A new structure for the product development organization, called the virtual research and development organization, is proposed. The model overcomes traditional communication and coordination hurdles of functional organizations, yet builds strong competencies by creating a product development organization with fluid membership and a stable core development group. The virtual research and development organization is an excellent example of an organization that facilitates organizational learning. Virtual research organizations provide a superior alternative to traditional R&D and cross-functional models of new product development by improving the overlap between innovation and organizational learning. The primary benefit of the virtual research and development organization is the leveraging of experiences and application of this knowledge to current development programs. By building on experience more effectively, firms should be capable of developing less costly, higher quality, and superior performing new products more quickly.

INTRODUCTION

Successful new product development is dependent on effective structuring and usage of resources within the firm. In studies of product development, researchers (Zirger & Maidique, 1990) have shown that a primary component of successful product development is the necessity for effective communication and coordination of activities between the functional groups, particularly Research and Development (R&D), engineering, manufacturing, and marketing. Unfortunately, traditional functional organizational structures which are used frequently by technology intensive organizations because of their strengths in promoting specialization and building of strong core competencies, often create subtle but fatal barriers to interaction across both internal units and external constituents. We propose a new structure for the product development organization which we call the virtual research and development organization. Our model overcomes traditional communication and coordination hurdles of functional organizations, yet builds strong competencies by creating a product development organization with fluid membership and a stable core development group. The proposed structure not only reduces obstacles to effective development of technically intense products, but also greatly facilitates learning across products.

TRADITIONAL R&D STRUCTURES

Attention is increasingly focused on the structure and integration of R&D organizations within technology intensive firms. In industries with significant technology content in their products, firms often have separate R&D groups that conduct advanced technology development, testing, and selection. Leonard Reich (1985) defines industrial research in this context as insulated laboratories who work toward advancing science and technology, yet are still responsive to long-term company needs. These groups are responsible for developing the next wave of technologies that will reinforce, or build, the firm's existing competencies to create new strategic opportunities for the firm. Technologies created within R&D are then passed on to product development groups that incorporate them into forthcoming product introductions. Because of the visionary and advanced nature of their work and the typical academia-like culture of these organizations, it is often difficult for firms to efficiently transfer new technologies to the firm's product groups and to learn from these advanced development groups. The structure of these R&D organizations relative to their primary supplier groups and product development teams further heightens integration and learning hurdles.

R&D organizations traditionally use one of two forms of organizational structure: centralized or decentralized structure. A centralized structure typically consists of a separate R&D group; one that has its own buildings, labs, and staff, that reports directly to the president and is structured as a cost center (Hax & Majluf, 1984). Within the R&D lab, scientists and engineers are usually grouped by core technologies or in basic science categories. Traditionally, R&D selects technologies to develop, in consultation with the product divisions, that extend existing technical capabilities or create new capabilities that can be the basis of radical new products. Once a technology's feasibility is proven, R&D looks within the product divisions for suitable customers to integrate the technology into a forthcoming product. When a product development team selects a technology from R&D for use in a new product, the R&D team passes on its accrued knowledge, largely embodied in a prototype or set of technical schematics, to the development team. It then becomes the developer's responsibility to finish the development process.

The decentralized R&D structure is very similar in internal structure and responsibilities to a centralized form, the difference being that each product group or Strategic Business Unit (SBU) has its own R&D organization (Hax & Majluf, 1984). In this form, technology selection can be more focused toward specific product applications since R&D is aligned with specific product groups. The R&D organization is also located in closer geographic proximity to the SBU allowing effective information sharing. R&D priorities are often selected in consultation with the SBU senior management and product managers. Finally, R&D activities can be funded in part by closely aligned SBUs.

Neither of these R&D structures are without problems. Centralized R&D groups can easily lose sight of product applications and customer needs. As organizational layers are added between the customer and R&D, R&D groups can misinterpret market requirements, miss, and even sometimes resist changes in the environment that could make existing skills obsolete. Additional organizational bureaucracy often results in R&D groups proposing technologies that are cost prohibitive or products that are not suitable, usable, or needed by the customer.

Centralized R&D groups spend significant time and resources selling their proposed technologies to SBUs in the organization (Souder, 1987). As new technologies are discovered and proven feasible, R&D units will search for product groups that could potentially use the new technology in a product application. Successful placement of new technologies is dependent on how well the R&D groups understand SBU needs, capabilities, priorities, and the level of cooperation between R&D and the SBU product development groups. If a SBU home is not found, a new technology will often languish within R&D, not exploited until an SBU champion emerges.

Finally, centralized R&D groups have major difficulties transferring new technologies to SBU product development teams. Transfer can be problematic because gaps in technical expertise and product applications knowledge often exist between the two parties, geographical distances exacerbate communication, and incentive systems impede cooperation.

An excellent example of these problems occurred in the International Business Machines (IBM) Corporation. IBM experienced all the classic problems associated with a centralized structure: IBM's technological know-how was so stagnant that IBM's R&D and product development groups could not see what could be done with data entry systems (Morris & Ferguson, 1993). The recent 'vital to IBM' program recognizes that transitions between R&D and product development are crucial to productivity (Coy, 1994).

In comparison, decentralized R&D and product development structures are significantly better at targeting R&D efforts toward specific product areas, yet have their own unique set of problems. First, separate R&D groups may make it more difficult to achieve a critical mass of technical talent in a specific technology or discipline. A critical mass of scientists and engineers are needed to achieve cutting edge developments. Second, distinct R&D groups are not recognized or rewarded for information sharing and, thus, must be encouraged to share knowledge and resources. Important technical discoveries from an SBU R&D group that has little use for the new technology can be lost by the organization even though these new technologies could potentially make a major contribution in other parts of the organization. Redundancy in research efforts across the organization can often occur with decentralized R&D structures because units have little incentive to cooperate and share know-how or discoveries. Since the SBUs compete against each other for resources it does not behoove an R&D group to give another unit a competitive advantage.

The case of the Dupont Corporation highlights the problems of a decentralized organizational structure (Wolff, 1986). Dupont had a decentralized development structure that assessed the value of new projects. The Polymer Products Development group was one of many SBUs within Dupont which conducted their own R&D and product development. The polymer group developed Selar, an important commercial resin, that was a major breakthrough in polymer technology. Unfortunately, the polymer group could not find a broad application of Selar. Furthermore, Selar was not directly related to any of Dupont's on-going businesses (i.e., other SBUs) at the time. Thus, without a champion in any of the existing SBU organizations, it was difficult to get Selar out of the R&D group and into a commercial applications SBU. From a macro perspective, the Dupont example also shows that SBU independence and interdependence is an important issue for selling projects. Important developments like Selar are costly if they are not utilized by other SBUs.

Decentralized and centralized forms also share common problems that limit their productivity and competitiveness in technology and new product development (Roberts & Fusfeld, 1988). Common to both centralized and decentralized R&D groups is the separation of R&D from product development. Separation of technology and product development organizations creates a learning barrier within the organization. With different organizations responsible for unique aspects of a product's development, experience tends to be captured and retained in functional groups, not organizationally. Retained learning within the functional groups impedes experience leveraging and sets the stage for reinvention of the wheel. Moreover, separation of R&D and product development creates a communication barrier (Souder, 1987).

Separation of the technology and product development functions have also resulted in development of radically different cultural and motivational norms. R&D organizations are often considered the intellectual elite within the organization. For example, a scientific instrument and electronics company would hire new engineers for their R&D groups categorizing them as lab quality engineers. Engineers who were not quite lab quality could be interviewed for other engineering positions in this company such as product or manufacturing engineering. Recognition and reward systems also accentuate the differences between R&D and engineering groups. R&D personnel are typically recognized and rewarded for their technical creativity (e.g., patents and papers). However, R&D's ability to link new technologies with product applications is often a secondary, not primary objective. Product development groups, on the other hand, are rewarded for their ability to make high quality products in a timely fashion that satisfy customer needs. These types of differences in cultures and goals contribute to the friction between development groups and enhances inefficiencies in the development process.

THE VIRTUAL RESEARCH AND DEVELOPMENT MODEL

The virtual research and development organization is substantially different from traditional forms of R&D and product development in several ways. It combines the R&D and product development functions into one organization and yet maintains some of the benefits of the corporate R&D function (see Figure 1). The virtual research and development organization consists of a cross-functional, crossdivisional product development team with R&D personnel that is organized for a specific project. The mission of the virtual research and development organization includes developing, assessing and selecting new technologies, as well as the traditional product development activities.

The virtual development team's structure is unique because it not only includes R&D personnel, it also integrates members from other SBUs. Team members from other SBUs enhance the cross fertilization of experience from related project experience that may be captured locally but not organizationally. We describe this proposed structure as a 'virtual' form because of the fluidity of team membership.

Characteristics

Enlarge 200% Enlarge 400%

The virtual research and development organization has several unique features which include: 1) its member composition, 2) the recycling of team members back into the organization, and 3) the use of senior core members from top management. Virtual research and development organizations acquire their members from a true cross-section of the organization (see Figure 2). In staffing the virtual development team, members are chosen from R&D, functional groups, and SBUs. Including team members from R&D to be part of the product development team is rarely done in today's product development teams. R&D personnel are often considered to be a special talent that should not be utilized on less technically challenging product development tasks. R&D persons may also be viewed as more specialized and less capable of contributing once a technology is formalized into a product application.

Virtual team members are chosen from multiple SBUs, not just from the SBU that is developing the project. Membership on a team is based on an individual's functional experience and technical training and not necessarily their most recent SBU assignment. An example would be SBU1 getting help from SBU2, SBU3, SBU4, and other SBUs. A major difference between the proposed virtual model and traditional forms is that team members truly represent the entire organization.

A second key feature of the virtual product development organization is the recycling of team members back to line functions. With traditional product development teams, once the project is complete, team members return to the functional groups. Virtual research and development team members can be reassigned to new operating divisions or to different functional groups. Placement for virtual team members is based on overall organizational requirements and does not restrict members to returning to their original SBU.

The third characteristic of the virtual model is the development of technical specialization centers which we refer to as the gray hairs. Gray hairs are core members of the team and are typically not project leaders since their function is to serve as a referent source of expertise. Gray hairs become specialization centers by broad involvement on a variety of new product development teams. Gray hairs serve as repositories of organizational memory, much like an historian, and disseminate learning through a variety of product program involvement. Gray hairs also provide a more impartial perspective of the product's development progress because their success is not tied to the project's success. Instead, gray hairs are evaluated based on their ability to consult and share lessons learned across projects.

Advantages

There are three advantages of the virtual research and development organization: 1) superior interpretation of customer requirements, 2) enhanced environmental scanning, and 3) broad dissemination of organizational information. Superior interpretation of customer requirements, characterized by the understanding and translating of customer requirements, includes a customer, as a source of articulated and unstated needs, and a developer, or supplier, of ideas to meet these needs (Van de Ven, 1986; von Hippel,1988). Understanding customer requirements is frequently listed as an initial activity in the product innovation life cycle (Griffin,1993; Maidique & Zirger, 1985; Pfleeger, 1991). However, management theory typically assumes much about fulfilling customer needs (Beckman & Mowery, 1993). Researchers and practitioners alike have identified many cases where customer ideas and needs have gone unfilled because of the developer's inability to understand true customer requirements (Peters, 1987; Peters & Waterman, 1982; Tushman, Newman, & Romanelli, 1987). Differences between customer's and developer's perceptions of the new product occur either in 1) identifying needs and potential uses of new technology, 2) translating needs into a product or service, and 3) understanding the relative importance of specific features, costs, or timing of the development (De Meyer & Van Hooland, 1990; DeMillo, McCracken, Martin, & Passafiume, 1987). For instance, the customer and developer can have divergent understandings of basic customer needs. Differences with regard to cost goals, schedule limits, and testing procedures can also emerge. Because developers are external to the customer, it is difficult for developers to understand, or accept, what the customer really wants (Burns & Stalker,1961; Zuboff, 1988). Beckman and Mowery (1993) found that adequately assessing customer needs was the most critical task of successful product definition. These authors suggest that developers understand the features the customer wants and the problem the customer is trying to solve.

Interpreting customer needs and then selecting technologies to satisfy those needs is increasingly becoming a major part of the product development teams' role as they use a higher percentage of black box supplier components. Product development teams must be increasingly adept at evaluating new technologies and technical capabilities presented by first tier suppliers, and then be able to integrate those subassemblies across a number of suppliers. Since first tier suppliers often have specialized technical capabilities that surpass the buying organization's skills, it is critical the development team have internal experts that can regularly assist in technology evaluation. A virtual research and development team is well suited for this task with its 'gray hair' core members who have the technical depth as well as cross product experience and expertise. The broad team membership further helps the virtual team develop products that meet customer needs.

Enlarge 200% Enlarge 400%

Another important advantage of the virtual research and development organization is its enhanced scanning capability of the organization's environment. External scanning is needed to identify and acquire knowledge about new technologies. External scanning is an important function for hightechnology companies where there is rapid technical change (Eisenhardt, 1989; 1990). Firms that have the best information can gain an advantage over competitors.

Treacy and Wierseman (1993) in their study of Johnson and Johnson (J&J) show how superior scanning capabilities can lead to competitive advantage. The authors describe how the J&J organization learned about a Copenhagen ophthalmologist who had discovered a way of producing disposable contact lenses inexpensively. By quickly purchasing the rights to this technology, before it was even completely developed, J&J was able to create an advantage, albeit a risky one, over competitors with the Acuvue product.

Virtual research and development organizations provide a superior scanning function because members come from diverse backgrounds which creates an advantage when gathering information about technology breakthroughs and new product applications (Tichy, Tushman, & Fombrun, 1979). Granovetter (1973) proposed that this innovation diffusion comes from the virtual member's social network which is more expansive than traditional R&D social networks.

The most competitively significant advantage of the virtual organization is its superior information dissemination capability. The dissemination of organizational information addresses how information, once acquired, is diffused within the organization. Unfortunately, traditional R&D and product development organizations do not learn well from product development experience. Weinberg (1990) observed from his discussions with industrial leaders that knowledge acquired by teams in the development process is lost when the team disbands.

In contrast, virtual research and development organizations provide structural mechanisms for organizational information dissemination. By having a composition of lower-level (i.e., operational level) to higher level (i.e., senior, core membership) members, information is made available across the organization. Since members come from a cross-section of functional and divisional groups, information gets disseminated and coordinated during product development. By having information disseminated within senior levels, top management can selectively transfer information back down into other parts of the organization. Certainly, a critical question is whether this structure can be successfully implemented in product development organizations. We address this question in the next section by describing a virtual research and development model as used by the United States Air Force (USAF).

AN EXAMPLE OF THE VIRTUAL ORGANIZATION

The Air Force has traditionally employed a hierarchical, organizational structure, with a centralized R&D division, to develop new technologies. However, in 1986, in response to Congressional inquiries on Department of Defense acquisitions, the USAF restructured the way it conducted R&D. The change was precipitated by the Packard Report (Packard, 1986) which was the result of an organization-wide audit of military R&D activities. The report concluded that the misapplication of human resources in R&D was problematic in the development of new military systems. The report stated that personnel policy was the keystone of virtually all of these reforms. One response to the report was the creation of an organizational structure we have referred to as the virtual research and development organization.

Members of the virtual research and development organization include non-core members, gray hairs and customer representatives. Non-core members include functional experts like engineers, accountants, managers and contractors. These experts typically come from matrix organizations that manage traditional personnel duties. Experience levels range from first-time developers to long-time careerists. Each home matrix organization determines the duration the member will stay in the organization based on the Air Force's needs and the person's own desires. The organization has little influence in determining how long non-core members stay in the organization.

Gray hairs also come from home functional organizations, but do not have specific project performance responsibilities. Instead, they are given a broad program charter on macro program issues encountered in the development process. Gray hairs normally stay with a program until it is completed and are, therefore, tied to the program longer than non-core members. The program director, the senior person in charge of the development, normally relies on gray hair advice since they have a deeper and broader experience base. The residency of gray hairs on projects until project completion makes it difficult for gray hairs to assimilate new technical or product information from other programs.

Customers are also members of the virtual development team. Customers are rotated in and out of the organization frequently, generally every two years, for several reasons: 1) to keep their home knowledge-base current, 2) to inform their own organization about the development, and 3) to provide new customer representatives who are closer with current customer requirements. The organization usually has no control over customer personnel decisions. An outcome of this process is that the customer representative's role is typically one of observer and not participant in the development.

An interesting aspect of the virtual research and development organization is the use of Air Force reservists. Many non-core members and gray hairs are also part-time customers. Because customers in this Air Force example are the operational units (e.g. fighter, bomber, and transportation groups), team members coming from these units have a first hand understanding of the new product's future uses which assists in development. Since the customer has had numerous people involved during development, the product's transition to its user group is facilitated and improved.

An example of this structure is now described. A virtual development structure was used in the design and development of a major weapon system, the Short Range Attack Missile (SRAM).

Typically, new products needs are generated by changes in the Air Force's readiness and response requirements. The customers are the fighter, bomber, airlift, refueling, radar, satellite, radio, and space groups. These various customer groups are similar in function to SBUs within a commercial organizations. In the SRAM example, the customer is the bomber group for a next-generation, short-range missile. One of the initial steps before starting development work on a new missile is to get Congressional support. While this step differentiates this example from commercial developments, it is not a critical aspect to our example.

Upon acceptance of a need for a short-range missile, the Air Force forms a virtual research and development organization to manage the development. One approach to developing new missile applications could be to lower the weight of a current missile to meet limitations in bomber capabilities. In this scenario, Air Combat Command (ACC), the primary customer of missiles, begins the development process with a requirement for lower missile weight. ACC employs Air Force Material Command (AFMC), the Product Center for the Air Force, to conduct the development effort. ACC is capitalizing on AFMC's knowledge base as a developer and integrator of aircraft and missile systems. On a project such as SRAM, AFMC works in conjunction with ACC, the end user, to determine what needs exist and how to best proceed in designing, developing, and producing a new missile system. AFMC would then form a virtual research and development organization for concept development to project completion.

The virtual research and development organization's responsibilities include 1) understanding ACC general needs and translating these needs into product capabilities, 2) selecting a developer or a set of developers to satisfy these needs, and 3) managing the development process. AFMC's virtual organization serves as an interpreter of ACC's need and translates ACC's needs into a language that developers' understand. The AFMC virtual organization interprets ACC's requirement(s) by first understanding all the ramifications of lowering missile weight and then translating these requirements.

In the interpretation process, AFMC relies on the gray hairs as information sources to help them understand the ramifications of lowering missile weight. The gray hair linkage is especially critical given the magnitude and the diversity of the USAF's overall R&D program and the need for cross service cooperation. In this case, gray hairs come from AFMC functional group leaders and basically provide the R&D perspective to new product developments. At the project's completion, gray hairs typically return to their functional homes and serve as policy makers and disseminators of project lessons learned until new development opportunities appear that match their particular expertise.

THEORETICAL FRAMEWORK AND IMPLICATIONS

We define organizational learning as the process of improving organizational actions by changing the range of team behavior in the product development process (jelinek, 1979; Huber, 1991). The purpose of organizational learning is the sharing of knowledge within the organization so that organizational performance, specifically new product development success, can be improved. We suggest that the virtual research and development organization, or team, provides an alternative way to learn that improves the likelihood of new product development success. Other well-known development models are similar but not equivalent in their use of resources or efficacy in improving the learning process. Our virtual model shares some common features with a "skunk works" or "venture team" model of product development (Kotler, 1997). Similarities between the two models include strong project leadership, drawing team members from multiple operating departments, assignment of responsibilities to team members only for the team's activities, budget and task autonomy, capability to evaluate and reward team performance-independent of functional influence, and the transient nature of team structure.

However, the virtual research and development organization is unique in several important ways. "Gray hairs" are actively involved as a team technical resource. In addition, team membership is crossfunctional, cross-division, and, in our case, crossservice (e.g., cross-organizational). A third unparalleled characteristic of the virtual research and product development is the integration of R&D personnel with the product development team. Finally, customers are involved in the development process. Thus, these unique aspects of the virtual research and development model captures lessons learned more effectively than traditional structures. Ultimately, products are more successful in the virtual research and development model because of the broader and deeper experience base and recycling of members. Zirger and Maidique (1990) found that a strong experience base-founded on past product, market and technical experience-leads to new product development success. Organizational, or team, learning provides a way to access that experience base, thereby, improving new product development.

More information is gathered with a virtual team because of the broader representation of both internal and external constituencies in the organization. This diversified team builds not only on their own individual experience base, but also gathers information from their personal network of contacts which are likely to be different from a traditional team's contacts. Organizational learning is improved because the team is rapidly understanding cause and effect relationships due to faster feedback cycles with more information processing.

Information dissemination occurs both within the team and across the organization with the virtual team structure. With highly cross functional and cross divisional membership, information is more likely to be widely shared. The broadness of information dissemination not only increases organizational learning, but also opens up team members to new information (Katz & Tushman, 1981). Cohen and Levinthal (1990) referred to this as a firm's 'absorptive capacity' or the ability to acquire and use new information in the R&D process.

Daft and Weick (1984) state that the development of a shared understanding, by knowledge acquisition and information distribution, leads to similar interpretations of lessons learned. Team members develop a shared interpretation by troubleshooting and problem solving issues together and then widely sharing their findings with each other. A shared interpretation is particularly important for a highly cross functional and diverse development team because members are likely to bring heterogeneous perspectives based on their varied prior experience.

The depth of development experience also improves interpretation of customer requirements in the new development process. Teams with greater product development experience better understand customer requirements. Experience is represented by the number of years in new product development, the number of years in industry, and formal education. Experience is valuable because it helps managers to troubleshoot more quickly and effectively, and avoid reinventing the wheel. Two measures of depth of experience on the virtual team are non-core team member and gray hair experience. Both measures give an indication of the experience depth, however, it is the presence of gray hairs that in part distinguishes virtual research and development organizations from traditional structures.

Gray hairs act in a counseling role on the development team (Eisenhardt, 1989). In this study of high velocity environments, top management sought the advice of 'counselors', or gray hairs, who were senior, experienced managers and engineers acting as resident experts. These counselors were able to provide advice to top management so that tough issues, especially in novel domains could be quickly resolved.

In addition, the diversity of the product development team also improves the interpretation of customer requirements. Diversity is defined as team membership that includes cross-functionality, cross-SBU, multiple levels within the organization with customer, and supplier membership. The more representation from these levels of the organization, the more diverse the team membership. Ancona and Caldwell (1992) found that when more functional representatives were on a team, there was better management-rated performance.

Finally, the recycling of product development team members improves the dissemination of information within the organization. Recycling refers to a team member's rotation within the organization. Recycling incorporates new ideas into the development process as team members are replaced (Daft and Lengel, 1986). Sussman and Dean (1992) found that rotation frequency between design and manufacturing groups provided the benefit of sensitizing each group about its system constraints. The idea of recycling is important in disseminating lessons learned because it provides a mechanism for members to learn about new products at multiple levels within the organization.

IMPLICATIONS OF THE VIRTUAL MODEL

The virtual model presented in this article is generalizable to other settings. The major difference between the Air Force example and commercial ventures is that the virtual model is based as a cost center while most commercial projects would likely be based as profit centers. However, since R&D is typically managed as a sunk cost, regardless of the financial framework, this difference is less pronounced when compared to the similarities between the Air Force case and commercial interests. Thus, from a corporate portfolio perspective, the virtual research and development model more effectively utilizes corporate resources in meeting corporate needs.

Another important implication of this model is that it supports current trends toward development partnering and strategic alliances (Wheelwright & Clark, 1992). The virtual research and development model incorporates diverse personnel while concurrently fostering the development and usage of core member resources. In an era where corporations are right-sizing to meet internal and external constraints, this model allows firms to use their own and key partner resources more effectively. Clearly, organizational learning is enhanced when broadbased and diverse information is shared (Nonaka, 1991).

CONCLUSION

The virtual research and development organization is an excellent example of an organization that facilitates organizational learning. Virtual research organizations provide a superior alternative to traditional R&D and cross-functional models of new product development by improving the overlap between innovation and organizational learning. The primary benefit of the virtual research and development organization is the leveraging of experiences and application of this knowledge to current development programs. By building on experience more effectively, firms should be capable of developing less costly, higher quality, and superior performing new products more quickly.