Know_Innov

17
Knowledge Reuse for Innovation Lynne P. Cooper Olivia E. Neece Ann Majchrzak

description

Knowledge Reuse for Innovation Lynne P. Cooper Olivia E. Neece Ann Majchrzak Olivia E. Neece informs ® Knowledge transfer is the process through which knowledge acquired in one situation is applied Department of Information and Operations Management, Marshall School of Business, University of Southern California, Los Angeles, California 90089-1421, [email protected] doi 10.1287/mnsc.1030.0116 © 2004 INFORMS Claremont Graduate School, Claremont, California 91711, [email protected] 174

Transcript of Know_Innov

Knowledge Reuse

for Innovation

Lynne P. Cooper

Olivia E. Neece

Ann Majchrzak

MANAGEMENT SCIENCEVol. 50, No. 2, February 2004, pp. 174–188issn 0025-1909 �eissn 1526-5501 �04 �5002 �0174

informs ®

doi 10.1287/mnsc.1030.0116©2004 INFORMS

Knowledge Reuse for Innovation

Ann MajchrzakDepartment of Information and Operations Management, Marshall School of Business, University of Southern California,

Los Angeles, California 90089-1421, [email protected]

Lynne P. CooperJet Propulsion Laboratory, California Institute of Technology, Pasadena, California 91109-8099, [email protected]

Olivia E. NeeceClaremont Graduate School, Claremont, California 91711, [email protected]

This study was conducted to better understand the knowledge reuse process when radical innovation (e.g.,experiments to prepare for human exploration of Mars) is expected. The research involved detailing the

knowledge reuse process in six case studies varying in degree of innovation. Across the six cases, a six-stagereuse-for-innovation process was identified consisting of three major actions: reconceptualize the problem andapproach, including deciding to search for others’ ideas to reuse; search-and-evaluate others’ ideas to reuse;and develop the selected idea. Findings include (1) the need for an insurmountable gap in performance tostimulate the decision to reuse others’ knowledge; (2) the critical importance of an adapter to bridge the ideasource and recipient; (3) three layers of search-and-evaluate activities in which the first layer of scanning tofind ideas to reuse and the last layer of detailed analysis of ideas are bridged by a layer of brief evaluations ofideas assessing the presence (or absence) of targeted information about each idea; and (4) the differential useof metaknowledge about each idea to facilitate proceeding through each search-and-evaluate layer. In addition,reusers in the more (versus less) innovative cases redefined problems at the outset in nontraditional ways usinganalogies and extensions, rather than accepting the preexisting problem definition; used a substantially broadersearch strategy with a greater variety of search methods; and worked more closely with adapters during thelatter stages of the reuse process.

Key words : knowledge management; knowledge transfer; innovationHistory : Accepted by Linda Argote, former department editor; received March 1, 2001. This paper was withthe authors for 8 months for 2 revisions.

Pete, the project manager, read the Announcement ofOpportunity, which called for the development of aninstrument that autonomously detects and measuresdust devils on Mars. Dust devils are notoriously diffi-cult to predict and yet carry with them enough forceto upset equipment, and dust so fine that it poses ahazard to future human exploration of the Red Planet.As a 20-year veteran of space mission development,Pete considered the use of “standard” (if there is astandard for Mars) meteorological solutions but theydidn’t provide the advance-warning capability. More-over, it wasn’t “sexy” enough for the NASA sponsors,so transfer of known practices wasn’t feasible. Peteconsidered developing a solution from scratch, but theproject resources didn’t allow the time or money. SoPete embarked on a search for ideas that he could reuseand adapt to innovate. What did Pete do to find thoseideas, how did he evaluate them when he found them,and what are the implications of Pete’s behaviors anddecisions for knowledge management? These are thequestions this paper addresses.

IntroductionKnowledge transfer is the process through whichknowledge acquired in one situation is applied

to another (Argote and Ingram 2000). We adopta broad definition of knowledge consistent withprior research: explicit knowledge such as drawings,analytic results, and scientific journal articles, aswell as tacit knowledge such as insights, intuition,and implied assumptions (Beccerra-Fernandez andSabherwal 2001, Grant 1996, Kogut and Zander 1992,Polanyi 1966, Teece 1981). Knowledge transfer cangenerally be subdivided into knowledge sharing (theprocess by which an entity’s knowledge is captured;Appleyard 1996) and knowledge reuse (the processby which an entity is able to locate and use sharedknowledge; Alavi and Leidner 2001). We are focusedon knowledge reuse.In this paper, we are interested in knowledge

reuse for the express purpose of facilitating thedevelopment of radically innovative solutions. Inno-vative solutions are defined as solutions that repre-sent creative (i.e., novel and useful) ideas that areimplemented (Amabile 1996). Radical innovation isdifferentiated from incremental innovation by involv-ing discontinuous development where unprecedentedimprovements or performance features are achieved

174

Majchrzak, Cooper, and Neece: Knowledge Reuse for InnovationManagement Science 50(2), pp. 174–188, © 2004 INFORMS 175

(Leifer et al. 2000). From a knowledge reuse perspec-tive, reuse for radical innovation is the exploitationof existing diverse ideas previously unknown to theinnovator when creating a new product or service(Armbrecht et al. 2001). Kogut and Zander (1992)describe this ability of the firm to generate new com-binations of existing knowledge as combinative capa-bilities. Grant (1996) argues that such a capability isa strategically significant resource to a competitiveorganization.Given the importance of understanding how these

new combinations of knowledge are created, iden-tifying ways to facilitate innovators’ search andreuse behavior is an appropriate objective. If inno-vators limit their search for solutions to their cur-rent personal knowledge base or existing networkof sources, the extent to which radical innovationis achieved will be limited (Leifer et al. 2000, Clarkand Fujimoto 1991). When innovators reuse others’knowledge which was previously unknown to them,the creativity envelope is expanded beyond a smallset of individuals (Armbrecht et al. 2001). Thus, aknowledge management system that expands thecreativity envelope improves the research and devel-opment process through quicker access and move-ment of new knowledge. Moreover, improving theuse of knowledge in innovation has benefits to thepractical field of knowledge management. Accordingto a 1997 Ernst & Young survey of executives (citedin Holsapple and Joshi 2000), innovation is seen asthe area of greatest payoff from knowledge manage-ment, even though such efforts to date have mostlyfocused on operational productivity improvements(Davenport et al. 1996).

Existing Literature onKnowledge ReuseThere are several frameworks in the literature forunderstanding knowledge reuse. Grant (1996) cate-gorizes these frameworks into those that focus onknowledge acquisition (or replication) and those thatfocus on knowledge integration. For example, stud-ies on spillover effects of knowledge between relatedresearch programs, product generations, and manu-facturing organizations, as well as studies of best-practice transfers (see review by Argote 1999) focuson how a “recipient organization” acquires and appliesthe knowledge of the “source organization” in aneffort to replicate the essential elements of the source’sknowledge. Szulanski (2000) provides an exampleof a study with this “knowledge reuse as replica-tion” (KRR) focus. Grant (1996), however, arguesthat knowledge acquisition is not necessarily an effi-cient approach when the objective is radical inno-vation. With radical innovation, knowledge is not

only acquired but also integrated across disparatesources of specialized knowledge. While acquisitionmay benefit from Nonaka’s (1994) conversion oftacit into explicit knowledge, for example, and fromBrown and Duguid’s (1991) communities of practices,Grant argues that knowledge integration requiresdifferent—as yet underresearched—mechanisms.In addition to Grant’s claims, there is research from

the literature on innovation to indicate that the frame-works and findings from the KRR literature may actu-ally restrict, rather than facilitate, effective reuse forinnovation. As one example, knowledge reuse is oftenthought to increase with initial shared experiencesbetween source and recipient (Argote 1999, Brownand Duguid 1991, Hansen 1999, Kogut and Zander1992, Nonaka 1994). That is, the more familiar you arewith a source, the more likely you are to reuse thesource’s knowledge. However, research on innovationdiffusion (Rogers 1983), new product development(Dougherty 1992), creativity (Amabile 1996, Unsworth2001), and how people reuse knowledge when inno-vating (Swan 2001, Gray 2000) comes to the oppositeconclusion: divergence and lack of shared experiencesare critical for developing new ideas. For example,Hargadon and Sutton (1997) explain why IDEO is soproductive at repeatedly generating innovative ideas:employees identify solutions in other domains thathave nothing in common with their focal domain.Allen’s (1977) research demonstrates that innovatorsget creative ideas from a variety of sources, includingsearches of knowledge bases with unknown sources.Thus, knowledge reuse for radical innovation (KRI)may not rely on shared experiences between sourcesand recipients as completely as KRR.KRI may also differ from KRR in the evaluation cri-

teria applied to knowledge being reused. While anyknowledge is likely to be reinvented as it is reused(Argote 1999, Rice and Rogers 1980), adaptation ofknowledge is likely to be greater and more deliberatein radical innovation than in the transfer of best prac-tices. Does this greater need for adaptation affect howknowledge is evaluated? For example, is knowledgethat is being considered for reuse in a KRI processevaluated not only for its reuse potential but also forits adaptation potential?KRI may also differ from KRR in the content of

the knowledge being transferred. Szulanski (2000)and Zander and Kogut (1995) found that practiceswith clear cause and effect relationships that werecodified and trainable were more easily transferred.However, because radical innovation involves thetransfer and integration of largely tacit knowledge(Leonard and Sensiper 1998), the knowledge beingtransferred is likely to be ambiguous, incompletelycodified, and complex. This requires reusers in a KRI

Majchrzak, Cooper, and Neece: Knowledge Reuse for Innovation176 Management Science 50(2), pp. 174–188, © 2004 INFORMS

process to find ways to understand the tacit knowl-edge being transferred. Clark (1996) and Star andGriesemer (1989) have theorized that tacit knowl-edge transfer is facilitated by the use of shared arti-facts. These artifacts convey contextual informationabout the knowledge being shared, helping to clar-ify the meaning underlying ambiguous knowledge.This would suggest that physical artifacts may play aparticularly important role in the KRI process.Finally, we do not know what a staged model

for the KRI process looks like. Radical innovationproceeds less as stages from conceptualization to com-mercialization (as is found with incremental devel-opment) and more as sporadic trajectory changes inresponse to unanticipated events (Cheng and Van deVen 1996, Leifer et al. 2000). For example, in radicalinnovation, idea generation and opportunity recogni-tion do not occur at the front end as in incrementaldevelopment, but sporadically throughout and oftenin response to organization, technical, and market dis-continuities (Dougherty 1992). Thus, knowledge reusethat occurs within a radical innovation work processmay not proceed as a staged model flowing fromopportunity recognition to execution (Szulanski 2000)or tacit-to-explicit-to-tacit conversion (Nonaka 1994).For example, Thomke (1998) and Kelly (1970) suggestthat tacit conversion in the form of experiments occurthroughout the KRI process, not just at the beginningand end.This review suggests that researchers need to study

KRI in its own right by examining how knowledgeis reused during the actual work process of innovat-ing. In particular, we are interested in answering thefollowing questions: How is knowledge reused forradical innovation? Is this reuse process fundamen-tally different from previous studies depicting a KRRprocess?

Research DesignOur research question suggests a research designin which we build rather than test theory. More-over, our research design requires examining reuseas part of the actual work process of innovation,requiring ethnographic research methods (Blackler1995, Schultze and Boland 2000). Thus, we applyEisenhardt’s (1989) guidelines for theory-buildingcase study research in conjunction with the guidelinesfor hypothesis-generation case study research offeredby Yin (1984) and Klein and Myers (1999). Table 1summarizes our study design, comparing it to recom-mendations made by Eisenhardt (1989).

Case Selection. Our case study research involvedcollecting and comparing data from six cases of reusefor innovation at the Jet Propulsion Laboratory (JPL).JPL is a federally funded research and development

Table 1 Our Study Design Compared to Eisenhardt’sRecommendations

Eisenhardt’s (1989)recommended steps Our study design

(1) Getting started: Defineresearch question with apriori constructs.

Two research questions defined. Usedconstructs from literature.

(2) Select cases based onspecific population andsampling to replicate orextend emergent theory.

JPL specifically picked as innovativeorganization. Selected 6 cases ofsuccessful reuse to vary alongadopt/adapt continuum.

(3) Craft instruments topromote triangulationamong data sources andinvestigators.

Initial semistructured interviews withJPL knowledge management staffand innovators; identified cases witharchival data; examined archival datato derive initial thoughts on researchquestions; conducted structuredinterviews with at least 2 people percase; participant-observer in casesinvolved in interpreting data.

(4) Enter field in such a way asto overlap data collectionand analysis.

Collected verbatim transcripts; after eachcase, discussed notes to determine ifadditional questions, data, or caseswere needed.

(5) Analyze data within andacross cases.

Wrote up each case separately. Createdmatrices to identify patterns acrosscases.

(6) Shape hypotheses bylooking for replication notsampling logic; iterativetabulation of evidencefor each construct; refinedefinition of constructs.

As each case unfolded, discussednotes extensively to shape emergingrelationships among constructs; usednext cases to replicate or modifyemerging hypotheses.

(7) Enfold literature bycomparing results withconflicting and similarliterature.

Compared our theory with innova-tion, NPD, and knowledge transferliteratures.

(8) Reach closure about whento stop adding cases anditerating between theoryand data. Guideline is6–10 cases.

Stopped adding cases and iteratingwhen conclusions matched evidence,were practical, and interpretable toparticipants not involved in analysis.Ended with 6 cases.

center of over 5,000 employees (and on-site con-tractors) with a $1 billion budget, and is managedby the California Institute of Technology for theNational Aeronautics and Space Administration(NASA). Started in the 1930s, JPL has conceived andexecuted missions to use robotic spacecraft to exploreall of the solar system’s known planets (except Pluto).Thus, JPL specializes in developing technologies andconcepts that have not been used previously, i.e.,radical innovation.The six cases were selected from an initial pool

of 15 cases identified through interviews with man-agers at JPL. The pool of cases came from two projectsthat had been supported by internal JPL funds. Thesefunds were earmarked to develop detailed propos-als in response to a NASA announcement of oppor-tunity (AO). In addition, only cases involving reuse

Majchrzak, Cooper, and Neece: Knowledge Reuse for InnovationManagement Science 50(2), pp. 174–188, © 2004 INFORMS 177

Table 2 Brief Description of Six Cases

Case description∗ Degree of innovativeness Informants and informant role

Case 1. Lidar designTook notion of a Laser Radar (Lidar) prototype that was previously usedto detect landing hazards and significantly adapted it to operate as anearly warning system for detecting the presence of dust devils on Mars.

High (change to form, fit, andfunction)

Project Manager (Reuser)Scientist (Reuser)Engineer (Participant)

Case 2. Electrometer (ELM) designUsed concepts and principles of industrial devices that measureelectrostatic discharge on earth, and significantly adapted them tomeasure electrostatic properties of Mars dust and its interaction effectson materials (for equipment and space suits) for use on Mars.

High (change to form, fit, andfunction)

Engineer (Reuser/Adapter)Project Manager (Participant)Scientist (Participant)

Case 3. AFM designTook notion of an atomic force microscope (AFM) originally used in thesemiconductor industry to test surface smoothness, and adapted it tocharacterize particles on Mars.

Medium (change to fit and form;some change to function)

Project Manager (Reuser)Scientist/Engineer (Reuser)

Case 4. AFM tip arrayTook concept of multiple AFM tips originally used to increase scan speedby operating in parallel in semiconductor industry, and adapted it toinstead provide redundancy for operation on Mars through an array ofsingle replaceable tips for AFM.

Medium (change function and minorchange in form)

Scientist/Engineer (Reuser)Project Manager (Participant)

Case 5. Electrometer (ELM) materialsAdopted existing set of materials from Kennedy Space Center collectionfor use in electrometer. Actual materials as well as test data were alreadyavailable.

Low (change to fit and minor changein form)

Engineer (Reuser/Adapter)Project Manager (Participant)Scientist (Participant)

Case 6. Magnetic (MAG) patchesAdopted previously used Mars magnetic experiment to fit on new Marsmission in different size package.

Not innovative (minor changes to fit) Scientist (Reuser)

Engineer (Reuser)Engineer (Participant)

∗All but the first case came from Project #1.

of knowledge which was not initially known to theinnovator and that led to an innovation were consid-ered. The cases in the pool were stratified by degreeof innovation; innovation was defined as degree ofdiscontinuity in form, fit, or function from gener-ally accepted approaches to the same problems. Themanager of both projects along with the participant-observer made judgments about the degree of inno-vation in each case and the level of confidence thatthe current approach would continue beyond the pro-posal selection process into development.1 From thispool, we selected six cases where confidence in thesolution being continued into development was high,and which ranged from little innovation to substantialradical innovation. We stopped at studying six caseswhen theoretical saturation was achieved. This num-ber of six cases falls well within Eisenhardt’s (1989)suggestion of 3–10 cases required for theory building.The six cases are described in Table 2.

1 The JPL product development life cycle begins with conceptdevelopment and proceeds through preliminary design, develop-ment, integration and testing, and operations. To reduce risk, pro-posal teams strive to create as mature a concept as possible duringproposal generation. In the two projects we picked, concept devel-opment and preliminary design were part of the proposal develop-ment, after which the project team underwent a NASA competitiveselection process. Thus, the six stages of knowledge reuse occurredduring the concept development and preliminary design activitiesand, for Project 1, continued after selection.

In sum, by selecting six reuse cases arrayed alonga continuum of innovation, we believe we have metYin’s (1984) call for replication logic in case selection.In addition, we have met Eisenhardt’s (1989) call forselection to be based on a population (i.e., as the pop-ulation of cases of reuse for innovation) that controlsfor extraneous variation. The six cases came from thesame organization, JPL, thus controlling for organi-zational culture that encourages innovation (Amabile1996). The six cases came from two similar projects,thus controlling for task differences, a factor foundby Becerra-Fernandez and Sabherwal (2001) to affectknowledge reuse. Finally, all six cases were led by thesame manager, thus controlling for the important roleplayed by project managers in new product develop-ment efforts (Clark and Fujimoto 1991).

Data Collection. Building on a tradition in theinnovation literature of using retrospective tracerstudies (Rogers 1983), data collection focused ondeveloping a detailed timeline for each case.These timelines were developed based on reviewof documents (AO, final proposals, e-mails, andengineering notes) and repeated interviews withreusers in each case. A minimum of two informantsper case were used in addition to the archival infor-mation. The informants included reusers and teammembers participating in team discussions with thereusers.

Majchrzak, Cooper, and Neece: Knowledge Reuse for Innovation178 Management Science 50(2), pp. 174–188, © 2004 INFORMS

The interview protocol consisted of a set of struc-tured, open-ended questions asking informants todescribe the knowledge they reused, reasons whythey reused this knowledge, the problem that theywere trying to solve with the reused knowledge,the initial state of their domain knowledge for solv-ing the problem prior to finding the reused knowl-edge, a description of the reused knowledge (in termsof the form, fit, and function it was serving whenthey discovered it), how the reused knowledge wasdiscovered and evaluated, and what they did withthe reused knowledge. Because the events had takenplace a relatively short time before the interviewswere conducted, the informants were able to answerthese detailed questions. Because multiple informantswere used for each case, differences between theinformants arose; however, these differences weregenerally attributable to the different roles that thedifferent informants played, rather than to conflicts.Nevertheless, when differences arose, we shared thesewith the informants to determine if our interpre-tations of the differences were correct. Thus, themultiple informants allowed us to generate a morecomprehensive timeline of events than we could haveobtained from any single informant.2

In addition to the use of multiple informants, stan-dardized interview protocol, and standardized datacollection format of timelines, Eisenhardt (1989) rec-ommends that theory-building case study researchshould consider the use of multiple investigators withcomplementary insights. We were fortunate becauseour research team consisted of a senior faculty mem-ber, a doctoral student intern who was given permis-sion to join JPL’s knowledge management team for ayear and spend time at JPL learning the culture andcontext, and a development engineer who had par-ticipated in an operational (rather than innovation-generation) capacity on the two projects and thusserved as a participant observer. To avoid bias duringthe interviewing process, the participant-observer didnot conduct the interviews; however, her first-handexperience with the team provided a perspective nottypically obtained through interviews. An example ofa timeline (for the most-innovative case) is shown inFigure 1.3

ResultsBecause Szulanski (2000) provides one of the fewoperationalizations of stages in the knowledge

2 The timelines span different periods, but all start with conceptdevelopment and include preliminary design activities which cul-minated in a high-confidence commitment to a specific designapproach. During the study period, the instruments under develop-ment reached varying degrees of maturity ranging from conceptualdesign to actual hardware.3 The remaining timelines can be accessed on the senior author’swebsite at www-rcf.usc.edu/∼majchrza.

transfer process, and his process is conceptuallysimilar to Rogers’s (1983) well-respected innovation-decision process, we initially tried to code eachaction taken in the timeline by the stages offeredby Szulanski (2000): “Formation of transfer seed,”“Decision to transfer,” “First day of use,” and“Achievement of satisfactory performance.”During the coding process, we found that the time-

line actions did not fall neatly into Szulanski’s stages;nevertheless, there were sufficiently similar actionsacross the six cases that actions could be coded, thengrouped by precedence ordering. The actions for allthe cases were then displayed on a single chart toobserve commonalities across the cases. A summaryof this chart is shown in Table 3 and is described inthe remainder of this section.4

Reconceptualization StageFor each case, the reuse process began not withthe formation of a “transfer seed,” as suggested bySzulanski (2000), but with a definition of the problem.The participants had the option to narrowly interpretthe problem as portrayed in the NASA AO. In theleast-innovative case (mag patches), they chose thatoption. In the other five cases, however, the respon-dents chose to redefine the problem in a way thatwould benefit from an infusion of as-yet unknownideas, requiring the need for radical innovation. Forexample, in the most-innovative case (Lidar), the par-ticipants chose not only to characterize the meteoro-logical phenomenon as called for in the AO but alsoto create an early warning system. In the atomic forcemicroscope (AFM) design case, the reuser describeshis reconceptualization:

The AO called for measurement to be taken on par-ticles of <1 micron. The traditional approaches wererejected as imposing too many constraints (size, sam-ple type and preparation, high vacuum and volt-age requirements). So the scientist considered a newapproach, atomic force microscope.

Respondents offered several reasons why theybelieved radical redefinitions were necessary. First,these redefinitions were motivated by the competitivenature of the “marketplace,” in which proposed Marsprojects that were perceived as having greater impacton the scientific community were more likely tobe selected in the NASA competitive process. Therespondents had some belief that radically differentideas would offer the greatest benefits to the sci-entific community. Characteristics of the individual(i.e., recipient) also played a role in the redefinition

4 The specific pattern-matching tables on which this summary tableis based can be accessed on the senior author’s website at www-rcf.usc.edu/∼majchrza.

Majchrzak, Cooper, and Neece: Knowledge Reuse for InnovationManagement Science 50(2), pp. 174–188, © 2004 INFORMS 179

Figure 1 Example Timeline for Lidar Design

1.6 [S] R defines search to

include finding lidars with dif-ferent functions (sky, hazards

terrain) in different conditions

(stationary, scanning)

1.1 [C] R reads AO. Problem

is: “In 45 days develop a tiny

lightweight instrument that

will autonomously detect and

measure dust devils on Mars

to characterize strength and

frequency of hazard to later

human exploration.”

1.10 Project engineer's husband

serendipitously installs on his

computer initial results from

a prototype laser range finder

which provides info about

where rocks and hazards are

to autonomously guide alander during landing.

1.13 [BE] R asks himself:

maybe we can convert a

scanning laser range finder

from scanning for rocks to

scanning for dust devils (Dec

to adapt Alt#3).

1.18 [BE] R remembers that

AO says Canadian Space

Agency (CSA) willing to

contribute to mission. Has

idea CSA might be interested.

1.20 [IA] R examines data from Alt#3b

firm’s prototype and meets with firm

about Alt#3b. In-house ballpark costing

indicates possibility of cost overrun if

idea #3a developed (Alt#3a discarded).

1.24[S] R searches

internet and contacts

Canadian scientists

1.25 [IA] R examines

Alt#3b firm’s proposals for

prototype for Alt #3b anddetermines it’s too big/heavy

for size/mass of instrument

(dec to adapt Alt#3b).

1.19 [BE] R contacts Can. firm

for info and to see if interested.

1.4[C] R uses radar as an

analogy for the operating

principle: wants a system to do

for dust devils what radars do

for thunderstorms. Because radar

can’t measure dust, substitutes

Lidar because knows Lidar

measures dust.

1.9 [BE] Russian space

instrument not

available. (Alt#1discarded.)

1.11 Project engineer sug-gests to R to look at hus-band's data to see the kind

of data one gets from laser.1.12 [C] R looks at husband’s data

which visually indicates benefits of

concept of using laser range finder.

1.15 [C] R meets with team

expert on Lidar to determine

costs and risks of Lidar approach.

1.17 [BE] R discovers 2

prototypes for rock scanning

had already been built by 2

firms identified by buddy as

reputable: one in U.S. (Alt#3a)

and one in Canada (Alt#3b).

1.21 [IA] Team expert conducts

analytic studies of Alt#3b con-cluding meets “borderline” mass,volume, and power requirements.

1.22 [S] Team member examines

AO to get names to contact to get

CSA names. Contacts CSA and asks

them to contribute Lidar for Alt#3b.

1.26 [IA] R works

with team members to

come up with ideas to

make Alt#3b smaller

by integrating with a

camera from U of A.

1.28 [BE] U of A scientists suggest Alt#4.Team looks at Alt#4 but because Alt#3b

is “free,” Alt #4 considered fallback.

1.27 [IA] Team expert has several subsequent

meetings with Alt#3b firm to determine if

adaptations to Alt#3b can be made.

1.30 [F] Team expert andAlt#3b firm build separate

software models to improve

performance of Lidar.

1.5 [DS] R asks: can we do

it ourselves? (Alt#2

discarded as too expensive.)

1.2[C] R considers traditionalsoln: measuring inside weatherphenomenon using standardmeteorological solns. Decidessoln isn’t sufficient becauseit doesn’t provide advancedwarning, 3d imaging, or meas-ures of velocity, size, and accompanying phenomenon.

1.3 [C] R defines

problem to require

innovation: provideinfo on dust storms so humanswill have enoughinfo to understand

and predicthazardous weatherconditions.

1.16 [S] R goes to see

“old buddy” involved

in lasar range

finder project to

learn more about it.

1.8 R remembers

Russian experiment

on Mars 98 using

Lidar (Alt#1).

1.29[F] Team expert & Alt#3b firm work to

make adaptations to Alt#3b.

1.31 [F] Data

exchanged by

e-mail/phone to

converge on final

solution

1.7 [S] R contacts

engineers, scientists, rover

operations, researchers

via Internet, friends.

1.14 [BE] R asks other team

members to see husband’s data to

1.23 [BE] CSA agrees

if advocacy comes fromCanadian scientists.

Evaluate Alt#3.

.

because all participants considered themselves inno-vators, had knowledge about the scientific market fortheir work as well as the current technologies thatserve the market, and had histories of innovating (asevidenced by the patents they held). Thus, they hadsignificant interest in innovating as well as the requi-site experience in their field to know what was inno-vative and what was not. Project factors also appearedto influence the degree of innovation desired. Therewas only a limited amount of funding and riskthat was acceptable. Innovators, together with otherproject members, decided which ideas (and AO prob-lems) would be handled in a more-or-less innovativefashion. Finally, the organizational factor of JPLaffected which problems to redefine: JPL was per-ceived in the science community as having a spe-cial expertise in space instruments (which was thefocus of the redefinition for the three most-innovativecases), rather than in space materials (the focus in thetwo least-innovative cases).As a first step in the KRI process, then, radical

reconceptualization was necessary to set the stage forinnovation to occur, regardless of whether the solu-tion would eventually be reused or new. The radicalreconceptualization established a challenging vision—

a goal that excited, motivated, and directed the scien-tists’ efforts to strive for an unknown future state.Simultaneously with their radical redefinitions,

reusers in the more innovative cases developed con-ceptual approaches which were ambitious and nottied to the past; they used analogies and extensionsto anchor the concepts. For example, the reuser in themost-innovative case used the analogy of an earth-based thunderstorm warning radar to describe aninstrument to detect dust devils on Mars:

Our focus was on developing an instrument to studydust devils on Mars. But the problem was how do youget a machine to tell when the dust devil has arrivedand turn the machine on to take pictures and mea-surements. Well, I thought about radars at airports.Isn’t this what radars do? But radar can’t be used herebecause it has a much longer wave length and needsharder things like airplanes. Lidar, however, could beused. Point it at the sky, swivel it around, and tellwhether there is a dust devil. This is a novel use ofLidar.

To use analogies and extensions required reusersin the more-innovative cases to not only be knowl-edgeable about traditional approaches, but also to beaware of, and open to, nontraditional approaches that

Majch

rzak,C

ooper,an

dNeece:

Know

ledgeReuse

forInnovation

180Managem

entScience

50(2),pp.174–188,©2004

INFORMS

Table 3 Brief Summary of Actions Taken in Each Case

Reconceptualize Decide to Scan for reusable Briefly evaluate In-depth analysis Develop ideaproblem (C) search (DS) ideas (S) ideas (BE) of ideas (IA) fully (F)

Case 1. Lidar design (most innovative)R reads AO and rejects traditional

approach as insufficient.R redefines problem innovatively to

not just characterize meteorologicalphenomenon but also create earlywarning system.

R uses analogy (radar) to suggest newapproach.

R acknowledges tooexpensive to inventwithout reuse.

R searches alternative functions,providers, geographies, andconditions of Lidar; makescontacts through Internet, AO,colleagues.

R briefly evaluates 4 ideas forexistence of data, source,and adapter attesting toideas’ credibility, relevance,and adaptability.

R and adapter examinedata, models, andstudies on 2 ideas toidentify adaptationsrequired.

R and adapter generateand evaluate prototypesof possible adaptations.

R and adapter worktogether to implementadaptations into finalsolution.

Case 2. ELM designR reads AO and rejects traditional

approach.R redefines problem innovatively

as creating a new measurementdevice.

R uses analogies (pressure gauges,robot arm) to suggest newapproach.

R checks with expert to evaluateimplications of concept forgenerating useful science.

R recognizes currentexpertise insufficient.

R searches seemingly unrelatedindustries (semiconductor,ESD, British textiles) usingInternet, catalog, and expertsto generate ideas aboutelectrometers.

R finds a handheld electrometerwith which to experiment.

R connects with electrostaticcommunity.

R briefly evaluates 4 ideasfor existence of drawings,source, and adapterattesting to ideas’ relevance,credibility, and adaptability.R decides will adapthimself.

R (as reuser and adapter)conducts experimentsand analyses of dataand models for6 prototypes of Alt #2b.

R generates and evaluatesprototypes.

R (as reuser andadapter) meets withsources (KSC and EScommunity) to createfinal solution.

Case 3. AFM designR reads AO and rejects traditional

approach.R considers new approach (AFM) not

used in space before.R conducts quick experiment to

evaluate concept as feasible.

R decides time too shortto invent ownsolution; needs to useexternal AFMmanufacturer.

R uses variety of methods(Internet, phone, visits) tosearch outside spacecommunity to find adapter andadaptation ideas.

R briefly evaluates 5 ideasfor existence of prototypes,source, and data attesting toideas’ credibility, relevance,and adaptability.

R and adapter examinedata and models ofAlt #4c.

R and adapter generateand evaluateprototypes.

Weekly teleconsbetween adapter andreuser to create finalsolution.

Case 4. AFM tipsR reads AO and defines problem as

the challenge of needing toautonomously change tips (notdone previously).

No traditional approach available.R considers new approach

(replacing tips).

R decides time too shortto invent ownsolution.

R uses variety of methods(Internet, phone, visits) tofind anyone doing tip arrayresearch even though notfor Mars.

R briefly evaluates 3 ideasfor existence of prototypes,data, and source attestingto ideas’ credibility,relevance, and adaptabilityapplications. R decides willadapt himself.

R (as reuser and adapter)examines data, models,and prototypes for2 ideas (Alt #1a, 1b).

R finalizes tip arraydesign in discussionswith source.

Majch

rzak,C

ooper,an

dNeece:

Know

ledgeReuse

forInnovation

Managem

entScience

50(2),pp.174–188,©2004

INFORMS

181

Table 3 (Continued).

Reconceptualize Decide to Scan for reusable Briefly evaluate In-depth analysis Develop ideaproblem (C) search (DS) ideas (S) ideas (BE) of ideas (IA) fully (F)

Case 5. ELM materialsR reads AO and defines problem

innovatively as selecting materialsto embed in ELM.

Traditional approach of usingrepresentative materials doesn’twork.

Consider new approach (materials asreferences).

R decides time tooshort to invent ownsolution.

R uses variety of methods(catalogues, KSC database,visits) to find materials thatmay be used as referencefor Mars.

R searches seemingly unrelatedindustries (space suits,semiconductor, and Britishtextiles) to generate ideasabout electrostatic materials.

R briefly evaluates 2 ideas forexistence of source,adapter, and data attestingto ideas’ credibility,relevance, and adaptability.R decides must be ownadapter.

R (as reuser and adapter)examines data, models,and prototypes fromsource on 4 ideas(Alt #2, 3, 4, 5).

Generate and evaluateprototypes.

R has close discussionswith source (KSC) toyield concurrence onmaterial selection.

Case 6. mag patches (least innovative)R reads AO.R accepts definition of problem in AO

as conducting dust experiments inways done previously.

R accepts traditional Mars approachby basing experiment on magneticpatches.

R decides not to inventown experimentbecause mag patchesalready exist andinsufficient funds topursue alternatives.

R uses scientific journals to getnames of scientists who havedone traditional approachin past.

R verifies benefits, costs, andrisks of using mag patchesfrom scientific journals anddiscussions with source.

Source agrees to adapt.

R and adapters exchangespecificationrequirements foradaptations, jointlydevelopingprototypes andconductingexperiments.

Adapters transferadapted artifacts andlessons learned to R.

Summary of similarities and differences between more- and less-innovative reuse casesSame: Read AO; knowledge about

traditional approaches; generate aconceptual approach beforedeciding to search and proceedingto evaluate specific ideas.

Different: Rs in more innovative casesaware of nontraditional approaches;redefine problem innovatively anddevelop a conceptual approachwhich is ambitious, not tied to past,that postpones detailedconsideration of constraints.

Same: Decide not toinvent own solutionand to seek outreusable ideas wheninsurmountable perfgap exists.

Different: None.

Same: Engage in scanningbehavior to identify potentiallyreusable idea(s).

Different: Rs in more-innovativecases conduct broadersearches of nontraditionalcommunities, using variety ofsearch methods.

Same: Assess each idea interms of credibility,relevance, and adaptability;make assessments basedon existence of supportivedata without examining datain detail; identify need foradapter role.

Different: More-innovativecases evaluate multipleideas.

Same: Examine andmanipulate data,models, and prototypes;interactions with sourceor adapter to generateprototypes.

Different: None.

Same: Discussionsbetween reusers andadapters.

Different: More-innovative casesfocus discussionson jointtechnologydevelopment;less-innovative casesfocus discussion ontechnology transfer.

Note. Cases are listed from most innovative to least innovative.

Majchrzak, Cooper, and Neece: Knowledge Reuse for Innovation182 Management Science 50(2), pp. 174–188, © 2004 INFORMS

might lead to greater levels of innovation. For exam-ple, in the second-most innovative case (electrometer(ELM) design), the reuser generally knew of devicesthat measure electrostatic properties on earth. Thisinspired him to believe that there may be a way tomeasure electrostatic properties on Mars and thuspropose such a conceptual approach. He extended thebasic principles via an analogy to other measurementdevices such as pressure gauges, and by using a robotarm as an analogy for the motion of an astronaut onthe surface of Mars.In sum, at this early reconceptualization stage,

individuals needed to balance the apparent paradoxof suggesting wildly ambitious conceptualizations,while having confidence that someone, somewhere,would have an idea that would help them opera-tionalize their ideas.

Decision to SearchHaving reconceptualized the problem and approach,the respondents proceeded to the next stage: thedecision to initiate a search for reusable knowledge.Unlike Szulanski’s (2000) decision to transfer, this sec-ond stage involved individuals considering whetherto invent their own solution—a clearly preferredstrategy as inventors—or to examine others’ ideas forpossible reuse. For our respondents to even considerexamining others’ ideas required that they acknowl-edge that they did not have the personal expertise inthe required area and that their aim in searching oth-ers’ ideas was not simply to support personal learning(so they could invent), but to actually reuse another’sideas. We found that before engaging in a search forreusable knowledge, our respondents needed to expe-rience an “insurmountable performance gap,” result-ing from severe time and/or cost constraints. Forexample, in both the most- and least-innovative cases,inventing one’s own solution was deemed too expen-sive. In the words of the scientist involved in the mostinnovative (Lidar) case:

The major problem was the cost cap. Full-up develop-ment would have broken the bank.

In the other cases, there was insufficient time toinvent a solution. Only with the insurmountable per-formance gap were the respondents willing to admitthat they could not invent their own solution and thatthey would therefore consider reusing others’ ideas.

Search and Evaluate StageHaving decided to consider reusing others’ ideas,respondents proceeded into a stage of active search-ing. In this third stage of the knowledge reuse pro-cess, we found that three layers of search and analysiswere needed for reuse to occur:(1) Scanning the environment to become aware of

possible ideas,

(2) Conducting brief evaluations of ideas to deter-mine if the idea was worth pursuing, and(3) Conducting an in-depth analysis of the idea.

Scanning. Respondents in all six cases engagedin scanning behavior, during which ideas and indi-viduals with potential relevance to the conceptualapproach were identified. Reusers in the more innova-tive cases conducted much broader searches extend-ing into nontraditional areas, resulting in ideas thatcame from sectors beyond their immediate space com-munity, did not fit the immediate functional require-ments, or did not have the expected or neededform. For example, in the second-most innovativecase (ELM design), ELMs to measure the electrostaticproperties of dust (the conceptual approach) did notexist at the outset of the timeline. Therefore, the reusercould not simply look for the right ELM. Instead, hesearched for information about how ELMs work ingeneral and how various industries use ELMs. One ofthe ideas he discovered from this search was that theBritish textile industry worries a great deal about theelectrostatic properties of chair covers; consequently,the industry routinely uses ELMs to measure the elec-trostatic properties of various materials. Thus, a spacescientist developing an instrument for Mars reusedideas from the British textile industry! In the wordsof this scientist:

I worked by analogy. I looked around to see what oth-ers were doing in the field: � � � semiconductor indus-try, electrostatic discharge industry. [There are] a num-ber of companies that deal with clean room gar-ments; chair covers [that result in] minimal static buildup. [There was] some help from the textile industry[for example, an individual] � � � from the British textileindustry.

In addition to broad search criteria, reusers in themore-innovative cases used a wide variety of searchmethods ranging from the Internet to face-to-face vis-its, using both strong and weak ties. For example, inthe most-innovative case (Lidar design), strong andweak ties, formal “introductions” via the AO, tele-phone, e-mail, and the Internet were all used to findpeople and artifacts related to Lidar.

Conducting Brief Evaluations. Having scannedthe environment to become aware of others’ ideas,the reuser proceeded to briefly evaluate each selectedidea. Three criteria were applied to decide if an ideashould be either discarded or continued into an in-depth analysis: credibility (the idea is valid and repli-cable), relevance (there is some degree of match withproblem needs in terms of form, such as shape andmaterials; fit, such as size and weight; and/or techni-cal functionality), and adaptability (the extent to whichthe idea can be modified to fit the new problemwithin time and cost constraints). When evaluating

Majchrzak, Cooper, and Neece: Knowledge Reuse for InnovationManagement Science 50(2), pp. 174–188, © 2004 INFORMS 183

for adaptability, we found that reusers assessed notjust if the idea could be adapted, but more impor-tantly for them, who (source, recipient or a third party)would do the adaptation. During this layer of search,the reuser was often trading off the cost, capabilities,and interests of a source or third party to do the adap-tation against the cost, capabilities, and interests ofthe solo reuser. For example, in the AFM tip arraycase, one reusing scientist commented on the value offinding an adapter:

They had an operational system of tip arrays � � � and afabrication process to make them [even though it hadnot been used on Mars or for our purpose]. This isa huge step forward. We immediately knew that weshould team up with [the adapter] as it would savetime and energy.

To assess the credibility, relevance, and adaptabil-ity of the ideas identified during the scanning step,we found that reusers looked for data, models, proto-types, and other contextual cues (such as whetherthe concept had been flown in space previously).We refer to these cues as the metaknowledge forthe reused idea. Similar to metadata, meaning “dataabout the data,” which is used to facilitate retrieval(Heery 1996), we define metaknowledge as “knowl-edge about the knowledge,” which is used to facil-itate evaluation and use (e.g., to assess relevance,credibility, or adaptability). While types of meta-knowledge such as document author and date (e.g.,Tiwana 2002) were of value to the reusers in the sixcases, the types of metaknowledge that were morevaluable were physical artifacts such as data, models,and prototypes.Though metaknowledge was used to evaluate

ideas, there were too many ideas identified during thescanning step for metaknowledge to be closely exam-ined for each idea. Therefore, during the brief eval-uation phase, the reusers were primarily interestedin determining if the metaknowledge even existed;they inferred from the existence of the metaknowledgethat the idea would be judged positively. For exam-ple, when reusers learned that they could access thedata or prototype for an idea, they inferred that theidea was more credible than an idea without data—even before they analyzed the data. When reuserslearned that they could find someone (such as amanufacturer) able to adapt an idea, they judgedthe idea to be more adaptable than ideas withoutadapters—even before they spoke with the adapter.When reusers learned that contextual data describingthe constraints and environmental conditions underwhich the original knowledge was generated existed,they inferred that the idea was more reliable. Thisfinding suggests that reusers do not need or wantaccess to metaknowledge when initially evaluating

ideas for reuse; rather, they only want to be informedthat the metaknowledge exists and can be accessed ata later point in time. In the words of one reuser:

The key [at this point in the evaluation process] wasnot the availability of the instrument but [knowing] thefact that the instrument was in development.

Conducting In-Depth Analysis. Finally, ideas thatsuccessfully met the three criteria for use (credibil-ity, relevance, and adaptability) progressed into thefinal layer of search: in-depth analysis. The goal ofthis layer was to determine if any of the ideas beingconsidered could in fact be adapted to meet the prob-lem as formulated. In this final layer, reusers accessedand manipulated the metaknowledge to test it againstthe constraints and challenges of the problem. Forexample, a hand-held commercial ELM for the ELMdesign case, a software model for the Lidar case, andlaboratory prototypes for the AFM design case wereacquired and manipulated to provide the necessaryconfidence for the reuser to commit to the devel-opment approach. During this layer of activity, thereuser often needed to turn to the source and adapterfor advice and artifacts.

Summary of Three Layers. This pattern of threelayers of evaluation suggests that a reusable idea mustsuccessfully traverse “gates” to be eventually reused.It must first be found on the “scanning radar,” thenit must meet criteria expected during a brief evalua-tion, and finally it must continue to show promise ina more in-depth analysis. Thus, for knowledge to bereused not one, but three evaluations must take place.

Full DevelopmentIn the final stage, the continued ideas were developedand incorporated into a final solution. This stage ischaracterized by the full commitment of the team tothe chosen implementation approach. The work nowshifts from “Is this feasible?” to “Make it happen.”It is at this point that sharing common experiencesbetween reusers and either the source or adapterbecomes valuable. Shared experiences took the formof sessions in which prototypes were tested, reviewed,and improved. In the words of one reuser:

We worked with the people at Kennedy Space Cen-ter to design the electrometer experiments on variousmaterials. While working with the data was important,it was equally if not more important to have a lot ofdiscussions and meetings.

While shared experiences were important for allsix cases, the focus of the shared experience var-ied depending on the degree of innovation: for theleast-innovative case, the shared experience focusedon transferring best practices, while for the more-innovative cases, the shared experience focused on

Majchrzak, Cooper, and Neece: Knowledge Reuse for Innovation184 Management Science 50(2), pp. 174–188, © 2004 INFORMS

codevelopment of the solution. Moreover, who wasinvolved in these shared experiences varied acrossthe cases. Sources were only included in these sharedexperiences in four of the cases; in the other cases,reusers shared experiences with adapters.

DiscussionIn summary, Figure 2 presents a staged KRI processthat was discernible across the six reuse cases. Thisprocess included six stages:(1) reconceptualize the problem and approach for

innovation,(2) decide to search for reusable ideas,(3) scan for reusable ideas,(4) briefly evaluate reusable ideas,(5) conduct in-depth analysis on reusable ideas and

select one, and(6) fully develop reused idea.Within each stage, reusers involved in the more-

innovative cases behaved differently from thoseinvolved in the less-innovative cases. Thus, eventhough the unique culture of JPL may seem to limitthe generalizability of the findings, radical inno-vation is by its very nature unique (Leifer et al.2000). Developing atom-sized computers or biotech-nology innovations require equally unique culturesand individuals. As revelatory cases (Yin 1984), then,these six cases from JPL provide the opportunity todevelop new theory about knowledge reuse for inno-vation that focuses attention on reuse for all radi-cal innovation processes. Therefore, the utility of thisresearch should be judged by the degree to whichit fosters new insights and stimulates new ques-tions and propositions for future research (Eisenhardt1989). We use these questions and propositions tosummarize key points as well as to extend our find-ings into new areas.

Figure 2 Model of Knowledge Reuse Process for Innovation

Search and Evaluate

Reconceptualize

Problem for

Innovation

Decide to

Search

Scan Briefly

Evaluate

AnalyzeIn Depth

FullyDevelop

Awareness of traditionaland nontraditional

Experience

insurmountable

performance gap

Access t ometaknowledgeConduct broad,

nontraditional search

Shared

experience with

adapter

Awareness that

meta-knowledge

exists

A Process Model for KRI. Our findings first sug-gest a process model for KRI. This model is composedof six stages starting with Reconceptualize the problemand ending with Develop the idea. That is, despite thenonlinear, chaotic nature of radical innovation, a six-stage model was identifiable. This suggests our firstproposition:

Proposition 1. An individual who proceeds throughall six stages in the manner described in Figure 2 is morelikely to reuse others’ ideas in ways that foster innovationthan individuals who skip any stage or perform any stagedifferently from how it is portrayed in Figure 2.

The uniqueness of the JPL context suggests thattesting this proposition is worthy of future study.In addition, we examined only cases in which reuseactually happened, as opposed to cases in whichinnovation occurred without reuse, or failed to occurbecause of improper reuse. As such, we cannot con-clude from our findings that these six stages are bothnecessary and sufficient for fostering reuse. Futureresearch that makes such comparisons is needed.

Three Levels of Search. One of the features of ourstaged process is the three levels of search: initialscan for possible reusable ideas, brief evaluation, andin-depth analysis. Clearly, such a three-tiered strat-egy benefits innovation by allowing innovators to beexposed to a diverse range of inputs quickly beforethey converge on a single idea, but it may be ineffi-cient if a known point solution is desired, such as withKRR. This raises a proposition for future research:

Proposition 2. Reusers who are interested in knowl-edge reuse for innovation have a greater need to proceedthrough the three layers of scanning, brief evaluation, andin-depth analysis than reusers interested in replication.

Role of Adapters. Our research found that thereare three criteria used to decide to reuse knowl-

Majchrzak, Cooper, and Neece: Knowledge Reuse for InnovationManagement Science 50(2), pp. 174–188, © 2004 INFORMS 185

edge for innovation: relevance, credibility, and adapt-ability. While the first two criteria have been foundelsewhere (e.g., Szulanski 2000), the criterion of adapt-ability has not been mentioned in the literaturepreviously (although Rice and Rogers suggest, asearly as 1980, that more research attention shouldbe paid to the reinvention process). We found thatthe absence of credible adapters, or the absence ofways to quickly determine if adapters are availableand credible, can create a barrier to reuse for inno-vation. This suggests, then, that if an idea is to bereused for innovation, adapters need to be readilyidentifiable. Thus, instead of simply storing an idea ina knowledge repository, names of agents (which maybe manufacturers, institutions, or other researchers)willing and interested in adapting this idea shouldbe stored as well, or included as part of a discussionforum on the idea. This notion needs further testingand thus we propose:

Proposition 3. An innovator who considers how anidea will be adapted and who will do the adaptation is morelikely to reuse ideas to facilitate innovation than reuserswho ignore these issues during the reuse process.

Role of Metaknowledge. Our findings that meta-knowledge plays a different role at the different stagesin the KRI process is another area for future research.We confirmed previous research that metaknowledge(describing context, credibility of source, etc.) abouta potentially reusable idea is important to reusers(Markus 2001). We also confirmed previous researchthat embodying this metaknowledge in physical arti-facts such as models, data, and prototypes instead oftext facilitates understanding and reuse (e.g., Star andGriesemer 1989). The additional insight provided byour research is that metaknowledge is accessed differ-ently at different stages in the reuse process. We foundthat when reusers first became aware of an idea theywanted only to know that the metaknowledge on theidea existed. It was only later when reusers conductedin-depth analyses of the idea that they accessed andmanipulated this metaknowledge. This suggests thatto facilitate knowledge reuse for innovation, the deci-sion about what metaknowledge to capture shouldbe based on which metaknowledge would provide,by its mere presence, some evidence for the credi-bility, relevance, and adaptability of the idea. Thisis a notion worthy of future research and thus wepropose:

Proposition 4. Ideas that are structured to indicatethe presence (or absence) of metaknowledge will be morereadily considered by reusers during the KRI process thanideas that are not so structured.

If this finding that awareness of metaknowl-edge is more important than acquisition at early

stages in the knowledge reuse process generalizesto other sites and research studies, this could sug-gest how information overload might be avoided.Reusers can be initially provided with a checklistof what metaknowledge exists on the idea, allow-ing access to the metaknowledge when the reuserreturns to the idea for a more detailed analysis.This may also suggest that metaknowledge can beconstructed and presented in a way to facilitate orinhibit knowledge reuse. Borrowing from Culnan(1983), we call this structuring the “chauffeuring”of knowledge through the reuse process, with meta-knowledge serving as chauffeur. Just as a technology-transfer professional is able to chauffeur an idea frominitial awareness to implementation (Rogers 1983),or Toyota’s knowledge-sharing network is able tochauffeur Toyota suppliers from awareness to adop-tion of best practices (Dyer and Noveoka 2000), prop-erly constructed metaknowledge might be able tochauffeur an idea through the knowledge reuse pro-cess and increase its chances of being reused. Meta-knowledge, viewed in this way, becomes the “bound-ary object” that links the source to the reuser (Starand Griesemer 1989). To facilitate reuse, each ideacould be coupled with its metaknowledge, organizedaround evaluation needs, and layered either for mereindication of presence or for access and manipulation.

Factors Affecting Reuse. We found many factorsthat affected KRI, including searching nontraditionalcommunities of practice, use of a variety of searchmethods, weak ties as well as strong ties, and sharedexperiences with sources and/or adapters (but pri-marily only at the end of the process, not at the begin-ning as found by Szulanski 2000). Additional researchon these factors is needed to confirm that they arespecific to KRI.We also found that innovators were motivated

to consider reusing others’ knowledge only when(1) they confronted a problem that was insurmount-able with their current knowledge and resources,(2) they reconceptualized the problem and approachto require an ambitious new perspective, and (3) theybelieved that existing ideas were likely to be foundsomewhere that would be useful. While the effectsof the first two factors have been found previously(Gupta and Govindarajan 2000, Osterloh and Frey2000), the third factor has important implicationsfor building a theory of KRI. Why one innova-tor believes ideas exist that would be useful, whileanother believes the opposite, is not clear. It maybe, as Allen (1977) observed, that those who believethere are existing ideas that would be helpful havebeen exposed, through experience and networking,to a wide range of inputs. Alternatively, as Leiferet al. (2001) recently observed, these individuals mayhave been exposed not just to a range of inputs but

Majchrzak, Cooper, and Neece: Knowledge Reuse for Innovation186 Management Science 50(2), pp. 174–188, © 2004 INFORMS

to a particular type of input, an “opportunity rec-ognizer.” These are individuals who recognize thebusiness potential of an innovator’s idea and whomotivate the innovator to pursue his or her ideas,including sharing them with others. These individu-als are different from traditional technology gatekeep-ers or brokers because they are focused on helpingsources pursue their own interests. Innovators whomake contact with an opportunity recognizer may, inthe course of a discussion, learn enough about whatvarious other sources may be doing or what vari-ous business opportunities may exist to increase theirbelief that an idea useful for their approach is likelyto exist. Our findings suggest that the notion of anopportunity recognizer can be applied to the KRI pro-cess. In this role, the individual helps innovators rec-ognize both that a conceptual approach has validityand that reusable ideas are likely to be found with areasonable search strategy. This raises a propositionfor future research:

Proposition 5. Innovators who are encouraged to pur-sue innovation through reuse by opportunity recognizersare more likely to reuse others’ ideas than innovators with-out that encouragement.

Role of Project-Level Decisions. Our findings sug-gest that the decision to reuse others’ ideas to solve aparticular problem needs to be examined with respectto decisions made about other problems in the sameproject or organizational unit. In our sample, theamount of innovation desired by a reuser for a par-ticular case was constrained in part by the amount ofinnovation incurred by the entire project. Too muchrisk in too many areas might damage the viabilityof the project. For example, in the least-innovativecase (mag patches), the reusers intentionally chose tolimit the innovativeness of the solution due to thedegree of risk already present in other aspects of theproject. This suggests that knowledge reuse researchshould broaden its focus beyond characteristics ofthe specific best practices being transferred or thespecific knowledge and participants involved in thetransfer process. This broadening of focus should takeinto account the web of decisions being made by theindividual, project colleagues, and project managers.Just as Argote and Ingram (2000) suggest that thetransfer process needs to be understood within anembedded network of elements, our findings suggestincluding projectwide decisions into this network,especially for innovation. We phrase this implicationas a proposition:

Proposition 6. A key determinant of reuse for inno-vation in any one case will be the degree of innovationincurred in other cases within the same project or organ-izational unit.

Reconciling Our Findings with Previous Re-search. We argued at the outset of this paper thatSzulanski’s (2000) model was unlikely to provide agood fit with an innovation environment because themodel was more linear in flow, and more determinis-tic about the knowledge being transferred, than is typ-ically experienced in radical innovation. Our findingssupport this initial argument by identifying a modelquite different from that of Szulanski.There are, however, commonalities between our

model and that of Szulanski: they both begin withthe recognition of a gap, end with integration of thetransferred knowledge into an organizational routine(or finalized solution), and include characteristics ofthe recipient as important factors affecting the trans-fer process. Despite these similarities, there are manydifferences: Szulanski says little about the searchand evaluation activities that were so critical to ourreusers, and the role of source and recipient are dif-ferent (in our study, recipient characteristics such asan openness to nontraditional approaches were criti-cal early in our process, and shared experiences withthe source were needed at the end, while Szulanskifound the opposite to be true).The differences between our model and Szulanski’s

may be caused by Szulanski’s assumption that meta-knowledge and alternative reusable ideas were gath-ered prior to the decision to transfer, and didn’tmodel it. We offer an alternative explanation forthe differences—and one we find more generative oftheory building. We suggest that KRI requires moreattention to the search and evaluation stages becausethe idea that is transferred is not committed to untillate in the process. It may be that, when integrationand innovation are the drivers of the process, thereuse process focuses on how the problem is con-ceptualized and what alternative reusable ideas canbe identified. As such, characteristics of recipientsare critical because the recipients define the level ofaspiration (March and Shapira 1987) that motivatesthem to find innovative ways of achieving the desiredobjective, even if it means not inventing it themselves.In contrast, KRR focuses on the practices being trans-ferred, and thus recipient traits may be less importantearly on.This suggests that the two models and the two

streams of research on knowledge reuse (for innova-tion and for replication) can be reconciled by stringingthem together into a single model. Our process modelcan be used to describe early stages of the reuse pro-cess and Szulanski’s model can be used to describelater stages. That is, early in the reuse process wheninnovation may be desired the reusers devote effortsto problem definition and search activities. Later inthe process when routinization of the selected idea isdesired the best practices underlying this idea (if best

Majchrzak, Cooper, and Neece: Knowledge Reuse for InnovationManagement Science 50(2), pp. 174–188, © 2004 INFORMS 187

practices exist) can be transferred by removing asmany of Szulanski’s barriers as possible. By stringingthe two models together, we may be able to explainthe difference in the impact of recipient and sourcecharacteristics between the two studies. It may be thatthe recipient’s and source’s characteristics are influen-tial at several points in a complete reuse process: earlyin the innovation phase and later in the routinizationphase for the recipient, and the inverse for the source.This suggests that future research should consider theentire lifecycle of the transfer process—from innova-tion to routinization—to understand the process bywhich transfer occurs.

ConclusionIn this study, we have offered several research ques-tions and propositions to stimulate future research onKRI. We have suggested that reusers in a KRI pro-cess act differently from those in a KRR process. Theunique context of JPL raises a challenge for futureresearch to test the generalizability of these propo-sitions. In the JPL context, the reusers redefined theproblem to provide the greatest value to the customer(scientist and funder) relative to competitors. We pro-pose that these same drivers exist in most industries;we would expect to see similar results in other con-texts of radical innovation. We found that reusers inthe JPL context balanced the paradox of identifyinga nontraditional untested conceptual approach to theproblem against the need for risk reduction by pick-ing only those approaches in which they had someconfidence that someone, somewhere, would have arelevant idea. We propose that risk reduction is amajor factor in most radical innovations and thuswould expect to see similar results in future studies.In the JPL context, we found reusers engaging in threelevels of search and evaluation that require a broadsearch of nontraditional communities of practice andthe ability to quickly scan metaknowledge. Creativityresearchers (e.g., Amabile 1996) have long argued forthe need to search in nontraditional communities; theuse of metaknowledge to do this searching quicklydeserves further study. Is the set of metaknowledgewe identified a highly contextualized one? Finally,we found in the JPL context that reusers sought outadapters (who were not often the source) to bridgethe gap between a source’s original idea and the finalsolution. Whether this finding is generalizable is aninteresting question. In the JPL context, the reuser andadapter were often not the same person because skills,funding, and experience for the two were quite differ-ent. In other fields and industries, the two roles mayoverlap. Is there some characteristic of a disciplinethat drives adapters’ and reusers’ roles?

In addition to generating research questions forfuture work in this area, our findings are also prac-tical. First, due to the importance of being able to usemetaknowledge of potentially reusable ideas, suchmetaknowledge that quickly communicates credibil-ity, relevance, and adaptability needs to be capturedand presented. Second, because reusers for innova-tion need to define problems broadly, be aware ofboth traditional and nontraditional approaches, con-duct broad and nontraditional searches, and use avariety of search methods, our research suggests thatorganizations should consider providing training andincentives to their innovators in these areas. Finally,new roles such as adapters, chauffeurs, and opportu-nity recognizers need to become part of the commu-nity of practice encouraging knowledge reuse.In conclusion, this research is intended to ground

knowledge transfer and reuse research in a relativelyunexplored context: innovation. In this context, asGrant (1996) so aptly explains, the intention is notspillover, replication, or acquisition, but recombina-tive integration. Knowledge is clearly being reused,but how? This study is an effort to address this ques-tion and stimulate future research.

AcknowledgmentsThe research described in this paper was carried out bythe Jet Propulsion Laboratory (JPL), California Institute ofTechnology, under contract with the National Aeronauticsand Space Administration. The authors thank the MECAand MITCH teams and the Knowledge Management Projectteam at the JPL. The authors also thank Rajiv Sabherwal,M. Lynne Markus, T. Ravichandran, and the anonymousreviewers for their helpful comments. An earlier version ofthis paper received the 2001 Academy of Management BestPaper Award for the Organizational Communication andInformation Systems Division.

ReferencesAlavi, M., D. E. Leidner. 2001. Review: Knowledge management

and knowledge management systems: Conceptual foundationsand research issues. MIS Quart. 25(1) 105–136.

Allen, T. J. 1977.Managing the Flow of Technology: Technology Transferand the Dissemination of Technological Information Within the R&DOrganization. MIT Press, Cambridge, MA.

Amabile, T. M. 1996. Creativity in Context. Westview, Boulder, CO.Appleyard, M. M. 1996. How does knowledge flow? Interfirm pat-

terns in the semiconductor industry. Strategic Management J.17 137–154.

Argote, L. 1999. Organizational Learning: Creating, Retaining, andTransferring Knowledge. Kluwer, Norwell, MA.

Argote, L., P. Ingram. 2000. Knowledge transfer: A basis for com-petitive advantage in firms. Organ. Behavior Human DecisionProcesses 82(1) 150–169.

Armbrecht, F. M. R., R. B. Chapas, C. C. Chappelow, G. F Farris.2001. Knowledge management in research and development.Res. Tech. Management 44(4) 28–48.

Becerra-Fernandez, I., R. Sabherwal. 2001. Organizational knowl-edge management: A contingency perspective. J. ManagementInform. Systems 18(1) 23–55.

Majchrzak, Cooper, and Neece: Knowledge Reuse for Innovation188 Management Science 50(2), pp. 174–188, © 2004 INFORMS

Blackler, F. 1995. Knowledge, knowledge work and organizations:An overview and interpretation. Organ. Stud. 16(6) 1021–1046.

Brown, J. S., P. Duguid. 1991. Organizational learning and commu-nities of practice: Toward a unified view of working, learning,and innovation. Organ. Sci. 2 40–57.

Cheng, Y.-T., A. H. Van de Ven. 1996. Learning the innovation jour-ney: Order out of chaos? Organ. Sci. 7(6) 593–614.

Clark, H. 1996. Using Language. Cambridge University Press,New York.

Clark, K. B., T. Fujimoto. 1991. Product Development Performance.Harvard Business School Press, Boston, MA.

Culnan, M. J. 1983. Chauffeured versus end user access to commer-cial databases: The effects of task and individual differences.MIS Quart. 7(1) 55–67.

Davenport, T. H., S. L. Jarvenpaa, M. C. Beers. 1996. Improvingknowledge work processes. Sloan Management Rev. 37(4) 53–65.

Dougherty, D. 1992. Interpretive barriers to successful productinnovation in large firms. Organ. Sci. 3 179–202.

Dyer, J. H., K. Noveoka. 2000. Creating and managing a high-performing knowledge-sharing network: The Toyota case.Strategic Management J. 21 345–367.

Eisenhardt, K. M. 1989. Building theories from case study research.Acad. Management Rev. 14(4) 532–550.

Grant, R. M. 1996. Prospering in dynamically competitive environ-ments: Organizational capabilities as knowledge integration.Organ. Sci. 7(4) 375–387.

Gray, P. H. 2000. The effects of knowledge management systems onemergent teams: Towards a research model. J. Strategic Inform.Systems 9 175–191.

Gupta, A., V. Govindarajan. 2000. Knowledge flows within multi-national corporations. Strategic Management J. 21 473–496.

Hansen, M. T. 1999. The search-transfer problem: The role of weakties in sharing knowledge across organization subunits. Admin.Sci. Quart. 44 83–111.

Hargadon, A., R. I. Sutton. 1997. Technology brokering and inno-vation in a product development firm. Admin. Sci. Quart.42 716–749.

Heery, R. 1996. Review of metadata formats. Program 30(4) 345–373.Holsapple, C. W., K. D. Joshi. 2000. An investigation of factors

that influence the management of knowledge in organizations.J. Strategic Inform. Systems 9 235–261.

Kelly, G. A. 1970. Behaviour is an experiment. D. Bannister, ed.Perspectives in Personal Construct Theory. Academic Press, NewYork, 255–269.

Klein, H. K., M. D. Myers. 1999. A set of principles for conductingand evaluating interpretive field studies in information sys-tems. MIS Quart. 23(1) 67–94.

Kogut, B., U. Zander. 1992. Knowledge of the firm, combinative

capabilities, and the replication of technology. Organ. Sci.3 383–397.

Leifer, R., C. M. McDermott, G. C. O’Connor, L. S. Peters, M. P.Rice, R. W. Veryzer, M. Rice. 2000. Radical Innovation. HarvardBusiness School Press, Boston, MA.

Leonard, D., S. Sensiper. 1998. The role of tacit knowledge in groupinnovation. California Management Rev. 40(3) 112–131.

March, J. G., Z. Shapira. 1987. Managerial perspectives on risk andrisk taking. Management Sci. 33(11) 1404–1418.

Markus, M. L. 2001. Toward a theory of knowledge reuse: Typesof knowledge reuse situations and factors in reuse success.J. Management Inform. Systems 18(1) 57–93.

Nonaka, I. 1994. A dynamic theory of organizational knowledgecreation. Organ. Sci. 5(1) 14–37.

Osterloh, M., B. S. Frey. 2000. Motivation, knowledge transfer, andorganizational forms. Organ. Sci. 11(5) 538–550.

Polanyi, M. 1966. The Tacit Dimension. Doubleday, New York.Rice, R. E., E. Rogers. 1980. Reinvention in the innova-

tion process. Knowledge, Creation, Diffusion, Utilization 1(4)499–514.

Rogers, E. 1983. Diffusion of Innovations, 3rd ed. The Free Press,New York.

Schultze, U., R. J. Boland. 2000. Knowledge management tech-nology and the reproduction of knowledge work practices.J. Strategic Inform. Systems 9 193–212.

Star, S. L., J. R. Griesemer. 1989. Institutional ecology, “transla-tions,” and boundary objects: Amateurs and professionals inBerkeley’s museum of vertebrate zoology, 1907–39. Soc. Stud.Sci. 19 387–420.

Swan, J. 2001. Knowledge management in action: Integratingknowledge across communities. Proc. 34th Hawaii Internat.Conf. System Sci. Oahu, HI.

Szulanski, G. 2000. The process of knowledge transfer: A diachronicanalysis of stickiness. Organ. Behavior Human Decision Processes82(1) 9–27.

Teece, D. 1981. The market for know-how and the efficient interna-tional transfer of technology. Ann. Amer. Acad. Political Soc. Sci.458 81–96.

Thomke, S. 1998. Managing experimentation in the design of newproducts. Management Sci. 44(6) 743–762.

Tiwana, A. 2002. The Knowledge Management Toolkit. Prentice Hall,Upper Saddle River, NJ.

Unsworth, K. 2001. Unpacking creativity. Acad. Management Rev.26(2) 289–297.

Yin, R. 1984. Case Study Research. Sage, Beverly Hills, CA.Zander, U., B. Kogut. 1995. Knowledge and the speed of the transfer

and imitation of organizational capabilities: An empirical test.Organ. Sci. 6(1) 76–92.