughclinicalobservation. In other words, SEDASYS™ would be able to determine if apatientwas inpain,andhence inneedofan increaseddoseofpropofol,whenthepatientwasunabletoindicatepain.Depending on the cost and development timetable of these and other

advances, new patient physiology measurement methods, individually or incombination,promisematerialprogressintherangeofproceduresSEDASYS™will be able to address.SEDASYS™enjoys a40percent cost advantageoverAP-administered propofol as it is,whichmeans that even if, in the short run,these improvements raise thecostofaSEDASYS™systemformorecomplexprocedures, the device will still likely be less expensive than the alternative.However, thesetechnologiesarealmostallelectronicinnature,sotheywillbesubject todramaticandpredictablecostdeclinesover timeaswell,whichwillonlyincreasethelong-runattractivenessofSEDASYS™.

DRUGS

In the simplest terms, there are three types of anesthesia drugs: sedative-hypnotics, analgesics, and amnestics.With sedative-hypnotics you “don’t careand don’t remember” what has been done to you because your state ofconsciousness has been altered. These drugs are typically sufficient formanaging procedures, like GI endoscopy, where there is mild to moderatediscomfort;propofolisasedative-hypnotic;Valiumisanotherfamiliarone.Analgesicdrugsareadministeredso theprocedure“doesn’thurt.”Themore

powerful analgesics, such as morphine, fentanyl, and Demerol, are typicallyadministeredintravenouslyandhaveageneraleffect.Localanesthetics,suchas

novocaineand lidocaine,or regionalanesthetics (e.g., spinalsorepidurals)cangreatlyreducetheneedforgeneralanestheticsandsedative-hypnoticsandallowpatients to remain fullyconsciousandawareofwhat isgoingon,even if they“don’tfeel”it.Relativelyinvasivesurgeries—likecholecystectomy(gallbladderremoval)—

aretypicallydoneusinggeneralanesthesia:patientsareunconscioustothepointthatpartsoftheirautonomicnervoussystemshutdownandtheycannolongerbreatheunassisted.Moresuperficialprocedures—forexample,plasticsurgeriessuch as liposuction, scar revisions, and face lifts—often use a combination oflocal analgesics and sedative-hypnotics: the painmanagement is fairly simple,buttheproceduresaresufficientlybloodythatbeingfullyawareofwhatisgoingonispotentiallytraumatic.Finally,outpatientprocedures,suchasvasectomies,areusuallydoneusinglocalanestheticsonly:patientsneedmerelynotfeelwhatis happening; a cloth drape is typically sufficient to mitigate any emotionaldistress.Today, SEDASYS™ is able to provide low to moderate levels of sedation

usingpropofol,and,forstrategicreasonsdescribedatlengthabove,istargetedattheGI suite.As the systemproves itsworth and reliabilitywith this sedative,thereiseveryreasontobelievethatovertimeitwillbeabletoexpanditslabeltoincludeabroaderrangeofsedative-hypnoticsandanalgesicsthatcouldallowSEDASYS™tomanageabroaderrangeofproceduresinhealthyadultpatients.In addition to expanding its label to other existing drugs, in a sort of positivefeedback loop, asSEDASYS™becomes successful itwill create amarket fornewhypnotics and stimulate the development of newdrugs optimized for usewithSEDASYS™.Inaddition,SEDASYS™createstheoptionforphysicianstocombinethelow

tomoderate levels of sedation it can deliverwith hypnotics and analgesics inways thatallow them toperformabroader rangeof surgical interventions thattodayarecurrentlyperformedonlyunderdeepsedationwithAPsinattendance.This would in fact constitute a “return” to the focus of the painmanagementfield of some years go. Historically the emphasis had been on determiningoptimal combinations of drugs with different effects, and all but the mostinvasive surgerieswereperformed thisway.Sincemostphysicians are trainedand credentialed in the administration of hypnotics and analgesics, theautomation of the delivery of moderate sedation through SEDASYS™ vastlyincreases the range of procedures that could be performed without an APpresent.

NONINVASIVESURGERIES

Finally, a long-term trend in surgery is the development of less invasiveinterventions. Laparoscopic methods are increasingly the standard of care formany procedures. There is growing interest and scientific evidence to suggesttheselessinvasiveapproachesmaynotrequiredeeplevelsofsedationorgeneralanesthesiainordertobeperformedsafelyandeffectively.Moderatesedation,ormoderate sedation in conjunction with the appropriate local or regionalanesthetics,maybe sufficient tomanage a broad rangeof procedures,makingthemviablecandidatesforSEDASYS™.MuchliketheexpansionoftheSEDASYS™frontierthankstoanexpanding

labelandnewdrugs,thisdramaticshiftinaccesstoandaffordabilityofalargenumberof surgical proceduresdependsnoton improvements inSEDASYS™,but on improvements in medical devices and surgical techniques. However,unlike the advances in sedative-hypnotics and analgesics,which are hoped forandlongerterm,thetrendtowardlessinvasivesurgeriesiswellestablishedandasclosetoasurethingasoneislikelytofind.

NOPAIN,LOTSOFGAIN

Putting it all together, the productivity frontier for SEDASYS™ is highlylikelytoexpandalongboththepriceandnonpricedimensionsofperformance.Its price will certainly come down as electronics decrease in price and asSEDASYS™ismanufacturedatscaleforaglobalmarket.Itsabilitytoprovidethelevelofsedationrequiredforanincreasingrangeofprocedureswillallbutcertainly increase thanks to advances in patient-physiology monitoring, anexpandedlabel,thedevelopmentofnewhypnoticsandlocalanalgesics,renewedresearch interest inmultimodal approaches to painmanagement, and the trendtoward less invasive surgeries. Note especially that these trends reinforce oneanother,butnoneisdependentuponanyoftheothers;thatis,iftheyallhappen,theimpactsaresynergistic,butanyoneofthemcanhappenindividuallyandstillmateriallyexpandthefrontieroftheSEDASYS™device.Incontrast, theproductivityfrontierforAPsislikelytoexpandonlyslowly.

ThevariablecostofhavinganAPislikelyalwaystobehigherthanthevariablecostofanelectronicallyautomatedsolution.Andeveniflower-costlaborcoulddothejobfromaremotelocationatcomparablevariablecost(whichishighlydoubtful), the capital costs of that kind of “remote medicine” solution willalmost certainlyoutweigh the capital costsof aSEDASYS™system.And the

abilities of APs will improve only as quickly as the medical profession candefineanddisseminatenewstandardsofclinicalpractice.

FIGURE35:THEEXPANSIONOFTHESEDASYS™PRODUCTIVITYFRONTIER

The expansion of the SEDASYS™productivity frontier along the verticalaxisisafunctionofpricedeclinesinthekeytechnologiesthatmakeupthesystem, particularly the PC but also the electronic components used formonitoring patient physiology. Since the cost position of the APs isexpected to be relatively unchanged—medical school is not getting anycheaper—the relative cost position of SEDASYS™vs. APs improves at anearlyconstantrate,asshownin and .Improvements in patient-monitoring technology, approval for use with

morepowerfulsedatives,useincombinationwithhypnoticsandanalgesics,improvements in these complementary pain management drugs, and the

trendto less invasivesurgeries thatdonotrequiredeepsedationallpushout the SEDASYS™ frontier along the horizontal axis by enabling thesystem to be used for a wider range of surgical interventions. Theseimprovementsmanifestthemselvesunevenly,asshownby and .

Asoflate2010,SEDASYS™hadnotyetsecuredFDAapproval.Butithasbeenapproved foruse inother jurisdictions, for example, theEuropeanUnionand Canada. With these market footholds, and given how well SEDASYS™fares against the tests for Disruption, it would be cowardly not to make anunequivocalpredictionofDisruptivesuccessforthisinnovation.J&Jhasarealopportunitytopursueboththe“market-creating”strategydescribedabove(drivedown cost, increase ease of use, and take the technology to nonconsumingmarkets) and a Disruptive strategy aimed at any medical procedure where,thankstobetterdrugsorlessinvasivemethods,mildtomoderatesedation,aloneor in combination with hypnotics and analgesics, provides sufficient painmanagement.Eachholdstremendouspromise.IexpectthatwithinfiveyearsSEDASYS™willbewellonitswaytoreaping

thebenefitsofcreatingandDisruptingmarketsforsedation.

CHAPTERSEVEN

INVESTINGFORINNOVATION

Noteverysuccessful innovationneedbeDisruptive.Hereisaviewonthemanagerialrequirementsandrisk/returntradeoffsassociatedwithdifferenttypesofinnovation.

Have you ever heard, or said, anything like “Ideas are easy. It’s theimplementation that’s tough”? or “We have lots of great innovations.We justcan’tseemtodoanythingwiththem”?Ifso,youareingoodcompany,foritisacommon sentiment expressed by both practitioners and researchers into thechallengesofinnovatingsuccessfully.The evidence to support this belief seems tome to be somewhat thin. I am

simplynotawareofanystudythatexploresinarigorouswaytherelativeimpactof “great ideas” and “great implementation” on eventual success. TheapplicationofDisruptiontheorytotheIntelportfoliorevealedthatforty-oneoftheforty-eightinitiativesfundedbyNBIwereexpected tofailsimplybyvirtueof the nature of the ideas. Implementation overcame what Disruption wouldcharacterize as bad ideas exactly once out of forty-three attempts, andimplementation failed to deliver when success was expected twice out of sixattempts. Tomy eye, that looks like a relative dearth of good ideas (that gotfunded) and a first-rate implementation capability. In short—and no surprisehere—it appears that innovating successfully requiresgreat ideas and the rightapproachtoimplementation.Thebadnewsisthatadviceonpreciselyhowtoimplementideasofdifferent

types is based exclusively on explanatory frameworks: there has not been asystematic test of the predictive power any alleged strategy-structurecontingencies.However,theabsenceofevidencedoesnotremovetheobligationtomakeadecisionintherealworld.IfyouwanttoapplyDisruption,youneedtomakechoicesabouthowtoorganizeandactontheprescriptionsofthetheory.In what follows, therefore, I will offer my observations on what I have seenworkandwhatIbelievetheimplicationsforactionofDisruptionare.

THELIMITSOFDISRUPTION

TheexperimentaldesignusedtotestthepredictivepowerofDisruptionrequiredstatingtheimplicationsofthetheoryinhigh-contrastterms:Incumbentssucceedwhen they launch sustaining innovations; entrants succeed with Disruptions.Failure awaits incumbents who attempt to Disrupt or entrants with sustaininginnovations. These categorical statements are obviously not universally true:even within the Intel portfolio there are counterexamples. There is onesustaining/entrant success and two sustaining/incumbent failures; and theCeleronprocessor(anIntel innovation,butnotpartoftheportfoliousedintheexperiments)wasaDisruptive/incumbentsuccess.Despite thesecounterexamples,however, thepredictive successof this stark

representation of Disruption is what led to the reformulation of the theory interms of “productivity frontiers” in chapter 3. This restatement of what aDisruptionis—anewfrontierthatappealstodifferentcustomersorcontextsofuse that then expands thanks to advancements in enabling technologies—formalizes Thurston’s experimental definition and explains how Disruptionreallyworks.Butitcomesatacost.BygivingDisruptionthekindofprecisionrequiredfor

demonstrable improvements in predictive accuracy, certain competitive battlesthatonemightliketoclaimasDisruptionsarenowbetterexplainedusingothertheories.For example, no analysis of innovation would be complete without some

discussionofApple.Thecompanyhasenjoyedaremarkabledecade-longrunofsuccess,drivenfirstbytheiPod,thenbytheiPhone,andnowbytheiPad.AreanyoftheseDisruptions?ThefirstiPod,launchedinOctober2001,enteredwhatwasinfactanalready

crowded and hotly contestedmarket. The portable-music space had long beendominated by Sony, which had been successful with the cassette-basedWalkmanandtheCD-basedDiscmanandthecompanywasapioneerinportabledigital music with its ill-fated MiniDisc. There were large entrants, such asSamsung with its Yepp MP3 player, and small entrants, such as DiamondMultimediawith its RioMP3 player. For the iPod to beDisruptive, it wouldhave todrawanew frontier that incumbentswouldbeunable tocopywithoutcompromising their strategies, and it would have to appeal to customers thatincumbentshadlittleinterestincourting.

Noneofthisapplies.Applereliedondisk-drivetechnologytoprovidemuchhigher storage capabilities than the flashmemory employedbyothers, but theiPodwasmuchmoreexpensiveandmuchlarger,andsoinsomewaysmuchlessconvenient. There was nothing about the other companies’ strategies thatprecluded them from adopting this approach, and so the iPod was not a newfrontier but insteadmerely a very different spot on the same frontier. (UnlikeSouthwest, which by relying exclusively on the 737was doing something nootherairlinecouldcopywithoutcorruptingitsexistingstrategy.)Consequently, Disruption would have predicted that Apple was unlikely to

succeed.Butitsucceeded:theiPodsoldwellandestablishedenoughofaleadinthe marketplace for digital music players that Apple was able to launch theiTunesstorein2004.Thecombinationofapopulardeviceandahugeselectionof legal digitalmusic created a juggernaut that dominates the industry to thisday.Sowas the iTunes store aDisruption of conventional record retailing?Not

really. It was a revolutionary innovation—a giant leap into the “white space”beyond the established price/nonprice curve for music retailing, and one thatincumbentmusicretailersfoundverydifficulttocopyeffectively.ButitwasnotaDisruption:itappealedtothesamecustomersandcontextsofusethatrecordstoreswerecontesting,andtheiTunesstorediditwithbetterselection(200,000itemsondayone!),moreconvenience(online),equalsoundquality,andamoreappealing price, thanks to unbundling the album and using a uniform pricingstructure(whichhasbeenmodifiedonlyslightlysince).Thiswas“moreforless”intheeyesofcustomersandhasnoneofthehallmarksofeitherthelow-endornew-marketfootholdsthatincumbentshaverelativelylittlemotivationtodefend.Tobe sure, the iPod/iTunes combinationwas a verydifferent businessmodel,but it delivered better performance targeted at exactly the same markets thatincumbentscoveted.ThatisnotDisruption;thatiswalkingintoacrowdedbar,picking a fight with the biggest guy there, and knocking him over with onepunch.Neither was the eye-popping success of the iPhone driven by Disruption.

Apple was once again an entrant, this time attacking successful smartphoneplayers such as RIM and Nokia. Apple’s proposition was that it was goodenoughalongvariousdimensionsofperformance,suchase-mailandsecurity,toappeal to mainstream customers and vastly superior in ways that othersmartphoneplayershadsimilarlytargetedasimportant:amusicplayer,Internetbrowsing,mapsanddirections,photosandvideos,andsoon.ReleasedinJune2007,thedevicewasaninstanthit—andahitwithallthesamecustomersthatincumbentswerestronglymotivatedtodefendorpursue.Synchronizingthrough

iTunes software, just like the iPod, and replicating the success of the iTunesstore with the Apple App Store, the iPhone made a bold and enormouslysuccessful leapbeyond theexisting frontier intoa space thatother incumbentshavesincebeentryingtoemulate.It is certainly the case that the increasing functionality of smartphones is a

possibleDisruptiontopersonalcomputers,butthatisnotentirelyrelevanthere.The smartphone market itself was well established, and the key performancedimensions thatwoulddrivegrowthwerewellunderstoodbybothparticipantsandobservers.Apple’s entrywasnotbuiltonan insight that incumbentswerestrategically precluded from adopting or a focus on customers or uses thatincumbentshadnointerestindefending.Instead,Applehadonceagainmarchedintoabar,pickedafightwithanothersetoftowntoughguys(NokiaandRIM),andhas,intheeyesofmany,becometheonetobeat.(Disruption is not entirely absent as a potential driver of some of Apple’s

successinthesmartphonespace,however.TheAppStoreisapossibleDisruptorofthesoftwareindustry.Ithasbecameaplatformforsmallapplicationspricedat rarelymore than a few dollars, rather than the shrink-wrapped packages ofmoregeneral-purposesoftwarecosting$20,$30,orsubstantiallymore.Shouldthese apps find a way to expand their functionality without concomitantincreasesinprice,thiswillhavebeenatrueDisruption.)OnlyApple’siPadseemstobeatrueDisruptor.The“tabletPC”markethas

beenaroundforawhile,with tentativeexploratorystepsbyavarietyofmajorPC manufacturers. But unlike the portable-music business, there has been noclear incumbent, noobvious functionality that customersvalue, andnoclearlydominantbusinessmodel.Therehasbeenno“bar”forAppletomoseyoninto,notownstrongmantopickafightwith.Inthisnewmarket,Apple’ssuccesshasbeenbuiltonitsabilitytoconnectwithcustomersinwaysmanycompaniescanonly envy. Apple has therefore created a sizable and profitable niche for theiPad.Since thisproduct’scapabilitiescanbeexpected to improvedramaticallyandrapidlythankstotheelectronictechnologiesthatpowerit,theiPad’sfrontierwill almost certainly expand inways that will likelymarginalize and perhapseventuallyreplacethepersonalportablecomputer.ThatistextbookDisruption.Microsoft has had few out-and-out “failures” as defined above—that is,

venturesremovedfromthemarket—butsomeofthecompany’smajorinitiativeshave struggled mightily. In the mobile phone business, theWindows Mobilesmartphone operating system, launched in 2003, was touted as a crediblecompetitor to theNokiaSymbianandRIMBlackBerryOSsoftwareplatforms.For an entrant with a sustaining strategy, a retrospective application ofDisruptionpredictsfailure,evenforacompanyasformidableasMicrosoft,just

as itwouldhaveforApple.Microsofthasnotfailed in thisspace, insofaras ithasnotabandonedthefield,butitsperformancehassurelybeendisappointingtomany:WindowsMobile has never enjoyed a market-leading position, and itssharehasbeenerodingsteadilyforthelastseveralyears.Despitethesesetbacks,Microsoft, seeming to recognize the importance of the smartphone as acomputingplatform,hasperseveredandlaunchedtheWindowsPhone7in2010to generally favorable reviews. Nevertheless, sustaining innovations are onlypredicted to succeedwhen launchedbysuccessful incumbents.And that statushaseludedMicrosoftinthisspace.Ofcourse,Microsoftmayyetdefy theodds in thesmartphoneOSbusiness,

justasApplehasdone,andinfactMicrosofthasbeensuccessfulasanentrantwith sustaining innovations in other markets, perhaps most notably the gameconsolebusiness.Launchedin2001,Microsoft’sXboxcompetedhead-onwithdominant products fromSony,Sega, andNintendo.Subsequent generations ofthe device, especially theXbox 360 and various other sustaining innovations,such as Xbox Live and Xbox 360 Live, have established the console as asignificantsuccess—flyinginthefaceofDisruption’spredictions.More fascinating still, Microsoft’s Kinect motion detection module for the

Xbox,introducedinlate2010,couldwellprovetobeaDisruptiveinnovationinthegameconsolebusiness.Thisdeviceallowsgamerstodispensewithagamecontrolleraltogether,usingonlytheirownbodymotionstooperatetheon-screenaction.CompetitorshavebeendismissiveofthequalityoftheKinect’sgamingexperience, but primarily because they are evaluating its precision andresponsiveness rather than the sense of total immersion a “device-free”interactioncanoffer. Inaddition,akeymarketsegmentforKinect,at least fornow,seemstobeyoungerchildren,typicallyalessattractivesegmentintheeyesofotherdevicemakers.ShouldKinectimproveinwaysthatmaintainthissenseofimmersionevenastheinterfacebecomesmoresophisticated—somethingthetrajectoryofhardwareandsoftwareimprovementswouldappeartomakeallbutcertain—Microsoftwill haveDisrupted amarket inwhich it is an incumbent,againinamannerinconsistentwiththepredictionsofDisruption.In each of these counterexamples (Apple and the iPod, iTunes, and iPhone;

MicrosoftandtheXbox)Disruptionfailsnotonlyasapredictivetoolbutalsoasa compelling explanation for the outcomes. Other theories are more suitable.Intriguingly, thenotionof “platformmediatednetworks” (PMNs)has a strongclaimon laying bare the underlying causalmechanisms.34 In each case, a keydriverofsuccessseemstohavebeenasuperiorabilitytoconnecttwosidestoatransaction using a superior platform: the iPod/iTunes combination connectedrecording artists with their fans; the iPhone connected application developers

with users; and Xbox connected game developers with gamers. Christensendiscovered Disruption by studying how entrant companies toppled powerfulincumbents; it could be that PMN is a second way for this to happen. Asresearchers in this space explore the explanatory power of this framework,definethecircumstancesofitsapplication,anddemonstrateitspredictivepowerinthosecircumstances,thearsenalofempiricallyvalidatedtheorieswillexpandaccordingly.

THEINNOVATIONPORTFOLIOMATRIX

The existence of these and other counterexamples does not undermine theconclusionsdrawninearlierchapters:Disruptionmateriallyimprovespredictivepower and has compelling explanatory power. But it is not a theory ofeverything, as these cases illustrate. Other types of innovation can succeedbrilliantly, and there is no justification for being a “Disruption bigot.”Rather,whatisrequiredisamorecompleteunderstandingofhowtomanageeffectivelyalldifferenttypesofinnovation.Thisleadstothenotionofaninnovationportfolio.Disruptioncanbeusednot

merelytopredicttherelativechancesofsuccessofdifferentinnovationsbutalsoto determine the nature of the diversity of an organization’s innovationinitiatives.This allows a company to calibrate themagnitude of investment ineach type of innovation according to an organization’s growth objectives,strengths,weaknesses,opportunities,andthreats(goodold-fashionedSWOT).Anymeaningfulportfolioanalysismustcapturedimensionsofdifferencethat

matter.When constructing a portfolio of investments, for example, fewwouldsuggestpursuinganalphabeticaldistribution,withonetwenty-sixthoftheassetsinvested in securities beginning with each letter of the alphabet. A moregenerallyacceptedapproachallocatesassetsbasedon,amongother things, theinvestor’s time horizon, tolerance for volatility and risk, and required rate ofreturn.Sincethepursuitofinnovationisatleastinpartafinancialinvestment,these

dimensions are relevant here as well. However, this means that accuratelyassessingthetimehorizon,risk,andpotentialreturnofindividualinvestmentsisreallywhat thegame isallabout,and it is inprovidingadditional insightherethatDisruptioncanbeespeciallyhelpful.BycategorizinginnovationinitiativesaccordingtowheretheyfallonwhatIwillcalltheInnovationPortfolioMatrix(IPM),acompanycannotonlyassess its levelof innovationdiversificationbutalsothinkmuchmorecarefullyabouthowtoincreaseitschancesofsuccessineachquadrantbymanagingeach typeof innovationappropriately. In thecharton this page, each type of innovation is characterized by awell-known gamewithrelevantsalientfeatures.

CHESS

Sustaining/incumbent innovation is typically a battle amongwell-resourced,andveryoftennearlyequallyresourced,combatants.Justasinchess,therulesofcompetitionarewellspecified,andcompetitorshaveasharedunderstandingofthose rules. Victory is essentially a computational problem, a function ofunderstandingyourcompetitors’capabilitiesandhowtheystackupagainstyourown.Whenyoucomeoutaheadinthatassessment,yourprobabilityofsuccessis relatively good.Andwhen you do not, youwill probably be able to tell inadvanceandsoatleastavoidbattlesyouarenotlikelytowin.Going beyond the metaphor, the investment required for a successful

sustaininginnovationasanincumbentwilldependlargelyonthemagnitudeofthe breakthrough you seek and how quickly you hope to realize its benefits.Innovationisdefinedasachangethatbreakstradeoffs.Breakingthembyalittlebit—an incremental innovation—might require relatively little investment, andthe payoffs will most likely be commensurate. Breaking them by a lot—abreakthroughinnovation—likelydemandssignificantinvestmentinordertoreapbigrewards.Typically,thetimehorizonforapayoffisrelativelyshort.Intheinnovationspace,itisacceptedpracticetoestablishateamdedicatedto

the development and commercialization of an innovation. This serves tofacilitate the necessary break with established practices in the ongoingoperations,which, in the interestof efficiency,necessarily resist change. (Thisresistancetochangeisnotapathologybutanessentialvirtue.)Thekeyvariableis the degree of functional completeness and autonomy enjoyed by theinnovation team. Does the new team have its own marketing unit? Its ownR&D? Its own production facilities? Its own distribution channels? It issufficiently difficult to make categorical statements that we are left with theratherunsatisfyingobservationthattheunitshould“havewhatitneeds.”Butasageneralrule,teamsplayingchess(incumbentspursuingsustaininginnovations)willneedrelativelylittlefunctionalcompleteness,typicallyfocusingononlythatdimensionofperformancetheyareseekingtoimprove.Since sustaining/incumbent efforts are, by definition, very close to the

interests of the existingmainstream organization, relatively little autonomy isrequired.Typically,someformofseparatemeasurement isappropriate—say,aprofit-and-loss statement that captures the costs and benefits generated by theinnovation.Butfull-scaleseparation—ofthekindrequiredbyDisruptiveefforts,for example—is rarely called for. By extension, this means that the lines ofcommunication between the team leading the innovation effort and themainstreambusinessneedtobewellmaintainedandlikelycrossseveral levelsof the hierarchy: the innovation team and themainstream businesswill likely

havealottolearnfromeachother,andfrequent,substantivecontactwillalmostcertainlyservetheinterestsofboth.

CHECKERS

Disruptive/entrantinnovationhasaverydifferentprofile.Akeyfeatureofthegame of checkers is that players are required tomake any jumps available tothem.Consequently,key tovictory is forcingopponents to takeyourpieces—somethingthatintheshortrunappearsintheirinterest—inawaythatultimatelyundermines their strategic position. Similarly with Disruptions: what allowsentrants tomarchupmarket is the relative inabilityof incumbents tomountaneffectivecounterattackintheearlystagesofaDisruptiveincursion,thankstothesuperior economic attractiveness of continuing to servemainstream customerswithsustaininginnovations.ItisbytargetingunattractivemarketsegmentsthatDisruptors are able to exploit this difference in economic attractiveness,essentially compelling incumbents to act in ways that are contrary to theirlonger-run interests. A clear understanding of their business model and theimprovement trajectory of the technology that drives outward the frontierdefinedbythatbusinessmodelallowsentrantspursuingaDisruptivestrategytoassesswhattheirchancesforsurvivalarewithbetteraccuracythanentrantswithasustainingapproach.CentraltosuccessfulDisruptioniscreatingafundamentallynewproductivity

frontier, one that is profitable serving segments that are unattractive toincumbents. An empirical fact (but not a theoretical necessity) is that thebusinessmodelsrequiredtoservethesesegmentscantypicallybedevelopedandlaunchedwithrelativelylittleinvestment,atleastwhencomparedtothecostofsustaininginnovationsingeneral,andcertainlywhencomparedtobreakthroughsustaining innovations. The catalogue of successful Disruptors is populatedalmostexclusivelybysmall,near-start-upcompaniesthathadlimitedinvestmentcapital and had to turn a profit quickly in order to stay afloat. Consequently,Disruptive innovations can often be launched with relatively little up-frontinvestment, even though the long-run growth and value creation is oftenenormous. Although the time it takes to realize these rewards varies, at leastDisruptions, properly understood, have a relatively predictable time horizon,sinceitispredicatedontheevolutionofkeytechnologies,whichveryoftencanbeestimatedwithreasonableaccuracy.An innovation team pursuing Disruption requires a much higher degree of

functional completeness than a sustaining/incumbent team. Since Disruption

requiresacompletelynewbusinessmodel—inordertocreateanewfrontier—enabled, typically, by entirely different technologies appealing to completelynew customer segments, it is not uncommon for a Disruptive entrant to be astand-alone business (even if owned by a larger corporation). Would-beDisruptorsoftenhavetheirownproductdesign,distribution,andmarketingandcanevenrequiretheirownwell-developedhuman-resourcepolicies,informationtechnology infrastructure, and separate physical locations. And since theDisruptionistargetingamarketthecorporateparentdoesnotcurrentlycompetein,theneedforcommunicationbetweentheDisruptiveteamandthemainstreamoperationsisrelativelylow.Certainly one can imagine circumstances where there are key resources

resident in the larger corporate parent—human, technological, or otherwise—that could prove valuable to the innovation team. In fact, when identifyingtargetstoDisrupt,companiesshouldlooktoexploitestablishedstrengths(moreon this below). But in such cases the key is to “borrow” those resources, not“share” them: without clear separation from the mainstream business, thegravitationalpulloflargerand,intheshortrun,morelucrativeopportunitiesissimply too strong. Resources critical to the success of the new venture seeminvariably to be allocated to the needs of the mainstream business. One canconceive of having the willpower to resist, but in practice it is simply toodifficult.Oneimplicationofthisfunctionalcompletenessandorganizationalautonomy

is that communication between the innovation team and the mainstreamorganizationisnotapriority.LaunchingaDisruptiveattackonanewmarketisfundamentally a portfolio play, a way to create new growth opportunitiesthroughintelligentdiversification.

BINGO

Goodbingoplayersareabletoincreasetheiroddsbyplayingalargenumberofcardssimultaneously.Theyscaneverycardquicklyandaccuratelyandnevermissanopportunitytoplaceamarkerwhentheirnumbercomesup.Butluckisthedefiningfeatureofawinner—anovicewithonecardcanstillwin,ifitistherightcard.Thebestyoucandoisbewillingtoplaceenoughbetsandbeabletomonitor them closely enough to improve your chances ofwinning. Sustainingattacksonnewmarketsaremuchthesame:aninnovatorwithmoreresourcesismorelikelytosucceed,andonethatcanmonitorcloselyandaccuratelyshiftsincustomerpreferences(theanalogtoone’snumbercomingupinbingo)willbe

abletodeploythoseresourcessuccessfully.

FIGURE36:THEINNOVATIONPORTFOLIOMATRIX

Evenso,theoddsofgettinga“bingo”aredistressinglylow,whichisallthemoreconcerningsincesuccesstypicallydemandsmaterialup-frontinvestment:it is often difficult to displace successful incumbents with only incrementaladvancesover existing solutions.This lowprobabilityof success and thehighinvestment requirements are complementedby ahighlyunpredictable time lagbeforeanyeventualpayoff.Asustainingentrant,afterall,isattemptingtobreaktradeoffsthatdefinecompetitionforsuccessfulincumbents,somethingthey,too,are working toward.Will consumers like your “new and improved” version?Howlongwillittaketoconvincethemtomaketheswitch?Howvigorouslyandsuccessfullywillyour competition respond?Thesecriticalquestionscannotbeansweredwithanythinglikethelevelofrigorandprecisionpossibleinagameofchessorcheckers.It isdifficult tocharacterizewithanyconfidencetherequisiteautonomyand

degree of communication between an innovation team within a companypursuingasustaining/entrantstrategyandthemainstreamoperationsfocusedon

the markets the company currently serves. Apple’s iPhone division seems tohavehadfairlycloseconnectionstotherestofwhatwasgoingonatApple:thedeviceleveragediTunes(whichhadbeendevelopedfortheiPod),incorporatesmuchoftheiPod’sfunctionality,andistightlyintegratedintotheMacpersonalcomputerecosystem.Thissuggestsrelativelylowfunctionalcompletenessandahigh degree of communication between the units developing the iPhone andApple’s established businesses. In contrast, it would appear that the XboxdivisionatMicrosoft—alsoasuccessfulentrantwithasustainingstrategy—wasvery nearly functionally complete and was largely separate from the groupresponsible for developing the next generation of the Windows operatingsystem.Thismadegoodsense,giventhelargelyseparatetechnologicalhurdleseachwas seeking to overcome and the relatively low level of synergy, in theshortrunatleast,betweenthetwodivisions.Theabsenceofaclearpatterninhowbesttomanagesustainingattacksbyan

entrantishardlysurprising,giventhelowprobabilityofsuccessinthisquadrant.Itisessentiallyunchartedterritory:successisunlikely,investmentdemandsarehigh, the time horizon is uncertain, and there are few proven rules governinghowtomanageit.Otherthanthat,Mrs.Lincoln,howwastheplay?

BACKGAMMON

For all the challenges and difficulties associated with the other threequadrants, they are at least conceptually straightforward with respect to theirobjectives: sustaining innovation in your core markets means improving theperformance of your existing business; entering new markets, whether viasustaining or Disruptive strategies, is about creating new growth.What couldpossibly be the argument in favor of launching innovations that fall into thisfourthquadrant—businesses intended toDisruptyour existingoperations?Therhetorically stirring exhortation that one should destroy one’s own businessbeforesomeoneelsedoesseemstometoignorethebenefitsthatmightaccruetosomeonewhowas able to determine amore nearly optimalway to time suchseeminglyself-destructivebehavior.For instance, Intel’s launch of the Celeron processor would not havemade

sensewhilethecompanywasstillreapingthebenefitsofagrowingbusinessandits would-be competitors had shown no ability or desire to launch their ownDisruptiveattacks.ItwasonlyafterAMDandothershadfoundtheirfoothold—but before they had begun their upmarket march—that Intel’s self-Disruptionbecameastrategicallysoundpath.Inotherwords,onehardlywantstoemploya

scorched-earthpolicywhentheenemyhasyettomusteritstroops.Self-Disruption is perhaps best seen as a defensive measure, launched at a

timewhenDisruptionbycompetitorsseemsmerelypossible,sincebythetimeitis clear that a Disruptive threat looms it is too late to respond effectively.Consequently, businesses with a mandate to Disrupt their core businesses arenecessarily subject to a higher degree of uncertainty than Disruptive attackstargeted at newmarkets: there, one can bet on the inability of incumbents torespondeffectively,andsopursueDisruptiveattacksbasedonaclearsenseofwhenandhowDisruptionwillbepossible.(There is the problem ofwhat to do if incumbents inmarkets you hope to

enterwithDisruptiveattacksadopttheadviceofferedhereandsetuptheirownself-Disruptiveunits,therebycuttingoffyourDisruptiveopportunities.Althoughfewthingswouldmakemehappierthanfortheseideastoprovesopopularandcompelling thatyourcompanyfaces thisproblemin thenearfuture, I fear thisobjectionwillremainatbestaslimpossibilityforsometimetocome.)Inshort,yourwillingnesstolaunchinnovationsthatarepotentiallyDisruptive

toyourcoreoperationsisafunctionoftheprobabilityofotherslaunchingsuchattacksandyourownorganization’stoleranceforstrategicrisk.Onecan,thankstoDisruption theory, assess these riskswith a higher degree of accuracy thanwith,say,sustaining/entrantstrategies(characterizedaboveasagameofbingo)andplaythegameinawaythatachievesthedesiredlevelofexposure.Successis therefore a function of understanding the range of possible outcomes,assessing the likelihood of each, and having a high degree of self-awarenessregardinghowmuchriskyouarewillingtotolerateandhowmuchyouareabletoinvesttoreducethatrisk.Itis,inotherwords,agameofbackgammon.Theprobabilityofsuccessofaself-Disruptive innovationcanbe lower than

thatof aDisruptive entry into anewmarket, thanks to the earlier stage in therelationshipbetweenapotentiallyDisruptivebusinessmodeland theevolutionof the underlying technologies that could fuel that Disruption. For example,Intel’srear-guardself-DisruptionwiththeCeleronwasaneffectiveresponsetoanexisting,ifearly-stage,competitor.Althoughtherewereimportantaspectsofthe technology to work though, the company understood the broad outline ofwhatitwasdealingwith.Other potential self-Disruptions are much longer term. For example,

Disruptions in medical diagnostics based on personalized medicine throughindividualgenotypingremainhighlyspeculative:thepaceandnatureoftherightkinds of improvements in genetic mapping and our understanding of therelationshipsbetweengenotypeandspecificdiseasesisfarlesswellunderstoodthan the pace and nature of change in microprocessors, software, or wireless

communications. Some pharmaceutical or medical devices firms are keepingvery close tabs on personalized medicine, however, for it could well be aseedbedforthesortsofinnovationsthatmightDisrupttheirexistingoperations.The language of options-based investing is especially relevant here.At one

level,anyinvestmentthathasstagestoitcanbeinterpretedthroughanoptionslens: an initial investment creates the possibility of making subsequentinvestments should the initial ante indicate that theventurehaspromise, or ofwithholdingadditionalfundsifitdoesnot.There are, however, at least two types of optionlike investments: growth

options and strategic options.35 Growth options create the possibility of new,profitablerevenues.Strategicoptionscreatethepossibilityofcompetinginyourcoremarketsinafundamentallydifferentway.Forinstance,Xboxcanbeseenas having both growth option and strategic option value forMicrosoft. It is anewbusiness that is successful in itsown right.And it creates a foothold inaspace that is potentially a new platform for software and personal computing,which is one way of characterizing Microsoft’s core business. Indeed, thecompany’spersistenceinfightingforameaningfulpieceofthesmartphoneOSmarket can also be seen as a testament to the importance of a viable strategicoptioninthisspace.Applecanalsobeseentobelaunchingnewbusinessesorpotentiallymaking

acquisitionswithaneyetogeneratingaportfolioofstrategicoptions.Asofthefall of 2010, rumors were beginning to make the rounds that Apple’s $50+billionincashwouldbedeployedtoacquireDisneyorSonyoranynumberofother companies in order to give it a substantial position in the entertainmentindustry, something some observers see as critical to expanding Apple’sdominanceinconsumerelectronics.Entering new markets, whether via sustaining or Disruptive strategies,

typicallyhassignificantgrowthoptionvaluebutcanhavestrategicoptionvalueaswell. Incontrast,becauseself-Disruptionscanultimatelycannibalizeagoodportion of the mainstream business, they are predominantly strategic optionswith perhaps some growth option value should it turn out to be able to takemarketshareorgrowthemarket.Thismeansthatself-Disruptionshavemanyofthe characteristics of Disruptive/entrants but should be managed quitedifferently.In practical terms, the up-front investment when playing backgammon is

relativelysmall,as it isforgrowthoptionDisruptions in theDisruptive/entrantquadrant. Also, functional completeness is a requirement so that the initiativecan stand alone, thereby permitting the mainstream organization to competeaccording to established rules of its game (the “chess” quadrant). The time

horizon will be somewhat less predictable than for Disruptive/entrants andtypicallylongerthanforsustaining/incumbentinnovations.The autonomy such initiatives require is only moderate, for if a strategic

optionpursueswhateverinitiativesleadtobusinesssuccess,itispossiblethatitcouldevolve inways thatcompromise itsability tofendoffDisruptiveattackson the core operations. For instance, if Microsoft’s Phone 7 smartphoneoperating system were to target exclusively a low-end niche, as did DollarGeneral in the discount retail space,WindowsPhone 7would no longer havestrategic-option value for Microsoft’s core computing OS business, since itwould no longer be exploring the capabilities and functionality thatmake thephone a powerful complement to, and perhaps Disruptor of, the personalcomputer. For similar reasons, the degree and nature of the communicationbetweenthepotentialself-Disruptionandthecorebusinessneedtobesignificantandcarefullymanaged.Askeyenvironmentaluncertaintiesareresolved,andtheneed to exercise or abandon the strategic option of self-Disruption becomesclear, a coordinated response is critical to calibrating the investment in eachbusiness.

FIGURE37:MANAGERIALCONTINGENCIESFORTHEQUADRANTSOFTHEINNOVATIONPORTFOLIOMATRIX

BALANCINGTHEPORTFOLIO

I cannot offer a categorical prescription on the right absolute or relativeinvestmentineachtypeofinnovation—toeachofthefour“games”intheIPM.Theresearchnecessarytouncovertheconsiderationsthatwoulddeterminewhatthe allocation should be or to explore the consequences of different levels ofinvestmentindifferenttypesofinnovationovertimehassimplynotbeendone.However,itseemsclearthatformostorganizationsthedangerisspendingtoo

much time and effort on sustaining innovation in their home markets. Thepressurestokeepupwiththerelentlessdemandsofimportantcustomersandtorespondtotheunceasingattacksofmotivatedcompetitorscaneasilyswallowupeveryminute and every dime. The challenge is not how to do enough in thisspacebuthowtoavoidlettinginitiativesherecrowdouttheabilitytoinvestinother“games”withdifferentpayoffstructures.Perhaps key to maintaining perspective on the relative importance of

sustaininginnovation toan incumbent is recognizing its limitedupsideandthenatureofthegrowththatpursingDisruptivepathsoffers.Inotherwords,dialingback sustaining initiatives is not about sacrificing thepresent to the futurebutratheramoreaccurateassessmentofthepayoffsandrisksassociatedwitheachtype of innovation. Diminishing returns set in far sooner than most realize.Disruptivegrowthcanhappenwithfargreatercertaintythatmosttendtothink.And Disruptive threats can materialize far more quickly than most want tobelieve.Once there is an acceptance of a need to put at least some eggs in other

baskets, hard data to support specific levels of investment in other types ofinnovationislacking.However,ithardlyseemsastretchtopositthathavingnoinvestments inDisruptive innovations is thewronganswer. Ihaveyet toworkwith an organization that did not have within its existing complement ofinitiativessomethingthatheldwithinittheseedsofaDisruptiveentryintonewmarkets.Similarly,rareindeedistheincumbentorganizationthatneednottakeseriously one or more potentially Disruptive threats. Consequently, everycompany is almost certainlywell served by placing a consideredwager on atleast one or two games of checkers (Disruptive/entrant) and backgammon(Disruptive/incumbent).It is certainly a defensible position that playing a little bingo

(sustaining/entrant) is not an entirelybad idea, either.But the structureof this

gamesuggeststhatwhenenteringnewmarketswithsustaininginnovationsonemust clear a particularly high bar. The need to avoid the well-documenteddecision-makingbiasesofoverconfidenceandoptimismisespeciallyacute:itissimply too easy to believe that you have a better mousetrap. Intel certainlythought so with every one of the thirty-eight businesses it launched withsustaining innovations into new industries… and it failed thirty-seven times.Thatdoesnotmeanyoucannotdobetter,butbeforeyougodownthisroad,myrecommendation is to ask yourself if you are really sure you have got awinner…andaskyourselfthatquestionatleastthirty-eighttimes.Asmoreislearnedabouthowtobesuccessfulwitheachtypeofinnovation,

therisk/returnprofileofeachwillchange,and therelativeemphasisplacedoneachtypeofinnovationwillchangewiththatnewknowledge.Butfornow,themostpromisingpathsappeartoliewith,inorder,sustaininginnovationsinyourcurrent markets, Disruptive innovations in new markets or as defensivemeasures, and sustaining innovations in new markets. There is no theoreticalnecessityforthistoremainforeverthecase,butitiswherethedatapointusfornow.

CHAPTEREIGHT

ORGANIZINGFORINNOVATION

Whetherusedtoshapeaspecificprojectoranentireportfolio,Disruptionrequires a break with some very popular approaches to innovation.Perhapsmostimportant,usingDisruptionallowsinnovatorstomoveawayfromtheprofligacyof“fail fast tolearn”modelsandinsteaddeliberatelyshapeinnovationsinwaysthataremorelikelytobesuccessful.

With a rough idea of how intensely to focus resources—both financial andhuman—ondifferenttypesofinnovation,thenextchallengeistodeterminehowbest to develop and pursue initiatives in each sector. When it comes toinnovation inestablishedcompanies, theragehas longbeen to tryandachievethis by emulating the behaviors of entrepreneurs and the swashbucklinginvestorswhobackthem.Afterall,itwasnotDEC’sinternallydevelopedAlta-Vista that dominated the search engine space but the venture-backed Google,andthisisbutoneofhundredsofsalientexamples.Andsotheexplicitbeliefformanyis that theecosystemthatflourishedalongMassachusetts’sRoute128inthe 1970s and 1980s and in SiliconValley formuch of the last four decadesshouldandcanbereplicatedinsidecompanieswithsimilarefflorescence.Theresulthasbeen,inmanycompanies,aseriesof“ideahunts”orthelaunch

of“innovation tournaments”designed toelicitor find ideas residentwithin,orforthatmatterbeyond,theboundariesofthecompany.Theseideasareevaluatedandthebestonesfunded,withthenewventurestypicallygivenhighdegreesofautonomytopursuetheirstrategies.Whydosomanycompanieshavesuchadifficulttimeadaptingthebehaviors

thatworksowellintheventurecapitalecosystemtothecorporateenvironment?I believe that the root cause lies in some very real differences betweenestablishedcompaniesandstart-upsthatnoamountoforganizationalredesignorprocesscreativitycanovercome.Bywayofexample, consider thatwhen fishabandoned the sea for the land

and evolved into amphibians, reptiles, and eventuallymammals, they gave upgillsfor lungs.Someland-dwellingmammalssubsequentlyreturnedtothesea,and some of those evolved into whales and dolphins. Whatever the provenbenefitsofgillsinextractingoxygenfromwater,theseaquaticmammalscouldnotgobacktothatsolution:toomanyotherhard-to-changecharacteristicsweredependent on lung-based breathing, and it proved easier for other elements of

their physiology to adapt—e.g., body mass and myoglobin concentrations intheirblood—thanforthemtobacktrackacrossmillionsofyearsofevolutionandrecovertheirgills.Thatiswhycetaceansholdtheirbreath.When we argue that corporations need to act more like VCs, we are, in a

manner of speaking, arguing that in order to return to “the sea” (innovatesuccessfully) corporations should “recover their gills”—replicate the VC-likeinvestingenvironmentsthatspawnedthem.Ihavecometobelievethatcorporationscannomoredothismetaphorically

than early cetaceans could do it literally. Consequently, the key to achievingbetterresultsliesnotinthecloseapproximationoftheventurecapitalecosystembutinafundamentallydifferentapproachtoaninnovationstrategy.

VARIATION→SELECTION→RETENTION

Theinnovationarisingfromtheseeminglysymbioticrelationshipbetweenstart-upsandventurecapitalinvestorscanbeunderstoodintermsoftheveryfamiliar“variation, selection, retention” model of evolution through natural selection.Life in all its forms has manifested relentless innovation in adapting to orrecoveringsuccessfullyfromhighlyvariedandchangingdemandsateverylevel—fromplanetaryconvulsions toperturbationsof thesmallestecologicalniche.Such an outcome has been driven by the nearly endless variety of life on theplanetandthemercilessselectionpressuresthatweedoutthosewhocannotkeepup, which serves to preserve what works until something that works bettercomesalong.First comes variation: there is a steady supply of new businesses seeking

capital.Thesewould-beworldbeaterspitchtheirconcepts—oftenaccompaniedbyworking prototypes and even a year or two ofmarket trial—and investorsselectfromthatsmorgasbordwhattheyseeasthemostpromisingopportunities.Yettheytakecarenottotrusttheirabilitytopickwinnerstoocompletely:mostconstruct a fairly broad portfolio of investments, accepting that some (many?most?)oftheirpickswillendintears.Most venture funds impose selectionpressures almost immediately, titrating

additional investment dollars based onwhether or not a venture is able to hitdefinedmilestoneswithinspecifiedperiodsoftime.Newbusinessesarealmostalwayskepthungry,spurredonbythepromiseofrichesiftheyreachlarge-scalecommercialsuccess.Andwoetoanythatshouldmisstheirinterimtargetsbytoowideamarkorsimplynotrisetothetopoftheheap:portfolioinvestingmeansthat the increasing capital requirements of themore successful ventures in thebroodprovideacompellingreasontocull theweakwithoutmercyorregret inordertofreeupcapitalandfeedthestrong.Finally,retentiontakestheformofthe“liquidityevent”:sellingmostorallof

theequitytothepublicmarkets,relievingtheventureinvestorsofanymaterialinterest(relativetotherestoftheirportfolio)inthelong-termsuccessorsurvivaloftheircreationandgeneratingthecashrequiredtobeginthecycleanew.Hoping to replicate these outcomes, many corporations try to create and

commercialize innovative ideas using a similar variation-selection-retentionmodel, but they are perforce compelled to use different mechanisms at eachstage. They pursue variation typically via large-scale innovation tournaments,

usually setting up some form of Web-based infrastructure to solicit ideas asbroadlyaspossible.Theexplicit belief is thatyouneverknowwhere thenextgreat idea will come from, and so the indispensable first step is to turn overeveryrockinthehopeofuncoveringadiamondintherough.This works for venture capital investment boards because they get to see

reasonably well-developed concepts, typically without having had to spend adimeonthem.Ideasaredeveloped,prototypesfunded,andevenleadcustomerssecured by the entrepreneur at the cost of nights and weekends sacrificed,vacations foregone, and second mortgages assumed. This serves to keep theinvestors’expensesdown,butitisalsoamaterialearly-stageselectionpressure,ensuring that only those ideas that have fired the imaginations and earned thefull-bloodedsupportoftheircreatorsevercomeupforconsideration.In contrast, when an established company seeks to liberate those nascent

blockbusterideasthatallegedlyliefallowinitsfieldsofcubicles,itgetstopayfor all that sweat equity in either official on-the-clock time or diverteddiscretionaryeffort.Realizingthis,evenifonlyimplicitly,mostcorporateinnovationtournaments

solicitonlythemostbare-bonedexpressionsofideas.Thiskeepscostsdown,butwithcripplingconsequences.First,thereisnomechanismtotestthepassionandcommitmentoftheiradvocatesinanymeaningfulway.Asaresult,mostoftheideasthatshowuparenotreallyworthmuchatall,mostoftentakingtheformof“wouldn’titbeneatif…”conjecturesthatareseenbythosewhosubmitthemaslow-cost lottery tickets providing a quick ascent up the corporate ladder ifsuccessful,withnocareerdownsideforbeingoverlooked.Oftenoverwhelmedbythehighnumberoflow-qualitysubmissions,corporate

innovation tournaments very rarely turn up anything terribly innovative and,worse, end up making most participants feel ignored or shunned; after all,AmericanIdolshattersfarmoredreamsthanitfulfills.Andsowhatstartedoutas a way to build enthusiasm for innovation and drive long-term corporategrowthendsupalienatingmostoftheparticipantsandconvincingtheinnovationskepticsthattheywererightallalong.Attemptstoformalize“selection”oftentaketheformofastage-gateprocess.

The idea is to fund a relatively large number of initiatives at a low level andmake subsequent funding contingent upon meeting ever-more-exigent criteriafor technological development or marketplace acceptance. The standardmetaphorisafunnelwithawidemouthandasteadilynarrowingaperture,outofwhichemergesasmallnumberofbigsuccesses.Unfortunately, the key assumption behind a successful stage-gate process is

often violated, namely, that some initiatives are actually killed off. In theVC

world,theprimarymeasureofsuccessisfinancial,andcuttingone’slossesjustmakes good sense.On the other side of the equation, failure is hardly a goodthing,but it rarelydisqualifies an innovator fromsubsequent attempts. Indeed,many entrepreneurial success stories attribute the insights that led to eventualvictorytothelessonslearnedfrompreviousbitterdefeats.Muchhasbeenwrittenabouttheneedforcorporationsto“failfast”andcreate

a culture of learning rather than punishment. But the inevitable reality ofcorporations—for,asissooftenthecase,theirgreatestweaknessistheirgreateststrength—is that they are political, social organisms, not purely economicsystems.Whentheyfunctionwell,theycangenerateandsharetacitknowledgebasedonrelationshipsbuiltontrust.Theirpathologiesarealsowellknownandunfortunately seemingly unavoidable: the long memories, the jockeying forposition, decisionmaking by consensus, and a host of other ills that result inprecisely the kind of risk aversion and incrementalism that leaves themvulnerabletothestart-upstheyallegedlyseektoimitate.Theresult?The“funnel”becomesa“tunnel,”andtheonlyprojectsthatever

getfedintothestage-gateprocessarethosethatareallbutguaranteedtomakeitout the other end—which is, unfortunately, very different from guaranteedcommercial success, for the politics of corporate innovation do not mandateresults,merelyallthenecessaryapprovalsalongtheway.Finally, theVCworld’s liquidityeventhasnomeaningfulanalogforgoing-

concerncorporations.VCfundsarenotdesigned tobuildmaterialandreliablerevenuegrowthwithsolidcashflow;theyaredesignedtobuildabusinessthatcan be sold. In a corporation, however, selling off a new venture, even at amaterialprofit,isoftenseenasasignoffailure.Yetretainingasuccessfulnewbusinesswithinanexistingcorporatefoldisadouble-edgedsword: thegrowthand profitability are terrific, but who is not keenly aware of the dangersassociatedwiththeincreasedcomplexitybornofdiversity?Inshort,theventure-backedmodel’ssuccess,likethatofevolution,requiresa

profligacy with the most precious commodities that few corporations cancomfortably contemplate, let alone accept. And so corporations often end upimportingthesurfacefeaturesofavariation-selection-retentionmodel,securingmuchofthecostandlittleofthebenefit.Corporateinnovationprogramsneedtoaccept that organizations cannot replicate or even meaningfully imitate theventurecapitalapproach.Variationis tooexpensive,selectionis toopoliticallyfraught,andretentiontoomanageriallycomplex.

FOCUS→SHAPE→PERSIST

Sohowmightacorporationlearnhowtoholditsbreath,toadaptitslungstoalife in the sea, to innovate consistently and successfullywithin the constraintsthatdefineit?Thefirststepistofocusthecorporation’seffortsonopportunityareasthatare

strategically relevant. Rather than going dynamite fishing with an innovationtournament, begin by identifying those high-level spaces that will define thefutureofyourindustry.Using,forexample,ascenario-basedplanningexercisethat can often be conducted in two to three weeks, it is typically possible toidentify a manageable list of where the most meaningful opportunities arelikeliesttolie.The process should not be driven by what will necessarily be highly

speculativefinancialanalysesbutratherbuiltuponanassessmentofwhatsortsofopportunitiesbothhavethepotentialtoreshapeyourindustryandarerelevanttoyourorganization.So,forexample,J&J’sbeliefintheviabilityofautomatedsedation arose out of a belief that the frequency and intensity of anesthesia-assistedsurgerieswouldincreaseasaconsequenceofanagingpopulation,evenas cost-containment pressures on healthcare generally would reward solutionsthat reduced cost without compromising quality. The company’s long-term,large-scale involvement inbothmedicaldevicesandpharmaceuticalsgave it acrediblelevelofrelevantexpertise.Theintersectionofanidentifiedmarketplaceopportunity with the company’s capabilities made automated sedation anexcellentcandidateforstrategicfocus.

FOCUS→SHAPE→PERSIST

Shifting from “selecting” ideas to “shaping” them is perhaps the largest breakfrom existing standard practice.Most organizations have any number of ideasfloating around, some of which keep resurfacing year after year without everseeming to get anymeaningful traction or being successfully dispatched. Thereasonisveryoften that there isakernelof truthsomewhere in the ideabut ithasprovenimpossibletodevelopaviablebusinessplan.Insuchinstances, it isoftenhelpfultoexaminetheunderlyingconceptfrom

theperspectiveofsustainingversusDisruptiveinnovation.Afterall,ideasdonotcome prelabeled; whether a concept is commercialized in a sustaining orDisruptive way is often a matter of strategic choice. Consequently, there istypicallyanopportunitytoshapeanideasothatitmeetstherequirementsofoneormoreofthecategoriesintheIPM.Take,forexample,PCAS,asmallOakville,Ontario—basedcompanythathas

developed MedCentre, a device for the remote dispensing of prescriptionpharmaceuticals. Resembling nothing so much as a bank ATM, the deviceconsistsofavault storingan inventoryof frequentlyprescribedmedications,avideoandvoicelinktoalivepharmacistinaremotecounselingcenter,scanningequipmenttoenablereviewoftheprescriptionandthecreationofanelectronicrecord,andacapacitytoprocessreimbursementorpatientpayments.Thedevicehasafootprintofabouttwenty-foursquarefeet.IsthisaDisruptiveinnovation?ThereisnothingintrinsicallyDisruptiveabout

anything,MedCentre included.WhatPeterSuma,DonWaugh, and the restofthe team at PCAS were faced with is what every entrepreneurial team faces,knowinglyornot,whethertheyareastand-alonestart-uporacorporate-backedventure: how to shape their ideas in ways that maximize the likelihood ofacceptablesuccess.WhenyouexploreboththesustainingandDisruptivepathstomarketforaconcept,itisfarlikelierthatthepathwiththegreatestprobabilityofsurvivalcanbeidentified.In thecaseofMedCentre,PCAShadbothsustainingandDisruptiveoptions

to consider—as is the case for just about every innovation. For example, inCanada,thedominantretailpharmacycompanies(e.g.,ShoppersDrugMartandPharmaPlus, which are similar to Walgreens or CVS in the United States)operate extensive chainsof retail drugstoreoutlets that sell consumerproducts(makeup,personalhygiene,etc.)andawidevarietyofsundries(greetingcards,

snackfooditems,magazines,etc.)inadditiontodispensingprescriptiondrugs—theserviceforwhichtheirbrandsareperhapsbestknown.Trafficinthesestoresisdriveninlargepartbytheprescriptiondrugservice,butprofitabilityisalsotoavery largedegree a functionofpurchasesof all theotherproducts they sell.Thestoresthereforefaceabindingconstraint:gettingpeopleinthestoremeanshavingawell-stockedandefficientpharmacy,whichisbothcapital(floorspaceandfacilities)andlabor(pharmacists)intensive;butprofitabilityisafunctionofdevotinglesscapitalandlabortothiselementoftheirbusinessmodel.PCAS could seek to develop MedCentre as a sustaining innovation for

incumbent retail drugstores. IfMedCentre could offer as broad a selection ofdrugsas existingpharmacieswithonlya fractionof the floor spaceandmuchlowerlaborcosts(thankstoeconomiesofscaleintheremotepharmacist-staffedcounseling center), the devicewould break a defining tradeoff of the existingdrugstorebusinessmodelandexpanditsfrontier.PursuingthispathsuccessfullywouldrequireMedCentretoclearaveryhigh

performancebar.Drugstorestypicallyhaveanextensiveinventoryofdrugsandare able to provide pill counts that are specific to each patient. They can alsocarry liquid formulations that require refrigeration and provide compoundingservicesoraddflavoringstomedicationsforchildren.However,currentversionsofMedCentrecarryonlyarelativelylimitedrangeofdrugsinpillform,andthefirst models carried those only prepackaged in the top two or three mostfrequently prescribed quantities. Now, it turns out that 90 percent of allprescriptions can be filled with only 116 different drugs in three pill counts,whichfallswellwithinMedCentre’scapacityof2,500drugstock-keepingunits.Andso, just asanATMcannotcopewith thecomplexityof tasks that abankteller can but is capable of dealing with the majority of day-to-day bankingtransactions, MedCentre is “good enough” for significant, if less demanding,tiersofthemarket.Consequently,developingMedCentreinawaythatwasclearlysustainingto

traditional retail drugstores would require overcoming two hurdles. First,MedCentrewouldhavetobegoodenoughtoaddressthefullrangeofservicesthat the established players currently serve with the asset-and labor-intensivemodel. Getting the technology up to that standard so that a pharmacy needaccept no decrease in performance promised to be expensive and time-consuming.In addition, adrugstore adoptingMedCentre—evenaversionofMedCentre

asgoodasatraditionalpharmacy—wouldlikelybetacklingchangesintheretaildrugstore business model that could have some potentially unsettlingconsequences.Thanks to the efficiency of theMedCentre, customers typically

complete their entire transaction in a single session.They interact in real timewithaqualifiedpharmacistviathevideolink.Thepharmacistprovideswhateverconsultationisrequired(e.g.,alertingpatientstoanypossibledruginteractions),and ifeverything is inorder thepharmacist signalsMedCentre todispense thedrug from inventory. (If the drug is not carried, MedCentre can signal thedrugstoretodispatchitbymail.)Incontrast,drugstoresoftenhavepatientsdropoff their prescription, which is then filled, typically in about ten to fifteenminutes.Thiswait timecancontribute todrugstoreprofitability, forcustomerswillusuallybeginwanderingabout the store,pickingup itemshere and there,and so not infrequently turn a quick stop at the drugstore into a half-hourshopping trip. CouldMedCentre’s efficiency—which frees up floor space foradditional products—reduce the likelihood that customers actually shop at thedrugstore? In other words, for all MedCentre’s benefits as a sustaininginnovation, achieving a clear-cut, unambiguous breakthrough is almost alwaysverydifficulttodoandtaintedbydeepuncertainties.For these reasons, conversations with executives at different major retail

drugstore chains have had very different outcomes. Some have been openlyhostile, seeingMedCentre as a threat to their installed asset base.Othershavebeenmore enthusiastic, for although they recognize the possible threat to thecurrent retail pharmacy assets they also see the opportunity to own additionalchannels of distribution, placing the MedCentre devices in more convenientlocations,suchastheworkplaceordoctors’offices,inordertowinbackmarketsharelosttoalternativedistributionchannels.Analternative strategy is todeployMedCentrewith retailers that arenot as

dependentonimpulsepurchasesduringthetraditional“waittime,”thatarelessinvested in an extensive retail pharmacy asset base, and that typically do nothave as demanding a set of criteria for their pharmacies.Grocery stores (e.g.,LoblawsinCanadaorPublixintheUnitedStates)andotherlargeretailers(e.g.,Walmart)wouldappear tofit thisdescription.For theseretailers, theprovisionofpharmacyservicesisawaytomaketheirstoresasnearlyaone-stopshopaspossible.Customerstendtogothereprimarilybecausetheyareembarkingonarelatively significant shopping excursion, and the need to drop off theprescription and then go back and pick it up is actually an inconvenience tocustomers. (Personally,more thanonceI’vedroppedoffaprescription,carriedonwithmyshopping,andthengottenhome,only torealizeInevercompletedthe “drug buy.”) Neither are these retailers likely to carry as extensive aninventory of drugs nor provide sophisticated compounding services. Theseretailersnotinfrequentlyruntheirpharmaciesasbreak-evenbusinessesdesignedto increase traffic and in fact see investments in them as a burden, a “cost of

doing business,” not a stand-alone profit center. This is clearest when thepharmacy hours do not match the grocery store hours because of the muchhigher-costlaborattachedtokeepingthepharmacyopen.For these retailers MedCentre could well be just what the doctor ordered.

MedCentre reduces the asset burden of having a pharmacy and allows thepharmacytobeopenwheneverthegrocerystoreis,thankstotheeconomiesofscalearisingfromtheremotecounselingcenter.MajorsupermarketchainshaveexpressedsignificantinterestindeployingMedCentre,andcertainlythismarketcouldbehugeforPCAS—precisely thekindofniche thatcouldcreatea largeandprofitablebusiness.It is worth considering, however, whether this application would reward

subsequentimprovementsintheMedCentretechnologyinwaysthatcouldfuellong-term and truly Disruptive growth. After all, if grocery stores and otherretailers wanted to compete vigorously for the mainstream retail drugstorebusiness,itiswellwithintheirreachtomakethoseinvestments,yetforthemostpart they have not. Consequently, it is possible that by deployingMedCentreprimarily in grocery stores the device would be used to create a “strategicinnovation”—the sortofDollarGeneral—type solutiondescribed inchapter4.Only in this caseMedCentre could find itself relegated to a niche—even if alarge and profitable one—not because of any limitations in its technology butbecause of a relative lack of customer desire for the sorts of innovations thatwouldallowMedCentretomovebeyondthatinitialmarket.The strategic question is whether optimizing MedCentre for this customer

segmentwould limitMedCentre’s improvement trajectory andhence constrainthecompany’sgrowthpotential.Ifthissegmentofcustomersisnotmotivatedtosee MedCentre improve in important ways, would PCAS soon find itself astationary target for other automated dispensing device providers?Would thecompanybetradingshort-termsuccessforlong-termmisery,seeingitscurrentlyhighly differentiated position erode and ultimately ending up locked in a lessprofitablecost-cuttingbattlewithequallywell-resourcedcompetitors?The key strategic point here is that the customers one initially chooses to

focusonoftendetermineone’slong-rungrowthopportunities.Itiscriticalnottofocusexclusivelyoncustomersthatarelesslikelytovalueanongoingstreamofinnovations that can extend your initial strategic differentiation. It is for thisreason that PCAS has refused requests for exclusive deals with grocery storechains,evenwhensuchofferscameintheearlycash-strappeddaysofthestart-up.AthirdalternativeistodeployMedCentreastheengineofatrueDisruption.

There is significant“nonconsumption”ofdrug-dispensingservices inavariety

oflocationswheresuchservicescouldbeofparticularvalue.Forexample,manypeople get a prescription from a doctor in either the hospital or the doctor’sofficeandthenarefacedwiththeinconvenienceofeitherfillingitatthenearestpharmacy or combining it with other errands. A MedCentre installed in thehospital or doctor’s office would make this chore much more convenient. Asimilarscenariomightplayoutinworkplaceinstallations.Intheseapplications,however, theexpansionofMedCentre’scapabilities servesprimarily to reducethehealthcarecostsincurredbyinsurers.TheymightvaluegreatercapabilityintheMedCentre device, but there aremany other competing priorities in theselocations.Doctors’officesandhospitals areworriedabout, amongmanyotherthings, electronicmedical records, new regulations, capital investment in newdiagnosticsdevices,andsoon.Andworkplacesaretypicallylookingforwaystodevote fewer resources to healthcare, not more—even if the investment doesoffer a cost savings. Consequently, they might be only marginally moremotivated than the supermarket pharmacies to Disrupt retail drugstores, andhaving customers that will pay you to innovate is key to preserving andextending advantages based on differentiation. What we really need is anapplicationwhereaMedCentrecustomerwouldbedirectlyrewardedby,andsorelatively more eager to pay for, an expansion in MedCentre’s productivityfrontier.MinuteClinicisquicklyestablishingitselfasanalternativehealthcaredelivery

channel. Often set up in shopping malls, MinuteClinic and other alternativehealthcaredeliverychannelslikeitprovideamorelimitedrangeofservicesthanaphysician’soffice(andofcoursevastlymorelimitedthanatahospital)andaregenerallystaffedbysomeonelesswelltrainedandlessskilledthanaphysician.However,forwhattheycandotheyprovidestate-of-the-artservice,enabledbystandardized,easy-to-usediagnostictestsandtreatmentprotocols.Forexample,a typicalmenuofservicesatMinuteClinic includesvaccinations,minor illnessandinjuryexams,basicscreeningsforconditionssuchasdiabetes,hypertension,andobesity,andsoon.A MedCentre device allows alternative healthcare delivery channels like

MinuteClinic to expand the scope of the services they offer in a way that iswhollyconsistentwiththeirbusinessmodel:thereislimitedneedforadditionalfloorspaceandnoneedtoaddexpensivelabor.Inshort,asMedCentreexpandsitsfrontier,thesechannelsexpandtheirfrontiers.Consequently,theselocationswouldlikelyplaceahighvalueonthesortsofimprovementsinMedCentrethatwoulddrive itsupmarketmarch into themainstream retailpharmacybusiness.Forexample,subsequentversionsofMedCentrehaveaddeda“pickandpack”capability,whichallowsthedevicetodispensecustomizednumbersofpills(and

with greater accuracy than human pharmacists). Consequently, the vault nowstockspillsinbulk,whichisafarmoreefficientuseofspace,andcandispensealargerrangeofmedicationsinanypracticalpillcount.MinuteClinicseemstometo be the kind of customer that would also highly value other improvements,suchastheabilitytodispensepreparationsinliquidorsuspension,andperhapsevenrudimentarycompounding.Thesecapabilitiesmaybejustoverthehorizon,givenMedCentre’sincreasingexpertiseinroboticsandtherateofadvanceinthefield generally—but are likeliest to be developed and deployed rapidly onlywhenthereareimportantcustomerswillingtopayforthem.Inotherwords,themoststrategicallyvaluablecustomerforMedCentreisthe

onethatseesMedCentreasfuelforitsownDisruptionoftraditionalhealthcaredelivery channels—doctors’ offices, hospitals, and retail pharmacies. It is onethingtohelppowerfulplayerssuchashealthinsurerstocutcosts;itissomethingelseagaintoserveasanenablingtechnologythatdrivesafull-scaleDisruptioninhealthcare.In summary,MedCentre has both sustaining andDisruptive applications for

established retail pharmacies. Some drugstore chains see the need to embracethis technology to avoid being steamrollered by it, while others are highlyresistant. This makes courting such companies a challenge. Supermarkets canmake their established businessesmore cost effective, and existing healthcaredelivery channels can be more efficient with MedCentre. They are thereforemore attractive customers than the retail pharmacies but might actually limitMedCentre’s long-termgrowth andPCAS’s ability tomaintain its competitivedifferentiation.Themostattractivestrategiccustomeristheonethatwillrewardthe kinds of innovations that allow MedCentre to realize its full Disruptivepotential.ThereismoretoMedCentreandPCASthanIhaverelatedhere,butjustthis

briefoverviewprovides an illustrationofhow the shapingprocess canunfold.The fundamental technological insight—automated drug dispensing—is notsomething that necessarily implies a specific approach to commercialization.WhatDisruptiontheoryoffers,whenappliedwithinaparadigmofshaping,isaway to identify and consider all the options so that whatever trajectory onechooses,itisatleastchosenonpurposeratherthanbydefault.

FOCUS→SHAPE→PERSIST

The“retention”modelofVC-typeinvestingisbuiltaroundpruningawaywhatfailsasquicklyaspossible,thensellingwhatworksattheheightofitsvalue.Foroperatingcompanies,however,thegoalisnottocreatesomethingofvalueandthensellitbuttofindaviablebusinessandrunit.Themechanismforcapturingvalue is not the one-time sale of equity but cash flow from operations over aperiodofyears.Theabilityofoperatingcompaniestosellthesekindsofassetsistypicallymuchmore limited—they lack the expertise, access to the right dealflow, and the right sort of priorities. After all, in most operating companies,sellingabusinessisasignoffailure,notsuccess.Inaddition,VC-typeinvestingtypicallyplaysoutoveramuchlargernumber

of investments than operating companies are able to make. A VC fund’stoleranceforfailureissimplymuchhigherbecausetheonlythingbeinginvestedismoney.Inoperatingcompaniespeopleareinvestingtheirpoliticalandsocialcapital,theircareersandreputations,andinmanyinstanceseventheirsenseofself-worth. Consequently, the price of abandoning something goes far beyondthe lossofanyopportunity torecoupsunkcosts. Inaddition, it is justpossiblethatthepeoplewhoworkinestablishedcorporations—ratherthanpursuingtheirown entrepreneurial opportunities—are simplymore risk averse than the folkswhobuildbusinessplansandseekventurefinancing.Nomatterhowloudlyitisproclaimed that “nothing is a failure so long aswe learn,” it appears to be animmutable factoforganizational life thatbeingpart of an effort thatgets shutdown,nomatterthecircumstances,isalwaysablackeyetosomedegree.This implies thatcorporationsseekinggrowthby launchingnewinnovation-

drivenbusinessesmustchoosewhichhorses tobackmuchmorecarefully thanVCinvestors(astheycannotbacknearlysomany),andtheymustridethemformuchlongerinordertorealizethevaluetheyhavecreated.In light of this, I propose an amendment to the oft-repeatedmantra of “fail

fast.” Instead, “learn fast” and use what you learn to improve your existinginitiativeinwaysthatavoidfailureasmuchaspossible.Thatmightsoundfacile,butinmyexperiencethe“failfast”bumperstickerdoesinnovationanenormousdisservice:weare tempted to think that theunpredictablenatureof innovationprovides a screen for incompetence, and would-be innovators infer thatinnovationnecessarilyimpliesfailure—somethingtheywouldratheravoid.Instead,when innovation is guided by the principles ofDisruption, the key

assumptions of the business model can be specified with uncommon clarityuncommonly early in the development and commercialization process. Thisallows a corporation to persist in its pursuit of an underlying innovation orinsight, quickly converging on the strategy that is likeliest to succeed. This isvery different from what many take away from individual failures—which isvery often little more than “that business didn’t succeed.” A determinationultimately to prevail, however, need not be quixotic but insteadmay enable asystematicandmethodicalmarchtowardawinningformulation,onethatbreakstherighttradeoffsintherightsortsofways.

WHOCARES?ANDABOUTWHAT?

WhyshouldacompanypursueDisruptiveinnovation-drivengrowthbeyonditscoreatall?If that10percentsuccessrateat Intel isanyindication, innovationgenerallyisagamble,andevenwithDisruptiontheorytoguideyou,thebestIcanclaimisthatyoursuccessratemightclimbbyfivepercentagepoints.Withthosekindsofodds,whatisthepoint?“There is none” appears to be the view of commentators calling for

establishedcompaniesto“sticktotheirknitting”andavoidthefraught-even-if-sometimes-fruitful attempts to build new growth businesses from within anexistingcorporation.Thedesired inference seems tobe that itwouldbebetterforshareholdersifeachcompanysimplyranoutitsowntrajectoryofsustaininginnovation, thenwillinglysuccumbedinabenigncorporateversionofLogan’sRun.The problemwith this view is that the benefits of competition can only be

securedifparticipantsinthegreatgameactuallycompete.Individualwildebeestdonot throw themselves into the jawsof the crocodilesbecause theyperceivethemselves to be weak or in some other way unfit. Rather, that a givenwildebeestgets caught is the evidence that thatwildebeestwasunfit. In short,whetherornotsomethingis“good”isafunctionofitssurvivalasdeterminedbyacontestinwhichallcombatantsputforththeirbestefforts.Anyparticipantthatforegoes any means at its disposal to survive not only undermines its ownsurvivalbutcorruptsutterlythelargercompetitiveprocessofwhichitisapart.ThisisastrueintheshoppingmallasitisontheSerengeti.Asaconsequence,themechanismthatallowsstockholderstoprofitfrombackingwinnersdepends

upon the willingness of faltering incumbents to resist their own demise asvigorouslyaspossible.If you accept that it is a worthwhile goal for corporations to pursue

innovation, it isworthsayinga fewwordsabouthowthevarious levelsof theorganizationshouldtakeonverydifferentbuthighlycomplementaryrolesinthepursuit of an effective innovation program. The key is adopting a definingconcept of strategy and innovation: constraints. Each level of the hierarchyshould focus on defining the constraintswithinwhich the level below itmustfunction.Withthisinmind,theboard’sresponsibilityisnottoworryaboutthespecifics

ofaninnovation.Rather,itfallstotheboardtomakethehardchoicesaboutthelevelof investmentanorganizationwillmake ineachof the fourquadrantsoftheIPM.Howresourcesareallocatedacrossthefourtypesofinnovationwillbea key determinant of the organization’s overall innovation risk profile. Acompany with too much emphasis on “chess” (sustaining/incumbent) will bevulnerable to Disruptive attacks; a firm with too much wagered on “bingo”(sustaining/entrant)willbe subject to thevagariesof fate;acompanywith toolittlecommittedto“checkers”(Disruptive/entrant)willbeunabletodrivelong-termgrowth;andsoon. It is theboard’s job, then, toconcentrateon the focusphase.Indeterminingtheresourcestobecommittedtoeachtypeofinnovation,the

boarddefinesakeyconstraintontheseniormanagement:itmustfindeffectiveways to employ those resources within each quadrant. This means seniormanagementturnsitsattentiontotheshapingphasebydefiningthescopeofthesearchineachquadrant.Broadlyspeaking,whataretheproductmarketsthatthecompanywillcompetein?Whatcustomersegmentswillbetargeted?Definingthese questions in a way that keeps operating managers and technologistsfocused on the right sorts of targets is central to actually getting the kind ofinnovation appropriate to each quadrant. In other words, senior managementmustdefine the tradeoffsamong thedifferentproductivity frontiers thatdefineeachinnovationtype.

FIGURE38:RESPONSIBILITIESFORINNOVATIONBYHIERARCHICALLEVEL

Forexample,anorganization’sbusinessmodeltypicallydefinesits“floor”interms of the market segments it can serve. Time and again, companies thatcommercializeinitiallyexpensivetechnologieswithahigh-costbusinessmodelareunabletoridethecostreductionsintechnologyovertimeintolowertiersofthe market. In colloquial terms, technologies trickle down, but organizationstrickleup.Consequently,akeyresponsibilityofseniormanagementistoensurethateffortsintendedtodrivelong-termgrowththroughDisruptiveinnovationinnewmarketstargetthelowestpossibletiersofthemarketinordertomaximizethe duration of the upmarket march. Similarly, when pursuing sustaininginnovation,howfaroutmusttheexistingfrontierbepushed?Or,whenhedginga strategic risk,whatalternative frontiersmust theorganizationbeprepared to

defendagainst?Finally, operating managers have as their primary task developing the

business plans that put the new venture on its desired path. Success is then afunctionofthewillingnesstopersist informulatingandtestinghypothesesandincorporating feedback quickly and efficiently in the pursuit of the specifiedgoal.Inpracticethismeansdeterminingwhatarethemostappropriatetradeoffsamong price and nonprice constraints on the frontier defined by seniormanagement.Withthisdovetailingofcascadingconstraintsfromoneleveltoanotherbased

on the principles of Disruption, an organization can claim to have in place arationalanddemonstrablyeffectiveinnovationstrategy.

EPILOGUE

CREATIVECREATION

Economist Joseph Schumpeter’s signal contribution to economicswas to putinnovation at the core of economic development. He saw beyond the field’sobsession with marginal cost analysis and price-based competition,understanding that this fixationwasmerely a function of the ability tomodelsuch features of the landscape cleanly, with little regard for their ultimatelysecondary,andperhapseventrivial,roleinshapingcompetitionamongfirms:

[I]n capitalist reality as distinguished from its textbook picture, it is not[price] competition which counts but the competition from the newcommodity,thenewtechnology,thenewsourceofsupply,thenewtypeoforganization… competition which commands a decisive cost or qualityadvantage and which strikes not at the margins of the profits and theoutputsoftheexistingfirmsbutattheirfoundationsandtheirverylives.36

Schumpeter memorably labeled the outcome of this type of competition“creative destruction,” a term that has become for many synonymous withinnovation.TheproblemwithSchumpeter’smodel,however,isthattheprecisemechanisms by which one achieves the sort of innovation that is truly“creativelydestructive”haveremainedlargelyablackbox.Wesimplyhavenotbeen able to gain anymeaningful traction on how one achieves the “decisivecostorqualityadvantage”thatstrikesatthe“verylives”ofincumbents.Indeed,commentators on Schumpeter’swork note that for him, the entrepreneur is “asocial miracle in the precise sense of the word: an event beyond the laws ofnatureandsociety.”37Andsoalthoughthecentralityofinnovationtoeconomicadvance is certainly correct, it is difficult to get much further when themechanismofinnovationisdescribedsimplyasa“miracle.”Nevertheless,Schumpeterprovidesatleastoneclueastothesortofthingthat

mightrisetotheleveloftherequisitesortofmiracle.ForSchumpetercapitalistentrepreneurship and technological progress are “essentially one and the samething,”thefirstbeing“thepropellingforce”ofthesecond.38AsDisruptionhasbeendefinedandexplainedand testedhere, technologicalprogress is similarlythe“propellingforce.”Anewbusinessmodelservestocreateanewproductivityfrontier, but technological progress is what expands that frontier in ways that

lead to Disruption. And Disruption leads to the creative destruction thatSchumpeteridentified.Finally,Schumpeterdemonstrated that the synthesisofentrepreneurshipand

technologicalprogressledtothedestructionoftheoldthankstothecreationofthe new (hence, creative destruction). That is, he explained how the“destruction” half of his model works. Unfortunately, the mechanisms ofcreationwereleftlargelyunexplained.If there is a broader significance to looking at innovation generally and

Disruption in particular in theway I have described it here, I hope it is that Ihave shown that Disruption has, for now at least, the most valid claim ondescribinghowthecreationthatleadstothekindofinnovationthatSchumpeterdescribed actually happens. It is not by pushing out the existing productivityfrontierinamannerthatis,inalllikelihood,whollyreplicablebyone’sequallyendowedcompetitors.Rather, it iscreatinganentirelynewbusinessmodel—anewformoforganization,touseSchumpeter’sterm—thatmakespossiblewhatincumbents can, almost literally, not evendreamof—not because it is beyondtheir imaginationsbutbecause it isbeyond theirdesires.Theoverthrowof theoldorderisuponuswhenentrepreneursridethewavesoftechnologicalprogressthatexpandthenewfrontier theyhavecreated inways that incumbentscannotrespondto,evenwhentheyfinallyrealizethattheymust.Inotherwords,Disruptionprovides an explanationof creativecreation: the

how, if,when, and how long of the kinds of innovations that have repeatedlyremadetheeconomiclandscapeintheserviceofthegeneralweal.

ACKNOWLEDGMENTS

ThecommitmentofIntelCorporation tocomputerscience isunquestionedandobvious. The company’s long-term investments in expanding the frontiers ofhuman understanding in the design, manufacture, and use of microprocessortechnologyhavemadeInteloneoftheiconiccorporationsofourtime.Lesswellappreciatedisthecompany’slong-standingcommitmenttopushing

outthefrontiersofknowledgeinmanagementscience.ThecompanyhasbeenalongtimecollaboratorwithClaytonChristensenoftheHarvardBusinessSchoolandRobertBurgelmanoftheStanfordGraduateSchoolofBusiness.Christensenhas explored the concept of Disruptive innovation, while Burgelman hasinvestigated theprocessesbywhichstrategy is formulatedand implemented inlarge,complexorganizations.Bothstreamsofworkareamongthebestregardedin theworld of business research, in large part thanks to Intel’s openness andwillingnesstocollaborate.To my mind, perhaps the most unusual aspect of Intel’s support for

Christensen’s, Burgelman’s, and now this research has been the company’swillingness to share a verynearly complete picture of its attempts to innovateandgrow.Mostcompaniesthatendupbeingacknowledgedinabusinessbookarebeinglaudedalmostexclusivelyfortheirmerits,soit ismucheasiertoseewhy they would allow their stories to be told. But Intel has been much lessguardedthanmostcompaniesIamawareof,oftenwillingtoshareitsmistakesas well as its triumphs. That is somethingmuch rarer and farmore valuable.From the company’s early struggles in themid-1990s launching its ultimatelysuccessful Celeron line ofmicroprocessors to its failed initiatives in the early2000s with videoconferencing, toys, and a variety of Internet-based services,Intelhasbeen,asfarasIknow,uniqueinprovidingawindowonitsexperiencesdevelopingnewbusinesses,wartsandall.This book is possible only because of that samewillingness. JerryBautista

andRichWykoff,currentor formergeneralmanagersof Intel’sNewBusinessInitiatives(NBI)division,havemadeavailablebusinessplansandsurvivaldataonalargeandrepresentativesampleoftheinitiativeslaunchedbythecompanyfrom approximately 1998 to 2007. Those datamade possible the experimentsthatarethefoundationfortheprimaryconclusionsIargueforhere.TherearenosubstitutesforthedataIntelhasprovided;thisresearchwouldnotexistwithoutthem. I am also grateful to Ellen Recko at NBI for her guidance andcontributions.Johnson& Johnson has also supported thiswork in significantways.Mike

Gustafson and Ken Dobler, both of whom worked extensively on theSEDASYS™ project profiled in chapter 6, were unstinting in giving of theirtimeandexpertisesothatImightdoaserviceablejobofexplainingtherelevantattributesofwhatIbelievewillbeasignificantcontributiontohumanwelfare.Beyondthat,theopportunitytoworkwiththeSEDASYS™teamfromsomeofits earliest days in 2004—when the technologywas still being developed andwhen the strategy that seems so obvious now was still being shaped inCincinnaticonferencerooms—wasaprivilegeandoneofthehighpointsofmyprofessionallife.TheworkofThomasThurston,whowasfirstexposedtoDisruptionwhenpart

of Intel’s NBI group and is now an independent consultant and founder ofGrowthScience International,was the catalyst for this book.He compiled theinitial data set on Intel’s new business development efforts and demonstratedthat Disruption theory could be shown to have true predictive power. I amgratefulthatThomaschosetosharehisresultswithmeandtocollaboratewithmeonsubsequentexperiments.My list of collaborators in the preparation of this book includes the usual

suspects:MumtazAhmed,DwightAllen,KenHutt,LauraMartin,andHowardWeinberg,allofwhomhelpedmedevelop theargumentsand revieweddrafts.For this project, addMeredithAmdur,WhitneyDelich,AndyHenderson, JeffJohnson,andBiffWruszek to that list; Ihope theywillbesimilarlywilling tohelpinsimilarwaysonthenexteffort.MatthewLarsoninvestedeasilydozensofhoursofhispersonaltime(sorry,Matthew,stillnoutilizationcodeforthis!)in developing the Web-based surveys used for the experiments reported inchapter 2. Mary Caputi, Florence Evina-Ze, and Jeff Schwarz did yeomanservicecollectingthedataonvariousindustryexamplesandinparticularontheevolution of Southwest’s route structure, while Ruben Gavieres found therelevantcomparativeefficiencydataonairframesthatclinchedthecasemadeinchapter3.Finally, I thank, not surprisingly, Clayton Christensen. He has not merely

permitted but actively encouraged my work on Disruption. This theory isClayton’s intellectual“baby,” themostsignificant fruitofhisacademiccareer.Discovering the theory of Disruption required a great mind. He hasenthusiastically supported my efforts to extend and eventually reformulate asystemofanalysisthathecreatedandintheprocesshasbeengraciousbeyonddescription.Thatrequiresastillrarertrait:agreatheart.

Mississauga,March2011

APPENDIXA:

DELIMITATIONSANDLIMITATIONSOFTHEEXPERIMENTS

The experimental design employed here is a longway from perfect, and it isworthbeingclearaboutwhatthesefindingsdonotdemonstrate.Someofwhatthe experiments do not show is a function of explicit choices in design (thedelimitationsof the study).Otherconsiderationsbearingon thepersuasivenessandgeneralizabilityofthefindingsareaconsequenceofunavoidableconstraints(thelimitationsofthestudy).Icannotcrediblyclaimthat thediscussionbelowexhaustseithercategory,butthesearetheissuesthatoccurredtome.

DELIMITATIONS

TestingthepredictivepowerofDisruptionversusthenullisonlyafirststepindemonstratingDisruption’ssuperiorityasapredictiveframework.ThenextstepistocompareDisruptionwithothertheoriesofinnovation.Suchanexperimentwould require essentially two treatment groups, one that was instructed inDisruption and another thatwas instructed in some other theory.An unbiasedhead-to-headcomparisonofthiskindwouldrequiretheinvolvementofequallyexpertinstructioninthetheoriesthatweretobecompared.IntheabsenceofsuchevidenceIhavetriedtobecarefulaboutclaimingthat

Disruptionisbetterthanothertheoriesofinnovation.WhatIbelieveIcanclaimis that Disruption has credible evidence supporting the claim that it improvespredictiveaccuracy.Thatis,Disruptionworks,butwhetheritworksbetterthanothertheoriesIcannotsay.However,ifnoothertheorycan,withequalorbetterjustification, claim that itworks too, then forpracticalpurposes the claim thatDisruption theoryworks amounts to the claim, for now at least, that it worksbest.Whenhasaventurebeen“launched”?Bythetimemostinvestorsarepouring

tens ofmillions of dollars into a new company, the company is hardly “new”anymore: it has been experimenting and refining its business conceptcontinuously, and sometimes for years. There is a great deal of informationavailableuponwhichtoassessitslikelihoodofsuccess.However,theventuresin theportfolioused in theexperimentswereassessedonlyon the informationavailable at a relatively early stage of funding: the Seed Approval Meeting(SAM)stage,whichconsistedofseveralhundredthousandtoperhapsasmuchas slightlyover$2million.Consequently,wecannotknowon thebasisof theexperimentalevidencehowmuchofanimprovementDisruptionmightprovidewhenappliedatadifferentstageintheinvestmentcycleor,indeed,ifitwouldofferanyimprovementatall.Inparticular,onewouldexpecttheretobealowerpercentageofultimatefailuresatlaterstagesoftheinvestmentcycle,andsothe50percentimprovementinpredictiveaccuracyseenwithanearly-stageportfoliowould likely not apply to a portfolio composed exclusively of later-stageinvestments.Thedependentvariable(survival)wasnotdefinedassurvivaloveraspecific

period of time: if a venture was alive at the time of the experiments it wasdeemed to have survived. The case studies indicated when each venture had

been funded, and so participants in the experimentswere able to consider theperiodoftimeoverwhichtheywerepredictingaventuretohavesurvivedwhenmakingtheirassessments.For example, for the experiment conducted in 2010 at Ivey some of the

ventureshadbeen funded in1999, andparticipantswould thereforeknow thatthey were predicting survival over an eleven-year period. Participants in the2008 experiments would have been predicting survival for the same set ofbusinessesoveranine-yearperiod.(Infact,therosterofsuccessesandfailuresdidnotchangebetween2008and2010.)Aprecise definitionof survival—for example, surviving for five years after

SAMfunding—couldhavebeenused.However,specificparametersonsurvivalare unavoidably arbitrary. Some initiatives will receive funding over longerperiodsoftimethanotherssimplybecauseofthenatureoftheindustrieswithinwhichtheycompete.

LIMITATIONS

As noted in chapter 3, there are noDisruptive/entrant ventures in the sample.This limits severely the ability to claim that Disruption theory accuratelypredictsthesurvivalofDisruptorsbasedontheexperimentalevidence.The accuracy ofDisruption could not be tested against the accuracy of the

NBI selection criteria because the only ventures that could be included in theexperiments were those on which actual survival data were available. It ispossiblethatifDisruptionhadbeenusedtoselectwhichinitiativestofundfromthe full population of ideas that NBI considered, a great many more failureswouldhavebeen funded.Oneway toovercome this limitation is toassess thesame population of unfunded business plans using two different frameworks,thenactually fund thosepredictedbyeach framework to surviveandseewhathappens.Ihavelittlehopethatsuchanexperimentispractical.

APPENDIXB:

IMAGEILLUSIONS

Belowisthetextofthedisguised,summarizedcaseforImageIllusionsasusedin the experiments reported in chapter 2. The full library of cases used in theexperimentsconsistedofforty-sevenmoresuchcases.

MARKETNEED

In 2001, digital cameras, camcorders, photocopiers, fax machines, and otherimagingdevicescontained“brains”intheformofcomputerchipsthatprocessedimages.ThesecomputerchipsweretypicallydesignedbyOEMssuchasXerox,Kodak, and Canon. Manufacturing was then outsourced to any number offoundries.Despitemanyinnovationsintheimageprocessingniche,whenOEMschose

betweencomputerchipsforimagingdevicestheyhadtomaketradeoffsbetweenperformanceanddesignflexibility.Therefore,therewasapotentialopportunityfor anyone who could design chips that were both high-performance andflexible.

BUSINESSUNIT

ImageIllusionswasanentirelynewbusinessunitthathaddevelopedanimageprocessorcalled“Cactus” thatwasbothhighperformanceand flexible.Cactuscouldbesold tomanufacturersof“document imaging”devices(ex.photocopymachines) such as Xerox and Kodak. Long-term, Image Illusions hoped toexpandCactusintootherimagingmarketssuchasdigitalcameras,digitalvideocameras, and other computer-related imaging devices. Direct competitorsincluded a large number of entrenched chipmanufacturers in theUS, TaiwanandotherpartsofAsia.Nevertheless,ImageIllusionsfelttheycouldsellCactusatapremiumduetoitshighperformanceandsuperiorproperties.

TEAM

ImageIllusion’sfoundingteamincludedmanyhigh-techveteransincludingtheformermanagerofanadvancedimagingteamataFortune500technologyfirm.He was joined by a Hardware Architecture Design Manager with years ofhardwaredesignexperienceandleadership,andaSystemEngineeringManagerwho was Chief architect for a division that developed image processingalgorithms,systems,CMOSsensorsandadigitalimageprocessor.Their OperationsManager hadmanaged a department of 110 programmers

andanalysts inhigh-tech supplyplanning, and theMarketingManagerwasanElectricalEngineer fromCornellUniversitywith anMBA fromGeorgiaTechwhoseexperienceincluded8yearsoftechnologyproductmarketing.

FINANCIALS

ThemarketfordocumentimagingASICsandMediaProcessorswasexpectedtobebetween4–5millionunitsin2003,growingtoaround6millionunitsin2006.Of that totalmarket, Image Illusions expected to sell (in abase-case scenario)around140,000unitsin2003,rampingtoaround745,000unitsin2006.TheASP[averagesellingprice]foreachprocessorwasprojectedtobe$60in

2003, dropping gradually to $35 in 2006. Furthermore, gross margins wereprojectedtobeover200–300%.Therefore,inabase-casescenariothebusinessexpected around $8M in 2003 revenues, growing to over $26M in annualrevenues by 2006.Other projections put revenues at over $160M annually by2006,dependingonmarket sharegainsand industryconditions.Thebase-caseinternally calculated 4-year NPV was just over $8M, while more aggressiveprojectionsplacedthatNPVataround$100M.

APPENDIXC:

EXPERIMENTALMETHODANDSTATISTICALANALYSIS

Thebasicexperimentaldesignisrelativelystraightforward.Studentsaregivenaset of cases to make predictions on. They are then instructed in Disruptiontheory.Theyarethengivenasetofdifferentcasestomakepredictionson.Wecompare the results of the “before” and “after” predictions to assess whetherinstructioninDisruptiontheoryhasimprovedtheirperformance.Threedifferentexperimentswererun:oneatHBS,oneacrossMITandHBS,

andoneatIvey.Eachisdescribedhere,andadetailedstatisticalanalysisoftheresultsisprovided.

HBS

ClaytonChristensenhasbeendevelopingacoursebuiltaroundtheprinciplesofDisruption theory for over ten years. It is called “Building and SustainingSuccessful Enterprises,” or BSSE. It is a second-year elective course and hasbecome,overtime,themostpopularelectiveatHBS:over80percentofsecond-yearstudentstakeit,anditnowhassevensectionstaughtbyfiveprofessors.Inthisexperiment,webeganonthefirstdayofthecourse,providingtoeacha

Weblinktoasetofsixcaseschosenessentiallyatrandomfromapopulationofforty-five funded cases.We constrained the randomness by insisting that one-thirdofthecasesetshadnosuccesses,one-thirdhadonesuccess,andone-thirdhadtwosuccesses.ThismeantourexperimentaldatahadmoresuccessesintheoverallpopulationthanthesetofcasesfromIntel,whichincreasedtheabilityofour experiment to observe statistically significant changes in accuracy:with alowfrequencyofsuccesstheoddsofgettinganextremeresult—goodorbad—bychancealonegoesupconsiderably.Over the course of the next two days—and before their second class in the

course—students read the cases online, made their predictions, and had anopportunitytoprovidesomecommentsonthereasonsbehindtheirpredictions.We requested that students read andevaluate the cases independently and thattheynotdiscussthecasesortheirpredictionswithanyoneelse.The treatment in this experiment is taking the first-semester offering of

Christensen’sBSSEcourse,whichranfromSeptembertoDecember2009.Thecourse focuses on Disruption theory and does not employ the stereotypicalinductive case-study approach thatHBS has historically been known to favor.Specifically,eachofthetwenty-sixcoursesessionsisdesignedtodescribeandillustrateaspecificelementofDisruptiontheory.Thistypicallytakestheformofa reading (anarticleorachapter fromThe Innovator’sSolutionor someotherpublication), a lecture from Christensen, and a case written with the explicitintentofillustratinghowthatprincipleexplainsaparticularoutcome.

FIGURE39:RESULTSOFHBSEXPERIMENT

Those students still enrolled in the courseupon its completion inDecemberwere then sent a secondWeb link to a set of six different cases and asked tomakeasecondsetofpredictions,withexplicitinstructionstousetheprinciplesthey had learned in the course when making their calls. In order to avoidselectionbias,onlythosestudentswhostayedinthecoursefromSeptembertoDecemberwereincludedinthecontrolandtreatmentgroupsthatwereanalyzed.Herearetheresults,asoutlinedinfigure39:Thecontrolgroup,asexpected,didnobetterthanrandom,withafrequencyof

success thatwas no different from the underlying frequency of success in thepopulationofcasesets.ButafterthreemonthsofinstructionfromeitherClaytonorRayGilmartin,thestudentsweresimilarlynobetterthanrandomchance.Wewould need an alpha value of nomore than 0.1 to have a reasonable level ofconfidence that thestudentshaddonebetter thanchancealone,yet thecontroland treatment groups had alpha values of 0.51 and 0.48, respectively.Neitherwas there any difference between the two sections: Christensen’s andGilmartin’ssectionsdeliveredessentiallyidenticaloutcomes.

MIT/HBS

In thefallof2008Christensenwas teachingaseminaronDisruptiontheoryatMIT. Taking advantage of the beginning of this seminar, Thomas had theparticipantsactasthecontrolgroup.Eachparticipantpredictedtheoutcomesofsevencaseschosenatrandomfromthesamesetofforty-fivecasesthatweretobeusedthefollowingfallintheHBSexperiment.Shortlyafter the launchof theseminar,Christensen’sBSSEcoursebeganat

HBS. Upon the completion of the foundational module in “pure” Disruptiontheory,Thomashadtheparticipantsmakepredictionsonasetofsevenrandomlychosencases.Thiswasthetreatmentgroup.Ideally, of course, the control group and the treatment group are as nearly

identicalinmakeupaspossible.Here,theyconsistedofgroupsofMBAstudentsfrom different schools with very different reputations: MIT is known to bestronglyquantitativeandtoattractmore“engineering-minded”applicants,whileHBS tends to have a more qualitative, case-driven bent. It is difficult todeterminewhether thesestereotypesare true,orwhether theywouldmakeanydifferencetothisexperimenteveniftheywere,butitissomethingtoconsider.The treatment in this case consisted of the first six sessions of the BSSE

course,whichfocusonthe“nutsandbolts”ofDisruption—essentiallychapter2ofTheInnovator’sSolution. In thesesessionsconceptssuchas“overshoot”byincumbents, the“inferior”natureofDisruptive technologies, and themarkedlydifferent performance benchmarks—and hence organizational autonomy—appropriate for Disruptive new ventures are explored. In essence, these earlysessionscoverall(andonly)thematerialthatThomasdistilledintohisdecision-makingalgorithm.Hereishowtheydid,asillustratedinfigure40:

FIGURE40:RESULTSOFMIT/HBSEXPERIMENT

The control group does 2.1 percent better than chance alone, a result thatcoziesup to statistical significancewith alphavalue=0.12butdoesnotquitecleartheconventionalhurdleof0.1.Withasmalleffectsizeandatbestmarginalsignificance,theconclusionthattheoutcomeisnodifferentfromchanceseemsthe most reasonable one. The treatment group had a success rate that was,coincidentally,thesameasthatofthecontrolgroup,butsincethefrequencyofsuccess in the underlying population was lower this time, the result wasstatisticallysignificantlydifferentfromchancealone:the3.7percentincreaseinaccuracywasenoughtomatter.

IVEY

TheMIT/HBS experimentwas the first of the three experiments Thomas ran.The encouraging results of that effort led to the HBS experiment. The Iveyexperimentwas run inFebruaryof2010, less thanamonthafter the resultsoftheHBSexperimenthadbeenanalyzed.Thecontroland treatmentgroupsconsistedofsecond-yearMBAstudentsat

IveyenrolledinanOperationsManagementelective.ThestructureofthesetsofcasesstudentsevaluatedwasthesameasintheHBSexperiment.Studentswereallseatedinasinglelecturehallandgivenforty-fiveminutestoreadandmaketheir predictions on the six cases in the set assigned to them. Students wereaskedtodothisindividuallyandwithoutdiscussingtheircaseswithothers.Theroomwassupervisedwhiletheyperformedthistask.Afterthestudentssubmittedtheirpredictions,Ideliveredaforty-five-minute

lectureonDisruption,focusingonlyonthoseelementsofDisruptiontheorythatwere relevant to Thurston’s decision-making algorithm. I then instructedstudentstouseonlythisalgorithmwhenmakingtheirpredictionsonthesecondsetofcaseseachwouldreceive.Following a fifteen-minute break, studentswere given forty-fiveminutes to

read and evaluate a set of six cases, again with the request that this be doneindividually.Thiswasdoneinthesamesupervisedlecturehall.This was the cleanest of the three experiments. The control and treatment

groupsconsistedofallandonlythesamepeople.Participantsreadandevaluatedtheircasesentirelyindependently.Theyallreceivedexactlythesameinstructionindisruptiontheory,instructionthatwasdesignedtogivethemthesamesortsofinsights that Thurston had developed over the course of his self-study in thesubject.Thehopewasthat,despiteitsveryshortduration,itmightneverthelessbeatleastmarginallyeffective.Hereishowtheydid:

FIGURE41:RESULTSOFIVEYEXPERIMENT

TheIveystudentsdemonstratedverystronglysignificantimprovement,goingfromaneffectsizeandsignificancelevelthatwerealmostidenticaltothoseofthecontrolgroupatMITtoaneffectsize1.7percentagepointsgreater—ora46percentincrease—andalevelofstatisticalsignificancetwoordersofmagnitudegreater.

COMPARINGTHERESULTS

Theanalysisabovepermitsdirectcomparisonsofonlythecontrolandtreatmentgroupswithineachexperiment.The relativemagnitudeandsignificanceof theresults across the three experiments suggests a descending order of Ivey,MIT/HBS,HBS,butourintuitionsareeasilymisledinsuchmatters.Luckily,wecando someadditional statistical analysis to assess the results across all threeexperiments in amanner that controls for differences such as sample and theincidenceofsuccessintheunderlyingpopulationofcasesineachexperiment.Alogisticregressiononallthedatayieldsthefollowing:

FIGURE42:LOGISTICREGRESSIONONALLEXPERIMENTALDATA

TheomittedvariableinthisregressionisfortheMIT/HBSexperiment,soall

thecoefficientsshouldbeinterpretedintermsof therelativeimpactofagivenvariablecomparedtothatbaseline.However, theestimatesare“rawvalues”oflog odds of making a correct prediction in a particular treatment and areconditionalonthevaluesofalltheothercoefficients.ThatmakessortingouttheimpactoftraininginDisruptiontheoryinaparticularexperimentverydifficult—whichiswhythoseresultsarepresentedseparatelyabove.ThekeyhereistonotethatintheIveyexperimenttheparticipantswereless

accurateoverallthantheMIT/HBSparticipants(theIveycoefficientisnegative)buttheinteractiontermIveyxPSisstronglypositive.ThismeansthattheIveystudentswerebetteratpredictingsuccessesthantheMIT/HBSparticipants.Thisfindingisconsistentwiththemethodsusedinchapter2toestimatethefinancialimplicationsofthedifferenttreatmentsinvolved:themoduleonDisruptionleftthestudentsbetterabletopickthelosers,whiletheIveytreatment(aone-hourlecture)leftthestudentsbetterabletopickthewinners.ItistemptingtointerprettheHBSxPRinteractionterminasimilarway,but

in that case one needs to add in the three-way interaction (HBS x PS x TR),whichisnegative,alongwiththestronglynegativeHBSmainterm.The main conclusion supported by this analysis is that the ability to pick

winnerssuccessfullyafterhavingreceived training inDisruption theorycanberankedIveythenMIT/HBS,whiletheHBStreatmenthadnoimpactonstudents’predictiveabilities.

APPENDIXD:

DIVISIONALREVENUESGENERATEDBYDISRUPTIONATJOHNSON&JOHNSON

Determiningwhether or not a given innovationhas taken aDisruptive path tosuccess can be donewith different degrees of precision.A careful analysis ofcost structures, markets served, the drivers of performance improvement, andhowtheinnovationbreakstradeoffsamongdifferentdimensionsofperformanceis themost thoroughapproach.Butfor thepurposeofgettingaroughsenseoftherelativeimportanceofDisruptiontotherevenuegrowthofJ&J’sdivisions,itisperhapssufficientmerelytoassessthemajorproductsofeachdivision,theirinitialmarketsandperformanceprofiles,andhowtheyultimatelybecamemajorcontributorstorevenue.In its 10-K filings J&J breaks out revenue by division, and within each

division according to the products and product groups identified in the tablebelow. (There are occasionally changes in how revenue is reported, but thesechanges do not materially complicate this analysis.) Each product or productgroupwithineachdivisionhasbeentaggedas“Disruptive”or“sustaining,”andabriefexplanationgivenalongwithqualificationstothiscategorization.Wherethere is doubt, the “tie goes to the Disruptor”: the hope is to overstate theimportance of Disruption to the Consumer and Pharmaceutical divisions andunderstate it for Medical Devices and Diagnostics (MD&D), since this is astrongertestofthehypothesisthatDisruptionwillbemostimportanttoMD&D,thenConsumer,thenPharmaceuticals.

FIGURE43:TYPESOFINNOVATION-DRIVERGROWTHATJ&J

NOTES

PROLOGUE

1. Here is a reasonable bibliography of books that define, employ, andelaborateonDisruption:

Christensen, Clayton M. The Innovator’s Dilemma. Boston: HarvardBusinessSchoolPress,1997.Christensen,ClaytonM.,andMichaelE.Raynor.TheInnovator’sSolution.Boston:HarvardBusinessSchoolPress,2003.Christensen, Clayton M., Scott D. Anthony, and Erik A. Roth. SeeingWhat’sNext.Boston:HarvardBusinessSchoolPress,2004.Christensen, Clayton M., Michael B. Horn, and Curtis W. Johnson.DisruptingClass.NewYork:McGraw-Hill,2008.Anthony,ScottD.,JosephV.Sinfield,MarkW.Johnson,andElizabethJ.Altman. The Innovator’s Guide to Growth. Boston: Harvard BusinessSchoolPress,2008.Christensen, Clayton M., Jerome H. Grossman, and Jason Hwang. TheInnovator’sPrescription.NewYork:McGraw-Hill,2009.Anthony, Scott D. The Silver Lining. Boston: Harvard Business SchoolPress,2009.Dyer, Jeff, Hal Greyerson, and ClaytonM. Christensen. The Innovator’sDNA.Boston:HarvardBusinessSchoolPress,2011.Christensen,ClaytonM., andHenry J.Eyring.The InnovativeUniversity.SanFrancisco:Josey-Bass,2011.

2.MassimoPigliucci,NonsenseonStilts:How toTellScience fromBunk(Chicago:UniversityofChicagoPress,2010).Seeespeciallychapter2.3.Forexample,ifyouhaveahitrateof95percentIdoubtverymuchthattheapplicationofDisruptiontheorycanmakeit100percent.Butifthatisyour success rate, youdonotneed anyhelp!The specific rangeof “baselevelsofsuccess”towhichthesefindingsapplyisnotsomethingmydataspeak to. So, in the interests of conservatism, I suggest only that to theextent that you are doingmaterially better than 10 percent overall as aninvestor in a broad portfolio of de novo business start-ups, Disruption

theorywilllikelyresultinasmallerimprovement.4.PedroFerreira,“GeneralRelativity:TheDarkUniverse,”NewScientist,June30,2010.

CHAPTERONE

5.SeeMichaelE.Raynor,MumtazAhmed,andJamesGuszcza,“Survivalof the Fattest,” Deloitte Review, 2010. (Available atwww.deloittereview.com.)6.SeeHarvardBusinessSchoolcase“IntelNBI:IntelCorporation’sNewBusiness Initiatives (A),” No. N9–609–043, product number 609043,October6,2008.7.Thurstonisnottheonlyonefrustratedbythisstateofaffairs.Twobookshaveof late criticized thegeneral stateofpopularmanagement science—see Jeffrey Pfeffer and Robert I. Sutton, Hard Facts, Dangerous Half-Truths & Total Nonsense: Profiting from Evidence-Based Management(Boston: Harvard Business and School Press, 2006) and Philip M.Rosenzweig,TheHaloEffect:…and theEightOtherBusinessDelusionsThat Deceive Managers (New York: Simon & Schuster, 2007). In myconversationswithother researchers I find that they invariablyagreewiththeseagentsprovocateurs,whileinvariablyexemptingtheirownworkfromsuchcriticisms.8. Atul Gawande, The Checklist Manifesto (New York: MetropolitanBooks,2009),pp.162–70.

CHAPTERTWO

9.ThurstondidnotincludeIOL,theventureinwhichhehadbeeninvolved,in his sample. Neither did he include ventures that had been recentlyfunded, inorder togiveeachnewventurea“chancetofail.”Therewasamedian of just over seven years of data available for the ventures in theportfolioThurston created, although some ventures had failedwithin thattimeperiod.The“survivor”withtheleastavailabledatahadthreeyearsofoperationsunderitsbeltasof2009,whentheexperimentswereconducted.10. Even so, there remained some points of disagreement among someexecutivesatNBIregardingwhichventurescountedas“alive”or“dead,”“Disruptive”or“sustaining.”Wehaveusedhere thedeterminationsmadeprior to revealing the actual outcomes and assessing the impact ofDisruption theory on predictive accuracy. Any recategorizations made inlightoftheseresultscannotbeconsideredimpartial.11.Thereisaseeminglyunavoidabletemptationtoattempttoevaluatethis10.4percentnumber; is thatgoodorbad?Isimplydonotknow:noothercompanythatIamawareofhassharedsimilardata,andcertainlythereisnomeaningfulpopulationofcomparablecompaniesthathaveprovidedthekindofdata required tomakeameaningfulcomparison.Butperhapsyoucannot help yourself and you need some way to think about that 10.4percent number.Consider the following: this survival rate is for businessproposals that received SAM funding—something akin to “angelinvestment.”The later-stageBAMfundingmightbeanalogous toventurecapital investing. The long-run success rate for VC funds is generallyreported(bydintmoreofrepetitionthanofharddata,asfarasIcantell)tobe around 10 percent. Tracking the survival rate of a portfolio from anearlier,andhencemoreuncertain,stageofinvestmentnecessarilyimpliesasuccessratethatisequaltoorlessthanthatofVCinvesting(ifweassumeeveryinvestmentpassesthroughbothstages).Consequently,itseemstomethatthemostreasonableassumptionisthatIntel’s10.4percentsurvivalrateisexemplary.12.SinceinthisthoughtexperimenttheNBIexecutiveswouldbelookingat a portfolio of initiatives their “preneuralized” selves had approved,wecan expect that they would approve a far greater percentage of theinitiatives in this sample than in a sample of initiatives that includedproposalsthathadbeenrejected.

CHAPTERTHREE

13. Michael E. Porter, “What Is Strategy?” Harvard Business Review,November–December1996.14.Asadefinitionalaside,Iuse“businessmodel”asitisatermingeneraluse.ItakeittobesynonymouswithPorter’snotionofan“activityset”andmakethissubstitutioninwhatisotherwisearehearsalofPorter’sargument.15. S. C. Buttrick. “Southwest Airlines—Company Report.” Kidder,Peabody&Company,Inc.,November30,1989.16. Clayton C. Christensen and Michael E. Raynor, The Innovator’sSolutionop.cit.,chapter2.17. JodyHofferGittell,TheSouthwestAirlinesWay:Using thePowerofRelationships to Achieve High Performance. (New York: McGraw-Hill,2003). See also Kevin Freiberg and Jackie Freiberg, Nuts! SouthwestAirlines’ Crazy Recipe for Business and Personal Success (New York:Texere,2011).18.The737-200hadadirectoperatingcost(DOC)perrevenuepassengerkilometer (RPK)of justunder4¢.The747,DC-10,andL-1011allhadaDOC/RPK of about 3¢ given an appropriate range. See Joosung J. Lee,StephenP.Lukachko,IanA.Waitz,andAndreasSchafer,“HistoricalandFuture Trends in Aircraft Performance, Cost and Emissions,” AnnualReviewofEnergyandtheEnvironment26(2001):167–200.19. Competing jets offered wider fuselages and greater flexibility inconfiguration for things like galleys and classes of seating—features thatwerecrucialtoestablishedairlinesbutirrelevanttoSouthwest.20.Tomyknowledge,onlyoneincumbentairlinehasmanagedtolaunchasuccessfulLCCdivision.Qantas’sJetStarsubsidiarywasestablishedasanentirelyautonomousdivisionandhasbeenfreetodefineanutterlydistinctbusiness model. Consistent with the predictive framework described inchapters 1 and 2, it is this autonomy that has allowed it to survive andthrive.

CHAPTERFOUR

21.IsadoreSharp.FourSeasons:TheStoryofaBusinessPhilosophy(NewYork:Penguin,2009).22.GeorgeStalk,PhilipEvans,andLawrenceE.Shulman,“CompetingonCapabilities: The New Rules of Corporate Strategy,” Harvard BusinessReview,March–April1992.23. See Harvard Business School case No. 9–607–140, “Dollar General(A),”revisedNovember19,2007.

CHAPTERFIVE

24.EverettM.Rogers,Diffusionof Innovations,5thed. (NewYork:FreePress,2003).25. There is a never-ending efflorescence of work designed to helpmanagersacceleratethetransitionofaproductfromthebirthphasetothegrowth phase. Perhaps the most well-known body of work in this areacomescourtesyofGeoffreyMoore,e.g.,CrossingtheChasm(NewYork:HarperBusiness,1991).AsIreadit,Moore’swork,andmostothersinthisvein, tend to describe the conditions underwhich productswillmake theleaprather thanseekingtopredictwhensuchtransitionswillhappen.Theimpact managerial choices might have on the timing of successfulDisruptionwillbeaddressedlaterinthischapter.26.Itisworthaskingiftheseinsightswillremainvalidifeveryoneacceptsthemandactsuponthem.Fromtheperspectiveofcapital-marketinvestors,theywill not.Makingmoney in the stockmarketmeans buying the rightthingbeforeeveryoneelserealizesitistherightthingtoo.Andifeveryoneusesthesamepredictivemodel,everyonewillcometothesameconclusionat the same time. But if the history of ideas is any indication, the morenearly trueDisruption theory is, the longeryouwill beable to exploit itspredictivepowerbeforetherestoftheworldcatcheson.27.“IntelCorporationin1999,”StanfordBusinessSchoolcaseNo.SM-70.28.ConversationwithClaytonChristensen.29.TheflightdistancefromNewYorktoLondonisjustundersixthousandkilometers.NewYorktoHonoluluisjustundereightthousandkilometers.Both routes are entirely feasible in the 737-900with the necessary safetycertifications.

CHAPTERSIX

30.FreibergandFreiberg,op.cit.31. Evelyn TatumChristensen and Richard Tanner Pascel, “Honda (B),”HarvardBusinessSchoolcaseNo.9–384–050.32.This aspectof theSEDASYS™example is explored inmybookTheStrategy Paradox: Why Committing to Success Leads to Failure … andWhattoDoAboutIt(NewYork:Currency/Doubleday,2007).33.LetterfromtheAmericanGastroenterologicalAssociation(AGA)totheU.S.FoodandDrugAdministration,May15,2009.

CHAPTERSEVEN

34.T.R.Eisenmann,G.Parker,andM.vanAlstyne,“StrategiesforTwo-SidedMarkets,”HarvardBusinessReview84,no.10(October2006).35.SeeMichaelE.Raynor,TheStrategyParadox,op.cit.

EPILOGUE

36. JosephA. Schumpeter,Capitalism, Socialism andDemocracy (1942;NewYork:HarperColophon,1975),pp.84–86.37. Herbert K. Zassenhaus, “Capitalism, Socialism and Democracy: The‘Vision’ and the ‘Theories,’ ” inSchumpeter’s Vision, as quoted in “TheCreativeDestroyer:Schumpeter’sCapitalism,Socialism,andDemocracy,”reviewessaybyThomasK.McCraw,Project 2000:SignificantWorks inTwentieth-CenturyEconomicHistory.38.Schumpeter,op.cit.

TableofContentsCoverOtherBooksbyThisAuthorTitlePageCopyrightFOREWORD:FromArttoSciencePROLOGUE:TheFive-Percentage-PointSolutionPARTI:PredictionCHAPTERONE:AProblemofPredictionCHAPTERTWO:BetterbyHalfPARTII:ExplanationCHAPTERTHREE:HowCHAPTERFOUR:IfCHAPTERFIVE:WhenandHowLongPARTIII:ApplicationCHAPTERSIX:DeliberatelyDisruptiveCHAPTERSEVEN:InvestingforInnovationCHAPTEREIGHT:OrganizingforInnovationEPILOGUE:CreativeCreationACKNOWLEDGMENTSAPPENDIXA:DelimitationsandLimitationsoftheExperimentsAPPENDIXB:ImageIllusionsAPPENDIXC:ExperimentalMethodandStatisticalAnalysisAPPENDIX D: Divisional Revenues Generated by Disruption at Johnson &Johnson

NOTES