' Academy of Management Execulive, 2002, Vol, 16, No. 3

Organizational designs for R&D

Geiaidine DeSanctis, Jeiiiey T. Glass, and Jngiid Moiiis Ensing

Executive OverviewResearch and development is becoming increasingly business-oriented, and corporate

reliance on new technology and innovation is greater than ever. How can R&D activitiesbe organized to yield the greatest value for the enterprise? A study of 14 leadingtechnology-intensive companies in six industries illustrates how three organizationaldesigns are being used to manage distributed, flexible R&D organizations: (I)decentralized. (2) networked, and (3) integrated. Decentralized designs limit the role of acentral R&D group: they direct R&D resources more toward products and markets ratherthan basic scientific activities. Networked designs push the boundaries of R&D outsidethe confines of a central R&D group and the firm as a whole. Integrated models usesophisticated communication linkages to tie centralized, science-based activities with thebusiness needs of the corporation. Within the three design approaches, there are designvariants, each with its positive and negative tradeoffs. We show how some designs forR&D are more effective than others and provide suggestions for how companies in searchof an ideal organization design for R&D can select among possible configurations topromote adaptive, value-oriented R&D organizations.

How can R&D activities be organized to yield thegreatest value for the enterprise? In an era whencorporate strategy and survival can dependheavily on a firm's ability to innovate rapidly, withmaximum business impact and rigorous cost con-trol, organization design issues have moved to theforeground.' The dilemma is classic: how to simul-taneously yield value from centralized and decen-tralized forms of organizing; but the context is new:a heightened pace of innovation, reduced productlife cycles, a global marketplace, growing relianceon intellectual property rather than hard-core tech-nologies, and the massive infusion of electroniccommunication systems for creating and coordi-nating knowledge work.^ Organizations are mov-ing from reliance on hierarchy and rigidity of struc-ture to new forms that are flatter, cross-functional,and dynamic.

How can R&D activities be organized toyield the greatest value for theenterprise?

Within this context, R&D activities are increas-ingly business-oriented and capital-intensive.Firms seek to be both innovative and efficient.

global and local, disciplined and flexible. Master-ing paradox is paramount.^ The head of R&D forGlaxoSmithKline recently put the issue this way:"We need to be big and small at the same time."^More than ever, companies seek value from theirR&D initiatives. The potential payoffs from organ-izing R&D effectively are enormous and the costs ofineffective organization structures extremely high.Executives confront a quandary: how to constructan ideal organization design for R&D among a seaof possible configurations.

Trends and Tensions

Historically R&D operations were centralized, andthis organization design persisted long after WorldWar II. Even as organizations grew, diversified,and spread across the globe, R&D remained cen-tralized for many companies. Well-documentedproblems, however, arose out of centralized struc-tures, among them weak links between R&D effortsand meeting the needs of customers and productlines. Product development cycles were perceivedto be too slow and R&D costs too high. R&D wasviewed as overly scientific and out of touch withthe business enterprise. By the late 1980s, mostlarge technology companies had decentralizedR&D operations, breaking down (though not neces-

55

56 Academy of Management Executive August

sarily eliminating) corporate laboratories andmoving R&D activities closer to business units.^ Amove from a central R&D (CRD) design to a busi-ness unit R&D (BU-R&D) design shifted the strategyof R&D away from one of hope toward one of ac-tion—that is, from heavy emphasis on pure re-search to heavy emphasis on product development.^

Without doubt, R&D in technology companieshas become more business relevant. But organiza-tion design choices for R&D are not without con-troversy. Classic tensions in R&D managementpersist. (See Figure 1.) Companies continuallystruggle with how best to reconcile the competingpressures associated with organizing R&D aroundscience versus organizing R&D around products ormarkets. Executives wish for the benefits of bothscience and business-oriented R&D; however, de-veloping both centralized and business-unit R&Dgroups is costly, and coordination needs becomemore complex as R&D activities spread throughoutthe enterprise.'' The organization design dilemmais that decentralized structures and formal ac-countability for R&D are more likely to bring morenew products and incremental innovations, butmajor technology advancements are more likelywhen R&D is centralized and informal.^ These ten-sions are evident in industries as varied as chem-icals, electronics, aerospace, communications, andPharmaceuticals.

Central R8ED

(CRD)

• Services the needs of thecorporation

• Permits long-range thinking

• Brings frame-breaking sciencelo product innovation

• Encourages risk taking

• Attracts leading-edge scientistsand engineers

• Emphasizes research

• The balance of power withinthe management structuretends toward centralization

Business Unit R&D(BU-R&D)

• Services the needs oithe market or productline

• Pushes short-termresults

• Develops technology inresponse to businessneeds

• Promotes productsuccess

• Attracts business-savvytechnologists

• Emphasizesdevelopment

• The balance ol powerwithin the managementstructure tends towarddecentralization

FIGURE 1Competing Organization Design Pressures

in R&D

Companies continually struggle withhow best to reconcile the competingpressures associated with organizingR&D around science versus organizingR&D around products or markets.

Insights from Leading Companies

We examined 14 leading technology companiesfrom six major industries with the goal of identify-ing the various organizational structures that firmsare using to arrange their R&D units. We soughtout large, profitable, multinational firms withstrong, established reputations for technologyleadership and significant R&D investments. All ofthe firms rely on technology to support their corebusiness strategies, and all have been operationalfor at least 10 years. On average, the firms westudied spend over $1.7 billion annually, or 8% ofrevenue, on R&D. They employ over 122,000 peopleand generate approximately $303,000 of sales an-nually per employee. With one exception the com-panies have a history of maintaining large corpo-rate laboratory facilities and attracting the bestscientists and engineers in their industries.^ Weidentified the dominant organization designs forR&D in these firms. We compared and contrastedthe firms on some basic financial parameters andthen documented in detail how centralization-decentralization tradeoffs were managed via or-ganization design. Because theorists emphasizeparadox in new organizational forms, we were es-pecially interested in the structures used to meetbasic science and product development needs si-multaneously.^°

Organization Designs for R&D

We observed that three general models are beingused today for design of R&D: (1) decentralizedmodels, (2) networked models, and (3) integratedmodels. Decentralized models place R&D close toor within BUs, limiting the role of any centralizedR&D function. Networked models link BUs to R&Dsources inside and outside of the organization. Inte-grated models combine centralized with decentral-ized R&D structures and use special communicationmechanisms to tie these structures together. All ofthe models address current pressures to simulta-neously meet the competing R&D goals of long-termresearch and short-term product development.

Figure 2 summarizes some basic financial andrelated parameters and shows the dominant organ-ization design associated with each of the firms in

2002 DeSanctis, Glass, and Ensing 57

Company'

FEABAverageLPCAverageGIMTDXHAverageOverall

Average

Organization Designfor R&D

DECENTRALIZEDDECENTRALIZEDDECENTRALIZEDDECENTRALIZED

NETWORKNETWORKNETWORK

INTEGRATEDINTEGRATEDINTEGRATEDINTEGRATEDINTEGRATEDINTEGRATEDINTEGRATED

Annual R&DExpenditures^

22793524

39492

15723648130810201992193050461297267480770135

17841768

AnnualRevenue^

3390023200151282363

186483014739150

845825918

100469816678648

262731844148091300

3451528140

TotalEmployees

251900234000

70400150000176575i41600lGlOOO

1500085867

293000291067

31800645003902925000

2000106628122164

Firm's Percent of IndustryR&D Expenditures

37.04%5.96%7.92%0.34%

12.82%6.17%

16.51%1.73%8.14%

31.37%8.54%

16.37%4.53%

10.18%2.56%0.41%

10.57%10.69%

Firm's Percentof IndustryRevenue

13.70%3.53%

11.61%1.41%7.56%4.59%

18.67%1.29%8.18%

40.60%12.43%4.12%4.00%8.80%2.87%0.18%

10.44%9.13%

ValueRatio*

01.4702.5320.8127.9213.1805.2511.2049.7222.0604.4205.0007.9213.7022.1344.75

216.1944.8730.93

' Companies E, L, C, T, and I are in the information and electronics industry. Companies P, D, and M are in the chemicals industry.Companies F and G are in the machinery and equipment industry. Companies B and X are in the medical/pharmaceuticals industry.Company A is in aerospace, and Company H is in the industrial materials industry.

^Millions of U.S. dollars.^Millions of U.S. dollars." The ratio is calculated as ((Firm's % of Industry Revenue/Firm's % of Industry R&D)' l,000,000)/total employees. The larger the ratio,

the more revenue the firm generates Irom R&D expenditures, relative to industry averages and controlling for firm size.

FIGURE 2Organization Design, Size, and Financial Measures for the 14 Companies

our study. The integrated model is the most popu-lar, and this is probably due to its many advan-tages relative to the other design types, as weshall discuss below. Note that there is no dominantorganization design type as a function of industry.For example, firms in the information and electron-ics industry fall across the three organization de-sign categories.

The data in Figure 2 reflect a few general pat-terns. Firms that adopt the decentralized approachto organization design have smaller R&D budgets,generate less revenue, and employ more peoplethan firms adopting the other two organization de-sign models. Firms that adopt the network organi-zation design are smaller, relatively speaking, interms of number of employees, and spend more onR&D than firms using the decentralized and inte-grated models. Firms using the integrated designare mid-range in the size of their R&D budgets andnumber of employees, yet they generate the high-est corporate revenue, on average. To account forvariations in R&D budgets and revenue across in-dustries, we calculated for each firm its percent-age of overall industry R&D expenditures and itspercent of industry revenue." So, for example.

firms using the decentralized designs spend anaverage of 12.8% of their industry's R&D budget,yet generate only 7.6% of industry revenue. In con-trast, firms using integrated designs expend about10.5% of industry R&D budgets and generate aboutthe same portion of industry revenue.

As a simple measure of each firm's ability togenerate value from R&D within its industry, wedivided each firm's percent of industry revenue byits percent of industry R&D expenditures. To ac-count for firm size, we then divided the result bythe firm's number of employees.'^ The larger thisratio, the more revenue the iirm generates fromR&D expenditures, relative to industry averagesand controlling for firm size. The data in Figure 2are displayed in order of increasing values of thisratio for each organization design type. On aver-age, integrated designs are associated withgreater value generation from R&D, but it is impor-tant to note that there is wide variation in valueratios within each design category. Looking acrossthe three right-hand columns of Figure 2, it seemsthat integrated and network designs are associ-ated with lower cost and greater value generationfrom R&D relative to decentralized designs. The

58 Academy of Management Execufive August

variation between companies within each organi-zation design category, however, suggests thatspecific implementation of the models variesacross firms. Value generation is possible withineach organization design approach, perhaps de-pending on the specific ways in which firms applythe models. With the data in Figure 2 as a back-drop, we can now look more deeply into the organ-ization designs of the 14 companies.

Integrated and network designs areassociated with lower cost and greatervalue generation irom R&D relative todecentralized designs.

(1) Decentralized Models

Four of the fourteen firms maintain a decentralizedapproach to R&D management, though their strat-egies and approaches to decentralization suggestthree forms. (See Figure 3). Companies A and Badopt the first form, whereas companies F and Euse the second and third forms, respectively. Notethat companies A and B are more successful in

their use of the decentralized approach—in termsof achieving value relative to their industries. (Seethe value ratios in Figure 2.)

Decentralized designs hold in common the factthat they limit the role of a central R&D group. Inthe purest form of decentralization, there is no cen-tiai R&D at all. All R&D activity takes place withinthe BUs. R&D groups across BUs may or may not becoordinated through a central office, such as theChief Technology Officer. This purely decentral-ized approach puts R&D close to the customer orproduct and is consistent with what Miles andSnow labeled a defender strategy.'^ Scientific andengineering expertise is directed toward the spe-cific product and/or market domains of the busi-ness, and any innovation effort tends to be highlytargeted. If R&D budgets are held relatively low,monies can be directed toward projects with highpayoff, thus yielding value from R&D activities.Despite these advantages, there is little learningor leveraging of R&D capabilities across the enter-prise as a whole, because the R&D group of onebusiness unit is not likely to interact with the R&Dgroups of other BUs. Research support for firm-wide strategic initiatives is difficult if not impos-

NoCRD

BU 1R&D

BU2R&D

BU2R&D

BU2R&D

Dispersed

Chief TechnologyOlficer

Segmented CRD

BU 1

BU2

BU3

BU4

CRD

BU 1 R&D

BU 2 R&D

BU 3 R&D

BU 4 R&D

Figure Legend

= person

( I = department or division

= cross-functional team

= computer system

FIGURE 3Decentralized Models for R&D Organization Design

2002 DeSanctis. Glass, and Ensing 59

sible. Economies of scale from R&D spending arenot available, since laboratories and scientists areduplicated and not shared across the enterprise.

Companies E and F have attempted to overcomesome of the downsides of the no-CRD approach butwith mixed success. Company F adopts a seg-mented CRD model in which R&D personnel asso-ciated with specific markets or products are con-solidated into a central location but operateindependently of one another. CRD is essentially ashell operation, a physical location and/or logisti-cal unit for managing R&D personnel and relatedresources. The major advantage of the segmentedmodel over the no-CRD model is that it increasespossibilities for knowledge transfer across theR&D groups of the various BUs. Co-location canfacilitate informal collaboration and joint problem-solving. Further, this model offers cost advantages.Scientists and engineers in co-located laboratoriescan be managed together, thus streamlining hir-ing, training, and other personnel costs. On theother hand, co-located BU-R&D groups risk becom-ing out of touch with the BUs that they were de-signed to serve.

Company F is not achieving high value from itssegmented model, perhaps because the firm's R&Dstrategy is not consistent with its organization de-sign. Company F is pursuing what Miles and Snowwould call a prospector strategy; the firm is veryaggressive with regard to searching for opportuni-ties beyond immediate customer and productlines, and the BU-R&D groups are having troublestaying aligned with BU needs due to their distantlocation from them. R&D expenses are very highrelative to industry averages, and value from R&Dexpenses is difficult to achieve. Like the no-CRDapproach, a segmented model may be better usedwith a defender approach to R&D strategy, since inthe segmented model R&D remains organizedalong product or market-oriented lines.

The dispersed mode! for R&D management scat-ters R&D throughout the enterprise, anywhere andeverywhere there is deemed to be opportunity andbusiness need. Company E in our sample exempli-fies this model. BUs that need and can afford R&Dcreate R&D groups. Some BUs may have more thanone R&D group. Other BUs may have no R&D.Cross-unit product-strategy groups form on an as-needed basis to develop and/or support new tech-nologies. In the dispersed model, there may or maynot be a CRD. Where a CRD exists, its role is not somuch to coordinate R&D efforts as it is to take onspecial projects that fulfill particular product-development needs that cannot be met inside ofthe BUs. In Company E there is no central coordi-nation or control unit for R&D. Rather, there is

continual (though not monitored or required) infor-mal information exchange among the many, diffuseR&D activities. Informal information exchange oc-curs via electronic discussion groups, seminars, andworkshops made available to technical employees.

The dispersed model for R&Dmanagement scatters R&D throughoutthe enterprise, anywhere andeverywhere there is deemed to beopportunity and business need.

The advantage of the dispersed approach is thatit fosters a culture of product innovation and ex-perimentation. The model fits a prospector ap-proach to R&D strategy. As one manager in Com-pany E put it, "New product ideas come fromanywhere and everywhere." The problem, ofcourse, is that ensuring adequate information ex-change is difficult when operations are scattered,and good ideas and research resources can besquandered in the fray. Recognizing this potentialweakness. Company E has established an "ideasdatabase" into which technology developmentideas from throughout the enterprise can be placedfor later use by others. Still, the reliance on aninformation system to integrate R&D information,without an organizational unit or other authorityrole, is problematic. The dispersed model risks be-coming decentralization run amuck, and althoughvalue through growth may be achieved in the longrun, it is an extremely expensive approach to or-ganization design.

(2) Networked Models

Much is written today extolling the value of theso-called network organization design in whichthe walls of corporate divisions or BUs—or indeedthe organization itself—are opened, allowing fluidexchange of information across corporate and BUboundaries.''^ For R&D, networked models facili-tate transactions-on-demand in which R&D re-sources are applied when and where they areneeded, regardless of where the technical capabil-ities lie^whether inside a CRD, inside a BU, oroutside of the organization. Networked models arecompatible with the prospector approach to R&Dstrategy, as they deliberately aim to link basictechnology developments to business needs. Firmsadopting a networked approach create and nurturea set of relationships to link basic technologysources to business demands. As business needschange, new relationships are established, and so

60 Academy of Management Executive August

there is flexibility in R&D ventures. We observedthree networked models. (See Figure 4.)

The internal-maiket model relies on informal in-teraction between CRD and BUs whereby BUsscout for relevant projects within CRD, and CRDgroups, in turn, seek BU sponsors for basic scienceinitiatives. Specific CRD groups are not necessar-ily designated to work with specific BUs; rather,linkages are developed depending on how a re-search group's technology or expertise can be ap-plied to the BU. In Company L, where we observedthis model in action, no single person within CRDis assigned to manage CRD-BU relationships; theincentives for all of the researchers are designed toencourage proactive, targeted relationships withBUs. In turn, BUs are free to approach researchersin CRD with requests for technical assistance orfull-blown research projects. Project milestonesand costs are negotiated to meet the specific needsat hand. Extensive CRD-BU interaction is promotedvia face-to-face meetings, regular visits to projectteams, job rotation between CRD and BUs, confer-ences and discussion groups around research ar-eas, and various social functions. Ongoing inter-action between BU and CRD personnel increases thelikelihood of successful initiatives. An incentivestructure can also be used to reward employees whotake initiative to cross the BU-CRD boundary.

The acquisition model is based on forming R&Drelationships with entities outside the firm. Tech-nologies are imported in accordance with thefirm's strategy and specific needs, and then inte-grated into the firm's existing product-develop-ment efforts inside BUs. The acquisition model in-cludes the purchasing of technology in thebroadest sense, whether research data or methods,R&D staff, or even entire organizations surround-ing the technology. Several companies in our sam-ple use the acquisition approach on a limited basisto acquire risky, cutting-edge technologies fromsmall and start-up firms. Only one company westudied—Company C—uses acquisition as itsdominant organization design. The acquisition ap-proach allows it to keep its R&D costs relativelylow and its value ratio high as acquisitions sub-stitute for a corporate CRD function, transferringexternal R&D resources to BUs as they are needed.The advantage of this outreach model is that itreduces the internal need for scientific staff andlaboratories and, more importantly, facilitatesrapid fulfillment of specific R&D needs. It also re-duces the risk of R&D investments since acquisi-tions can be made after desired technology devel-opment milestones have been achieved. On theother hand, identifying and developing externalrelationships require special business and legal

Internal-Market Model

Extended-Enterprise Model

Acquisition Model

(Company 1]

R&D done in acquisitions

Corporate R&DPartnerships

Figure Legend

University R e s e a r c h ^ C D = P^'^™I [ = department or division

= external entity

FIGURE 4Networked Models lor R&D Organization Design

2002 DeSancfis, Glass, and Ensing 61

skills. Further, the company is dependent on avibrant external marketplace to meet its internalR&D needs.

The exfended-enferprise model pushes the net-work notion further to include formation of R&Drelationships with a wide set of internal and exter-nal partners. For example, CRD may form alli-ances or joint R&D ventures with government lab-oratories, universities, or other companies; theR&D department of one BU may form an agreementto support the R&D needs of another BU; or the CRDmay work with university laboratories to supplyspecific R&D needs of several BUs. The idea is toform R&D relationships that meet H&D needs, nomatter where the R&D source might be located.CRD may broker these relationships for the BUs,but BUs are not bounded by CRD and can formrelationships with external parties or with one an-other on an as-needed basis.

There has been a general trend among technol-ogy companies toward external sourcing of R&Dfrom universities, start-ups, private labs, alliances,and so on.'^ Among the firms we studied. CompanyP is using the extended enterprise as its dominantorganization design for R&D. Many others practiceexternal sourcing on a limited basis, relying moreheavily on internal CRD and/or BU-R&D resources.The findings from our interviews suggest thatthe extended-enterprise model requires organiza-tional incentives that foster coordination and col-laboration efforts across organizational bound-aries. For example, both BU and CRD staff shouldfeel free to identify and create research contractsor partnerships without concern for turf wars whendealing with a common external lab or researchsite.

There has been a general trend amongtechnology companies toward externalsourcing of R&D from universities, start-ups, private labs, alliances, and so on.

In addition, effective external partnerships re-quire recognition of parties' common and differingneeds, and the ability to coordinate technologyplans and negotiate conflicts as they arise. Sophis-ticated legal, technical, and business skills areneeded to manage a wider range of relationshipsthan in the internal market or acquisition models.But the payoff can be high for companies that areable to manage a multitude of ventures. The ex-tended-enterprise approach is particularly attrac-tive for simultaneous support of basic scientificresearch along with fast-paced growth; and rela-

tionships can be scaled back in times of economicdownturn.

In a sense, networked models replace the tradi-tional, centralized CRD organization with moderntwists: either an open CRD or location of basicscientific activities outside of firm boundaries.Successful implementation of networked modelsrequires a host of corporate capabilities—monitor-ing and support of the internal and/or externalR&D marketplace, careful review and selection ofavailable technologies, and strong internal social-ization and leadership skills to broker relation-ships. R&D success is dependent on the network ofrelationships. Networked models offer high poten-tial value, but they are tricky to implement, espe-cially as dominant organization designs.

(3) Integrated Models

Integrated organization designs for R&D incorpo-rate both short-term product development initia-tives and basic science initiatives with long-runpotential. Integrated models practice a "mixedmode" strategy for innovation, and CRD takes on akey governance role in implementing this strategy.The strong governance role of CRD differentiatesfirms with integrated organization designs fromcompanies with dominantly decentralized or net-worked models. CRD is not necessarily large insize, but it takes on an R&D leadership role for thecorporation. CRD serves as an important hotbed ofscientific energy and as a hub in the wheel of R&Defforts throughout the enterprise. There is a strongphilosophy that "technology belongs to the corpo-ration, not the business units." The goal of CRD isto help BUs "steal freely," as one manager in ourstudy put it, by developing cross-unit technologyplatforms and meaningful information sharing inorder to "look for exploitations and assure that thecompany will not be blindsided by new technologydevelopments." CRD may have some laboratoriesthat are isolated to pursue basic research, but CRDas a unit is anything but isolated. It has vibrantties to the BUs. CRD links BUs with one another,and it links firm R&D efforts with the company'soverall, strategic goals.

Among integrated models, the typical CRD in-cludes dedicated research project teams, laborato-ries, and functional support groups. Some com-panies have only central R&D facilities (e.g..Companies G, I, M, T, and X), whereas others alsohave BU-R&D groups (e.g.. Company D and Com-pany H.) Both approaches can be effective. Forexample. Companies X and H are yielding greatvalue from R&D, yet they have different balancesof BU and CRD resources for R&D. Company X has

62 Academy of Management Execufive August

a large CRD unit with many laboratories and func-tional groups. CRD is responsible for supportingthe R&D needs of the corporation and promotingnew product development, both short term andlong term; almost no R&D is conducted inside ofBUs. In contrast. Company H conducts R&D ven-tures largely within specific BUs or via cross-BUR&D projects rather than inside CRD. The CRD unithas no laboratories, but it does have a visible,prestigious group of managers and scientists whoset leadership direction and facilitate firm-widescientific support for R&D ventures. In both Com-pany X and Company H, CRD is directly funded bythe corporation and not from a "tax" or charge-back system to BUs on a project basis.

Within the integrated-design approach, thechoice of whether to develop CRD as a large orsmall entity is secondary to the decision of how tocoordinate extensive corporate-wide R&D efforts.We observed two general approaches to structur-ing interaction between CRD and the BUs. (SeeFigure 5.) Team-based coordination involves for-mal mediating groups that consist of BU and CRDpersonnel. For example. Company H has standingtech teams and tech services groups that includeBU members and CRD members organized aroundspecific types of technology. Usually these teamsinclude members from multiple BUs. As other ex-amples. Company M has joint BU/CRD groups as-sociated with each major product line, and Com-pany X creates ad hoc project teams with CRDpersonnel to address specific BU research needs.

Liaison-based coordination occurs when manag-ers within CRD are given designated responsibil-ity for linking CRD to the BUs. As an example.Company G appoints relationship managerswithin CRD to serve as the interface between CRDand BUs. There is one designated relationshipmanager per BU. These relationship managershave no line responsibility, but they are providedwith enormous implicit authority by reporting tothe head of CRD. They are an integral part of thestrategic planning for CRD, and each relationshipmanager is evaluated in part by the success ofjoint CRD-BU projects. Relationship managers whohave been transferred to CRD from a BU view theposition as a rotation that is good both for the BUand CRD. Other relationship managers have beenlong-time CRD researchers who are virell informedabout corporate R&D capabilities.

Regardless of whether teams or individual liai-sons serve to mediate CRD-BU relationships, asthey work with a particular BU on technical issuesthese facilitators become knowledgeable aboutR&D strategy and specific BU needs. This outcomeencourages a high probability of matching BUneeds to CRD capabilities and R&D activities tocorporate strategic directions.

Along with a formal coordination structure, con-stant communication is a cultural imperative infirms adopting integrated organization designs.Most firms that we studied use R&D informationsystems that are accessible by both BUs and CRD.The information collected for these systems may

Team-based coordination Liaison-based coordination

Cross-unittechnology teams

f R&D information support system J

Figure Legend

^ ^ = person

I = department or division

. = cross-functional team

I = computer system

FIGURE 5Integrated Models

Relationshipmanagers I R&D information support system J

2002 DeSanctis, Glass, and Ensing 63

include market intelligence, technology discussiondatabases, information on competitors' R&D initi-atives, research data, R&D project tracking sys-tems, and so on. Though it may be tempting forcompanies to rely on these systems in place ofmore traditional forms of communication, wefound that the most successful firms rely asheavily on people-based communication mecha-nisms as on computer-based systems. For exam-ple. Companies D, G, H, and I use R&D informationsystems support, but they take care to design andsupport mediating groups and managers forknowledge transfer. Further, they promote a cul-ture of communication via multiple methods, suchas bulletins, conferences, and discussion groups.Information systems are used to support the R&Dorganization design, but they do not replace it.

Along with a formal coordinationstructure, constant communication is acultural imperative in firms adoptingintegrated organization designs.

A Value-Driven Approach

Figure 6 summarizes the major advantages, dis-advantages, strategy, and other implementationguidelines for the three major models of R&D organ-ization design. Reviewing these models, we findthat there is no "one best" organization design.Indeed, a host of factors might dictate the specificmodel implemented by any given firm, such asfirm goals, resources, existing infrastructure, mar-

OrganizationDesign Advantages Disadvantages

Preferred R&DStrategy

To implement torsuccess . . .

DECENTRALIZED

NETWORK

INTEGRATED

• Supports a businessorientation in R8ED

• Focuses R&D efforts oncurrent customersneeds

• Short-term benefits fromR&D investments areemphasized

Can be used to supportbasic scientific researchat lower cost thantraditional CHDEffective matchingprocess puts researchresources when andwhere they are neededRelationships areflexible and can beexpanded or reduced asresearch or businessneeds change

Mixed mode of businessand science orientationin R&DSupports both short-and long-term R&DprojectsHelps link R&D to thestrategic direction o(the technology-basediirm

Difficult to shareknowledge or leverageR&D capabilities acrossthe enterpriseResearch support for lirm-wide strategic initiativesis diificullEconomies of scale inR&D facilities, personnel,etc., are difficult toachieve

Requires business, legal,and social skills fordeveloping effectiverelationships betweenparties in the networkNetwork dependencies arecreated, thus increasingrisk if parties withdraw ordo not meet commitmentsTends to be more costly ifnetwork reaches onlyinside the firm (internalmarket model)

Requires cross-functionalteams or designatedrelationship managers tocoordinate CRD-BUrelationshipsCRD can be high relativeto most decentralized andnetwork models

Defender:Innovation istargeted towardimprovingexisting productsor searching fornew technologieswithin specificproduct or marketlines

Prospector: newproduct andmarketopportunities areaggressivelypursued;exploration ofnew technologiesis wide in scope

Analyzer: a mixedmode approachto strategy, withtargeted R&D forsome products ormarkets andaggressivepursuit of newopportunities inother areas

• Use when R&D needsare decidedly differentacross business units

• Avoid this approach if aprospector strategy is tohe pursued

• Keep R&D expenditureslow through carefultargeting of projectswithin business units

• Use to support growth innew markets or newproduct ventures

• Troll for valueopportunities that lieoutside of the firm(acquisition or extended-enterprise models)

• Invest in skills andprocesses to supporteffective management ofrelationships

• Invest in coordinationstructures andcommunicationmechanisms to link CRDwith BUs

• Empower CRD with agovernance role for R&D

• Build a reputation forR&D excellence that isrecognized inside andoutside the firm

FIGURE 6Summary of Advantages, Disadvantages, Strategy, and Implementation Guidelines for the

Organization Design Models

64 Academy of Management Executive August

ket opportunities, and so on. Most important is thatexecutives recognize the tradeoffs associated withthe design choices they make and take steps torealize specific design advantages and overcomepotential disadvantages. We synthesized the at-tributes of effective R&D management from acrossthe three design types found in our sample to sug-gest the general recommendations for organiza-tion design that follow.

(1) Strategic Mechanisms

Regardless of the specific corporate strategy, val-ue-driven companies adopt practices to link R&Dventures and progress with the corporation's stra-tegic goals. In Company H, for example, R&D ap-pointments are considered important and given anelite status. Company D has established an intel-lectual property department with intellectual assetmanagers who undertake valuation analysis foreach R&D project. The value of the technology tothe market and/or to the firm is assessed as aproject is approved and as funding is renewed.Company M uses information systems to bringR&D progress and updated technical informationto customers. Company A has a chief growth offi-cer who scouts for high-potential projects from BU-R&D units and then advocates them for executive-level sponsorship

(2) Executive-Level Attention

To extract value from whatever organization de-sign is implemented, the executive level of the firmdevotes significant attention to R&D. This attentionfeeds downward and outward into all R&D opera-tions. Executive involvement can be very heavy,such as in Company G where R&D projects of $1Mor more require executive council approval. Insome companies a corporate executive council de-termines all R&D initiatives. The CRD director mayalso be the chief technology officer and serve on anexecutive council (e.g.. Company G) or even on thecompany's board of directors {e.g.. Company P).Executive appointments and ongoing involvementin R&D reflect the recognition that nearly everyaspect of the company is technology-based. InCompany H the CEO sends out weekly reportsprepared by R&D, adding comments and stressingthe importance of various R&D activities to every-one in the company. As one executive put it, "Re-search is the basis of our business, not somethingwe do on the side."

(3) An Entrepreneurial. Communications-OrientedCulture

Value-driven organizations engage in continualinteraction between the BUs and R&D groups. Theynot only create relationship managers, projectteams, R&D information systems, and so on; theyuse these mechanisms constantly. This constantlevel of communication contributes to a sense ofpartnership between general managers and R&Dmanagers and reinforces the importance of inno-vation as a value in the firm. In this way, themanagerial matrix between BUs and R&D operateswell beyond the reporting structure and deepwithin the company.

This constant level of communicationcontributes to a sense of partnershipbetween general managers and R&Dmanagers and reinforces the importanceof innovation as a value in the firm.

(4) Putting Sharing Where It Matters

Value-driven firms overcome the isolation ofR&D from the rest of the company by promotingsharing where it matters, rather than haphaz-ardly or everywhere. For example, teams are cre-ated where the payoff is likely to be greatest, andchief technology officers or relationship manag-ers target high-probability projects. Most impor-tantly, formalities are replaced by expedited re-view procedures and ad hoc methods of selectingpeople for projects and getting work done withthe highest level of excellence possible. For ex-ample. Company G has replaced elaboratemethods for selecting and managing projectswith the use of simple evaluative spreadsheetsand project milestones that can be specified rap-idly, usually within a few hours. Company D hasreplaced pacing of project requests with R&Drequests on demand—whenever there is a need.One manager referred to this approach as a"pick-up basketball game" as opposed to the"formal league games" typical of previous years.

(5) Externally-Oriented R&D

Value comes from using capabilities both insideand outside of R&D, and both inside and outside ofthe firm, to create new technology-based options.Value-driven R&D organizations look for scientificand business opportunities outside traditionalR&D walls—to individuals, groups, companies.

2002 DeSanctis. Glass, and Ensing 65

and information systems both inside and outsidethe firm. Technology found in the marketplacemight be the basis of a new product line, provideincremental development, or fill in the missingpiece of an ongoing R&D effort. Though specificlaboratories and projects may operate in isolationfrom time to time, BU-R&D and CRD operations arerun with business objectives in mind, and manag-ers are customer or product oriented. They recog-nize the need for collaboration and engage in on-going scouting for new ideas and opportunities.

(6) Willingness to Mix and Match Structures

Varying R&D needs across a business enterprisecan require different degrees of CRD and/or BU-R&D attention, or different types of teams to coor-dinate R&D efforts. Dynamic economic conditions,corporate consolidations, or moves to spin off oracquire businesses likewise put pressure on R&Dorganizations to be flexible in their organizationdesigns and use different structural or communi-cation mechanisms as R&D demands change. Val-ue-driven companies meet these challenges byrecognizing that there is no one-size-fits-all ap-proach to organization design; they maintain acoherent overall structure for R&D but vary thespecific design attributes as needed. As one exam-ple. Company P in our study relies on an extendedenterprise model as its dominant organization de-sign but complements this with use of jointCRD/BU R&D teams for technology transfer. Bothtechnologists and general managers are willing touse different R&D structures for different productsor at different periods in time.

Achieving Value from R&D

To conclude, organization design does not directlydetermine corporate success, but it is one criticalfactor that comes into play in determining whethercompanies are able to yield value from their R&Dspending. Both science-oriented companies andbusiness-driven companies can use new organiza-tional forms to yield value. The integrated model isan excellent starting point for most firms as itaccommodates both product development and ba-sic science initiatives. From there, organizing R&Dfor value means pursuing centralized and decen-tralized R&D activities that collectively facilitatethe firm's overall strategic objectives.

To design mechanisms to accommodate scien-tific and business units in the firm, executives canreview the range of possible organization designoptions and identify the advantages and disad-vantages of each model in light of organizational

resources and goals. They can then take steps toimplement the kinds of structural, communication,and other mechanisms we have described to breakdown organizational boundaries and satisfy re-search and development needs when and wherethey arise. The result is an adaptive and progres-sive model of organizing for R&D directed towardachieving value for the enterprise.

Acknowledgment

Gerardine DeSanctis would like to Ihank INSEAD lor sponsor-ing her sabbatical leave during which this article wascompleted.

Endnotes

' Daft, R. L., & Lewin. A. Y. 1993. Where are the theories for the"new" organizational forms? Organization Science, 4(4): i-vi;and DeSanctis, G., & Fulk, J. (Eds.). 1999. Shaping organizafionform: Communication, connection, and community. NewburyPark, CA: Sage.

^ Drucker. P. F. 1999. Knowledge-worker productivity: The big-gest challenge. California Management Review, 41(2): 79-94;and Doz, Y.. Santos, J., & Williamson, P. 2001. From global tometanational: How companies win in the knowledge economy.Boston: Harvard Business School Press.

'̂ Child, J., & McGrath, R. G. 2001. Organizations unfettered:Organizational form in an information-Intense economy. Acad-emy of Management Journal, 44(6): 1135-1148.

•* Dyer, G. GSK considers spinning ofl its research units. Com-panies & Markets. Financiai Times. 25 January 2002, 15.

^Jankowski, J. E. 1998. R&D foundation for innovation, fle-search Technology Management. 41 (March-April): 20: and Lar-son, C. F. 1998. Industrial R&D in 2008. Research TechnologyManagement. 41(November-December): 19-24.

^ In this paper we use the term Business Unit (BU) to refergenerically to any product, customer, or other business-orientedgroup within the organization. Central R&D (CRD) refers to anystand-alone R&D group that reports to corporate headquartersrather than to a business unit. Business Unit R&D (BU-R&D)reiers to any R&D group that reports to a product, customer, orother business-oriented unit.

' Birkinshaw, I.. & Hagstrom, P. (Eds.). 2000. The flexible firm:Capability management in networked organizations. New York:Oxford University Press; Kreiner, K., & Schultz, M. 1993. Informalcollaboration in R&D: The formation oi networks across organi-zafions. Organization Studies. 14(2): 189-205; and Kuemmerle,W. 1997. Building effective R&D capabilities abroad. HarvardBusiness Review (March-April): 61ff.

^ Etflie, I. E., Bridges, W. P., & O'Keeie, R. D. 1984. Organiza-tion strategy and structural differences ior radical versus incre-mental innovation. Managemenf Science. 30 (June): 682-695:and Van den Bulte, C. & Moenaert, R. K. 1998. The effects of R&Dteam co-location on communication patterns among R&D, mar-keting, and manulacturing. Management Science. 44(11):S1-S18.

^ The exception is Company C. We conducted in-depth inter-views with key informants in each company. We also revieweddocuments and reports made available to us by the companiesand by publicly available sources such as annual reports, in-dustry publications, and SEC filings. The companies partici-pated in our study with the understanding that they would notbe identified by name.

66 Academy ol Management Executive August

'° Child & McGrath, op. cit.; and Heydebrand, W. V. 1989. Neworganizational forms. WorJc and Occupations, 16(3): 323-357.

" Industry figures are based on data provided by the Officeol Technology Policy of the U.S. Department of Commerce andthe Division of Science Resources Studies in the Directorate forSocial, Behavioral, and Economic Sciences ol the National Sci-ence Foundation. See U.S. Corporate R&D: Volume 1; Top 500Firms in H&D by Industry Category, September 1999; and U.S.Corporate R&D Investment, October 2001.

'̂ To aid interpretation of the value ratio, we multiplied eachratio by $1,000,000. which is the denomination of the revenueand R&D budget numbers shown in Figure 2.

'̂ Miles and Snow outlined four types oi corporate strategyfor innovation: reactors, defenders, analyzers, and prospectors.See Figure 6 for definitions. Miles, R., & Snow, C. 1978. Organi-zafionai strategy, structure, and process. New York: McGraw-Hill. This framework is widely used in research on corporatestrategy. For example, see Forte, M., Hoffman, J. J., Lamont, B. T.,& Brockmann, E. N. 2000. Organizational form and environment:An analysis oi between-form and within-form responses toenvironmental change. Sirafegic Management Journal. 21(7):753-773.

"'For example, see Ebers, M. (Ed.). 1997. The formation oiinter-organizational networks. Oxford: Oxford University Press;and Birkinshaw & Hagstrom, op. cit.

'̂ See R&D trends forecast for 2001. a report published by theIndustrial Research Institute, November 2000, htfp.V/www.irinc.or g I web I.

Gerardine DeSanctis is theThomas F. Keller Professor ofManagement in the FuquaSchool oi Business at Duke Uni-versity, Durham, North Caro-lina. Her interests are in thegeneral area of organizationdesign, especially the growingrole of information technologyin new forms of organizing. Sheholds a Ph.D. degree in busi-ness administration from TexasTech University. Contact: gd@mail.duke.edu.

Jeffrey T. Glass is the Joseph F.Toot, Jr. Professor of Engineer-ing and Co-Director of The In-stitute for the Integration oiManagemenf and Engineeringat Case Western Reserve Uni-versity. He received his Ph.D. inmaterials science and engi-neering from the University ofVirginia and his Global Execu-tive MBA from Duke University.Contact: jfg4@po.cwru.edu.

Ingrid Morris Ensing is a doc-toral student in management atthe Fuqua School of Business,Duke University. She holds anMBA from Kellogg GraduateSchool, Northwestern Univer-sity, with majors in organiza-tional behavior and strategy.She also holds a BA with honorsin economics from ClaremontMcKenna College. Her areas ofresearch include organiza-tional structure, organizationallearning, and knowledge man-agement. Contact: icm@maiJ.duke.edu.