August 1997 

The basic approach of this paper is to see knowledge as a resource which can be used to create gains out of the uncertainty facing the firm. We see the firm, following Casson (1997), as a specialised intermediator created by an entrepreneur to synthesise information routinely about different sources of volatility. Three sources of volatility can be identified. The first source of volatility is primary uncertainty which arises both from exogenous shocks, such as random acts of nature, unpredictable changes in consumer's preferences and external technological change, and from endogenous change resulting from the firm's own research and development. Superior knowledge about these areas of uncertainty is what enables the entrepreneurial firm to create and maintain profitable reallocations (Casson, 1982 and 1997). 

In large corporations there is a division of entrepreneurial (knowledge synthesising) labour (Casson, 1985; Carter, 1995). External uncertainty is monitored and internal developments are led by a number - possibly a large number - of key managers within the firm. This division of labour gives rise to secondary uncertainty which is the risk that individual managers will not combine their knowledge in an optimum fashion with their colleagues. Synthesising this information is the key task of the entrepreneur - expanded in large corporations into a supra[or multi]personal force which may be embodied in the Board of Directors, in cross functional or cross national teams, in committees or in strategic planning groups. 

A third, tertiary, type of uncertainty might arise from opportunism, "self-interest seeking with guile" (Williamson 1996. p. 56) if managers choose not to reveal this knowledge which they hold, or if they divulge incorrect or misleading information. 

It is secondary uncertainty which most concerns knowledge management. As Koopmans (1957) puts it, "In a rough and intuitive judgment the secondary uncertainty arising from a lack of communication, that is from one decision maker having no way of finding out the concurrent decisions and plans made by others (or merely of knowing suitable aggregate measures of such decisions or plans), is quantitatively at least as important as the primary uncertainty arising from random acts of nature and unpredictable changes in consumers' preferences" (1957, p. 162-63). 

This definition of secondary uncertainty encapsulates the 'knowledge management' problem. We concentrate on this issue because, like Koopmans, we believe it to be 'quantitatively more important' than primary or tertiary uncertainty. It requires us to abstract, for the moment, from the issue of primary uncertainty. The sources of volatility (exogenous shocks and internal change) are issues for the specialist (entrepreneurial) members of the firm, whose judgment is necessary to identify or plan the primary source of such shocks. Tertiary uncertainty - the issue of trust in key managers is dealt with in a later section. The model is shown in heuristic fashion in Figure 1 (overleaf). 

The firm's response to the three types of uncertainty corresponds to what we have elsewhere called the three organisational problems of acquiring information, appropriate coordination and achieving effective motivation (Buckley and Carter, 1996). Knowledge management concerns the internal mechanisms for coordination, that is for pooling the key information garnered by managers whose task it is to monitor external volatility and discover new opportunities. Our subject is thus the intermediate knowledge flows within the corporation. Our focus is on multinational firms and so the process will take place across national boundaries. In order to reduce this source of volatility, the study concentrates on U.S. owned firms with UK subsidiaries and UK owned firms with subsidiaries in the USA. Language barriers are similarly reduced by our selection. 

When the processes for gathering knowledge are decentralised, as described above, firms face secondary uncertainty. This uncertainty can be reduced by identifying what items of knowledge should be brought together, developing efficient and effective methods of combination and deciding where and with whom the combined knowledge should be held. Items of knowledge which increase the potential gains for the firm when combined together are 'complementary': dealing with the organisational problem of secondary uncertainty is a question of managing complementary knowledge. 

It is possible to conceive several distinct scenarios in which it benefits the firm to combine knowledge from different sources. The simplest is where knowledge in one part of the firm is of direct relevance to an action which must be taken in another part the firm. For example, knowledge of production technology (say costs), held by the engineering or production department, is relevant to the planned output decision which might be made by sales or marketing based on their own knowledge of market demand. This scenario is represented in Figure 2 indicating the transfer of Firm Member 2's knowledge to Firm Member 1, who acts on the basis of the combined knowledge. We will call this additive complementarity. 

A second possibility is that knowledge from one of the firm members acts as an input prior to the acquisition of knowledge in the second part of the firm (Figure 3). For example knowledge about consumer tastes or market competition might be important for choosing the areas of product research and development to undertake. We will call this sequential complementarity. 

A third case concerns knowledge the firm's own actions when there are interactions or externalities between them. These interactions may be positive (complementary) or negative in their impact on the outcome for the firm. Examples might include decisions on advertising and product quality, which can be the responsibility of different parts of the firm which each have specialist knowledge affecting their actions, but both actions affect consumers' perceptions of the its products. In such cases it may important to coordinate the actions to ensure the best combined outcome (Figure 4). We will call this complex complementarity. 

Within these simplified scenarios there is scope for a good deal of variety in the structures and processes that firms can adopt for knowledge combination. Consider additive complementarity, continuing with the example of combining knowledge of market demand and production technology. These areas of knowledge might be combined in different ways. Production might convey information about production possibilities and costs to managers in marketing so that marketing chooses the quantity it would like production to supply. Alternatively, marketing could give demand information to the managers in production for them to decide the firm's output. These two possibilities are represented in Figure 2 by switching the identities of 'Firm Member 1' and 'Firm Member 2'. Furthermore, the form in which knowledge is transferred is itself variable between, say, a full transfer in which Member 2 learns all that Member 1 knows, and a summarised version of Member 2's knowledge judged to be adequate for the decision that is to be made. The form of transfer will depend on characteristics of the knowledge being transferred - how readily it may be codified - and of the individuals concerned - for example what codes of understanding they share. A more detailed analysis of the case of combining two items of knowledge for a single decision is to be found in Carter (1995). 

Variations of structure and implementation are also possible for the Sequential and Complex scenarios. In the first of these, in which knowledge is an input for the creation of further knowledge, it may be that the transfer of the input K₁ can be a distinct stage which precedes the search for K₂. Alternatively Firm Member 2 may need to draw on K₁ during the search for K₂. In the third scenario, there are several ways in which knowledge of other actions in the firm might be integrated together. Firm members could coordinate their actions by informing one another of their intended action; or they could act independently but share the knowledge on which their actions are based; they could each share their knowledge with a separate "coordinator" who would choose the actions jointly; they could join together as a team to choose both of their actions jointly; or they could act in sequence, so that one of the agents knows the decision of the other and then decides their own action in the light of the prior decision. A two-decision scenario is analysed in Buckley and Carter (1996). 

Complementary Knowledge in Multinational Enterprise 

The internalisation of complementary knowledge transfer is one of the principal reasons for the existence of multinational enterprise (Buckley and Casson 1976, 1985). Internalisation is a response to several possible kinds of market failure. These include (among others): 

• The 'public good'-like character of knowledge, for which replication (transfer) costs are significantly less than initial production costs. Internalisation is required for efficient production and exploitation.
• Buyer uncertainty resulting from Arrow's 'paradox of knowledge' (Arrow, 1971). A prospective market purchaser of knowledge cannot assess its value until it is revealed, but the seller will not reveal it before a price is agreed.
• Bilateral monopoly. Potential users of knowledge in overseas markets may themselves control key resources including, of course, the acquisition of complementary knowledge.

Additive complementarity (Figure 2) is the simplest case and represents an important class of multinational enterprises. The parent (firm member 2 in Figure 2) has some knowledge or expertise (K₂), developed in its home market and with potential for creating gains for the firm in other national markets. K₂ may be one or more of: an invention, a product or process design, more intangible technological know how or expertise, or particular marketing assets and capabilities or distribution and selling skills. In overseas markets the potential gains are at risk without specialised knowledge of each particular market, indicating the need for complementary knowledge, K₁, concerning local requirements and conditions, to be acquired by the affiliate (Firm member 1 in Figure 1). The characteristics that concern us here are the following: 

• K₂ is has public good characteristics in the sense that its use by the affiliate does not affect its availability or value to the parent, either in its home market or in other national markets which the parent may choose to enter.
• The costs to the firm are confined to the resource costs of knowledge transfer and the costs of acquiring local knowledge by the affiliate. These costs will depend on both the characteristics of the knowledge concerned (for example, the ease with which it is codified) and also the characteristics of the parent and the affiliate. (See, for example, Teece, 1977; Kogut and Zander, 1992)

• The parent's knowledge does not have the same public good character as in the additive case. K₂ has been created in response to particular requirements of the affiliate's market and may have significantly lower value to the parent in its home market or in other national markets. However, it is also possible that the affiliate's input knowledge stimulates developments with a wider application so that K₂ has a high value to the firm.
• The resource costs are likely to be significantly higher than for the additive, since the costs of knowledge creation are higher than those of transfer.

Combining complementary knowledge can overcome secondary uncertainty, but what of tertiary uncertainty, the risk that firm members do not reveal the knowledge they have or that they choose to make use of it for their own ends? For Williamson such opportunism - "self-interest seeking with guile" - is a necessary precondition for the market failures which explain the internalisation of international markets, (the others being bounded rationality and uncertainty). But can we be more precise in explaining how internalisation deals with the problem of opportunism? 

The connection between parent and affiliates in a multinational enterprise can be viewed as an exchange relationship. This exchange constitutes an internal market in knowledge and in capital, even though these particular commodities are not traded by price. The interests of the parties in the exchange are not symmetrical and in consequence neither are their respective opportunistic risks. The parent supplies knowledge and financial capital to the affiliate in return for an expectation of either profit (if the affiliate is a profit centre) or of intermediate product (if the affiliate is a cost centre). The opportunistic risks faced by the parent are: 

• expropriation of knowledge by the affiliate
• 'shirking' or 'cheating' by the affiliate, say through under-investment in specialist local knowledge, extraction of excess perks or payments, low effort and so on.

• 'shirking' by the parent - under-spending on the resource costs of (Additive) knowledge transfer or on the acquisition of (Sequential) knowledge for the affiliate
• withdrawal by the parent from the affiliate's market.

The organisational problems of information, coordination and motivation, corresponding to the primary, secondary and tertiary uncertainty, each have associated transaction costs and benefits. Evaluating alternative organisational arrangements depends, in principle, on comparative analysis, asking questions such as: how does the value achieved through a particular approach to combining complementary knowledge compare with the value which would have been achieved with an alternative approach? And what are the variable costs associated with the current approach and what would they have been under the alternative? 

For theoretical purposes we have elsewhere compared the costs and benefits of a given organisational structure with a notional ideal of 'perfect information', 'perfect coordination' and 'perfect motivation' (Buckley and Carter 1996). We treated the goal of organisation design to be minimising the sum of information, coordination and motivational losses (due to departures from the ideal) and information, coordination and motivation costs, (resource costs of acquiring information and of coordinating and motivating within the organisation). 

This approach requires modification in the context of the current project in particular, because of the issues raised by the counterfactual problem i.e. what would be the feasible alternative arrangement. And if an alternative organisational form were available, the value and cost quantities which might be used for comparing one organisation with another are not susceptible to easy observation. The practice of firms is to choose organisational structures on the basis of advantages and disadvantages perceived and articulated verbally, expressing judgment and intuition rather than numerical measurements or forecasts. The costs and the benefits of alternatives are not generally evaluated by firms in quantitative terms. Furthermore, empirical work on transaction costs by independent investigators has universally inferred the existence of transaction costs from firms' observed organisational choices rather than by developing a method of direct measurement or estimation (Buckley and Chapman, 1997). 

While the main purpose of this study is descriptive - to present examples of approaches to knowledge coordination by firms (see below) - we wish also, as far as possible, to make an evaluation of the structures we describe. The approach we evolved in the interviews to the counterfactual problem might be termed 'perturbation of the observation', in which we took the observed organisational structures as a given starting point and consider potential benefits and costs of making changes to particular components of the arrangements. It was not possible to do this in every interview. However, this seems both to be a practical evaluative procedure for independent investigators to use, and also to resemble the form of evaluation that business decision-makers would be expected to use in practice. It is still subject to the quantification problem, and assessments of this kind are qualitative and judgmental in character. 

While the problems referred to above place limits on the form of evaluation that can be carried out for an individual processes, it may still be possible to draw conclusions about approaches to knowledge management by comparing several processes, either within a firm or between firms. In order to realise such comparisons the descriptions of individual processes need to be analysed for characteristics which are significant for the effectiveness and efficiency of the process. We consider these next. 

Characteristics of Knowledge Coordination Processes 

We adopt six headings for analysing the characteristics of knowledge processes and for comparing processes. These are as follows. 

1. Knowledge Characteristics 

The first characteristics to note are those concerning the firm's knowledge itself. Of these perhaps the most important is what the knowledge is about, that is what sources of volatility face the firm. These can be external, such as fluctuations in levels of demand, changes in customers' requirements, changes in competitor activity or internal, such as the outcomes of its own research and development. Knowledge of these areas will vary in the ease with which it can be transferred according to how tacit or articulable it may be. The way in which the knowledge is held will influence whether it may be at risk of being lost, say if individuals leave the firm. A further characteristic which may be important is the risk of other firms discovering the same knowledge, through simultaneous monitoring or even through imitation (Winter, 1987). 

2. Value Added from Knowledge Combination Process 

In being concerned with processes for combining complementary knowledge it is necessary to be explicit about the ways in which these add value for the firm, and what capabilities are dependent upon the combination process. 

3. Participants in Process 

Knowledge is held by individuals and any knowledge transfer process is dependent on the individuals within the firm who acquire knowledge, transfer it and receive it. The location of these individuals, geographically and organisationally (that is whether they are part of an organisational unit which is defined by product area or by function or by geography) may also matter. 

4. Knowledge Transfer Methods 

The variety of means by which knowledge can be transferred might be classified in three broad forms: personal communication (talking, meeting, email etc.), codified communication (reports, drawings etc.) and embodied transfer (e.g. as product or equipment). Firms are likely to use multiple methods simultaneously, depending on many factors including how codifiable the knowledge may be, how much detailed information is required, the shared knowledge of the participants and so on. The costs and effectiveness of the process of knowledge combination will be very dependent on the methods which the firm can use. 

5. Governance 

The measures taken by the firm to monitor and limit tertiary uncertainty due to opportunism will influence the success of knowledge combination. These include the accountability of the participants, the level of supervision by higher management and the possibilities for participants to have conflicting goals. 

6. Performance 

While it is not possible to quantify the performance of observed knowledge combination processes, for reasons discussed above, it is nevertheless possible to comment on the effectiveness, speed and costliness of alternative approaches. 

Aims 

The present study is an empirical investigation of ways in which international corporations organise the internal transfer of knowledge whose aim is to: 

• Identify examples of complementary knowledge and how and where it originates, within a sample of firms.
• Identify the organisation and methods used by firms to combine complementary knowledge.
• Consider what factors influence the firm's choice of organisation and means of knowledge transfer, such as the characteristics of the knowledge, where it is located in the organisation, the benefits and costs of exchange, the ease of transfer, questions of motivation etc.
• Consider the effectiveness of the knowledge transfer .
• Compare the approach used by firms with the 'business process' concept, particularly its interpretation as a process of combining complementary knowledge (Buckley and Carter, 1996, 1997).

Method 

The study was conducted through interviews with managers in a sample of six companies, three from each of two contrasting industries. The industries were chemicals/ pharmaceuticals and engineering. All firms in the sample comprised a parent and wholly owned subsidiaries based in the UK and the USA. The parent/ industry distribution was as follows: 
 

	Parent based in
	United States	United Kingdom
Chemicals/ Pharmaceuticals	Firm 1 Firm 2	Firm 3
Engineering	Firm 4	Firm 5 Firm 6

A number of managers were interviewed in each firm at middle and senior level, according to access. The intention of the interviews was to gather information about a representative sample of the firms' operations and not necessarily cover every activity. 

The aim of the research project was explained to interviewees as 

The findings from our interviews are summarised in Table 1 in the Appendix. The diversity and complexity of the information obtained necessitates a degree of simplification and stylisation. The classification of process types introduced earlier enables us to highlight the significance of different knowledge combination processes in the six firms, and the processes within each firm can be described using the characteristics discussed earlier. The significance of the types of knowledge combination process differs between the firms in the sample and firstly we discuss these differences and compare the characteristics of each type of process across the firms in the sample. 

Additive Knowledge Combination 

Purely additive knowledge combinations are not found in any of the firms - there is always a degree of 'sequential' activity, in which the parent undertakes a level of activity which is specific to each affiliate. But the degree of such affiliate-specific knowledge creation by the parent varies and is lower in three of the firms in our sample than in the other three. That is, the 'additive' character of knowledge combination is higher in these three, which are Firm 1, Firm 4 and Firm 5 (see Table 1 in the Appendix). Firm 1 is a manufacture of sterile disposable products for hospital use, Firm 4 manufactures filters for air conditioning equipment and Firm 5 produces chain for power transmission. Compared with the other firms in the sample (see below) these three are all concerned with segments of their respective market sectors where volatility is relatively low. In particular, all are using comparatively mature technologies and the rate of innovation and product change is less high than for the other three firms. 

There are, nevertheless, differences in the characteristics of the knowledge combination processes in these three firms, particularly in the degree of parental monitoring of the transfer process and the overall performance of the process. Ranked by the level of parental monitoring the differences can be summarised as follows: 
 

Firm 4	Lowest level of parental monitoring. Some problems in knowledge transfer through slowness, failure to adapt information to affiliate market requirements. Limited ability to recognise and adapt to these problems. Identifiable obstacles to transfer.
Firm 1	Affiliates have significant autonomy within framework of corporate systems and standards. No problems reported in knowledge transfer. Communications are generally good, and information can be obtained provided the affiliate knows where to ask.
Firm 5	Active monitoring through system of group and divisional (cross-border) committees (see below). Communication problems can be resolved through this framework.

The explanation of these differences may lie partly in differences in the individual historical developments of these firms, but there may also be factors in their market and technological contexts that have influenced these variations. For Firm 1, the most important product knowledge is the product specification, which can be largely codified in the form of engineering drawings. Much firm specific manufacturing know-how is transmitted in embodied form as specialised manufacturing equipment. The firm has therefore not developed the practice of transferring knowledge in ways which are particularly responsive to the affiliate's needs. Firm 1, on the other hand, is in an industry in which 'good pharmaceutical manufacturing practice' is essential for regulatory approval of its production facilities and its products. The methods for transferring this tacit knowledge have been integral to its process of internationalisation. Therefore, an effective knowledge transfer capability has become ingrained in the organisation. Firm 5's still more active practices of may have arisen because, for reasons discussed below, they have established a framework for active international coordination of the 'complex' type, and once in place this provides a means of monitoring all aspects of knowledge transfer within the firm. 

These observations indicate that an important requirement for effective transfer is for the parent knowledge provider to recognise the knowledge requirements of the recipient in order to provide what is appropriate, in a form that is appropriate and with appropriate timing. In the case of additive knowledge combination, by definition, the parent's knowledge has public good characteristics and there are limited resource costs of transfer. But transfer performance can be deficient if the parent and recipient do not have a cooperative team-like approach, that is if they do not recognise their common interests in effective knowledge combination. A team approach, focused on the gains of the particular knowledge transfer process, was clearly a characteristic of the examples of successful combinations which we observed. 

Sequential Knowledge Combination 

Having suggested that 'additive' combination is more significant for the three firms in less volatile environments, then sequential combination is more important for Firms 2, 3 and 6. Firm 2 is a large ethical pharmaceutical manufacturer, wholly dependent on continual discovery and development of new therapeutic compounds and Firm 3 uses the chemical technology in which it is a leader to develop and manufacture chemicals with applications in a wide range of industries. Firm 6 is engaged in contract design to leading automotive and aeronautical manufacturers. All three are at the leading edge of dynamically developing technologies. 

Once again, the knowledge transfer processes differ between firms and again these differences may be ranked by the level of formal parental monitoring. 
 

Firm 3	Lowest level of parental monitoring. The approach has been a conventional two way communication between affiliate and specialists n the parent company.
Firm 6	Where sufficiently important problems have arisen, parental knowledge has been combined with the affiliate's knowledge by establishing task forces.
Firm 2	Research and development projects, which are international in character, are managed through teams in each affiliate, with overall control in the hands of project management specialists based in the parent's HQ.

 

While these three are the firms for which sequential combination is most important, it is present in all six firms. It is perhaps least effective in Firm 4, which as noted above has relatively weak knowledge transfer capabilities. Firm 1 encourages a team approach to sequential knowledge transfer, although the initiative for establishing the process often lies with the operating subsidiary wishing to develop the combination. Firm 5 uses its international committee system to evaluate, choose and monitor product development proposals from affiliates. Again, we see that the character of these knowledge combination processes is different in each firm. 

Sequential combination calls for specific resource expenditure by the parent on behalf of the affiliate. For this to be effective, a degree of hierarchical overview is required for resource allocation and prioritising decisions. This overview has generally been from within the parent company as we see in the examples above. For Firm 3 the control takes place within the established functional hierarchy, for Firm 6 through the establishment of task forces and in Firm 2 through project management specialists. However, such parent-based control is increasingly insufficient to deal with the more complex interdependencies of the global firm, as discussed next. 

Complex Knowledge Combination 

The final form of knowledge combination is of growing importance. Two firms in our sample already have mature means of coordinating activities and knowledge between their operating companies. Two more have recently established organisational means for doing so, and are in the process of allowing these to develop and they are likely to extend the approach throughout their businesses. 

The two mature approaches are in Firms 5 and 6. Firm 5's system of international committees has already been mentioned. These exist at group level and also within individual product divisions. Their role is partly to manage manufacturing and transfer pricing in the face of currency fluctuations, but they also collect operational knowledge from individual firms on for purchasing and agree priorities for capital expenditure, product development etc. Firm 6 has rather different requirements. It is important for the firm's operating companies to be leaders in technological and market developments, and the firm has a highly active information centre (library) which both writes and disseminates technical reports on leading topics, but also collates and disseminates key market intelligence gathered from subsidiaries. 

Firms 2 and 3 are developing strikingly similar approaches to international coordination in the form of 'global business teams' and 'global business units' respectively. Both are marketing-based developments in which manufacturing is coordinated separately. While developing similar knowledge coordination processes the firms are otherwise at different stages of development. Firm 2 has been an established multinational company for many years, whereas attempts to coordinate the international activities of Firm 3 are more recent. In the case of Firm 3, there has been a deliberate decision to avoid establishing a separate 'staff' group. The teams are made up of line managers, not necessarily from the parent, whose role is to coordinate the activities of the marketing affiliates in each country to avoid cross border conflicts and externalities, to agree development priorities and to develop common marketing approaches for customers who also have a global presence. The aim of both firms is to achieve the best combination of local knowledge, central knowledge and the delocalised knowledge of the corporation as a whole. In doing so, both are shifting control away from regionally based structures towards product-based structures. In both firms the regional and product-based structures co-exist, so that there is a dual responsibility for profit. It remains to be seen whether this tension is creative. 

It is clear from the analysis of the results above that volatility arises both from the external circumstances of the firm in the sample and from their internal dynamics (the impact of research and development and other internal perturbations). The analysis and combination of these elements is at the core of our understanding of knowledge management. 

Our approach identified three types of combined knowledge: additive, sequential, and complex. These three types of process have six key characteristics - knowledge characteristics, value added from combining the knowledge, participants in the process, knowledge transfer methods, governance and performance. The wide differences observed in the six characteristics account for the disparate nature of the knowledge management processes even in a small sample of six companies. However, as Table 1 shows, our classificatory system allows us to identify the reasons for the extreme variability of solutions adopted to the problems of knowledge management. This should not be surprising - firms (even within the same broad industrial group) do arise from very different historical bases, do have vastly different administrative heritages, do face varying local (and national) conditions, are differently placed in the competitive hierarchy and in the global technostructure. It would be ridiculous indeed to expect uniform systems of knowledge management to prevail where the driving forces of such systems - the volatility faced by the firm -are so radically different. 

We can thus assess the efficacy of a firm's knowledge management system by the degree to which it is capable of identifying the key sources of volatility which face the firm, the degree to which a system of monitoring and synthesising this information is in place and the way that pooling of knowledge internalises the greatest potential externalities for the firm's long term benefit. 

The evaluative question is whether any of these firms might gain by making changes to their established knowledge transfer processes - for example, whether any of them would gain by adopting aspects of the processes of other firms. In several interviews, managers expressed great interest in other companies' efforts on knowledge management. There are examples where changes might be made which would improve performance. For example, in Firm 4 communications were rather rigidly hierarchical and questions were channeled upward rather than directly to the individuals who hold the relevant knowledge. And it is possible that the performance of the comparatively autonomous affiliates of Firm 1 might benefit from a higher degree of international coordination and comparison of the kind actively pursued by Firm 5. 

It is also clear that processes which are successful in one company may not transplant readily into another. For example, managers in Firm 3 expressed some surprise that the central information system of Firm 6 was successful, suggesting that in their own firm there were insufficient incentives for managers around the business to share information in this way. Also, as described in Table 1, Firm 2 has a an active process for transferring the parent company expertise and values to new affiliates. In contrast Firm 6 found that the least successful approach to knowledge transfer was to 'take the company operating manual'. It had more success from transferring a few selected individuals from the parent to the subsidiary and allowing them to develop the appropriate combination for the local requirements. 

The firms in our sample themselves recognise that their management of complementary knowledge is imperfect. In every case they are moving towards a system which they believe will more effectively tackle the problem of coordinating knowledge from different sources - hence the creation of 'global business teams', project management teams, 'global business units', centralised knowledge pools and group management committees. 

Is there then, or can there be, an optimal solution to the design of knowledge management systems? Our answer is yes, but it is likely to change as key sources of volatility change, and it is likely to differ between firms precisely because they face differing sources of volatility. However, this is not a recipe for nihilism or despair. A careful analysis of the knowledge process types and knowledge process characteristics enables the design of a system of knowledge management which will differ between firms, but will contain several generic types. 

Generic Strategies 

We can begin by focusing on the key process types. Additive knowledge processes require a 'local team' (local in the sense of focus rather than in place). Sequential require a more hierarchical overview and complex knowledge processes require an overall coordinating mechanism. All firms face all three types of volatility and so we should expect to see all three types of solution in the firm, and we do. We should thus expect a combination of focused team, hierarchical overview and a coordinating mechanism in place in each firm. Are there any rules as to which form should dominate? 

Each of our six knowledge process characteristics provides part of the answer to the issue of dominance. 

1. Knowledge Characteristics 

Of fundamental importance are the key sources of volatility which knowledge management is used to counteract. This may be internal or external, it may be more or less tacit, it may vary in degree of transferability or imitability and the required scale of response may vary. In research intensive industries, the R & D output may be a key source of volatility - 'driving' the knowledge management process. In other cases the key driver is the volatility arising from the changing consumer demand. 

2. Value Added from Knowledge Combination Process 

The motivation for 'knowledge management' strategy arises from the gains (profit) from combining pieces or sources of knowledge less the costs necessary to achieve this. Combining knowledge is , in a sense, a synonym for the benefits of internalisation (Buckley and Casson 1976). 

3. Participants in Process 

These are the acquirers and recipients of knowledge, who are constrained by the technology of 4. Knowledge Transfer Methods. The outcome of this constrained process is 5. Governance which has 6. Performance characteristics. 

Our framework has two drivers 

Knowledge management is perforce complex. Because there are three distinct types of processes, we expect three co-existing solutions. We can identify additive, sequential and complex knowledge processes and suggest that focused teams, hierarchical overview and overall coordination are the appropriate co-existing solutions. 

These solutions will exist in all companies but the particular combinations will be determined by the key drivers of the system, arising from the key sources of variability facing the firm and its (potential) gains from synthesising the knowledge necessary to exploit that volatility. The technological constraints on achieving such a synthesis are in decline and thus the gains from an improvement in a firm's knowledge management process in terms of improved governance and a better performance outcome are actually being sought by companies throughout the world economy.