Where and how to search? Search paths in open innovation

(1)

Where and how to search? Search paths in

open innovation

Henry Lopez Vega, Fredrik Tell and Wim Vanhaverbeke

Linköping University Post Print

N.B.: When citing this work, cite the original article.

Original Publication:

Henry Lopez Vega, Fredrik Tell and Wim Vanhaverbeke, Where and how to search? Search

paths in open innovation, 2015, Research Policy, (45), 1, 125-136.

http://dx.doi.org/10.1016/j.respol.2015.08.003

Copyright: © 2015 The Authors. Published by Elsevier B.V. This is an open access article

under the CC BY-NC-NDlicense

http://www.elsevier.com/

Postprint available at: Linköping University Electronic Press

http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-124530

(2)

ContentslistsavailableatScienceDirect

Research

Policy

j ou rn a l h om ep a g e :w w w . e l s e v i e r . c o m / l o c a t e / r e s p o l

Where

and

how

to

search?

Search

paths

in

open

innovation

Henry

Lopez-Vega

a,∗

_,

_Fredrik

_Tell

b,c

_,

_Wim

_Vanhaverbeke

d,e,f

a_Department_of_Management_and_Engineering,_Business_{Administration,}_Linköping_University,₅₈₁₈₃_Linköping,_Sweden b_Department_of_Business_Studies,_Uppsala_University,_P.O._Box_513,₇₅₁₂₀_Uppsala,_Sweden

c_KITE_Research_Group,_Department_of_Management_and_Engineering,_Business_{Administration,}_Linköping_University,₅₈₁₈₃_Linköping,_Sweden d_Faculty_of_Business_Economics,_Hasselt_University,_B-3590_Diepenbeek,_Belgium

e_ESADE_Business_School,_E-08172_Sant_Cugat,_Spain

f_National_University_of_Singapore,_Faculty_of_Engineering,_Singapore_117575,_Singapore

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received3September2012

Receivedinrevisedform22August2015 Accepted23August2015

Availableonline22October2015 Keywords: Searchspace Searchheuristics Organizationalsearch Innovationintermediaries Problemframing Boundaryspanning

a

b

s

t

r

a

c

t

Searchforexternalknowledgeisvitalforﬁrms’innovativeactivities.Tounderstandsearch,wepropose twoknowledgesearchdimensions:searchspace(localordistant)andsearchheuristics(experientialor cognitive).Combiningthesetwodimensions,wedistinguishfoursearchpaths–situatedpaths,analogical paths,sophisticatedpaths,andscientiﬁcpaths–whichrespondtorecentcallstomovebeyond“whereto search”andtoinvestigatetheconnectionwith“howtosearch.”Also,wehighlighthowthemechanisms ofproblemframingandboundaryspanningoperatewithineachsearchpathtoidentifysolutionsto technologyproblems.Wereportonastudyof18openinnovationprojectsthatusedaninnovation inter-mediary,andoutlinethecharacteristicsofeachsearchpath.Explorationofthesesearchpathsenriches previousstudiesofsearchinopeninnovationbyprovidingacomprehensive,butstructured,framework thatexplainssearch,itsunderlyingmechanisms,andpotentialoutcomes.

1. Introduction

Organizational search is central to classic and contempo-rary innovation theories (Laursen, 2012; Nelson and Winter, 1982). However, while firms in search of “new combinations” (Schumpeter,1934)buildonaccumulatedexperience,theyarealso cognitivelyconstrainedbypreviouschoicesandresource commit-ments,potentiallyresultinginmyopia(LevinthalandMarch,1993) andhighR&Dexpenses.Segmentsoftherapidlyexpanding discus-siononopeninnovationhaverevisitedandrevitalizedtheroleof searchininnovation(c.f.AfuahandTucci,2012;FelinandZenger, 2014;Laursenand Salter,2006).Akey ideainopeninnovation isthat firmsshouldexploitsearchoutsidetheconfinesof their organization(c.f.Westetal.,2014),makingthesearchfor exter-nalknowledgeanimportantmanagerialtask(LaursenandSalter, 2006,p.147).Searchforexternalknowledgeisarguablyquite com-plexanddifficult,involvinguncertaintiesandcharacteristicssuch asthetacitness,complexity,rivalry,andindivisibilityof knowl-edgewhichmaynotbeconducivetoitsdetectionandtransfer(c.f. ZolloandWinter,2002).Despitethiscomplexity,searchhasbeen

∗ Correspondingauthor.

E-mailaddresses:henry.lopez.vega@liu.se(H.Lopez-Vega),

fredrik.tell@fek.uu.se(F.Tell),wim.vanhaverbeke@uhasselt.be(W.Vanhaverbeke).

analyzedprimarilybyusingone-dimensionalconstructssuchas localvs.distant(KnudsenandSrikanth,2014),whichseldom rec-ognizehowdifferentheuristicsinteractwiththesolutionlocation (NickersonandZenger,2004).

This paper explores the dynamics and direction of search. We suggestthatorganizational searchinvolvestwodimensions (GavettiandLevinthal,2000).Theﬁrstreferstowheretosearch, i.e.,thelocationofsolutions–localordistant–inrelationto cur-rentlyavailablesolutions(KatilaandAhuja,2002;Levinthaland March,1981).Thesecondconcernshowtosearch,andwhichsearch heuristicstoapply,i.e.,experientialorcognitivesearch(Gavettiand Rivkin,2007;NickersonandZenger,2004).Sofar,researchonopen innovationfocusesmostlyonwhere tosearchinagivensearch space(Garrigaetal.,2013;LaursenandSalter,2006;Piezunkaand Dahlander,2015),andseveralscholarslamenttherelativelysmall attentiongiventohowsearchtakesplace,andwhatalternative searchheuristicsareappliedinopeninnovation(FelinandZenger, 2014;JeppesenandLakhani,2010).

Combiningthe“where”and“how”dimensionsofsearch,we propose a framework for firms’search for external knowledge thatencompassessituatedsearchpaths,analogicalsearchpaths, sophisticatedsearchpaths,andscientificsearchpaths.Inpursuing thesesearchpaths,firmscanexploittwomechanismstoidentify solutionsinideaandtechnologymarkets:first,aproblemframing

http://dx.doi.org/10.1016/j.respol.2015.08.003

(3)

mechanism(Baeretal.,2013;Kaplan,2008;VonHippelandVon Krogh,2015)thatinvolvesfocusingonandarticulatingthe prob-lemasatechnologyneedbeforeitsdissemination,andsecond,a boundaryspanningmechanism(FlemingandWaguespack,2007; RosenkopfandNerkar,2001)thatinvolvesrecognizingand con-nectingthetechnologyneedtoa specificcrowdof technicalor scientificsolvers.Weaddressthefollowingresearchquestions:(1) Whatarethecharacteristicsandobjectivesofeachsearchpath? (2)Howdoproblemframingandboundaryspanningmechanisms facilitatetheidentificationofsolutions?

Our ﬁndings draw on18 open innovation projects tostudy howinnovationintermediaries(c.f.Chesbrough,2006;Roijakkers etal.,2014)helpclientsﬁndpotentialsolutionstotheir technol-ogyproblems,1 _by_selecting_a_search_path_and_exploiting_search

mechanisms.Ourresearchinvolvesanembeddedcasestudy con-ducted ata leading innovationintermediary-NineSigma. Anew breedofinnovationintermediaries(e.g.,NineSigma,InnoCentive, Yet2.com)is offeringservicestoassist ﬁrms intheirsearch for externalknowledge and intellectualproperty(IP) management. Wefocus onintermediariesthat facilitateconnectionsbetween ﬁrms(knowledge-seekers)pursuingsearchforsolutionsandideas intechnologymarkets,andaglobalnetworkofsolution-providers suchasR&Dlaboratories,universityfaculty,andspecialist compa-nies(Boudreauetal.,2011;JeppesenandLakhani,2010;Siegetal., 2010).

Ourfindingsmakeseveralcontributionstotheliterature. Theo-retically,weproposeasearchpathframeworktoclarifyandexplain complexsearchpatternsandchoices,andtoextendtheoriesinthe innovationliteraturethatbuildonthesearchframeworksuggested byMarchandSimon(1958).Wealsoproposetwonewtypesof search–analogicalandsophisticated–asimportantsearchoptions. Finally,weconnecttheproblemframing(Baeretal.,2013;Kaplan, 2008;VonHippelandVonKrogh,2015)andboundaryspanning (Fleming andWaguespack, 2007)literatures topropose mecha-nismsrelatedtothesolutionofproblemswithinthesesearchpaths. The paper is structured as follows: Section 2 presents the two search dimensions, discusses the search mechanisms, and describesthefourproposedsearchpaths.Section3discussesthe researchdesign anddatacollected.Section4describeshowthe searchforsolutionstoproblemsisassociatedwithourfoursearch paths,and examinestheuseofproblemframingand boundary spanningmechanismsin18openinnovationprojects.Section5 discusseshowtheselectionofaspecificsearchpathinfluencesthe identificationofsolutionstoproblems.Section6presentsourmain conclusions.

2. Literaturereviewandframework

2.1. Thesearchforsolutionstoproblems

Organizationssearchforalternativesolutionstoproblemswhen current routinesfail to produceresults that match the organi-zation’saspirations (Marchand Simon, 1958).The screening of alternativesolutionsand taskdecompositionare major compo-nentsof theproblem-solvingprocess (March andSimon, 1958, p.178).2 _For_cognitive_reasons,_{“problemistic} _search”₍_Cyert_and

March,1963, p.120–122) tendstobebothsimple-minded and biased,causingorganizationstosearchlocally, inthevicinityof

1 _Hereafter_referred_to_as_{“problems.”}

2 _March_and_Simon₍₁₉₅₈_,_pp._178–179)_also_discuss_randomness_and_the

hier-archicalstructureofproblem-solvinginsearch.Weacknowledgethatinnovation involvesmuchmorethanjustsearchforsolutionssinceitrequiresknowledge inte-gration,implementation,anddiffusion,marketacceptance,etc.butinthisstudywe focusonthesearchproblem.

alreadyidentified solutions.Levinthaland March(1981,p.309) describethisas“refinementsearch”,which“emphasizesrelatively immediate refinementsin the existing technology, greater effi-ciency,anddiscoveriesinthenearneighborhoodofthepresent activities.”

However, when a problem cannot be solved using current routines,theﬁrmisforcedtoinnovatebydevelopingnew knowl-edge. “Innovativesearch” (Levinthaland March,1981)includes distantsearchfornewtechnologies,basedonnewcombinations of knowledge(Carnabuci and Operti,2013; Schumpeter, 1934). Thesubsequentliteratureonsearchandinnovationinvestigates thepropertiesandoutcomesofreﬁnement-orientedlocalsearch (exploitation)vs.innovation-orienteddistantsearch(exploration) inmoredepth(Laursen,2012;March,1991).Also,theseanalyses focusonthelocationofalternativesrelativetocurrentbehavior and“theelementsthataretobesearched”(GavettiandLevinthal, 2000,p.114).Below,weshowthatthissearchproblemcenterson thequestionofwheretosearch.

2.1.1. Searchspace:wheretosearch?

Firmslookingforsolutionstoproblemssearchamong combi-nations ofknowledgein asearchspace(Knudsenand Srikanth, 2014).Howdoesthefirmknowwheretostart?Byenvisagingthe searchspaceastherelativedistancefromthefirm’scurrent knowl-edgebase,search maybelocal,i.e.,in thevicinityofthefirm’s currentknowledge,ordistant,i.e.,fartherawayfromthefirm’s cur-rentknowledge.Inpractice,knowledgecategoriesandknowledge combinationsneedtobedeterminedinadvance.Knowledge cate-goriescanberepresentedbytechnologicaldomains(e.g.,internal combustion,electronics,bioenergy,etc.),industry classifications (e.g., automobiles, consumer retailing, telecommunications), or scientificfields(e.g.,electromagneticwaves,particlephysics, opti-mization).However,itiscrucialthatthefocalfirmunderstands wheretheappropriateknowledgeis“stored”(e.g.,inindividuals, organizations,theories,patents,products,etc.)inorderto effec-tivelysearchforit.

Organizationsprimarilysearchintheproximityofexisting rout-ines and previoussolutions (Levinthal and March,1993;Stuart andPodolny,1996).Therefore,whenconductinglocalsearch, orga-nizations lookforsolutions thatbuildonknowledgealready in use. Although local search decreasesthe probability of finding novelsolutions,itincreasesthechancesoffindingandacquiring workable solutions. In contrast, distant search entails knowl-edgerecombination(FlemingandSorenson,2004;Rosenkopfand Nerkar,2001),whichmayprovideopportunitiestoidentify disrup-tiveinnovationsandachievecompetitiveadvantage.Buildingon Schumpeter’s(1934)seminalargument,knowledgerecombination andintegrationisaquintessentialelementofinnovativecapability (CarnabuciandOperti,2013).Distantsearchessentiallyinvolves thesearchforsolutions thatareunrelatedtothefirm’scurrent knowledgebase.However,organizationsoftenfilteroutsolutions basedondistantknowledge,preferringtoevaluatesolutionsfrom localknowledgesources(PiezunkaandDahlander,2015).

A mechanism that helpstobalance thelocal–distant search spaceisboundaryspanning.Althoughexploringwithinthe bound-aries of the firm’s technological domain may satisfy a specific technologyneed,boundaryspanninginvolvingadistant technolog-icaldomainhelpsidentifynewwaystosolveproblems(Rosenkopf andNerkar,2001).Mostfirmsemploymechanismsthatfacilitate theidentificationofshort-termsolutions,i.e.,localsearch,or poten-tiallonger-termbreakthroughs,i.e.,distantsearch(Hargadonand Sutton,1997).Understandingtheunderlyingsearchspaceisatthe heartoftheboundaryspanningmechanism(FlemingandSorenson, 2004),whichenablesinformationprocessing,interpretationand translationofknowledge,andnegotiationofcommonmeanings

(4)

amongheterogeneousparties andacrosscohesivetechnological boundaries(FlemingandWaguespack,2007).

InnovationintermediariessuchasInnocentiveandNineSigma areactorsthathavedeveloped specializedprocessestoconduct localanddistantsearchbyorganizinginnovationcontestsacross distinctknowledgeﬁelds(Roijakkersetal.,2014;Spradlin,2012). Theseproblem-solvingcontestsforgelinksbetween (knowledge-seeking)ﬁrmssearchingforexternaltechnologicalsolutions,and potentialtechnologicalsolution-providers(AfuahandTucci,2012; Chesbrough,2006).However,severalauthorsarecriticalofthe rela-tivelyscantattentionpaidtohowinnovativesearchtakesplaceand thealternativesearchheuristicsappliedinopeninnovation(Felin andZenger,2014;JeppesenandLakhani,2010).Below,wediscuss thisdimensionofsearchforexternalknowledgeasaquestionof howtosearch.

2.1.2. Searchheuristics:howtosearch

Thefocusinthecognitivedimensionofsearchisonthe heuris-ticsusedbyﬁrmstoevaluatealternativeswhenmakingdecisions (GavettiandLevinthal,2000,p.114),andtheprocessesinvolved (GavettiandRivkin,2007).Existinganalysesemphasizethe sepa-rationbetweenactionandcognition:“onepart[ofdecisionmaking] occursintheworldofcognitionandcompriseswaysof conceptual-izingtheﬁrmanditsenvironment.Theotherunfoldsintheworld ofactionandconsistsofmechanismsthatshapewhatacompany actuallydoes”(GavettiandRivkin,2007,p.420).

Experientialsearch is guidedby adesire fordirect feedback fromtrials(NickersonandZenger,2004).Animportantelement ofexperientialsearchisthe“on-line”evaluationofalternatives, andthesubsequentactiononimmediatefeedbackfromcurrent actionswithoutreﬂectingontheircauses(GavettiandLevinthal, 2000;ZolloandWinter,2002).Boundedlyrationalorganizations canactivateaccumulatedroutinesorprograms(CyertandMarch, 1963).NelsonandWinter(1982)consider“routinedevelopment” asexemplifyingexperientialsearchprocesses,oftendescribedas “learningbydoing”(Pisano,1994).

Cognitive search involves the use of representations and abstractionsinthesearchforsolutions(Gavetti,2012).Cognitive searchemploys“off-line”evaluation(GavettiandLevinthal,2000), and learning-before-doing (Pisano, 1994). Such representations contributetothedevelopmentofheuristicsofvariable sophisti-cation(Grandori,2013;NickersonandZenger,2004).Ratherthan allowingenvironmentalstimulitofeedbackdirectlyintoaction, cognitivesearchallowsforthefeed-forwardofinformationinto existingornascentrepresentations.Theseoff-linerepresentations canbestudiedandthenlaterenacted.Modelscanbedeveloped usingrepresentationsandsystemsofsymbolswhichallowcausal inferences which may be applicablemore generally (Zollo and Winter,2002).

Asearchmechanism facilitatingsearchheuristicsis problem framing,or problemformulation (Baeret al.,2013; VonHippel andVonKrogh,2015).Problemframingfacilitatesthe interpreta-tionofassumptionsandexpectationsrelatedtoaspecificproblem, andprovidesguidancefortheidentificationofpotentialsolutions (Kaplan,2008).Itcantaketheformofknowledgearticulationand codificationtoproducewrittendocumentationintheformoftools, manuals,etc.todescribenewproposalsandperformance implica-tions(ZolloandWinter,2002),takeplaceon-lineoroff-line(Gavetti andLevinthal,2000),andinvolveheuristicssuchascase-basedor deductivereasoning(GavettiandRivkin,2007).

Theopeninnovationliteraturesuggeststhatinnovation inter-mediariescan create valuefor knowledge-seekersby explicitly stating a “tacit” technologicalneed in a document that canbe usedtosolicitsolutionsininnovationcontests(Siegetal.,2010; Spradlin,2012).Thisrequirestheuseofsearchheuristicsto iden-tifywhetheraspeciﬁcproblemrequiresexperientialorcognitive

search(Boudreauetal.,2011).Ithasbeenarguedthatinnovation intermediariesaremoresimilartoconsultantsthanbrokers,since theyengagemoredirectlywithﬁrmsbydesigningandsupporting theinnovationsearchprocess(c.f.Howells,2006).

2.2. Aframeworkofsearchpaths

Previous research considers search in relation toeither the search space or search heuristics. We propose that these two dimensions should be studiedsimultaneously: the“where and how”shouldprovideanintegratedviewoftheﬁrm’ssearchfor externalknowledge.Weextendtheresearchonsearchthatuses one-dimensionalconstructs(e.g.Tippmannetal.,2013,p.1870). We combinethesetwo dimensionsin relationtofourdifferent searchpaths:(1)situatedpaths,(2)analogicalpaths,(3) sophis-ticatedpaths,and(4)scientiﬁcpaths.Fig.1depictsatypologyof themaincharacteristicsofeachsearchpath.

2.2.1. Situatedsearchpaths

Situated search is the default model in many evolutionary theories of organizations and innovation which emphasize the developmentofroutinesthroughlocaltrial-and-errorrefinements (e.g.Cyertand March,1963;Nelsonand Winter, 1982).A situ-atedsearchpathoperatesonthebasisofexperimentationinthe vicinity of previous solutions that lead tofeedback on current actions,andverificationofgeneralizedbeliefsbasedonrepeated experiences and observations. Empirically, situated search is a component of the firm’s innovative activity, as demonstrated in Martin and Mitchell’s (1998) study of theUS product mar-ket for magnetic resonance imaging devices in 1980–86. The authors showthat incumbent firms searched in the vicinityof the dominanttechnologies, and subsequently introduced prod-uct designssimilar toexisting designs. Comparably, Stuartand Podolny (1996), using a network approach, analyze strategic partnering and the evolution of the technological positions of Japanesesemiconductor producerspatentingbetween1982and 1992.Their findings showthat the majorityof these manufac-turersoccupieddistincttechnologynichesoverthetimeperiod considered.

2.2.2. Analogicalsearchpaths

Analogicalsearchcanbedescribedasdrawingupon experien-tialknowledgefromdistantdomains toinformcurrent actions. Such search is enabled through the use of analogical reason-ing (Gentner, 2002). Analogical reasoning involves structural comparison between a base and a target domain (often with unrelated content). Applying the characteristics of a solution that spans domains can provide newinsights tosolve current problems.

Analogicalreasoninghasbeendiscussedinthecontextof strate-gic decision-makingand innovationanalysis(Gary etal.,2012). For example,Hargadon andSutton (1997)discussIDEO’s expe-rience of innovation through technology brokering. Analogical searchpathsthatmake useoftechnologybrokeringinvolvethe blending by engineers of existing but previously unconnected ideas(seealsoCarnabuciandOperti,2013;HargadonandSutton, 1997).

2.2.3. Sophisticatedsearchpaths

We associatesophisticatedsearch pathswithdeductive rea-soning.Deductivereasoninggeneratespredictionsandhypotheses thatevolvefromgeneralpremisestospecificapplicationsofamore generalsetofscientificknowledge(GavettiandRivkin,2007).This isacentralcharacteristicofsophisticatedsearchpaths,inwhich explicitgeneraltheories feedforwardinto representationsthat make specificclaimsregarding (local) potentialstatesof affairs

(5)

Analogical paths Objective: Recombination Characteristics: Using experiential knowledge from distant fields to feed back on current actions

Adjacent search concepts and related studies: Boundary-spanning search (Laursen, 2012; Rosenkopf and Nerkar, 2001), Technology brokering (Hargadon and Sutton, 1997), Recombinant search (Carnabuciand Operti, 2013)

Situated paths Objective: Trial-and-error refinement Characteristics:Experimentation, in the vicinity of previous solutions, that feeds back on current actions

Adjacent search concepts and related studies:Refinement search (Levinthaland March, 1981), Exploitation (March, 1991), Local experimentation (Gavettiand Levinthal, 2000), Local search (Stuart and Podolny, 1996; Martin and Mitchell, 1998)

Sophisticated paths Objective: Puzzle-solving

Characteristics: Using established theories to derive predictions that feed forward to representations.

Adjacent search concepts and related studies:Deductive reasoning (Gavettiand Rivkin, 2007), Technological trajectories (Dosi, 1982), Path-deepening search (Ahuja and Katila, 2004)

Scientific paths Objective: Breakthrough

Characteristics: Creating new theories to derive predictions that feed forward to representations. These, in turn, make general claims regarding potential futures. Adjacent search concepts and related studies: Innovation search (Levinthal and March, 1981); Exploration (March, 1991); Scientific search (Fleming and Sorenson, 2004)

Se

ar

ch

spa

ce

Dist

a

nt

Search

heuristics

Experiential

Cognitive

Local

Fig.1. Frameworkofsearchpaths.

(Johnson-Laird,2001).Dosi(1982)referstosophisticatedsearch pathswithintechnologicalparadigmsas“technological trajecto-ries.”Asalientfeatureofsophisticatedsearchpathsishowthey operate to achieve “path-deepening” search (Ahuja and Katila, 2004).TripsasandGavetti(2000)tracesearchdeepeningthrough increasingsophistication of representations in theevolution of Polaroid’ssearchactivitiesindigitalimaging.Polaroidcontinuously investedlargesumsinR&Danddevelopedradicallynew technolo-gieswithinaclearsearchpattern,inwhichnewproducts–camera systemsforbothconsumerandcommercialapplications–builton thesedevelopmentsandappliedthisnewknowledgetoPolaroid’s existingstrengthsinmanufacturing.

2.2.4. Scientiﬁcsearchpaths

Scientific search paths allow the discovery of theories and models which give rise to predictions that then feed forward intorepresentations.Theserepresentationsinturnmakegeneral claimsregardingpotentialfutures.Knowledgeacquisitionthrough exploratory(March,1991) or innovative(Levinthal and March, 1981)searchhasbeendescribedas“searching”inscience.Ahuja andKatila (2004, p.891) studytheacquisition oftechnological capabilitybyUS-basedchemical firmsfrom1979to1992.They showthatwhenthesearch spaceforrecombinationof existing technologicalsolutionsisexhausted,scientificsearchcanextend thesearch space through the incorporation of new theoretical buildingblocksandhypotheses regardingcause-and-effect rela-tionships.Similarly,FlemingandSorenson(2004,p.911),intheir analysisofcitationsinUSpatentsin1990,suggestthatscientific searchpaths“[provide]inventorswiththeequivalentofamap–a stylizedrepresentationoftheareabeingsearched.”

3. Dataandmethods

3.1. Sampleselectionandresearchsetting

TheselectionofNineSigma3 _was_based_on_an_initial_study_of

thebusinessmodelsof11ofthelargestinnovationintermediaries: InnoCentive, NineSigma, Yet2.com, TekScout, IdeaConnection, YourEncore, Innoget, BigIdea Group, InnovationXchange, Creax, and Ocean Tomo. We interviewed representativesfrom fourof theseintermediaries,andcollectedsecondarydatafortheother seven. The NineSigma intermediation processes are similar to those of other innovation intermediaries. However, in contrast toInnoCentiveandYet2.com, thebusinessmodel usedby Nine-Sigmadoesnotrequiresolution-providerstoselltheirIPoutright, andallowsbothpartiestonegotiatecontractualagreements,e.g., licensingand co-development.Moreover,the NineSigmamodel involvesknowledge-seekersreceivingresponsesandcontact doc-umentation from all solution-providers, not just the winning solution-provider.

NineSigma’s headquarters are in Cleveland, Ohio, with sub-sidiaries in Belgium, Japan, Korea, Australia, South Africa, and Brazil.Sinceitsfoundationin2000,NineSigmahasadvisedaround 350companies–primarilyFortune500ﬁrmsandlarge multina-tionalenterprises(MNEs)–engagedinresearchandnewproduct launches, arranged over 2500 open innovation projects, and receivedover35,000 uniqueproposals fromsolution-providers. NineSigmahasanexternalinnovationnetworkofover2million

(6)

solution-providersdistributed across16industry sectorsin 115 countries. The technology suppliersinclude global multi-sector businesses(ca.52%),universityandgovernmentlabs(ca.34%),and inventorsandconsultants(ca.14%).

3.2. Datacollection

Thedataforthiscasestudywerecollectedviainterviewswith NineSigmaprogrammanagers(PMs)andinnovationand corpo-ratemanagersfromknowledge-seekingfirms,andanethnographic field study conductedat NineSigma’s headquarters. We carried out 34 semi-structured interviews(see Appendix 1), each last-ingover an hour. Ofthese, 12 werewithNineSigma corporate managers and PMs to obtain a description of the intermedia-tionprocessandsearchpathcharacteristicsandmechanisms.On average, PMs had worked at NineSigma for sevenyears, a sci-entific background (PhD degree), and practical familiarity with industrial product-development processes. The other 22 inter-viewswith18knowledge-seekerrepresentativesfocusedonopen innovationprojectsandhowNineSigmamanagedthem.All inter-viewswererecordedandtranscribed.Intervieweessharedarchival information, e.g., diagrams and charts, to increase our overall understandingofthevariousinteractionsanddecisions.

Moreover,weconducted anethnographic studytohighlight path-relatednuancesthatintervieweesmightbeunableor unwill-ingtoshareininterviews.Fortwomonths,thefirstauthorwasa non-participantobserverofinnovationcontestprojects,wasparty toconfidentialconversations,andinteractedwithemployeesfive daysaweek.Thisauthortooknotesduringalloftheseinteractions, andlaterdiscussedthemwithNineSigmaprojectmanagers. 3.3. Analyticalapproach

We entered thetranscribed interview and observation field notes,videos,andarchivalinformation intoAtlas.tisoftwarefor qualitative analysis.Thisprocedureassistedin thedata organi-zationandcoding, andfacilitated analysisandinterpretationof specificactivitiesandmechanismsforeachsearchpath.We fol-lowed Strauss and Corbin (1998), interpreting and coding the intrinsicmeaningofthefoursearchpathsandspecificmechanisms. Weappliedtwotypesofcoding:openandaxial.

Thedataanalysiswasconductedintotwosteps.First,weapplied opencoding tothe information collectedfrom observations of theinteractions betweenknowledge-seekers andNineSigma,to distinguishthefour search paths.In thesecond step, we open coded theinterviews and observations ofinteractions between knowledge-seekersandNineSigmainordertoempiricallyidentify twogenericmechanisms:problemframingandboundaryspanning (seeSection4.1).Finally, weconductedaxial codingtoconnect each search path to a specific mechanism (see Section 4.2). In linewith previousresearch onmechanism-based explanations, we wereinterested in explicit and detailed empirical evidence tosupportourcausalgeneralizations foreach pathand mecha-nisminvolved.Usingthetypologyofsocialmechanismsproposed byHedströmandSwedberg(1998,p.22),weanalyzedthe prob-lemframingmechanismasanaction-formationmechanismthat remainsattheknowledge-seekerlevel,sinceitshowsaspecific combinationofcharacteristicsforeachtechnologyproblem.The boundaryspanningmechanismisatransformationalmechanism thatdisseminatesaninternallyspecifiedtechnologyneedtoa net-workofsolution-providers.

4. Searchpathsandsearchmechanisms

Section4.1describestheoverarchingpropertiesand relation-ships of problem framing and boundary spanning. Section 4.2

explainstheobjectiveofeachsearchpath,andhowproblem fram-ingandboundaryspanningoperateineachofthefoursearchpaths andhelpachievethepathobjective.

4.1. Searchmechanisms:problemframingandboundary spanning

Ineachsearchpaththemechanismsofproblemframingand boundaryspanningareusedtoobtainsolutionsfromideaand tech-nologymarkets.Byproblemframing,theinnovationintermediary canuseitsaccumulatedexperienceto(a)analyzethe knowledge-seeker’sproblem, (b) frameproblems in collaborationwiththe knowledge-seeker,and(c)defineaspecifictechnologyneed.By boundary-spanningtheinnovationintermediarycanlinkthe spe-cifictechnologyneedtoexternalsolution-providersinpreviously unidentifiedtechnologicalandscientificareas.ANineSigmasenior managerdescribedthesemechanismsasfollows:

Wetakeproblems.Wetakethemapartintoidentifiablepieces, butnotnecessarilyintothepiecesastheywouldbeapplied.So, wetaketheapplicationoutofitandlookatthepurescienceand thenwegoandidentify.Anotherofourcapabilitiesisthatwe canidentifypotentialproblemsolvers.So,itisnotpassive,itis notpostingonachat,itisnothavingawebsitefullofexperts whoaccepteverychallenge;we lookforspecificcapabilities relatedtoeverychallenge...We canarticulatetheneed,we canpushitouttofindwherethesolutionsmayexistandthen wehaveaprocessforbringingthosesolutions–intheformat ofaproposalperhaps–andthetwopartiestogether.

Inopeninnovationprojects,problem-framingrefersto articu-latingaknowledge-seeker’sproblemintoatechnologyneedbefore revealingittoexternalscientificandtechnologicalcommunities, i.e.,solution-providers.Thetechnologyneediscontainedina docu-ment,i.e.,arequestforproposal(RFP),whichallowstheinnovation intermediarytoassisttheknowledge-seeker’steamtoseparatethe problemintodecomposable parts.The objectiveis tohighlight, e.g.,businessopportunities,technologicalspecifications,possible approaches,IPspecificities,limitationsofthetechnology,project timing,andfinancialandevaluationcriteriatoobtainclear,concise, andcompellingsolutionstotheproblem.Knowledge-seekersuse cross-orintra-unitmeetings,committees,andtaskforcestocraft thelanguagethatmaybalancetheneedforcontextspecificityand thegeneralityoftheproblem.Thesemeetingswereaimedat ensur-ingtherequisitecommunication,collaboration,anddelegationof responsibilitieseithertodefineproblemsconcretelyortodetach themfromtheircurrenttechnologicalcontextsandinformational content.

Inaddition,innovationintermediariesuseboundaryspanning toidentifyand selectpreviouslyunidentifiedsolution-providers fortheknowledge-seeker.Priortobroadcastingthealready spec-ifiedtechnologyneed,theinnovationintermediaryisresponsible formatchingknowledge-seekersand solution-providersin local or distant search spaces. This involves matching technological domainsandpotentialareasofapplicationwithaspecificcrowd ofsolution-providers4 _not_known_to_the_{knowledge-seeker.}_The

innovationintermediaryidentiﬁesanetworkofsolution-providers thatmightbeinterestedin,andcapableof,respondingtothe spe-ciﬁctechnologyneed,butavoidstoomanyandtoowide-ranging responses.

Boundaryspanningincreasesthenumberofsolution-providers andtheirinterestintheproblembyclarifyingthebeneﬁtstobe

4_NineSigma_has_built_its_contact_database_over₁₁_years;_it_is_{cross-referenced}

andusedtocategorizesolution-providersaccordingtoexpertise.Eachtechnology needisbroadcastedtoapproximately15,000potentialsolution-providers.

(7)

Analogical paths

Objective: Search for medium-tolong-term solutions from technologically unrelated and previously unexplored domains

Expected type of innovation: New processes or technologies to improve

existing specifications

Characteristics: Operational innovation

Path selection criteria: Market creation, timescale of one year,

implementation in multiple products, involvement of multiple business

units

Examples: Processes for delaying the drying of water-based emulsions

(Sherwin Williams); innovative packaging to minimize the melting of

chocolate bars in warm climates (Kraft Foods); redefinition of a bed

mattress (Sealy); Alternative and more-efficient uses of copper, e.g., in

healthcare (International Copper Association); packaging products from

the military and oil industries (Ferrero Rocher); problems with optics and

photonics (Carl Zeiss); innovative methodology (BP)

Situated paths

Objective: Search for quickly implementable solutions to problems in

familiar technological fields

Expected type of innovation: Add-on technologies or information

Characteristics: Defensive innovation, quick win, operational tactics

Path selection criteria: Application inexisting technology problems and

markets, timescale of months, implementation in a single product

Examples:a natural-based sunscreen and aluminum-free deodorants

(Natura Cosmetics), new membrane technologies for a tire and rubber

company (Goodyear), technology efficiency improvements to cope with

government regulations (RheemManufacturing),newly proven

technologies for a healthcare manufacturer (Kimberley-Clark)

Sophisticated paths

Objective: Search for short-to-medium-term insights into visible market

and industry trends

Expected type of innovation: Additional added-value technologies for

existing products

Characteristics: Solutions for a market-entry opportunity, proof of

concept, pre-launch, consumer involvement

Path selection criteria: Application in existing markets; timescale one

year; implementation in multiple products

Examples: A foam component for electric shaving machines (Philips),

alternative use of roofing granulates in asphalt shingles (3M); alternative

chemistry methods to increase the sustainability and reduce the toxicity of

paints (AkzoNobel), new technology to augment physical greeting cards

with digital experiences (Hallmark Cards)

Scientific paths

Objective: Search for long-term breakthrough solutions in unrelated

scientific fields to problems that require either the recombination of

knowledge or the discovery of a distant scientific network

Expected type of innovation: Disruptive technologies, new business models

Characteristics: Strategic insights, ideation results

Path selection criteria: Application in new markets, timescale in years,

implementation over multiple business units

Examples: A washing machine that does not require water (ArçelikA.Ş.),

alternative sources of sodium for potato chips (PepsiCo), substitutes for

formaldehyde in hair products (L’Oreal)

Se ar ch sp ace Dist a nt Search heuristics

Experiential Cognitive

Local

Fig.2. Examplesofsearchpaths.

derived fromworking ontheparticular problem, and by solic-itingfor solutionsthat complement orsubstitute existingones. Moreover, contact with speciﬁc solution-providers allows the knowledge-seekertoestablishlinkswithindividuals,groupsand associations representing particular scientiﬁc and technological domainsorgeographicalareas,andenablesthedecisiontocontinue orabandonaproject.

4.2. Searchpaths

Based on the study of 18 different problems submitted by knowledge-seekers,Section4.2describeseachofthefoursearch paths.Wediscussthespeciﬁcobjectivesandcontingenciesrelated toeachsearch path.We thendescribehow themechanismsof problemframing and boundary spanning play out acrosseach search path.Since each search path sharesat least one search dimensionwiththeothersearchpaths,e.g.,situatedpathprojects will share a search space dimension with sophisticated paths, andsearchheuristicswithanalogicalpaths,theknowledge-seeker might select the wrong search path and the problem remain unsolved.Wethereforedescribethemaindifferencesbetweenthe describedpathandthetwoalternatives.Fig.2depictsthe differ-encesamongthesearchpathsandthestudiedcompanycases,and Table1presentsthecharacteristicsofthetwosearchmechanisms. 4.2.1. Situatedpaths

Theobjectiveofsituatedpathsistosearchforquickly imple-mentablesolutions toproblems in familiartechnologicalﬁelds,

whileapplyingknowledge-seekers’accumulatedexperiencetothe problem.Frequently,situatedpathproblemsinvolveincremental orminoradaptations,e.g.,operationaltactics,defensive innova-tion,orquick wins, whichhave specificprojectobjectivessuch asmethodsormaterials.Situated pathproblemsaredifficultto solvebecauseoftheirstrictspecificationswithregardto complex-ity,technologicalmaturity,viability,and narrowsetofpossible solutions.Suchproblemsrequireshort-termsolutionsthatcanbe implementedquickly,e.g.,within12weeks.Situatedpath prob-lemsin ouranalysisinvolvedcollaborationbetweenNineSigma andtheknowledge-seekerinordertoselectandspecifythe prob-lem,andevaluatetheproposedsolutions.Situatedpathprojects arepresentedinFig.2.

Anexampleofaknowledge-seeker’sproblemfeaturesNatura Cosmetics,whichwaslookingforanaturallybasedultraviolet(UV) absorbertoreplacepetrochemicaloilsinasunscreen.IntheNatura example,thesunscreenproblemwassolvedbyasolution-provider operatinginthelaundrysector.AseniorR&DmanagerfromNatura explained:

Itwasanavailablesolution,buthadnotbeenappliedinthe cosmeticandpersonalcareindustry.Nobodyhadthoughtabout thatkindoftechnology.Whenwelookforsolutions,sometimes wemisssolutionsthatwewerenotthinkingabout.Notexactly whatweweresearching,butarelatedtechnology,whichcould changetheproject.

Theidentiﬁcationofsolutionsrequirestheuseofproblem fram-ingandboundaryspanningmechanisms.First,problemframing

(8)

Table1

Howsearchmechanismsinﬂuencesearchpaths.

Problemframingmechanism Boundaryspanningmechanism

Objective:Separatetheproblemintodecomposableparts,andcrafta statementoftheprobleminlanguagethatbalancestheneedfor generalitywiththatforcontextspeciﬁcity

Objective:Matchthesearchscopeandaspeciﬁccrowdof solution-providerstoaspeciﬁcsearchpath

Outcome:Speciﬁesthetechnologicalorscientiﬁcproblemasa technologyneed,andstatesthisneedinwriting,i.e.,anRFP

Outcome:Establisheslinkswithscientiﬁcgroupsinnewtechnological domainsorgeographicalareas

Searchpaths Effectontheinterpretation ofassumptionsand expectationsrelatedtoa speciﬁctechnologyproblem

Boundaryconditions Effectonthebalancingofthe searchspacetoidentify short-termsolutionsor long-termbreakthroughs

Boundaryconditions

Situatedpaths Thelevelofabstractionand codiﬁcationisdeterminedto attractsolversofrelated technologicalﬁelds

Criteriaandspeciﬁcationsare ﬂexibletoavoid

disappointmentwithreceived solutions

Theidentiﬁcationof solution-providersinother geographicalareaswith implementablesolutions becomespossible

Solution-providershave experience,i.e.,theyhave alreadysolvedrelated problems.Ideagenerationis excluded

Analogicalpaths Theraisond’etreoftheproblem isunderstoodand“translated” intoalanguageother industriescanunderstand

Domain-speciﬁcapplications areunderstoodinordertolater identifyopportunitiesfor recombination

Technicaltermsthatare commontoindustriesand areasacrossknowledge-seeker boundariesareselected

Boundariesandpotential misunderstandingsaremade cleartosolution-providers

Sophisticatedpath Thescientiﬁcproblemis writtenusingacompeting problemformulationtoallowa scientiﬁcexplanationofthe problem

Theproperﬁltersareapplied toreceivereasonable responsesandnotmissouton opportunities—adifﬁculttask

Asmallnetworkof

solution-providersindifferent scientiﬁcﬁeldsaretargeted

Areducednumberofsolvers, i.e.,anexpertecosystem, comesfromdiverseareas

Scientificpaths Thegenericscientificneedis detailedandtheparticular expectationsarespecified usingacompetingproblem formulation

Contentinvolvescorporate, technicalandresearcher perspectives

Academicallyrigorous scientistsinbroadscientiﬁc areasaretargeted

Searchincludes

solution-providerswithout off-the-shelfsolutions

involvescodifying a technologyneed, and describing it sothat itappealstosolution-providersin thevicinityofprevious tech-nologicalsolutions to feedbackinto or solvea problem. It also involvessomelevelofknowledgecodiﬁcationandabstractionto engagesolversinadjacentindustries.Natura’sseniorR&Dmanager explained:

Thedescriptionisanimportantsuccessfactorbecauseifyou donotdescribeitwell,youwillhaveaproposalthatisalready beingusedinthemarket.Howyoudescribeitandthepossible approachesshouldnotlimitthese[proposedsolutions]soyou canreceiveapproachesnotpreviouslythoughtof.

Itis alsoimportantnot todeﬁnethetechnologicalneedtoo narrowly.ANineSigmaR&Dmanagerexplained:

Infact,it[toonarrowtechnologyproblem]onlycreates disap-pointmentbecauseexpectationsarehighandthecriteriaand specificationsaretoostrict,andifyouhave10specifications andyousatisfy9butnotthelastone,youhavenotfilledthem all.Andbydoingthatallyoudoiscreatedisappointment. Theboundaryspanning mechanisminvolvesidentificationof solution-providerswithexperienceinsolvingproblemsfrom simi-lartechnologicaldomainsthatareabletoproviderefinedsolutions. AseniormanagerfromNineSigmaexplainedhowhehaddesigned aspecificnetworkforasituatedpathprojectinanaturalsciences firm:

Whenyougetintolifesciencesorchemistrythatisveryfocused, usuallythesolutionsdonotcomefromengineeringorother areas.Inthosecases,wewillfindpeoplewhoareworkingin thosespecificfieldsbutareunknowntoourclients,incountries suchasIndiaandChina.Wetrytofindpeoplewhohavealready solvedtheproblem,andarenotjustgeneratingideas.Infact, wetendtostayawayfromideageneration.Wefindourclients havemoreideasthantheyknowwhattodowith,forthemost part.

Situatedpathsandanalogicalpathsdifferinthespecificationof problemsandidentificationofsolution-providers.Anexampleof ananalogicalpathisKraft’ssearchforaninnovativepackagingto minimizethemeltingofchocolatebarsinwarmclimates.While solutionstosituated pathproblemscomefromlocaltechnology fields,analogicalpathproblemsrequiredistanttechnologyfields tobesolved,e.g.,thoseinvolvingproductsaffectedbytemperature orlight.AninnovationmanagerfromKraftgaveanexampleofthe analogicalpath:

Iﬁndthem[NineSigma] bestattechnologydevelopment,in areaswherewearelookingforsolutionsthatisnotready avail-abletousthroughournormalchannels.So,wearelookingfor ideas,forsomeonewhohasanapproachtosolvingtheproblem. Itisnotshort-termatall;weusuallyuseNineSigmafor medium-tolong-terminitiativesandstrategicinitiatives.Rightnow,we arerunninginmeltingchocolate;itisdirectedtoabusinessunit butitisadifﬁcultproblemtosolve.Itisnotsomethinganyone knowshowtosolveatthispoint.

Sophisticatedandsituatedpathsdifferintheselectionof prob-lemsthatneedtobesolvedusingdistinctconceptsandtheories. Anexample ofa sophisticatedpath isAkzo Nobel’sproblemof ﬁndinganalternativechemicalmethodtoimprovethe sustaina-bilityandsafetyofpaint.AnAkzoNobelmanagerhighlightedthe characteristicsofsophisticatedpathproblem:

Oneareaislookingatenvironmentalorsafetyareas:theseare thingsaboutpaintthatarecurrentlynotsustainableornotsafe, orcouldgeneratesometoxicologyproblemsinyearstocome. Wesawtheproblembeforeitwasrecognizedpubliclyand legis-lationcameintoplacetosay,“Youcannotuseleadasachemical ingredient.”

4.2.2. Analogicalpaths

Ananalogicalpathis usedtoidentifynewmedium-to long-termprocessesortechnologiesfromtechnologicallyunrelatedand

(9)

previously unexploreddomains, which can berecombined and implementedtoresolvethepredeﬁnedproblem.Complex prob-lemsrequiringananalogicalpathtotheirsolutionattractgreater interestandsupportfromtopmanagementsincetheyarelikely tobeaimedatentrytonewmarketsandexploitationinmultiple products.

In Fig. 2, we identiﬁed the projects that employed analogi-cal search paths. The example of Sherwin William’s analogical search path to delay the drying process in water-based emul-sionsisillustrative.Thisoperationalproblemaroseasaresultof newgovernmentregulationsrequiringﬁrmstoreducethe con-tentofvolatileorganiccompounds.Sincetheregulationaffected anumberofindustries,thesolutionmightpotentiallycomefrom unrelatedindustriessuchasfoodorpaper,orarelatedindustry suchascosmetics,andmightinvolvenovelprocessesormethods suchasdelayedwater-retentiontechnologies.Thesolutioncriteria includeduseofenvironmentallyfriendlyandnon-toxicmaterials thatwouldbesafeforusers.

Selectionofthisspeciﬁcproblemrequiredtheinvolvementof aSherwinWilliamstechnologyscoutresponsibleforcoordinating prioritiesandresourceavailabilityfromdifferentbusinessunits, andmakinguseofmanagementtoolstodevelopaninternal, non-conﬁdentialneedslist.Inrelationtothedelayeddryingprocesses problem, an open innovation manager from Sherwin Williams explained:

Openinnovationisreallyaboutgoingoutandidentifying solu-tionstothoseideasthatyouhaveneverthoughtabout,from specificareas.So,Ithinkitisimportanttohavethecapabilityto beabletodefinebroadlywhatyourproblemis.Letmetellyou thatwewentbackandforththroughninerevisionsfrom dif-ferentRFPs,andevenmore.So,wearenottoosegregated.We want[solution-providers]tounderstandwhattheproblemis, butwedonotwanttodefineitinawaythatpeoplewillpasson itbecausetheydonotunderstandtheterminologyintheirarea. Youneedtohaveaclearunderstandingofwhatyourproblem isandbeabletodefineitinbroadenoughtermsthatyoucan gointomultipleareasandtothefundamentals.

Asmentionedbythisinterviewee,theproblemframing mech-anismentailedcollaborationbetweeninternaltechnologyscouts andaNineSigmaPMinordertodocumenttheproblemintheform ofatechnologyneedinordertomakeitunderstandablebyother sectorsandindustries. Itrequiredanunderstandingof domain-speciﬁcapplicationstoidentifyopportunitiesforrecombination. Forexample,SherwinWilliams’sknowledge-seekersdonothave routineinternalprocessesinplacethatallowthemtoconnecta problemtodistanttechnologyareas:

We areacoding company;Iknowhowtosaysomethingin my pinkcodinglanguage.Butthat couldlimittheresponses fromothersinthecosmeticsoroilindustrythatcansay“Ihave atechnologythathasthatfundamentalissue.”Gettingitback intothosefundamentaldeﬁnitionshasbeenastruggle.Ithink NineSigmapushestheirclients,sotheyreceivemoreresponses andsuccessfulresponses.Ontheﬂipside,itishardformeasa clienttowriteanRFPthatway.

Theboundaryspanningmechanisminvolvesidentifying oppor-tunities in distant technological domains or industries, and selectinganetworkofsolution-providerstoconnecttounrelated technologydomains.Toavoidpotentialmisunderstandingsacross technologicaldomains, boundary spanning includes the use of technicaltermscommontoseveraldifferentindustries.A Nine-Sigmamanagerexplained:

So, NineSigmatriestolookatproblemsobjectively:a“why” viewofthings.Weareabletotranslatethatneedintoalanguage

thatpeopleinotherindustriesmaybeabletounderstand,and wehavethiscapabilitytobeabletobroadcasttheseneedsto peopleweresearchontheInternet,databases,etc.

Analogical path and situated path problems differ in the dimensions of their search space, i.e., local or distant. While knowledge-seekersusean analogicalpathtoidentifynew pro-cessesandtechnologiestoenternewmarketsorimproveexisting products,asituatedpathisusedtosolveanarrowandlatent tech-nologicalprobleminonespeciﬁcproduct.Goodyear’ssearchfor newmembranetechnologieshighlightshowasituatedpath nar-rowsdownthepossiblesolutions:

WhatIwaslookingforwasmembranetechnologies,andIdid notknowwhothemainplayerswere,orthemaintechnologies. IwantedtousetheRFPtounderstandaboutothertechnologies thatwereoutthere.Iwantedtoexploreothertechnologiesthat couldreducemycosts,majorplayers,anythingelseoutthere, andputpressureonourinternalchemicalengineersforother possibilities.

Analogicalandscientificsearchpathsinvolvecross-unit iden-tificationandselectionofproblems,i.e.,heterogeneousteamsto frameatechnologyproblem.Analogicalpathsareusedtoidentify newprocessesormaterialstoimproveanexistingtechnologyor process.Scientificpathsareusedtosearchforradicalsolutionsthat maychangetheknowledge-seeker’sbusinessmodel.Anexampleof ascientificpathisL’Oréal’ssearchforasubstituteforformaldehyde initshaircareproducts.Anopeninnovationmanagerexemplified thescientificpath:

Wehaveaproblemandwetrytoworkthroughittosearchfor newideas,newsolutions,newconceptswhichcouldanswer thatproblembutalsocouldbeappliedtransversallyacrossthe company.Bydoingthat,wehaveaprocessofinspirationand creativity.

4.2.3. Sophisticatedpaths

Sophisticatedpathsareusedtosearchforshort-to medium-terminsights into visible market and industry trendsthat will add value to existing products.The characteristics of expected solutionsincludeapplicationinexistingmarketsandinmultiple relatedproducts:proofofconcept,pre-launchphase,and prospec-tiveinnovations.Asophisticatedsearchpathcanidentifypotential solutionsthataddvalueintheformofimprovementtoanexisting productorimprovedﬁrmperformance.

Someexamplesofproblemsresolvedbyusingasophisticated patharepresentedinFig.2.Anexampleisthefoamcomponent inPhilips’selectricshavers,whichwasatechnologyareainwhich Philipshadnoexperience.Inthecaseofthefoamcomponent,the solutionshadtoincludeasamplesolutionthatcouldbetestedin theproduct.Ultimately,onlyoneoftheproposedsolutions deliv-eredtherequestedfunctionality.APhilips’sengineerdescribedthe useofNineSigmaforsophisticatedpathproblems:

At the beginning, when Philips was hesitant about posting arequest, weactually startedwithholy-grailquestions that included things that were in people’s minds for years and seemedimpossibletosolve.Now,werecognizeitismoreabout technologysolutionsoutsideourﬁeldofexpertisethatwedo notactuallyhaveherein-house.Wethinkaboutthoseprojects thatwehavenotworkedonbeforeandalsoacceptthatwe can-notstartreadingpapersandgoingtoconferences,thatittakes twotothreeyearstolearnhowitmaywork,orevenjusttoset upaproposal.

Inasophisticatedpath,problemframinginvolvestheinnovation intermediaryformulatingthetechnologyneedinaratherabstract

(10)

waytoallowascientiﬁcexplanationoftheproblem.AseniorPhilips researchmanagerexplained:

Ifyouaskabroadquestion:“Whocanhelpmewithshaving foam?”youwillgetresponsesthataretoobroad.Atthesame timeyouneedtoavoidbeingtoorestrictiveandnotreceiving responses.Youneedtoapplysomefilters,sothatyoucanexpect reasonableresponsesbutnotmissoutonopportunities. Sincetheunfamiliarformulationofproblemsrequiringa sophis-ticated path result in difficulties for the knowledge-seeker in definingthescientificboundariestotheproblem,theboundary spanningmechanismiscrucialforidentifyingasufficiently nar-rownetworkofsolution-providers.Inordertoincreasethechances offindingasolution,theinnovationintermediarytargetsasmall numberofsolution-providersfromscientificdomainswith valu-ableexpertisefortheproblem.Forexample,oneofNineSigma’s corporatedirectorsexplained:

Thisissomethingonlyforverysophisticatedcustomers...We willcreateaspeciﬁcexpertecosystemofmaybe20peoplefrom verydiverseareas.Wemanageitthiswaywhenitisavery spe-ciﬁctargetedproblemthatwearelookingtosolveWeaskthem toprovideaverybriefbackground:“Whatareyour capabili-ties?”Doyouhaveanyparticularinterestinthetopic-yesor no?”

Whilesophisticatedandscientificpathsmightseemsimilar,the searchforsolutionstoaproblemdiffers.Oneexampleisthecase ofPepsiCo,whenitwassearchingforeitheralternativesourcesof sodiumormethodstoreducethesodiuminitsFrito-Laypotato chipsthatwouldstillmaintainasaltyflavor.Intheirattemptto identifyanewformulationforamicro-particularhaliteordrying techniquefromanunknownscientificfield,PepsiCoandNineSigma formulatedtheproblemtoincludeabroadrangeoftechnological domainsandbusinessmodelsdistantfromtheknowledge-seeker’s snackbusinessdomain.

Sophisticatedsearchpathsdifferfromsituatedpathsintermsof thematurityoftherequestedsolution,i.e.,whetheraquickwinor aproofofconcept.IntheNaturaexampleinvolvingan aluminum-freedeodorant,Naturarequestedsolutionsalreadymatureenough tobeimplemented,ratherthanearly-stagesolutions.Thisproblem requiredasituatedpathtoprovideashort-termsolutionderived fromatechnologicallyrelatedsector.Natura’sresearchmanager explainedtheproblemandtheabsenceofmaturesolutions:

Foroneoftheproblemsweweretryingtoﬁndanewwayto workwithdeodorants-notusingaluminumsalt.Currently,all theproductsusealuminumsalt,andbecauseofitstoxicity,we weretryingtoﬁndanotherwaytosolvethisproblem...Forthe deodorant,wehadalotofproposalsbuttheywerevery sim-ilartowhatwerethinkingandimaginingalready.Theproject failedbecausesolution-providersproposedsolutionsthatwere alreadytestedornottechnologicallymature.Itwasveryearly stageresearch.

4.2.4. Scientiﬁcpaths

The adoption of a scientific path is aimedat finding break-throughsolutionsfromunrelatedscientificfieldstoproblemsthat mayrequireeitherknowledgerecombinationorexploitationofa distantscientificnetwork.Theseproblemsspecifyinputandoutput parameters,whichdifferentiatesthemfrom“holygrail”problems thatarebeyondthescopeofinnovationintermediaries. Connect-ingtheproposedsolutionsfromunknownscientificcommunities tothefocalprobleminvolvesusingchartsandassessment matri-cestohighlighttheissuesofinteresttotheknowledge-seekerwhile bearinginmindthenoveltyoftheproblem.Someidentifiedcase examplesareprovidedinFig.2.Inoursample,onlyPepsiCo’shalite

problemwasresolved.Byformulatingtheproblembroadly, Pep-siCoreceivedasolutioninvolvinganewapproachtothecontinuous productionofhalitenanoparticles,i.e.,crystalsalt,fromaEuropean universityspin-off(amedicallab)workingontechniquestotreat osteoporosis.

Inscientificpaths,theproblemframingmechanismrecognizes thatspecificationof thetechnologyneedshouldbegenericand scientific;inthePepsiCoproblem, itremainedspecificinterms ofwhatPepsiConeededtomeetparticularmanufacturing require-ments,i.e.,asignificantreductioninsaltcontent.Byunderstanding theknowledge-seeker’sproblem,problemframingwasaimedat findingabreakthrough technologytofeed-forwardtonew rep-resentations of the problem. In another example, ArçelikA.S¸.’s technology need for a washing machine that doesnot require theuse ofwater was toobroadlydefined, and didnot explain whyanalternative solutionwasneeded.ANineSigmamanager elaborated:

TheangleoftheRFPwillgetadaptedaccordingtothe ﬁnal-itytheclienthasinposingthequestion...Ifyouarelooking forawashingmachinethatwasheswithoutwater,thenthey dosomethingcompletelydifferent.Itcomesfromtheanalysis beforehand,ifyoudonotdothatanalysisbeforehand,allyou haveattheendisdisappointment.

TheboundaryspanningmechanisminvolvedinPepsiCo’s prob-lemrequireda strongerfocusonthescientificcommunity,e.g., universitiesandlabs,ratherthanonindustrysolution-providers, since there was no expectation of an off-the-shelf solution. Solution-providers are expected toprovide long-termsolutions that offernovelalternative waysto solveatechnologicalneed. Moreover,theboundaryspanningmechanismidentifiesaspecific networkofsolution-providersthatwillneedtimetorespondto thechallengesince theymust evaluatethepotentialsuccessof thetechnologysolution.ANineSigmamanagerdetailedhowthe boundaryspanningmechanismwasusedinthehaliteproblem:

You can tell immediately where it stands. It is very much upstream-youcouldnotexpecttogetanythingofftheshelf, ready,justdropinaningredientandthat’sit.Inthis particu-larcase,whenweselectedthebroadcastpoolofcontacts,we focusedalittlebitmoreonthescientiﬁccommunityatthe uni-versitylevel,thelablevel,ratherthangoingtoindustrialparties. So,wemadesortofabalancebetweenthepurescientiﬁc com-munityandtheindustrialone.

Themaindifferencebetweenthescientificandsophisticated pathsistheknowledgedistancetodiversescientificfields,which,in thecaseofsophisticatedpaths,haveclearlyspecifiedinputand out-putparameters.Inscientificpaths,thescientificfieldsareunknown anddistant,whileinsophisticatedpaths,thetechnologyfieldsare localandaimtoprovidesolutionsthataddvaluetoexisting prod-ucts.For example,3M’salternative useof roofinggranulates to addresstheirpost-consumerasphaltshinglesproblemappeared tobeascientificpath.A3Mmanagerexplainedwhythisproblem requiredasophisticatedpathinstead:

Whatweweretryingtoevaluatewasthepotentialofa better-valuesolutioningoingallthewaybacktothebasicasphalt shingles,andwhetherthatwouldbemoreeconomically valu-able.Wedidnothaveenoughinformationtounderstandhow youwouldmakethatseparationandrecovery.Wedidnothavea good-enoughunderstandingoftheasphaltthatwouldbe incor-porated.

Scientiﬁc paths differ fromanalogical pathsin terms of the impactof therequestedsolution,i.e., whetherbreakthroughor disruptivesolutionsornewprocessesortechnologies.An exam-pleisSealy’sefforttoredeﬁneitsfoundationmattress.Asenior

(11)

processengineerfromSealydescribedthisprojectashaving spe-ciﬁccharacteristicsthatwarrantedananalogicalpath:

Mostprojectsareverylinearandincludeprocessimprovement stages.Thisisalong-terminitiative,andalotofcreativityhasto gointotheactualactivitiesofaprojectsuchasthis.So,itisnot onlyaboutanalyticalreasoning,butalsosometeam-building andfacilitatingskillsinordertomakesurethingsareorganized andthemomentumstaysthere.Itislongterm,andyoumakeit toapointwhereyourealizethatisnotthebestthingtodo,and eventhoughtimehasbeeninvested,youhavetomakesureyou haveenoughdatatoknowthatitisstillagoodinitiative. Tosumup,eachsearchpathhasspeciﬁcobjectives, characteris-tics,selectioncriteria,andexpectedtypesofsolutions(seeFig.2), andtheproblemframingandboundaryspanningmechanismsare applieddifferently(seeTable1).Inthefollowingsection,we dis-cussourﬁndingsinlightoftheextantliteratureonsearchandopen innovation.

5. Discussion

Thepresentstudyaddstoconceptualdevelopmentsrelatedto thebreadthanddepthofthesearchspace(KnudsenandSrikanth, 2014;LevinthalandMarch,1993),andinvestigateshowdecisions aboutsearchheuristics(Gavetti,2012;Grandori,2013;Nickerson andZenger,2004)canincreasethepossibilityofidentifyingnovel solutionstoproblems.Basedontheexistingsearchliterature(cf. AhujaandKatila,2004;FlemingandSorenson,2004;Gavettiand Levinthal,2000;Laursen,2012),weproposeadepictionofsearch inopeninnovationrepresentedbya combinationoftwosearch dimensions:thesearch space,whichcanbeeither localor dis-tant(Garrigaetal.,2013;LaursenandSalter,2006;Piezunkaand Dahlander,2015),andthesearchheuristics,whichcanbeeither experientialorcognitive(FelinandZenger,2014;Siegetal.,2010; Spradlin,2012).

Intheproposedtypologyof searchpaths,wefoundthatthe combinationofsearchspaceandsearchheuristicshelpstoexplain organizationalsearchinopeninnovationbyreinforcingthe impor-tanceofreﬁnementsearch (exploitation)and innovativesearch (exploration)(LevinthalandMarch,1981;March,1991),or situ-atedandscientiﬁcpaths,respectively.Weintroducetwopreviously neglectedsearchpaths(analogicalpathsandsophisticatedpaths). Moreover,combined withtheuseof innovation intermediaries (Chesbrough,2006;Roijakkersetal.,2014),weproposethatthe mechanismsofproblemframing(Baeretal.,2013;Kaplan,2008; VonHippelandVonKrogh,2015)andboundaryspanning(Fleming and Waguespack, 2007), in relation to the knowledge-seeker’s problem,areintegraltoeachsearchpath.

Theproblemframingmechanismoffersthepossibilityto for-mulatea technologyproblembyusing familiarterminology, or analternativeorcompetingproblemunderstanding(Baeretal., 2013).Morespecifically,ourstudyrevealedhowinnovation inter-mediariesinteractwithknowledge-seekingclientstoformulate theirproblems intoaddressable technologyneedsfor solution-providers.While thisdoesnot imply thatinteractions between intermediariesandknowledge-seekersleadtocompleteproblem decompositions,thearticulationintoanRFPzoomsinonspecific needsthatarebottleneckstoprogressintechnologydevelopment. Thecontributionofproblemframingistwofold:(a)itfacilitates theidentificationofcritical issuestosolveproblems,and (b) it articulatesadistincttechnologyneedthatcanbecommunicated toacommunityofsolversnotcurrentlyfamiliarwiththespecific problemsetting.

The boundary spanning mechanism consists of identifying potentialareas,i.e.,scientiﬁcandtechnological,aswellascrowds

of solution-providers that might be interested in solving the speciﬁc problem. This mechanism facilitates the knowledge-seeker’saccesstoadistantnetworkofsolution-providers(Jeppesen and Lakhani, 2010). Our ﬁndings suggest that boundary span-ningbyinnovationintermediariescontributes tobridgingthree typesofboundaries:(a)boundariesbetweenareasofapplication andknowledgedomains,(b)organizationalboundariesbetween knowledge-seekers and solution-providers and (c) boundaries betweenknownandunknownsolutions.

Theintersectionbetweentheproblemframingandboundary spanningmechanismsaddressestwogapsintheliterature.First, while distantsearch seemsbeneficial forbreakthrough innova-tions,thelargenumberofirrelevantsolutionsmakestheevaluation ofdistantsolutionscumbersome(PiezunkaandDahlander,2015). Weproposethatmorepreciseproblemframingcouldreducethe numberandheterogeneityofproposedsolutions,andincreasethe possibilityoffindinganapplicablesolution.Second,Jeppesenand Lakhani(2010,p.1031)reportthatuptotwo-thirdsofbroadcast problems remainunsolved. We suggest that themechanism of problemframingcouldbeusedtoprovideanalternative under-standingtoa specificproblemandalsoincrease thechancesof identifyinganacquirablesolution.

Thefoursearch pathsandtwo proposedsearchmechanisms have implications for scholarshipin organizational search gen-erally, and openinnovationin particular.While the benefitsof innovationintermediariesperformingsearchforexternal knowl-edgeinlocalanddistanttechnologicalfieldsinrelationtoproblems, orideationandexpertise-basedprojects,havebeendocumented previously(c.f.Siegetal.,2010;Spradlin,2012),relativelylittle attentionhasbeengiventohowsearchtakesplace,andwhat alter-nativesearchheuristicsareappliedinopeninnovation(Felinand Zenger, 2014;Jeppesen and Lakhani, 2010).These observations reinforcetheneedforamorerefinedunderstandingofhowsearch takesplaceinrelationtoopeninnovationprojects(Duetal.,2014) andproblemsolving(VonHippelandVonKrogh,2015).The empir-icalfindingsfromourstudyof18openinnovationprojectsshed lightonwhenknowledge-seekersandinnovationintermediaries employthefoursearchpathsintheirquestforexternalknowledge. Inthecase ofsituatedsearchpaths,theliteraturehighlights their benefits for “clearly defined, well-structured and simple problems(i.e.noncomplex)orsub-problems”(FelinandZenger, 2014, p. 921). Here, early stage or not-quickly implementable solutionsarenotdesirable.Analogicalsearchpathsinvolve rea-soning via recombination across different knowledge domains (Garyet al.,2012).Althoughanalogicalsearchpathsare benefi-cialfortechnology-developmentproblems,searchacrossdistant technologicalboundariesrequirescombinativeandcollaborative capabilities(JeppesenandLakhani,2010).

Whilesituatedandanalogicalsearchpathsaremore advanta-geous forproblems thatexploit feed-backfromtheproblemat hand,aroutinedevelopmentprocess,orlearning-by-doing(Nelson andWinter,1982;Pisano,1994),someauthorssuggestthatthe useofdifferentsearchheuristics,i.e.,problemredeﬁnition, decom-position,andknowledgeabstractionfromtheknowledge-seeker’s industrycontext,wouldimprovethechancesofobtaining success-ful solutionstomore complexproblems (Sieget al.,2010;Von HippelandVonKrogh,2015).

Sophisticatedsearchpathsareaimedatnovelsolutionsbased onconcepts,theories,andmodelsinadjacentdomains(Tripsasand Gavetti,2000).Theirsuccessdependsontheproblemformulation, whichrequires abstractrepresentationsordeductive reasoning. Weshowedthatinnovationintermediariescanbehelpfulpartners withsophisticatedpaths,theframingofproblemsinabstractterms, and the targeting of networks of potential solution-providers. Hence,sophisticatedsearchpathsaresuitablefordecomposable problemssincetheybuilduponpath-deepening(AhujaandKatila,

(12)

2004)andabstractreasoning,asthesedo notrequireextensive boundaryspanningduetocommonalitieswithcurrentsolutions.

Scientificsearchpathsareusedtoidentifysolutionstocomplex andnovelproblemsinvolvingsubstantialuncertainty.Thesetypes ofsearchpathsrequiretheory-drivenconjectures,whichinvolve searching and exploiting distantknowledge domains. Although somepreviousstudiessuggestusingfirm-hosteduserand inno-vationcommunitiestoengageexpertsolution-providers tofind inexpensive solutions to complex problems (Felin and Zenger, 2014), thesescenarios do not takeinto account expensiveand confidentialproblemswithverylowhitrates,suchasaneedfor alternativesourcesofsodium(PepsiCo)orsubstitutesfor formalde-hyde(L’Oreal).

6. Conclusions

In this paper, we analyzed how search takes place when knowledge-seekingﬁrmsuseinnovationintermediariesin open innovationprojectstosolvetechnologyproblems.Wesummarize ourmainarguments,primarycontributions,somelimitations,and suggestionsforfutureresearchas(1)problems,(2)paths,and(3) projects.

First,theﬁndingsreportedherecontributetorecenttheorizing onsearch in openinnovationthat uses a problem-solving per-spective(c.f.AfuahandTucci,2012;JeppesenandLakhani,2010; PiezunkaandDahlander,2015;VonHippelandVonKrogh,2015). Weaddtothisliteraturebyshowingthecombinedeffectsofboth searchspaceandsearchheuristicsdimensionswhen knowledge-seekers assisted by open innovation intermediaries search for solutionstoproblemsthroughinnovationcontests. Asempirical researchonopeninnovationintermediariesisscarce(Chesbrough, 2006;Roijakkersetal.,2014),wesuggestthatfuturestudiescan usethis settingtofurtherempiricallyexamine theviabilityand effectivenessofdifferentsearchoptionsinproblemsolving.

Second,weconcludethatthecombinationofsearchspaceand searchheuristicspreviouslydiscussedintheliterature(e.g.Gavetti andLevinthal,2000)makesfourdistinctsearchpathsavailable: situated, analogical,sophisticated,and scientiﬁc. 18 open inno-vationprojectswereexaminedandclassiﬁedaccordingtothese search paths.We argue that one-dimensionalsearch path con-structs (Tippmann et al., 2013) fail to recognize the range of searchpathsavailableinopeninnovation(FelinandZenger,2014). Moreover,thereislittleacknowledgmentintheliteratureof ana-logicalorsophisticatedsearchpaths,whileourstudyshowsthat theyrepresentimportantsearchoptions.Wealsoindicatedhow themechanismsof problemframing(Baeret al.,2013; Kaplan,

2008)and boundaryspanning(FlemingandWaguespack,2007; RosenkopfandNerkar,2001)operateinthesesearchpaths.Aswe merelyidentifythesesearchpaths,futureresearchcould exam-ine contingencies that inﬂuence their use, rate, and direction. This shouldinclude variables identiﬁed in therecent literature onsearchinopeninnovationsuchasproblemcomplexity(Felin andZenger,2014),incentivesforsolution-providerstocontribute theirsolutions(Boudreauet al.,2011), attentionof knowledge-seekers(PiezunkaandDahlander,2015),accumulatedexperience ofknowledge-seekerswithopeninnovation(Siegetal.,2010),and businessmodelsusedbyinnovationintermediariesorthird-party platforms(AfuahandTucci,2012;Howells,2006).

Third,despitethecharacteristicsofsearchpathsbeinggeneral, weconcludethatmuchsearchtakesplaceinprojects.Weargue thatempiricalobservationandanalysisofsearchpathsrequires in-depth,project-leveldatatoreveal boththesearch heuristics appliedandthesearchspacecovered.Inprojects,problemsare framedandboundariesarespanned.Ourﬁndingsdrawupononly 18openinnovationprojects,andwerecognizeampleopportunity forfutureresearchontheprojectleveltoprovidefurther expla-nations oftheprocesses andmechanisms inoperation(seee.g. Du et al.,2014).For instance, since successfulopeninnovation alsoinvolvestheintegrationofexternalknowledgesubsequentto search(Lakemondetal.,2014),futureresearchusingbothlarger andsmallersamplesofprojectscouldinvestigatetheconditions underwhichprojectgovernanceandprojectroutinesareapplied indifferentsearchpaths.

Acknowledgements

Theauthorsacknowledgeinsightfulconversationonthetopic ofthispaperwitheditorAshishAroraandtwoanonymous review-ers.Extensiveandconstructivecommentsonpreviousdraftswere providedbyLarsFrederiksen,KeldLaursen,AnnMajchrzak,Peter Murmann,AmmonSalterandJonathanWareham.Previous ver-sionswerepresentedinseminarsatAarhusUniv.,ErasmusUniv., LinköpingUniv.,Univ.ofNew SouthWales,Univ.ofZurich,and the Special Conference onOpen Innovation, London,2012 and theOrganizationScienceWinterConference,SteamboatSprings, 2012.Theauthorswanttothanktheseminarandconference par-ticipantsfortheirhelpfulfeedbackonearlierdrafts.Thesupport fromESADEBusinessSchoolandNineSigma,andthefunding pro-videdforthisresearchbyHandelsbanken’sResearchFoundations andRiksbankensJubileumsfondisgratefullyrecognized.Theusual disclaimersapply.

Appendix1. Interviewedcompanies

No.of interviewees Nameofthe organization Position No.of interv. Nameofthe organization Position No.of interv Nameofthe organization Position

1 L’Oreal OpenInnovation

Manager

1 Xerox OpenInnovation

Manager 1 Kimberly-Clark HealthCare Product& Technology Development Manager

1 CarlZeissAG SeniorManager,

ScientiﬁcAffairs

1 KraftFoods SeniorAssociate&

PrincipalScientist

1 Natura DirectorofAdvanced

Research 1 Rheem Manufacturing CompanyBPPLC, Reﬁning Technology

PrincipalEngineer 1 Ferrero Packaging

Development Director

1 SealedAir ResearchScientist

(OpenInnovation Manager)

1 ProcessToolsand

AnalyticsManager

1 SherwinWilliams TechnologyScout 1 International

Copper Association

AssistantDirectorof Technology

1 Sealy SeniorProcess

Engineer

1 TheGoodyearTire

&RubberCompany

SeniorR&D Associate 2 HallmarkCards, Inc. ProductInnovation Manager,Senior EngineerII