Acknowledgements: We would especially like to thank the Bosch Insititute at Carnegie-Mellon University for supporting this work. We would also like to thank the CIBEAR program at the Marshall School of Business at the University of Southern California, the Academy of Finland, Turku Graduate School of Computer Science, and Lund University. Special thanks are warmly expressed to Patrik Laxell and Stina Schonberg for their conscientious data collection effort. Finally, we would like to thank the participating companies who wished to remain anonymous. 
  
 

 The motivation for this study was to better understand shared knowledge creation (SKC) around business processes in the context of international corporations. The research design is a three country progressive field investigation (Finland, USA, and Sweden) with staged theory building and iterative hypothesis testing -- which are preceded by a field pilot study. The focus is on SKC around the new product development process in high technology companies. The paper develops a model of shared knowledge spaces that includes catalysts, organizational values, information technology infrastructure, and SKC processes. This paper presents the results of the pilot in the USA and the first stage of the field study (companies in Finland). The results help us to discover the nature of SKC spaces around business processes and how to better design them. Implications for the next stage of theory development and field test are discussed. Implications for practice are also derived. 

1. SHARED KNOWLEDGE CREATION IN AN INTERNATIONAL CONTEXT 

1.1 The Increasing Importance of Knowledge Management & Knowledge Creation 

It is widely stressed that a corporation's competitive advantage flows from its unique knowledge and how it manages knowledge (Nonaka & Takeuchi, 1995; Spencer, 1994; von Krogh & Roos, 1996). That observation rings increasingly true as we enter the complexity of the global electronic economy. Researchers have pointed out that for many corporations, their ability to create, share, and use knowledge will have a major impact on their competiveness in the future --- and some researchers state that the only sustainable competitive advantage in the future will be effective organizational knowledge creation and knowledge management (Nonaka & Takeuchi, 1995; Wikström & Normann, 1994). Said Nonaka: "When markets shift, technologies proliferate, competitors multiply, and/or products become obsolete almost overnight, successful companies are those that constantly create new knowledge, disseminate it widely throughout the organization, and quickly embody it in new technologies and products." (Nonaka, 1991). This has led to an interest in idiosyncratic knowledge that is valuable, rare, immobile, and exploited by a firm to give the firm a competitive advantage (Barney, 1991, 1994). 

1.2 Knowledge Management in International Corporations 

The challenge and opportunities of knowledge creation and knowledge management in international corporations are much more complex than those in corporations that operate in local markets. International corporations have much more complex structures, have more geographically-dispersed operations, must manage diverse cultures and markets, and have to deal with a greater multitude of political and regulatory issues. It is thus more difficult to create knowledge that can be deployed uniformly throughout the corporation, it is more difficult to create knowledge that can be adapted and integrated in different contexts, and it is more difficult to synthesize knowledge from heterogeneous source. Furthermore -- different cultural propensities generate different ways of sharing and creating knowledge. However, these challenges are balanced by opportunities to get new ideas from diverse sources, to use complementarities in skills and expertise, and to profit from various styles of knowledge creation within the same corporation. 

From a scholar's standpoint, studying knowledge creation and knowledge management in international corporations that operate in different countries also provides an excellent research design in that particular aspects of knowledge creation will be more salient and obvious in particular cultures than in others. The differences and variations in knowledge sharing styles observed across two collaborating business teams from different countries may bring to the surface new concepts that underly knowledge sharing that could not otherwise be unearthed. Thus, some characteristics of knowledge sharing in groups that may not be obvious in the USA will show up intensely in Sweden as a natural style. For example, the Swedish work culture is much more naturally group-oriented and participative than the US work culture. While a characteristic of knowledge creation that is taken for granted in Japan (such as the dominance of sharing tacit knowledge) will be perceived quite differently in a Finnish organization. 

For all the above reasons we find it very useful to use the international corporation as a context for studying knowledge creation and knowledge management. For purposes of this study, we have opted to do our field research in international corporations in three countries (Finland, Sweden, and USA) for reasons that are obvious from our affiliations. 

1.3 Shared Knowledge Creation around Core Business Processes 

Knowledge creation and knowledge management can be approached from different organizational perspectives. A survey of business practices and the business literature on the topic (El Sawy et. al, 1997) suggests that there are at least three conceptualizations of business knowledge in the research literature. These views are interrelated, but it is useful to understand their nuances as they lead to very different foci of attention. One view regards business knowledge as an object (or a thing) i.e. knowledge as codified, storable, and patterned information that produces insight. This view of knowledge highlights a perspective of knowledge management that focuses on building and managing knowledge repositories. A second view regards knowledge as a capability that has potential for future action. The capability view brings with it a perspective on knowledge creation and knowledge management that is centered around building core competences, understanding the strategic advantage of know-how, and the creation of intellectual capital. The third view conceptualizes knowledge as process. This latter view focuses on the processes of knowledge creation and knowledge sharing and their relationships to learning and collaboration among people in work settings. The process of knowledge management is thus one of triggering knowledge creation, enriching knowledge sharing, preventing knowledge loss, and augmenting knowledge synthesis. It is this third view of knowledge, knowledge as process, that we mainly adopt for purposes of this research study. Our focus is on the activity of shared knowledge creation in business settings. 

We define shared knowledge creation (SKC) as the set of processes (both cognitive and behavioral) that take place when a group of people create and shape new knowledge together. It is the together part which makes the difference in work settings. While SKC includes sharing of existing knowledge between people, it also implies that much new knowledge is created through simultaneously sharing existing knowledge. The nature and structure of SKC processes is not well understood and it is the purpose of this study to better understand it in order to improve it. 

Improved SKC is most useful in work settings when it is targeted directly at important business outcomes and put into action. This study focuses on studying SKC around core business processes that add value to the customer. Thus, improved SKC is only as practically useful as the extent to which it helps to make core business processes more effective. Studying SKC around business processes such that they can be improved can also be viewed as an emergent strategy for business process reengineering of knowledge-intensive processes (Davenport et.al., 1996). Thus changing or expanding the knowledge-creating capacity of a business process is a also a tactic for business process redesign. Typically business processes are redesigned by restructuring the process and its infrastructural environment or by changing the information flows around the process. Changing the SKC processes around a core business process provides another set of heuristics for redesign. 

1.4 The New Product Development (NPD) Process in High Technology Companies 

The core business process chosen for our study of SKC is the new product development (NPD) process in high-technology companies. There are several reasons for that choice. First, NPD is a business process that is highly knowledge-intensive and one of the key business processes for creating new organizational knowledge. Thus, the NPD process context is an ideal "laboratory" for studying SKC. Second, from a strategic learning perspective for the organization, NPD is the context from which it is most likely to transfer SKC methods to other areas of the organization. NPD is seen as a main driver of organizational renewal; it is a continuous process of knowledge creation, in which the organization is adapted to its changing environment of markets and technologies (Dougherty, 1992). Nonaka & Takeuchi (1995) say it most elegantly: "Organizational knowledge creation is like a 'derivative' of new-product development. Thus, how well a company manages the new-product development process becomes the critical determinant of how successfully organizational knowledge creation can be carried out." 

The choice of high-technology companies further accentuates both the knowledge intensity and the criticality of the NPD process. NPD in high-technology companies involves extensive research and development activities where much knowledge creation takes place. Furthermore, high-technology companies derive much of their competitive advantage through new products. They are also under growing pressure to produce new products faster and at lower cost, without a loss of quality, but with higher flexibility. In order for them to remain competitive in a global economy, they especially have to be effective and efficient in knowledge management around the NPD process. 

In NPD as well as in other core business processes shared knowledge creation plays a critical role, and thus provides both an interesting area for research as well as an excellent leverage point for improvements. Although this investigation focuses on NPD, its results and learnings are likely to be applicable to other core business processes where knowledge creation and knowledge management around the process are critical. 

 2. A Pilot Study 

A field pilot was carried out at the R&D Center of a telecommunications equipment manufacturer in the USA. We studied two NPD projects, a larger one in its development phase, and a smaller one in its definition phase. We interviewed the two project managers. We then conducted structured interviews with short face-to-face questionnaires with a sample of members from different discipline teams (10 from the larger project, and 4 from the smaller project). We did not want the interviews to last more than an hour and most of them took about 50 minutes. We investigated knowledge sharing practices, team dynamics, learning strategies, and identified some of the critical aspects of the NPD process at the local level. In preparation, both researchers attended the NPD procedures training workshop given to company employees. We also met briefly before the pilot with line managers, the project managers, and some of the team members to get their input and familiarize them with the research project. 

After two days of data collection (Thursday & Friday) the data was statistically analyzed on the weekend and on Monday we had group meetings with each of the two project teams. We presented our preliminary findings, got some feedback and input, and showed what kinds of results were possible. The pilot has helped us test some of our instruments and uncovered new concepts and questions. We gained some insights regarding the modes of knowledge sharing around NPD, critical aspects of SKC around the NPD process, and communication within and between teams (frequency, ease, imbalance, blind spots, technologies). 

The field pilot study we ran on the two NPD teams in the telecommunications equipment company indicated that apart from the tacit/explicit continuum, it appears that professionals can easily identify and operationalize a hierarchy of activities that they think can lead to SKC: informing (which is just passing information back and forth), coordinating (synchronizing activities and interdependent outputs), and collaborating (true joint problem solving). 

The pilot study also brought to the limelight the role of knowledge catalysts to SKC. The team may share knowledge among its members and with other teams, but there also knowledge catalysts (whether they are consultants, university research centers, industry associations, or even customers or suppliers) who often play a major role in amplifying SKC. In some cases it is apparent that the knowledge catalyst fuels SKC within the team. Initially, we thought that these were just boundary spanners, but it appears that their role is much more of a triggering one than just a spanning one. We found this to be quite intriguing because it has critical implications for how information systems for SKC should be designed: perhaps they should be designed around the knowledge catalysts rather than around the teams. This led to our second researchch question. 

From the pilot study we gained more understanding of the SKC process and formulated our research study objectives as seeking to answer these four main research questions: 

RQ1. How does SKC take place around the new product development process in international corporations? What are its various explicit and tacit modes? What characterizes best practices in SKC ? 

RQ2. Can knowledge catalysts accelerate or facilitate SKC between and within NPD teams ? How ? How does the effect of knowledge catalysts change depending on the strength of links between collaborating teams ? 

RQ3. Should information systems for improved SKC be designed primarily around knowledge catalysts? If yes, how ? If no, how should they be designed ? 

RQ4. What are the implications of RQ1, RQ2, and RQ3 for the effective use of information technologies for improving SKC ? 

We also realized at that point that we needed to develop a better theoretical model of SKC around business processes. 

3. A Conceptual Model of SKC around Business Processes 

3.1 SKC Spaces 

We have defined SKC earlier in the paper as the set of processes (both cognitive and behavioral) that take place when a group of people create and shape new knowledge together. It is the together part which makes the difference in work settings. While SKC includes sharing of existing knowledge between people, it also implies that much new knowledge is created through simultaneously sharing existing knowledgethat Shared knowledge creation is the set of processes that take place when a group of people create and shape new knowledge together. It is the together part which makes the difference in work settings. 

We also know from the pilot study that informing, coordinating, and collaborating are activities that comprise SKC. It is more than communication, information exchange, and coordination. It is much more than the sum of individual actions and cannot be done at the surface level. It is also more than knowledge exchange or knowledge conversion across tacit and explicit dimensions. It involves collaboration in which there is shared discovery and articulation. SKC involves creating a "shared space" or "shared bonds of interest" that allows the collective massaging of ideas and information in situations of ambiguity and change. SKC spaces are tacit and explicit embodiments of the richness of commonality among the members of a team around a knowledge creation endeavor (like NPD). How can the dynamics of this shared space be conceptualized ? Our model of SKC spaces is shown in Figure 1. 

Our model of SKC spaces is based a situated learning framework (Brown & Duguid, 1991; lave and Wenger, 1991). A situated learning framework focuses on what people do and what resources they use that are available to them. Tyre & Hippel (1997) have shown he physical context that people work in is very important for adaptive learning and sharing of tacit and explict knowledge. Hutchins (1990) has also shown how the problem takes its form based on the tools and resources based in the setting. SKC spaces evolve because complex problems can only be understood collectively and because codified abstract knowledge is usually insufficient for dealing with them. The social and cultural setting will also affect how and what people learn collaboratively (Brown & Duguid, 1991). Thus knowledgeable action involves a set of SKC processes and a perceived environment (which we divide into information technology (IT) infrastructure and organizational values/norms/procedures). We separate out IT infrastructure due to its increasing importance for collaboration in international corporations. The set of SKC processes involves the various modes of knowledge creation, sharing, and preservation and will be further defined through the empirical field data. 

In addition, to separating out IT infrastructure we also identify catalysts as a separate box in our conceptual model of SKC spaces based on our findings in the pilot. Catalysis occurs when entities outside a team trigger improvements in the work that is to be done in the team. Such improvements can for instance be the number of ideas and perspectives that are considered or focusing on key issues, or on speeding up knowledge creation. Von Hippel (1988) has argued similarly that the sources of innovation often lies outside a closely knit collaborating group. 

3. 2 SKC Spaces around the NPD Process 

The NPD process conceptually has several phases: developing product concept, design specification, detailed design, integration and testing, piloting,. and there are many iterations, and it is not a sequential process (cf. Leonard-Barton, 1995). Imporving NPD involves both developing better products and developing them in better ways i.e both product improvement and process imporvement. 

The NPD business process is primarily knowledge work in which ideas, innovations, and designs are jointly created, and in which team expertise and skill evolves and grows. There are various work activities that take place in a NPD team process that include design activities, administrative and clerical activities, and organizational and management activities. Around these activities is the SKC space for the NPD. As the NPD process is executed instances of SKC are generated through the SKC space. The impact of the SKC activity on process and product improvement is mediated by the nature of the organizational context as shown in Figure 2. This concept is important because it provides a way of getting at the capabilities for SKC in a team and how they related to enacted instances of SKC that ultimately influence the NPD process. Thus, by building better SKC spaces we can increase the potential for more effective instances of SKC and consequently better NPD processes. Thus, we would like to know the best practices for SKC spaces that result in the most improved NPD process. 

Figure 2 Model of How Different SKC Spaces Can Relate to the NPD Process through Enacted Instances of SKC
	NPD Product Improvement
Enacted SKC Instance	NPD Process Improvement
Nature of NPD Organizational Context
SKC space with potential  for generating SKC instances

4. Thinking Through the Research Design 

Having gained experience and better understanding from the pilot study and having developed a conceptual model of shared knowledge spaces, we were now ready to go forward with our field study. The field study would be used to better understand the nature of SKC spaces around the NPD process. The research strategy, research design and operationalization are discussed below. 

4.1 An International Progressive Field Investigation Design 

Choosing a knowledge intensive environment was the first decision. From the pilot study we already knew that one very good candidate in this respect would be the new product development process. There is plenty of new knowledge to be developed during such a process. There is also plenty of existing knowledge and skills as well as existing practices and methods as part of the process. New product development by definition is a creative process. It also normally involves highly educated people from several fields of expertise who have to communicate with each other. There are also external parties involved in such a process so that context certainly offers a challenging research environment. New product development process and knowledge sharing around that process was our choice. 

Selecting the countries for the comparative study was easy since the members of our research group come from the USA, Finland and Sweden. Indeed, even if the choice was natural it also offers a very interesting setting. All are democratic countries with similar management principles as it looks like at the surface level, but digging a little bit deeper shows differences in ways of management, the hierarchical structures, participation, and values. There are differences in the educational systems, job security, the power of labor unions and so on. 

Selecting and finding companies which we considered suitable and which were ready to participate in the study was, on the other hand, not the simplest thing. There were certain criteria we had put forth. We wanted to have two companies from each country and within each of these companies we wanted to have two NPD-projects to follow. The companies should be international high-technology companies. The new product development cycles of the companies were expected to differ from each other, --- one shorter and one longer development cycle was our preference. This would let us compare what influence the fast response requirement by the market had on the interaction between SKC spaces and the NPD process. We wanted two teams from each company that were at different stages of the NPD process -- preferably at opposite ends of that spectrum. The first companies fitting these criteria and willing to participate were the Finnish ones. One company was from the electronic measurement equipment industry and the other one from the pharmaceuticals industry. We could even select two projects which matched our criteria on mainly research versus development focus, so this was a good match. After these Finnish companies were selected we decided that the USA and Swedish companies should be in the same or very similar fields to make the data between the countries and companies comparable. We have even been successful in finding such companies, although the data collection in the USA and Sweden is still not completed. 

The data collection methods to be used in the field study involved studying available documentation, conducting preliminary interviews with core people, jointly deciding on who should participate in the study, designing and customizing the questionnaire for the company context, conducting the survey, doing fast analysis of data to select people for follow-up interviews, and then data analysis. We would also provide the results to each of the companies and get their feedback. 

The same structure and techniques will be used in all countries and companies. We decided to do the studies somewhat "one after the other" and not in parallel because we wanted to learn from one case before starting with the next. The experiences and findings direct the study to be carried out in the next country - still securing certain comparability of data. Such an approach helps to find out what works and what does not. It gives the opportunity to fine-tune and improve the research instruments between the studies - as well as to adapt them to the current context. The research strategy chosen was thus a three country progressive field investigation with staged theory building and iterative hypothesis testing. In this paper we will describe the Finnish study where the data collection effort has been completed. 

4.2 The Finnish field study 

We started the study in Finland since it was possible to find suitable companies there first, and there were two master students ready to start working on the data collection under the supervision by the main investigators. The data collection and feedback sessions for the companies have been completed. The research methodology and techniques discussed above were applied, starting with a pre-study, continuing with the questionnaire and interview stages and closing with feed back sessions. 

The pre-study was conducted through discussions with liaison persons, on the top-management level, and project managers. The goal was partly to "sell" the project to the company partly to learn as much as possible of the company. Documentation complemented this study phase. The questionnaire design was based on four sets of questions: 

1. Organizational factors affecting the new product development process
2. The new product development process
3. The role of information technology
4. The interaction around the new product development

4.3. The Pharmaceuticals Company (FL) 

One of the companies is an international pharmaceuticals company, mainly located in Finland. The traditional role of the company has been the production of drugs (licensed from other companies) for the domestic market. Lately, because the legal and competitive environment has changed, the company has started to concentrate more than previously on R&D and international markets. 

There are three levels of R&D at the company: research of original drugs, original product developments and product improvements. The organization for research is built around project groups. One of the project groups participating in this study is a product development project and the other one an original drug research project. 

In the product development project there are people from several departments. The number of core project members was estimated to 15 - 20 people, with up to 50 - 60 other people loosely coupled to the process. Altogether 23 respondents filled out the questionnaire at a meeting. Five respondents from the project plus one additional respondent who was knowledgeable of both the company's projects were interviewed. Two of them were specialists and three were managers. One person was from a support unit not directly involved in the project work. 

The other project participating in the study is an original drug research project. People from several disciplines and from several earlier phases of the project participate in the project. Altogether 60-70 persons are more or less involved. In this team we had the largest number of filled-in questionnaires (24), although the project had the lowest response rate. Five plus one individuals were interviewed as in the previous project: three managers and two specialists, and the one with knowledge from both projects. There were people from the project group and some who were in the line organization and in the support functions. The project manager was also amongst the interviewees. 

4.3 The Electronic Measurement Equipment Company (FS) 

The other company participating in the study is a Finnish-based company producing complex electronic measurement equipment and systems that include mechanical, electronic hardware and software components. These systems are sold to customers all over the world. There are around 50-60 employees working in NPD in a major program. These programs are divided into smaller projects working on specific parts of the composite system products. A typical size of a project would be 5-7 people. 

One such 'program' and one 'project' participated in our study. They are cooperating in developing a new product. The 'program' is situated at the company's facilities in Sweden. They build the functional part of the system. The 'project' is situated in Finland and they contribute with one module to be added to the part in Sweden. The responsibility of the joint project is in Sweden even if the company's main research site is in Finland. The two teams cooperate through both traveling and communications technology. 

The Swedish program is divided into three projects, one is developing the user-interface, one the mechanical part of the product, and one the software. There are 15-20 persons working in the program. There was a meeting where the group filled in their questionnaires. A total of 15 questionnaires were returned. Five people were interviewed plus the "link" as described before. Two of them were managers (1 program, 1 project manager) and three specialists, one of these working in a support function. 

The project situated in Finland is part of the technological platform department. The goal of the department is to build a base for future products. The project we studied is developing a module to complement the system being developed in Sweden. At the technological platform department there are altogether 20 people. Half of them are hardware designers and the other half software designers. About 10 people from the department are core members of the project we are studying, with five additional persons assisting at times. The work of the project members is very specialized, they work on their expertise areas either alone or in very small groups (2-3 people). The questionnaires were filled in at a meeting. Altogether 11 responses were received. Interviews were conducted with 5 project members as usual plus the person involved in the collaborative Finnish-Swedish project. Three of the interviewees were managers and two were specialists. 

A summary of the companies, the projects and their characteristics as well as the names to be used for them are presented in Table 1 (Laxell,1997). The names of the companies, FL and FS, and of the NPD groups, FLR, FLD, FSR and FSD, will be used in the presentation from now on. F stands for Finland, L for Long (cycle time), S for Short (cycle time), D for Development (focus) and R for Research (focus). 

In order to answer the research questions posed in Section 2 above, we identified four sets of issues through which results could be examined and hypotheses operationalized and tested:- 

• How knowledge is created, shared, and preserved from loss
• The nature of interactions around SKC and the role of knowledge catalysts
• The use of information technology in SKC
• The relationship between SKC and the performance of the NPD process

5.1 HOW KNOWLEDGE IS CREATED, SHARED, & PRESERVED FROM LOSS 

5.1.1. How Shared Knowledge is Created: 

It is useful to examine the interaction between syntax and semantics of the term shared knowledge creation (SKC) in order to understand the non-decompositional nature of that process. So, does the descriptor "shared" refer to the creation process or the knowledge ? In order to answer this question we need to examine some fundamental assumptions about knowledge --- in the context of business knowledge in particular --- and to make sure we do not confuse it with information. There is a difference between the transfer of information and the transfer of knowledge. Information transfer can be explained by a pure transportation metaphor implying that a sender can convey information by sending it to a receiver who can capture it in its form and use it. Knowledge transfer cannot be explained by a transportation metaphor as knowledge cannot be captured through transfer --- it has to go through a recreation process in the head of the receiver who learns. Thus, because knowledge can only be recreated then shared knowledge and a shared creation process for knowledge are one and the same. Thus, in SKC the product is the process. 

In examining Table 2 it appears that the most useful knowledge creation ocurred between meetings: the best ideas tended to be found between meetings (rather than during meetings) and that was also when the most progress was made. This is true across all 4 projects although this effect was more accentuated in some projects more than others. The table also shows that in comparing formal and informal meetings, the best ideas were generated in informal meetings. 

When team members were queried as to whether they created their best ideas working alone or working with others, the results show (column 4 of Table 2) that the best ideas were created mostly when working with others -- rather than working alone. Thus, it appears that there is an awareness of the value of SKC for generating useful knowledge. This is true across all four projects, although it is somewhat more accentuated with the FL project. This may be a reflection of the difference in work culture between the electronic measurement equipment company and the pharamaceuticals company. It may be worth noting that the FS company has more of a software development culture that perhaps also draws "hacker" types who are typically stereotyped to be loners. While we have not yet tested the statistical significance of this, it also appears that the effects are similar across research-focused projects compared to development-focused projects. 

In examining the dynamics of meetings among team members the results show that the most productive meetings occurred when participants tried to develop common views rather than keep their own individual views. This is true for all four projects. Cole (1989) has shown that there are cultural differences across different countries as to sharing behavior in small group work. In comparing the FSR group (which is Finnish) and the FSD group (which is in Sweden) -- and these are in the same company -- the results show that the Swedes scored significantly higher in terms of trying to develop common views. This is not surprising given the high value based on democratic participation in the Swedish work culture. We begin to see here the value of having an international research setting: it draws out concepts through dimensions that are accentuated differently in each setting. 

The findings from these data are shown in the box below. In combination they suggest that common collective knowledge is valued and that the setting and mode of interaction must support the sharing of tacit as well as explicit knowledge. 
 

Finding #1: Most useful knowledge is generated between meetings rather than during meetings  Finding #2: In comparing formal and informal meetings, the best ideas were generated in informal meetings  Finding #3: The most productive meetings are the ones in which participants tried to develop common views rather than kept their own views  Finding #4: Team members create their best ideas working with others rather than working alone  Combining these four findings, we come up with following hypothesis:  Hypothesis #1: The creation of useful collective knowledge is influenced by the structure of group meetings and the willingness to develop shared views during those meetings.

  

5.1.2 How Knowledge is Shared: 

It is very difficult to separate out the tacit component of SKC from the explicit one. Our literature review, pilot study, and company interviews suggested to us that we could at best try to identify SKC situations in which the tacit component was substantial. 

To further examine the role of tacit knowledge, we asked whether most of the skills and expertise were transferred through "telling" or "coaching." Telling is a more explicit mode of sharing knowledge, where as coaching has a much larger tacit component as it involves showing how things are done repeatedly and making corrections when necessary. Results in Table 3 show that telling was more dominant than coaching as a knowledge sharing mode. We initially thought it would convey a tacit versus explicit comparison and were surprised by the results, but apparently it was more of a "tell me once and I'll pick it 
up" versus "tell me, watch my mistakes, correct me, teach me, worry about my performance." 
 

Finding #5: Knowledge sharing is enriched by the use of graphics and the frequent use of examples  This generates the following hypothesis:  Hypothesis #2: Knowledge sharing is enriched by methods that allow the transfer of tacit knowledge.   Finding #6: Coaching is not viewed as a typical mode of knowledge sharing.

  
5.1.3 How Shared Knowledge is Preserved from Loss 

Collective knowledge needs to be preserved from loss so that it if needed it can be reused as a guidance for action, understanding, and further knowledge creation. Table 4 shows the results of questions related to methods of preserving shared knowledge from loss. 
  

There is also a difference between the two companies. FL is in the pharmaceuticals industry in which documentation is very important for legal reasons and they are held accountable. This is reflected in Table 4 in which FL relies much more on records than FS. 

Other questions related to goodness of practice in preserving knowledge from loss are shown in Table 5. The first column indicates that there is quite a spread in terms of non-implementation due to lack of follow-up, and that shows up most vividly between the development-focused groups in the two companies. The results also show that participants whether they document or not do not report forgetting ! Again, FLD with the highest requirement for accountability reports the least amount of forgetting. The participants generally reported that they do frequently use their records but it is not a very strong agreement except for the FSR project -- which does not document very well as can be seen in Table 4. 

The rightmost column in Table 4 is some indication of how the participants wished that things could be with respect to documentation. Comparing the rightmost in Tables 4 and 5: FL reports that they do generate meetings reports regularly and they do not seem to think that more documentation is needed nor that they can reduce it. FS on the other hand does report that more documentation would be helpful, especially FSR that reports very little documentation. 

- "We frequently use records"(-0.48), 

- "After meetings usually some kind of meeting report was generated" (-0.37) 

- "We collect a lot of records such as written notes, e-mail, sketches,etc.. (-0.42) 

- "To remember what was discussed during meetings we mostly relied on our records" (-.43) 

In combination these negative correlations suggest that people who document much and rely on their records do think that more documentation is helpful. There are two possible interpretations for this. The first is that they are satisfied with what they have. The second interpretation is that documentation is not very useful. The second interpretation is supported by interviews and is the most plausible of the two. 

The FSR data shows that even though they do not document very well, they appear to forget less than others who document more. Thus, they must have some other informal method of remembering that allows them to retain the knowledge at their fingertips more readily. In other words they document in an informal way but in a "high energy" mode that enables them to retain a high level of knowledge creation capacity on an on-going basis. While we have no other systematic data to confirm this, we find it is important enough to warrrant detailed probing in the next stage of the study. 
 

Finding #7: The more formal documentation and records are generated, the more skeptical people are about the usefulness of more formal documentation  Finding #8: Informal documentation can yield high preservation under certain conditions  Hypothesis #3: There are high energy modes of documentation that allow less reliance on records between meetings.

  

5.2 THE NATURE OF INTERACTIONS AROUND SKC AND THE ROLE OF KNOWLEDGE CATALYSTS 

5.2.1. The nature of interactions around SKC 

In Section 2 above we indentified three modes of SKC activities: informing which means passing information back and forth, coordinating which means synchronizing activities and interdependent outputs, and collaborating which means true joint problem solving. These SKC activities take place within and between groups in a project - as well as with external groups. The nature of this interaction infrastructure is an important dimension of SKC spaces. 

Krackhardt and Hanson (1993) have shown at the individual level that in informal networks, not receiving critically important information, and only one-way communication is a reason for job dissatisfaction. Furthermore, they have found that the quality not the quantity of communication determines the profitability of an organization. These arguments are applicable for knowledge sharing at the group level as well. Krackhardt and Hanson (1993) identify three types of networks: 1) the advice network for the "prominent players in an organization on whom others depend to solve problems". Such a network could be useful also for contacting catalysts, 2) the trust network for backing up in crisis. This type of network could be considered an expert network with local knowledge as well, and 3) the communication network for talking about work. This could be for all modes of SKC interaction. Studying communication network can reveal gaps in information flow as well as "bow ties" which both can be essential to discover for efficient interaction. 

 Table 6 -General management roles in the NPD projects (Laxell, 1997) 

Before delving into the interaction structures within and between groups, we also assessed the roles of project managers and general managers in each of the companies. The development oriented projects in both companies ranked distributing important information to the team as the management's most important role. Surprisingly, this was not true for neither one of the research oriented projects (see Tables 6 and 7). In the research-oriented groups, creating visions and selecting products to be developed were viewed as the manager's most important roles. This reflects the more professional managerial culture of development-oriented groups. 

We asked the participants about their interaction around the NPD process within the project, within the company, and with external groups. We asked them to report how frequent, easy and important they considered the interaction to be. We also asked them to report their most essential reason for communicating: informing or coordinating or collaborating. We then used network analysis software (UCINET and KrackPlot from Analytical Technologies) to graph interaction patterns. Figures 4 and 5 are outputs that show the interaction patterns between FSD and FSR and groups within the company. 

In Figures 4 and 5 the frequency (Freq) and ease of communication (Ease) are shown below the boxes for each group. For Freq the scale is 1 to 5 (1=never, 2=seldom, 3=monthly, 4=weekly, 5= more frequent). For Ease the scale is 1 - 7 where 1 is very difficult and 7 very easy. In addition the numbers on the arcs measure how important the group's own output is to another group, and how important the other group's output is to them. Standard deviations are shown in parentheses. We use the ratio between those two importance measures as an indication of asymmetry between groups in terms of interdependence. S is the number of responses. The most essential reason for communicating is shown in percentages for each group on the right of the box. 

These interaction patterns are very useful to the participating companies as a diagnostic tool. For example, in looking at the interaction patterns within projects we found that interaction seems to be easy, and the frequency rather low. Some assymetries in importance identify expected "supplier" type of relations. For example the output from the 'information services' group was found to be much more important to both FLD and FLR compared to their own output to that department (not shown here). However, unexpected hidden assymetries may be revealed through these interaction patterns. For example, FSR reports that 45% of the instances the most essential reason for communicating with the "Main Software" group in FSD (see Figure 4) is collaboration which is very high. However, an examination of the assymetry (4.0/2.5) reveals more of a one-sided supplier-type relationship. These inconsistencies signal trouble spots for the company to further investigate. 

In addition to the utility to the partcipating companies, the interaction patterns also reveal some interesting findings for our study. For example, it is surprising how infrequent all the groups interact with other groups within their own project ---- monthly to seldom. It is also surprising to find that informing is reported as the dominant mode of interaction. For example, informing is reported as comprising more than 50% of the instances of interaction within the groups in the FSD project. A question that arises is whether tighter collaboration would improve and speed up the NPD project, and whether SKC spaces should be redesigned to help do that. 

In Figures 4 and 5, the FSR group reported that they collaborated with two of the three FSD groups (25% and 45% collaboration), and mostly informed the third team (67% informing). However, the FSD group's understanding of the same situation was that they did not collaborate with any of the FSR groups but primarily informed them (64%, 63%, 67%% respectively). In a collaborative situation such inbalance in perceiving the interaction mode may create problems in SKC effectiveness. 

A regression was run to further examine the relationship between the degree of collaboration between groups and the importance of interaction. The collaboration variable (the dependent variable) was calculated from the number of 'collaborate' instances identified divided by the number of possible collaborating groups. The independent variables were importance, frequency and ease of interaction. The result was quite interesting: 

Collaborate = 12.6 + 9.96 * (Your Output) + 6.20 * (Their Output) - 9.03 * (Frequency) 

Your Output = importance of the project's output to counterpart (signifiacnt at .005 level) 

Their Output = importance of counterpart's output to the project (significant at .01 level) 

Frequency = frequency of interaction (significant at .07 level ) 

The result is highly significant (reliable at the .000 level, F=9.2 for the overall regression). 

The regression results show that the importance of the group outputs in both directions influenced the degree of collaboration. Thus, the higher the importance, the more likely the degree of collaboration. However, the frequent interaction between groups influences the degree of collaboration negatively (although significance is not high). Our conservative interpretation of the regression results is that frequency of interactions is not a necessary condition for collaboration. This finding at first sight appears to be contrary to the traditional findings in the communications literature. However, frequnecy of communication and the quality of interaction are not the same thing. It is also interesting to note that ease of interaction had no effects on the degree of collaboration. This could be interpreted to mean that if interacting is considered important for both parties they will overcome problems if necessary. 
 

Finding #9: Imbalance and assymmetry in interaction patterns for groups that need to collaborate within projects signals potential problems in SKC effectiveness.  Finding #10: There are assymetrical relationships in interaction patterns within NPD projects that indicate a traditional customer/supplier relationship rather than collaboration.  Finding #11: The degree of collaboration increases as the importance of mutual outputs increases.  Finding#12: It is unclear whether the frequency of interaction between groups influences the degree of collaboration, and whether the correlation is negative..  Hypothesis #4: The more the balance in interactions between groups and the more important the mutual outputs, the richer the SKC mode in terms of having more collaboration between the groups

  

5.2.2 SKC and the Role of Knowledge Catalysts 

As discussed in Section 2, knowledge catalysts are external groups that have the potential to improve SKC inside a team. Catalysts are always outside the team whose actions are studied. They can have an amplifying effect on SKC. Their role is more than boundary spanning and involves triggering the SKC process. The literature on social network effects on the extent of innovation diffusion (cf. Abrahamson and Rosenkopf, 1997) shows that many innovations diffuse through social networks linking individuals and organizations, and that these networks are segmented by internal boundaries. We augment this idea of boundary spanning in the context of SKC through the notion of knowledge catalysts as an important part of SKC spaces. These catalysts trigger and facilitate innovative ideas to diffuse via interaction, formal or informal. In NPD processes, innoavtive ideas are necessary and fast diffusion is important. 

On the other hand, examination of the interaction patterns with external groups gives a different story. Figures 6 and 7 show the interaction patterns between FSD and external groups, and FLD and external groups. The legend and numbers are the same as explained in Figures 4 and 5 above. Figure 6 and 7 show that the team do collaborate with their external sources (as high as 57% of all instances for the FLD team and clinical experts). Furthermore, the output of external sources is viewed as important, and this is especially true for the FLD team (as high as 6..2 on a 7 point scale). This contradicts how they report their sources of knowledge in Table 5. Thus, it appears that knowledge catalysts play a role in collaboration and SKC, but it is under-appreciated and explicitly not acknowledged by NPD teams. 
  

Finding#13: NPD teams perceive their sources of knowledge creation to be mostly within the team.  Finding #14: The interaction patterns of NPD teams with external sources indicate that substantial collaboration takes place.  Hypothesis #5: External catalysts influence the effectuveness of SKC processes but are not directly preceived by NPD groups that way.

  
One relationship that we still need to test is how the effects of knowledge catalysts on SKC processes varies depending on the strength of the coupling between collaborating teams. 

5.3 INFORMATION TECHNOLOGY USE AND SKC 

This section addresses how IT was used by the teams in SKC. The respondents were asked to rank the means (media) by which they most frequently interact with others (Table 9 below was used to collect the data). Rankings were done for "Groups within the project" (61 respondents did the ranking), "Groups within the organization" (n=60), and "External groups" (n=43). For each type of activity (interaction) - collaborating, coordinating, and informing - three tables were calculated (best tool, 2nd best tool, and 3rd best tool). The analysis was done for the three types of processes (interactions) - within the project, within the organization, and with external groups. 
 

For the different levels of interaction, please rank the means by which you interact with others. Please mark the most frequently used with a "1" , the next most frequently used with a "2" and the third most frequently used with a "3". 
 

EXTERNAL GROUPS	Organized face to face meetings	Informal face to face meetings	Telephone	Speaker phone	Fax	Email	Sharing printouts / paper copies	File sharing via common directories	Calendar system	Bulletin boards
Tools used for informing
Tools used for co-ordinating
Tools used for collaborating

Media richness theory (Daft & Lengel, 1986) can be used to discuss the findings. Simplified, the media richness theory suggest that in situations and tasks with high uncertainty and high equivocality a suitable media should be rich and in situations and tasks with low uncertainty and low equivocality a poor (lean) media is suitable. Hence, the theory suggests that there should be a fit between a situation/task and the used media. Based on the theory, it can be suggested that richer media should be used to a larger extent for collaborating and that less rich media should be used to a larger extent for informing. 

We start by looking at media use within the group and this is done for the total sample: 

• For collaboration the two most frequently used media are informal and organized face-to-face (f-t-f) meetings. This is not surprising since these media are very rich. Notable is that e-mail, a poorer media, is the third most used medium.
• For coordinating the three most used media are e-mail, informal and organized f-t-f meetings. There is a striking increase for e-mail.
• For informing the most used media is e-mail (ranked by one third as the most frequently used media). Informal f-t-f meetings, organized f-t-f meetings, and telephone are all used almost equally often.

Looking at the data for interaction within the organization we find more or less the same pattern as for interaction within the group. The major differences are that e-mail for all three types of interactions is used more, f-t-f meetings are used less --- and for informing, the telephone is used as much as informal f-t-f meetings and organized f-t-f meetings are used. What is a little bit surprising is that e-mail seems to be the major substitute for meetings. The media richness theory would suggest that the substitute would be a media that is closer in richness to meetings than e-mail, for example the telephone. 

For interactions with external groups we could expect that the tendency would be even more distinguished. Indeed, we see a continuing decrease in the use of f-t-f meetings and especially in informal meetings, but the these are not substituted by e-mail. Instead we see an increased use of telephone, fax, and sharing printouts/paper copies. The latter two are poorer than telephone. There are at least two explanations to the pattern. First, there are still quite a few organizations where the employees are not easily accessible via e-mail. Hence, instead of e-mail the other three media are used. Second, it can also be that these three media were used to a larger extent than they are today and that they have been substituted by e-mail. 

The responses of the teams indicated that the use of IT is highly valued in their companies. On a scale of 1 to 7 the response had means that ranged from 5.1 to 6.2 across the 4 groups. On the other hand, when asked about the fetaures of a dream tool that might help NPD in general and SKC around NPD in particular, the respondents were generally stumped except for identifying better file sharing features. 

When asked about the extent ot which the company provided sufficient training, the responses indicated that a little more training would be helpful. It is also clear that IT per se is not a silver bullet. IT has to be implemented and managed in a careful way to leverage and augment an organization´s SKC processes. None of the teams had a clear policy or plan for how IT should be used in their SKC processes. 
 

Finding#15: Information technologies are highly valued by NPD teams.  Finding #16: NPD teams are unable to identify features of an IT "dream tool" for SKC.  Hypothesis #6: Information technologies are necessary but not sufficient for improving SKC around NPD

  
Several limitations are worth noting. First, the means (media) were ranked according to how frequently they were used. This does not directly indicate how useful the respondents find the different means. How often a specific medium is used might also be affected by how easy it is to use the medium. Second, we did not control for how easy it was to use the different media. Some media require that those involved in an interaction use the medium (for example, e-mail and file sharing) but other media, like face-to-face meetings require a geographical and timely coordination. 

5.4 The relationship between SKC and performance of the NPD process 

5.4.1 Factors that slow down the NPD process 

Writers on NPD point out the importance of time and speed. Organizations have to speed up their NPD processes in order to stay competitive. This applies to organizations with short (e.g. electronics) as well as long (e.g. pharmaceutical) development processes. Hence we were interested in identifying causes for slow down in the NPD process. 

The respondents were asked to what degree certain causes affected the process -- a 7-point scale was used: "Totally disagree" to Totally agree". The data (Table 10) suggests the major causes for slow down of the processes and hence indicates area that can be addressed in order to speed up the processes. Overall, the data shows that the major reason for slowdown was that people were too occupied in other projects. 

As discussed in Section 5.4 IT was not used extensively in the teams. The data shows that information technology unreliability was not causing slow down to any major degree -- in fact it has the lowest overall score. Information overload does not seem to be a major factor -- it has the second lowest value. Furthermore, the effects of rules and regulations --- while not restrictive for any the groups -- were much more 

pronounced for FL compared to FS. 

The SKC spaces are likely to have an impact on NPD process performance - SKC spaces were presented in Section 3. Measuring NPD process performance is not easy and measuring the relationship between SKC spaces and NPD process performance even less so. Hence, we did not try to directly measure the relationship. Instead we identified the main goals of the NPD teams and then looked at the SKC spaces and their relationships to NPD process goals. This means that we are not assessing how well the teams were fulfilling their goals but what the goals were. Given that this is mainly an exploitative study the proceeding followed can be justified. An alternative approach could have been to via indirect measures assess how well the teams were fulfilling their goals. 

Based on previous writings four types of goals can be identified: 1) cost, 2) quality, 3) fast response management, and 4) accountability. In a cost driven NPD process the goal is to minimize the cost. In a quality driven NPD process the goals include high product quality, satisfying customer needs, and creative design. In a fast response driven NPD process the goals are to have a short time to market, have re-usable design, and gaining distinctive competencies for the future. In an accountability driven NPD process the goals include on time results and producing the result according to regulations. In most NPD processes all goal types are present but their importance will vary and we can expect that one or two of the goals are the dominating ones. 

From a list of items (goals) each respondent indicated the three most important end goals in the NPD processes. Using the answers, we for each team calculated a goal profile showing the importance of each of the goal types (Figure 8; data is normalized by percentage). 

There are both similarities and differences between the companies. Both companies have quality as an important goal (FS = 36%; FL = 35%) whereas fast response management is less important (FS = 19%; FL = 22%). In FS cost is an important goal but not so in FL (FS = 31%; FL = 11%). In FL accountability is important but not in FS (FS = 14%; FL = 32%). Breaking down the data to team level showed some intra-organizational differences (see figure). 

  

Finding #17: NPD slowdown is caused by people being too occupied in other projects, rather than infrastructural technology factors.  Hypothesis #7: Effective SKC Spaces must be matched to the management context of the NPD process

  
Unless an organization is working in a very stable business environment the ideal for an organization would be to have flexible NPD processes. These will give better performance. More flexible NPD processes means for example that activities (phases) previously done in sequence can be done in parallel, that is an increased efficiency. More flexible NPD processes can also increase the effectiveness, for example, by knowledge having an impact on the product´s final characteristics is constantly brought into the NPD process. This can be by extending the concept time and hence having the possibility to sense the market for a longer period before concept freeze. It can also be later in the NPD process by having broad internal as well as broad consumer testing. A flexible NPD process can be supported by a dynamically stable SKC space. This is a SKC space that is capable of serving a wide range of NPD requirements (dynamic) and at the same time it builds on long-term SKC process capabilities and the collective organizational knowledge (stable). 

6. Implications for Research and Practice 

6.1 Assessing the Value of an International Study 

The results in Section 5.1 showed that there were differences in SKC behavior between groups in Finland and Sweden even within the same company. Post-questionnaire interviews confirmed that view. There have been discussions in the literature that compare SKC behaviors in Japanese and Western work settings (Hedlund and Nonaka, 1993) and American, Japanese, and Swedish work settings (Cole, 1989). We have found it useful to use archetypes of SKC behavior as a lens to use for generating more hypotheses and as a communications tool to explain to the participating companies how their SKC styles may differ. We have successfully used a 4-way categorization (See Table 11) as a starting anchor for discussions of dimensions of SKC. We have somewhat caricatured the SKC styles of the 3 countries in the study and also added the Japanese context for further contrast. We expect to further refine this as the study goes to the next stages. In Table 11, the Finnish "Kekkonen" SKC style comes from the management style of the longtime President of Finland, Urho Kekkonen which is epitomized by "trust me, I know what is best for you." The Swedish "Lagom" style (which means collectively adaptive and tuned to be "just right") draws its roots from the democratic labor history of Swedish work culture. The USA "Cowboy" SKC style epitomizes the individualistic hero culture ideal of western cowboy roots. The Japanese "Kaizen" style is one of incremental continuous improvement that is based on collective input and consensus. 

Low Need for Consensus (Do it Myself) High Need for Consensus (Do It Together) 
 

FInnish "Kekkonen" SKC style	Japanese "Kaizen" SKC style
USA "Cowboy" SKC style	Swedish "Lagom" SKC style

6.2 Implications for the design of SKC spaces 

Although only two of the four phases of the study have been presented in this paper, there are findings that provide for the design of SKC spaces. In a NPD context this will give the possibility to generate different and more flexible, and hopefully more efficient and effective, SKC instances. The SKC spaces will be the backbone for flexible NPD processes. 

In order to make the research study as useful to participating companies as quickly as possible, we have used vivid comparative summary tables with the most important findings presented in a way that is conducive to identifying opportunities for the better design of SKC spaces. An example is shown in Table 12. 

After debriefing the participating company through the discussion of Table 12 and validating the results with the company, the following opportunities were indentified for each of the dimensions of SKC spaces: 

- Improving the SKC mode: Providing less but "high energy" documentation and common knowledge structures for in-between meetings (documentation for the collective mind) 

- Enhancing Information Technologies: Providing a common interface intranet. 

- Amplifying the Effect of Catalysts: Sending more Swedes to Finland, and having more external groups interact with FSR 

- Changing Organizational Values: Put more emphasis on informal knowledge sharing and the necessity to capture the knowledge created during informal meetings. 

The reserach design for this study is three-country progressive field investigation with staged theory building and iterative hypothesis testing. In this stage we have generated through one country a number of findings and hypotheses to be tested. They are boxed throughout the body of Section 5 and we shall not repeat them here. These findings have provided us with a better understanding of the nature of SKC spaces around the NPD process which we can take advantage of in the next stages in the other two countries. We continue to enhance the questionnaire based on our experiences and findings, and have opportunity to test some of our newly formulated hypotheses. We believe that this iterative method is a more solid and valid method of theory building. Furthermore, we are still able to provide interim recommendations to participating companies without having them wait until the entire study is finished. An NPD team may work from sun to sun, but a researcher's work is never done...