Dimensions of Organizational Knowledge-Bases
We now examine the relationship between the concept of an organizational knowledge-base and the three key innovation search processes. Understanding the implications of various knowledge-base dimensions becomes more meaningful, with the prior understanding of the innovation search processes in mind. We follow Walsh’s suggestion that researchers focusing on the knowledge structure construct in relation to management need to address several distinct issues. First, in accordance with his suggestion, we survey the prior research to ‘uncover the attributes’ of knowledge structures that managers use. Second, following Walsh (1995), we need to understand the implications of these knowledge structures for consequences of relevance to managers; specifically we focus on the relationship between these various dimensions of knowledge structures and organizational innovation performance. Finally, in keeping with Walsh’s third precept we try to ‘uncover the origins’ of organizational knowledge structures, i.e. understand the key determinants that shape knowledge structures in any organization. Going forward, unless it is contextually necessary, we use the term organizational knowledge-base everywhere as the term collectively captures all types of knowledge structures.
Research on organizational knowledge-bases has identified several key attributes or dimensions on which knowledge-bases differ. We focus on six such primary dimensions: size, content, veridicality, differentiation, integration, and embeddedness. However, in the course of the discussion we also mention some key ‘derived’ attributes of knowledge-bases such as decomposability, malleability, and relatedness. The distinction between the primary and derived attributes is that derived attributes emerge from the interaction or implications of primary attributes (see Figure 25.2).
Figure 25.2 Antecedents and Consequences (Dotted line relationships are not within the scope of this study)
Size
Knowledge-base size reflects ‘how much’ an organization knows. In the context of recombinant innovation, the size of a knowledge-base determines its recombination possibilities and, hence, its inventive potential (Fleming, 2001; Ahuja and Katila, 2001; Puranam, 2001). In the simplest conception the size of a knowledge-base should be a reflection of the resources committed by an organization to gain knowledge, for instance its research and development spending. However, the literature has identified several leveraging strategies by which firms increase the de facto size of their knowledge-bases by tapping into extra-organizational sources of knowledge. These include alliances with other firms (Kotabe and Swan, 1995; Lane and Lubatkin,1998; Mowery, Oxley, and Silverman,1996; Powell, Koput, and Smith-Doerr, 1996; Stuart, 2000; Rosenkopf and Almeida, 2003), investments through corporate venture capital units into start-ups (Dushnitsky and Lenox, 2005; Wadhwa and Kotha, 2006; Benson and Ziedonis, 2009), more effective organizational learning arrangements or location decisions that enable the firm to impound more of the knowledge created from prior searches (Zollo and Singh, 2004) or geographically proximate organizations (Jaffe et al., 1993, Almeida and Phene, 2004; Lahiri, 2005; Singh, 2009), investments in absorptive capacity that enable a firm to internalize more of the spillovers from public science (Van den Bosch, Volberda and de Boer, 1999; Lenox and King, 2004), from competitors and complementors (Bowman and Hurry, 1993; Schilling, 2002; Lichtenthaler, 2009), or investments in network ties that enable them to occupy favorable positions in inter-organizational networks and, thus, foster the absorption of knowledge spillovers (Baum, Calabrese, and Silverman, 2000; Ahuja 2000; Owen-Smith and Powell, 2004).
Grafting new knowledge onto their knowledge bases through acquisitions is another mechanism by which firms expand knowledge-base size (Ahuja and Katila, 2001; Ranft and Lord, 2002; Puranam, 2001; Puranam, Singh, and Zollo, 2006; Puranam and Srikanth, 2007). In this context researchers have identified that a key issue is the integration paradox (Ahuja and Katila, 2001; Puranam, Singh, and Zollo 2006; Ranft and Lord, 2002). For grafted knowledge-bases to enhance the recombinant capacity of the firm, the acquired and acquiring firm’s knowledge-bases have to be integrated. However, this integration may entail significant organizational disruption and may thus compromise innovative capability. Investigation of these ideas shows that whereas the absolute size of an acquired knowledge-base is positively associated with subsequent innovation (providing support for the recombination effect) large relative size of an acquired knowledge-base reduces innovation possibly by creating knowledge integration problems (Ahuja and Katila, 2001). Further, research suggests that as a given knowledge-base exploits its recombination potential, its need for recharging the knowledge-base with new elements that can be recombined grows (Fleming, 2001). Supporting this basic logic researchers find that firms facing recombinant exhaustion as indicated by their building on long chains of patents are more likely to invest in science, presumably as a means of adding new recombinant fodder (new elements) to their knowledge-base (Ahuja and Katila, 2004).
Merging knowledge-bases through acquisition integration may lead to fairly nuanced implications. For instance Puranam and Srikanth (2007) find that structural integration may enable firms to exploit the available existing knowledge (the empirical construct) but potentially at a cost to the development of the capabilities inherent in the acquired firm (knowledge in the latent sense). Puranam, Singh, and Zollo (2006) show that the effects of structural integration are contingent; firms that are early in the product development trajectory may be hurt by structural integration, but such effects may be attenuated for firms with more developed innovation trajectories. Puranam, Singh, and Chaudhuri (2009) indicate that under certain conditions coordination between the acquired and acquirer knowledge may be better achieved through mechanisms other than structural integration. Finally, it has been proposed that the nature of the knowledge (tacit or explicit) may also significantly affect successful integration of knowledge following an acquisition (Ranft and Lord, 2002).
The size of the knowledge-base is not only affected by the organization’s capability to acquire new knowledge, but also by its capability to unlearn (deliberately) or its tendency to forget (unintentionally). Unlearning is necessary when established knowledge can constitute a barrier to further learning and has to be removed to make room for new knowledge (Bettis and Prahalad, 1995; Lyles and Schwenk, 1992; De Holan and Phillips, 2004, 2010). Forgetting, or the unintended loss of knowledge elements, can be productive (when obsolete knowledge is lost) or unproductive (when useful knowledge elements are lost).
Content
Content refers to the subject matter of the knowledge-base or the identity of the individual elements of knowledge. It is ‘what’ the organization knows. The most obvious use of the content of a knowledge-base is in terms of recognizing its overlap with other knowledge-bases. Such an approach can be used to operationalize the concept of relatedness between organizational knowledge-bases of different organizations (Mowery, Oxley, and Silverman, 1996; Ahuja and Katila, 2001). Research suggests that in the context of acquisitions moderate degrees of overlap between knowledge-bases lead to productive innovation performance after an acquisition (Ahuja and Katila, 2001). Similarly, the risk of knowledge appropriation is also likely to be higher in the context of knowledge relatedness (Mowery, Oxley, and Silverman, 1996).
In recognizing what an organization knows it is important to note the possibility that an organization’s knowledge-base may extend beyond the organization’s boundaries (Rosenkopf and Nerkar, 2001). Firms often use leveraging strategies or absorptive mechanisms to expand their knowledge-bases beyond their organizational boundaries. Such strategies could be informal, such as know-how exchange (Von Hippel, 1978; Rogers, 1983), or formal, such as licensing (Teece, 1986; Arora, Fosfuri, and Gambardella, 2001). Indeed, research suggests that some of the most important inventions may occur when firms use knowledge from beyond their own boundaries (Rosenkopf and Nerkar, 2001).
Noting this possibility we can make a distinction between an organization’s knowledge from within its boundaries and that from beyond its boundaries. The latter may imply a broader search space available to the organization than the in-house experience would indicate; however, assessing the eventual value of such external elements of knowledge is more complicated. On the one hand, the mastery of knowledge situated outside the organization may be more limited and, further, the perceived value of internal versus external knowledge may differ with external knowledge subjected to a ’not invented here’ bias that may limit its actual use by the organization (Katz and Allen, 1982). Conversely, knowledge obtained from outsiders could be valued more than internal knowledge because it appears more special and unique (Menon and Pfeffer, 2003). More broadly it appears that the relative utility of internal versus external knowledge is contingent on a variety of factors such as team composition, the proportion/number of local members versus cosmopolitans (Haas, 2006), characteristics of the knowledge-base such as size, specialization, and codification (Schulz, 2003), and even the perceptions of (or lack of) a shared or superordinate social identity (Kane, 2010).
Many different types of knowledge can be distinguished in talking about the content of a knowledge-base, but we will make a special note of just two kinds—technical knowledge versus scientific (Rosenberg, 1982; Brooks, 1994). Although it would appear that in science-driven sectors of the economy a knowledge-base that is significantly devoted to science may be an advantage, research suggests the need for caution in drawing such conclusions. Science can increase the recombinant potential of a knowledge-base by the identification of new elements, or provide a cognitive map of the landscape that can be used for search and recombination (Fleming, 2001; Fleming and Sorenson, 2004). It can also help to guide trial and error or experimental search (Gavetti and Levinthal, 2000); and it forms a very important basis for cognitive search (Ahuja and Katila, 2004; Gavetti and Levinthal, 2000). Yet, in spite of these advantages an excessive commitment to science may be detrimental to innovation because the institutional ethos of science is fundamentally different from technology (Stern, 2004). Science is about solving abstract problems, basic enquiry, and broad dissemination; technology is about creating pragmatic artifacts, commercial utility, and value appropriability. The conflicts between these modes of operation can undercut the science payoff to a technological knowledge-base (Gittelman and Kogut, 2003).
Veridicality
Veridicality refers to the fit between the external or ‘true’ information environment of an organization, and the reflection of that environment in the organization’s knowledge-base (see Fiol and Lyles, 1985; Fiol, 1994; Walsh, 1995; Huber, 1991). A perfectly veridical organizational knowledge-base would imply that there is no difference between the reality of the true world and its image in the organization’s cognition. This attribute is important because managerial and scientific decisions in an organization depend upon the information context perceived by the decision maker, not on what the true state of the world might be outside. Although at first glance higher veridicality might appear to be an unmitigated virtue (Hogarth, 1980), thoughtful analysis by scholars suggests that such a simplification or judgment may be premature for several reasons.
First, as Starbuck and Milliken (1988) and Walsh (1995) note, managers may not be well served by having the complexity of the real world replicated inside the organization’s knowledge-base. Au contraire, functionally effective knowledge-bases may entail the suppression of some elements of knowledge and the highlighting of others. Perceptual filters that amplify some information and attenuate other information may be critical to managerial action (Starbuck and Milliken, 1988; Walsh, 1995). For instance, McNamara, Luce, and Thompson (2002) find that a cognitively parsimonious classification of the competitive environment is more predictive of firm performance than an in-depth representation of it.
Second, inaccurate knowledge-bases and templates can in themselves be the basis of learning (Weick, 1991; Yayavaram and Ahuja, 2008). Experiential or trial-and-error learning is often the way that complex realities are uncovered. Such realities often take the form of interdependence relationships between knowledge elements. In the absence of a veridical map of reality firms make assumptions about such interdependencies and act upon them. Their resultant experiments confirm or invalidate these assumptions. Under either circumstance this leads to an increase in the knowledge of the real world as either a true dependency is discovered or an assumed one is found to be falsified (Yayavaram and Ahuja, 2008).
Third, the existence itself of a true reality that is mirrored by firms’ knowledge-bases is questioned by some research streams which suggest that there isn’t such a thing as an objective reality, rather the world is subjectively defined (Berger and Luckmann, 1966) and different interpretations of reality coexist. Under these premises the utility of a knowledge-base does not depend on the accuracy with which it represents the external environment: rather it is situationally dependent and has to be evaluated on the basis of the actions that can be derived from it and the conditions of use (Weick, 1991; Walsh, 1995). If there is no such thing as a ‘true reality’ the idea of accuracy associated with the concept of veridicality can be used to identify the accuracy with which firms’ knowledge-bases refer to the ‘consensually defined environment,’ i.e. the environment that constitutes the standard for organizations (Krackhardt, 1990). Under these circumstances less veridicality may be associated with greater system-level variety and innovation (e.g. Huber, 1991) for two reasons.
First, in the absence of definitive veridical mappings of the external informational environment firms need to make their own simplifying assumptions about such mappings. Such assumptions may increase the salience of different knowledge elements differentially across organizations, leading some to develop some elements of knowledge more than others (Yayavaram and Ahuja, 2008). Further, at the firm level a cognitive representation that moves away from the consensually defined environment may be the basis of new actions and strategies that have not already been implemented by other firms in the industry and may lead to more or breakthrough innovations.
Second, firms may generate more variety and potentially increase their innovative capabilities by increasing the salience of different knowledge elements differentially within the organization. Indeed organizations do not necessarily rely on a unique shared cognitive map. Rather, several representations may coexist within the same firm. For instance, Huber (1991) suggests that the wider the variety of interpretations held by the organization’s various units, the wider the range of the organization’s potential behaviors. Similarly, Fiol (1994) finds that learning involves the development of new and diverse interpretations of events and situations and it is fostered when managers actively encourage the development of different and conflicting views of what is thought to be true, while striving for a shared framing of the issues that is broad enough to encompass those differences.
The literature on interpretation and sense making (e.g. Daft and Weick, 1984; Milliken, 1990) suggests that firms scan the environment to collect information and then implement a process of information interpretation, which involves translating events to develop shared understandings and conceptual schemes. Two features of this sense-making process are important in the context of studying organizational knowledge-bases. First, organizational politics, organization structure, and managerial actions may all influence the final ‘truth’ that is recognized within the organization (Yayavaram and Ahuja, 2008; Lant et al., 1992, Walsh 1995, Mezias and Starbuck, 2003, Beck and Plowman, 2009). Second, the interpretive process itself may vary across organizations depending on the assumptions that the organization makes about the environment (Daft and Weick, 1984; Gavetti and Rivkin, 2007). A cognitive process involving linear thinking and logic is more likely to be used if the organization assumes that the external environment is concrete, that events and processes are hard, measurable, and determinant (Aguilar, 1967; Wilensky, 1967; Daft and Weick, 1984); an experiential process is more likely to be used instead when organizations assume that the external environment is unanalyzable, for instance as in the case of uncertain and quickly changing environments (Perrow, 1967; Duncan, 1972; Tung, 1979; Daft and Weick, 1984). In these cases the interpretation process is likely to be ‘more personal, less linear, more ad hoc and improvisational’ (Daft and Weick, 1984). The use of a cognitive as opposed to an experiential search for the development of the representation of the environment is also going to be a function of the extent to which an organization is active in intruding into the environment, a factor which may be determined by the characteristics of the firm, such as age and size, or of the environment itself (Daft and Weick, 1984; Gavetti and Rivkin, 2007).
Differentiation
Differentiation of knowledge-bases has been variously defined as the number of dimensions in a knowledge-base (Walsh, 1995) or as splitting a knowledge base into clusters Lawrence and Lorsch, 1967; Yayavaram and Ahuja, 2008). It refers to the partitioning of a knowledge-base into two or more components and has been argued to provide several innovation relevant benefits. First, differentiation permits the separation of knowledge into discrete categories and in turn this can facilitate specialization (Brusoni, Prencipe, and Pavitt, 2001). Relatedly, differentiation in an organizational knowledge-base may be necessary to optimally utilize external knowledge (Brusoni et al., 2001). The embodiment of knowledge into products requires the integration of different types of knowledge (often) from different sources. Having an internally differentiated knowledge-base can help the organization to coordinate the knowledge-flows from all these extra-corporate actors. In this sense large corporations may often maintain broader and more differentiated technological knowledge-bases than would appear necessary from their product range (Patel and Pavitt, 1997; Granstrand, Patel, and Pavitt, 1997; Brusoni et al., 2001).
In contrast to the coordination argument proposed above, researchers have also raised the possibility of a control motive for building a differentiated knowledge-base (Tiwana and Keil, 2006). Since firms often need to combine technology from different sources into their own technology or products, developing a differentiated knowledge-base that spans domains beyond their own core technologies may be useful in its own right. Investments in such ‘peripheral’ technologies can help the firm control and better govern the relationships through which external knowledge in such technologies is brought into the firm. In particular Zander and Kogut (1995) talk about system dependence of the knowledge to refer to the extent to which knowledge is dependent on many different groups of experienced people for its production.
Differentiation in knowledge-bases can also lead to the possibilities of cross-fertilization. From a recombinant perspective knowledge-base differentiation should provide the possibility of combining high search scope with search depth (Katila and Ahuja, 2002) with positive consequences for innovation (Quintana-Garcia and Benavides-Velasco, 2008). Searching across the differentiated sub-units of the knowledge-base provides the potential for search breadth while searching within the sub-units of the knowledge-base provides an opportunity for developing search depth. Miller, Fern, and Cardinal (2007) point out that, in the context of large multidivisional firms, knowledge from outside the division but inside the corporation has greater impact than knowledge from outside the corporation or within the sub-unit. Knowledge-base differentiation can also occur along functional lines (Nerkar and Roberts, 2004; Brown and Duguid, 2001) providing another basis for cross-fertilization. From an experiential perspective, differentiation can also provide the basis of search by analogy. Breadth and depth of experiential knowledge increase the likelihood that search will be more effective in new and complex contexts (Gavetti, Levinthal, and Rivkin, 2005) allowing managers to derive a favorable set of policies and principles from experienced settings and apply them in the new contexts.
Analogical reasoning can also enable organizations to translate differentiation in the knowledge-base into superior innovative performance (Novelli, 2010). By increasing the number and diversity of knowledge inputs to which an organization is exposed, differentiation increases the organization’s ability to recognize more general patterns across variations, i.e. patterns of underlying constructs that minimize the distance across all the different variations observed (Novelli, 2010). These patterns constitute new knowledge that has the potential to be applied to contexts where such patterns have not been applied before. Thus, cross-fertilization via abstraction and superior generalization can become the basis of new product generation.
Additionally, differentiation can improve the performance of search for innovations through search simplification and narrowing (Yayavaram and Ahuja, 2008). In the context of recombinant search that we described above inventors create new products and processes by recombining elements or groups of elements into new syntheses. However, a key intermediate step is to recognize interdependencies between elements and group or cluster them on the basis of those interdependencies. This clustering helps to reduce the combinatorial complexity of the search process. However, identification of interdependencies between elements that can be used as the basis of grouping is itself difficult. Splitting up a knowledge-base into clusters, i.e. differentiating it, enables the recognition of interdependencies by limiting the number of elements that must be simultaneously studied (Yayavaram and Ahuja, 2008).
Integration
Integration refers to the building of connections across the differentiated components of a knowledge-base. Although integration of a differentiated knowledge-base may be desirable for a variety of reasons (see below), integration does not naturally or automatically or even beneficially always follow differentiation. The logic for integration following differentiation is fairly straightforward and broadly (though not universally) accepted (Postrel, 1998; 2002). The classical argument suggests that differentiation and integration are complements—to make the most of more differentiated knowledge integration mechanisms are needed (Grant, 1996; Nesta and Saviotti, 2004). While differentiation can enable an organization to collect information as a specialist in multiple areas, and possibly faster and deeper than a generalist would, integration of that knowledge is often required to produce successful applications (Lawrence and Lorsch, 1967). This basic intuition that differentiation should be accompanied by integration is, however, subject to several caveats. Integration may fail to follow differentiation because (1) it is needed but resources are not provided for it, (2) resources are provided for it but execution of integration is difficult, or (3) it is not needed or perceived to be not needed. Of these the first argument is relatively obvious, although no research appears to have examined it in detail. Several arguments suggest support for the second proposition. For instance, integration is hard because language differs across the differentiated components of a knowledge-base (Carlile, 2002). Similarly, integration can run into a legitimacy problem. Whereas the differentiated elements of knowledge are part of one hierarchy—the differentiated sub-unit that they belong to—and are regarded as legitimate within that hierarchy, integration by definition spans sub-unit hierarchies. Thus, it may be the case that integrative elements or arrangements simply do not get the organizational legitimacy, access, or attention that is required to make them effective.
The third argument against the occurrence of integration is probably the most controversial and the most interesting—integration may not necessarily be needed even in the presence of differentiation. The differentiation–integration duality draws attention to the fact that in any differentiated knowledge-base individual sub-units have to make resource choices in terms of investing in two different types of knowledge—investing in deepening their own specialist knowledge and investing in learning about the knowledge of others (Postrel, 1998; 2002). Postrel (2002) provides an interesting abstraction relating these two aspects of ‘specialist capability’ and ‘trans-specialist understanding’ with performance through the ‘design production function.’ Postrel’s formal analysis then provides a counter-intuitive finding—that integration and differentiation may not necessarily be complements; indeed for certain ranges of parameters in his model they are actually substitutes. His explanation for this effect is that the high capability of very effective specialists buffers the other from needing to know too much about how the first does his or her job. Alternately, if the specialist is of low capability then high levels of trans-specialist knowledge are helpful because the others know the limitations of the focal actor’s capabilities and adapt themselves accordingly.
Puranam and others highlight that strong integration may actually limit the potential and effectiveness of differentiated knowledge-bases (Puranam, Singh, and Chaudhuri, 2009). This suggests that the different mechanisms of integrating a knowledge-base may differ in their applicability to a given differentiation problem. It also draws attention to another important aspect of knowledge-base structuring, the balance between differentiation and integration. Following Simon’s lead, researchers have examined knowledge-bases in terms of their decomposability—a derived attribute that emerges from the combination of differentiation and integration. When a knowledge-base is differentiated and thus has several distinct components, integration mechanisms span and connect these differentiated sub-units. These spanning linkages or couplings can be distributed in several different patterns. When the couplings are pervasive the knowledge-base can be described as non-decomposable, when there are no couplings, the knowledge-base can be thought of as modular or decomposable, with each sub-unit being essentially stand-alone. Between these two extremes stand nearly-decomposable knowledge-bases—knowledge-bases where the differentiated sub-segments are connected by a few linkages (Simon, 1962).
Yayavaram and Ahuja (2008) demonstrate that in the context of the semiconductor industry, nearly decomposable knowledge-bases out-perform fully decomposable and non-decomposable knowledge-bases in terms of the utility of the inventions generated from them as well as the knowledge-base’s own malleability. Malleability, like decomposability, can be considered a derived property of a knowledge-base. It refers to the knowledge-base’s capacity for change. Similarly, in the context of alliances, Schilling and Phelps (2007) show that alliance networks that secure both high clustering and high reach positively affect innovative output.
Embeddedness
The last major dimension of an organizational knowledge-base examined in this review is embeddedness. Embeddedness refers to the degree to which the knowledge in a given knowledge-base is formal, observable, codified, or articulated versus informal, tacit, or organizationally embedded (Nonaka, 1991, 1994; Zander and Kogut, 1995; Haas and Hansen, 2007; Nonaka and Von Krogh, 2009). Tacit knowledge refers to the knowledge that is ‘unarticulated and tied to the senses, movement skills, physical experiences, intuition of implicit rules of thumb.’ Conversely, explicit knowledge is ‘uttered and captured in drawings and writing’ and has ‘a universal character, supporting the capacity to act across contexts’ (Polanyi, 1966; Nonaka and Takeuchi, 1996; Nonaka and Von Krogh, 2009). Embeddedness is related to knowledge teachability and observability, i.e. the degree to which capable competitors can copy it (Zander and Kogut, 1995).
Research also finds that highly embedded knowledge-bases can be expected to be ‘sticky’ (Szulanski, 1996; Ahuja 2002). Diffusion of knowledge from such knowledge-bases may be difficult both from the perspective of competitors trying to extract knowledge as well as the organization itself trying to use the knowledge in a different location or application. Research also suggests that even for highly explicated forms of knowledge, there eventually remains an embedded component that limits its mobility. Subtle evidence to this effect is also provided by Rosenkopf and Almeida (2003), and their colleagues, who find that employee mobility is connected with higher rates of cross-citation between organizations.