战略管理2

Strategic Management Journal, Vol. 17(Winter Special Issue), 27-43 (1996) EXPLORING INTERNAL STICKINESS: IMPEDIMENTS TO THE TRANSFER OF BEST PRACTICE WITHIN THE FIRM GABRIEL SZULANSKI The Wharton School, University of Pennsylvania, Philadelphia, Pennsylvania, U.S.A. The ability to transfer best practices internally is critical to a firm's ability to build competitive advantage through the appropriation of rents from scarce internal knowledge. Just as a firm's distinctive competencies might be difficult for other firms to imitate, its best practices could be difficult to imitate interncally. Yet, little systematic attention has been paid to such internal stickiness. The author analyzes intemal stickiness of knowledge transfer and tests the resulting model using canonical correlation analysis of a data set consisting of 271 observations of 122 best-practice transfers in eight companies. Contrary to conventional wisdom that blames primarily motivational factors, the study findings show the major barriers to internal knowledge transfer to be knowledge-related factors such as the recipient's lack of absorptive capacity, causal ambiguity, and an arduous relationship between the source and the recipient. The identification and transfer of best practices is emerging as one of the most important and widespread practical management issues of the latter half of the 1990s. Armed with meaningful, detailed performance data, firms that use fact- based management methods such as TQM, bench- marking, and process reengineering can regularly compare the performance of their units along operational dimensions. Sparse but unequivocal evidence suggests that such comparisons often reveal surprising performance differences between units, indicating a need to improve knowledge utilization within the firm (e.g., Chew, Bresnahan, and Clark, 1990).' Because internal transfers typi- Key words: internal stickiness, best practice transfer, knowledge transfer, knowledgement management, rent appropriation I Besides the published references, I have found up to 10: I gaps in performance in otherwise comparable units, and gaps of 2: I rather frequently. Personal communication with Robert Camp, a widely known benchmarking specialist at Xerox, confirmed that gaps of 200-300 percent are a typical finding in internal benchmarking efforts. CCC 0143-2095/96/S20027-17 ? 1996 by John Wiley & Sons, Ltd. cally are hindered less by confidentiality and legal obstacles than external transfers, they could be faster and initially less complicated, all other things being equal. For those reasons, in an era when continuous organizational learning and relentless performance improvement are needed to remain competitive, companies must increasingly resort to the internal transfer of capabilities.2 Yet, experience shows that transferring capa- bilities within a firm is far from easy. General Motors had great difficulty in transferring manu- facturing practices between divisions (Kerwin and Woodruff, 1992: 74) and IBM had limited suc- cess in transferring reengineered logistics and hardware design processes between business units (The Economist, 1993). Although strategic man- agement research has examined impediments to the transfer of best practices (i.e., organizational 2Such concern is typically expressed as the need to avoid the duplication of effort or to capture the benefits of the internal 'pockets of excellence' and the 'great ideas' that are implemented on a daily basis (see for example Xerox, 1992: 1-1). 28 G. Szulanski capabilities) between firms because such practices are seen as important drivers of firm performance (e.g., Prahalad and Hamel, 1990; Grant, 1991), impediments to transfer capabilities within firms have received little attention. This article reports the findings of a systematic empirical investigation of internal stickiness. The study analyzed internal stickiness of knowledge transfer and tested the resulting model by canoni- cal correlation analysis of a data set consisting of 271 observations of 122 best-practice transfers in eight companies. Contrary to conventional wis- dom that places primary blame on motivational factors, the major barriers to internal knowledge transfer are shown to be knowledge-related fac- tors such as the recipient's lack of absorptive capacity, causal ambiguity, and an arduous relationship between the source and the recipient. ANALYZING INTERNAL STICKINESS Definitions The transfer of best practice inside the firm has a concrete and fairly unambiguous meaning to practitioners. It connotes the firm's replication of an internal practice that is performed in a superior way in some part of the organization and is deemed superior to internal alternate practices and known alternatives outside the company. Practice refers to the organization's routine use of knowl- edge and often has a tacit component, embedded partly in individual skills and partly in collabo- rative social arrangements (Nelson and Winter, 1982; Kogut and Zander, 1992). The word 'transfer' is used rather than 'dif- fusion' to emphasize that the movement of knowledge within the organization is a distinct experience, not a gradual process of dissemi- nation, and depends on the characteristics of everyone involved. Transfers of best practice are thus seen as dyadic exchanges of organizational knowledge between a source and a recipient unit in which the identity of the recipient matters. The exchange of organizational knowledge consists of an exact or partial replication of a web of coordi- nating relationships connecting specific resources so that a different but similar set of resources is coordinated by a very similar web of relation- ships. In this sense, transfers of best practice could be conceived as replications of organiza- tional routines (Winter, 1995). Stages in the transfer process Intrafirm transfer of best practice is seen as an unfolding process consisting of stages in which characteristic factors not only appear in greater or lesser degree but also in a certain order of occurrence. Four stages are identified: initiation, implementation, ramp-up, and integration.3 Initiation This stage comprises all events that lead to the decision to transfer. A transfer begins when both a need and the knowledge to meet that need coexist within the organization, possibly undis- covered. The discovery of the need may trigger a search for potential solutions, a search that leads to the discovery of superior knowledge. Alternatively, the discovery of superior knowl- edge may reframe as unsatisfactory a hitherto satisfactory situation (cf. Rogers, 1983; Zaltman, Duncan, and Holbek, 1973; Glaser, Abelson, and Garrison, 1983). In the language of bench- marking, the discovery of superior results will reveal how good is 'best' and who is currently best (Balm, 1992). That discovery may be fol- lowed by a more focused inquiry into how those results are obtained. Once the need and a potential solution to that need are identified, their fit-that is, the feasibility of the transfer-is explored. As Teece (1976) found, that process often requires months of information collection and evaluation. The events that lead to the decision to transfer may follow an orderly sequence or one that resembles the working of an organized anarchy (Cohen, March, and Olsen, 1972). Implementation The implementation stage begins with the decision to proceed. During this stage, resources flow between the recipient and the source (and maybe a third party). Transfer-specific social ties between the source and the recipient are estab- I The stages model presented in this section builds on the insights of the rich empirical traditions of research on inno- vation diffusion (Rogers, 1983), social change (see Glaser et al., 1983, for a review), technology transfer (e.g., Teece, 1976; Galbraith, 1990), and implementation (e.g., Tyre, 1991; Tyre and Orlikowski, 1994). Exploring Internal Stickiness 29 lished and the transferred practice is often adapted to suit the anticipated needs of the recipient, to preempt problems experienced in a previous transfer of the same practice, or to help make the introduction of new knowledge less threatening to the recipient (cf. Rice and Rogers, 1980: 508- 509; Buttolph, 1992: 464). Implementation- related activities cease or at least diminish after the recipient begins using the transferred knowl- edge. Ramp-up The ramp-up stage begins when the recipient starts using the transferred knowledge, that is, after the first day of use. During this stage, the recipient will be predominantly concerned with identifying and resolving unexpected problems that hamper its ability to match or exceed post- transfer performance expectations. The recipient is likely to use the new knowledge ineffectively at first (cf. Baloff, 1970; Adler, 1990; Galbraith, 1990; Chew, 1991; Chew, Leonard-Barton, and Bohn, 1991), but gradually improves perform- ance, ramping up toward a satisfactory level. The ramp-up stage provides a relatively brief window of opportunity to rectify unexpected problems (Tyre and Orlikowski, 1994). Integration The integration stage begins after the recipient achieves satisfactory results with the transferred knowledge. Use of the transferred knowledge gradually becomes routinized. This gradual routi- nization is incipient in every recurring social pattern (Berger and Luckman, 1966; 53). As time passes, a shared history of jointly utilizing the transferred knowledge is built up in the recipient, actions and actors become typified, and types of actions are associated with types of actors. These shared meanings and behaviors facilitate coordi- nation of the activities, making behaviors under- standable, predictable (March and Simon, 1958; Nelson and Winter, 1982; Tolbert, 1987) and stable (Berger and Luckman, 1966). In this way, new practices become institutionalized. They pro- gressively lose their novelty and become part of the objective, taken-for-granted reality of the organization (Berger and Luckman, 1966; Zucker, 1977). Analyzing the difficulty of transferring practices within the firm The notion of internal stickiness connotes the difficulty of transferring knowledge within the organization.4 The point of departure for the analysis of internal stickiness is Arrow's (1969) classificatory notes on the transmission of techni- cal knowledge. Arrow observed that the capacity of a social conduit of knowledge is inherently constrained and hence social conduits are costly to use. Referring to Arrow, Teece (1977: 242) argued that the ease or difficulty of transferring technical knowledge is reflected in the cost of a transfer. More recently, von Hippel (1994) introduced the notion of 'sticky information' to describe information that is difficult to transfer, stickiness being reflected in the incremental cost of transferring the information. Cost, eventfulness and internal stickiness Cost could be a poor descriptor of difficulty, however. First, deciding exactly which portion of the cost of a transfer actually reflects difficulty- the increment-is a matter of conjecture without a base case-the cost of the same transfer without such difficulty. Systematically constructed base cases are rare, and past experience in transferring knowledge might be inadequate as a base case if prior transfers are not equivalent to the one under scrutiny. Moreover, experience is likely to be distorted by faulty memory, ex post embel- lishment of past events, and noncomparable trans- fer cost accounting. Second, cost might fail to discriminate between problems that are equally costly but qualitatively very different. Some prob- lems are resolved routinely or by prespecified contingency plans with relatively little effort from all but the most directly involved participants. Other problems involve participants whose atten- tion is not normally required, such as senior managers, to expedite the identification of pos- sible solutions and explicitly coordinate their implementation. This second type of problem is 4 In the strategy literature, the adjective 'sticky' has been used as a synonym for 'inert' (Porter, 1994), or 'difficult to imitate' (Foss, Knudsen, and Montgomery, 1995). Macroecon- omists use the term 'sticky price' to mean 'slow to adjust'. In the lingo of Wall Street, 'sticky' means 'difficult to sell.' 30 G. Szulanski likely to engage, and thus be noticed by, a broader range of participants. This second type of problem is also likely to be remembered as being relatively more difficult to resolve, at least by those who could not cope with them without assistance. Hence, problems that participants cannot handle on a routine basis are likely to evoke the greatest overall perception of difficulty. Whether or not problems are objectively difficult to resolve mat- ters little because perceptual processes, not objec- tive properties, affect organizational behavior (cf. Hellriegel and Slocum, 1974). The perceived dif- ficulty of a problem for the individual is what determines his or her reaction to it. Therefore, transfers that involve the most nonroutine prob- lems will be perceived as the most difficult, other things being equal. This suggests that the notion of eventfulness, the extent to which problematic situations experienced during a transfer are wor- thy of remark, is conceptually related to the notion of difficulty. Eventfulness has a universal base case: a transfer that is not at all difficult is unremarkable, is uneventful. The implication is that an organization equipped with effective rou- tines to handle all aspects of a transfer is unlikely to consider that transfer sticky. Eventfulness could be translated into an out- come-based descriptor of stickiness. If an organi- zation has effective routines to handle all aspects of a knowledge transfer, it should be able to specify milestones, budgets, and expectations for the transfer process rather accurately. To the extent that the transfer turns out to be sticky, requiring ad hoc solutions, some of those mile- stones are likely to be missed, budgeted cost will be exceeded, and some of the participants' expectations about the transfer will not be fully met. As in the case of cost, the outcome-based descriptor requires the specification of a transfer- specific base case in the form of ex ante expec- tations. A transfer-specific base case is not necessary, however, if the descriptor of stickiness is based on the process rather than the outcome of the process. Combining the notion of eventfulness with the stages model presented in the preceding section provides four different descriptors of stickiness, one for each stage of the transfer. The process model suggests that the problems encountered as the transfer unfolds will vary according to the stage of the transfer. During the initiation stage, problems will stem from efforts to identify needs, identify knowledge that meets those needs, and assess the feasibility of the transfer. During the implementation stage, problems will reflect efforts to bridge the communication gap between the source and the recipient or to adapt the practice to the recipi- ent's needs. During the ramp-up stage, problems will reflect the struggle to achieve satisfactory performance. Finally, during the integration stage, problems will reflect efforts to achieve and preserve routine use of the new knowledge in the recipient. The more these problems require participants to develop ad hoc solutions-that is, the more remarkable the problems are-the higher will be the perceived eventfulness of the transfer. ORIGINS OF INTERNAL STICKINESS Prior research suggests that four sets of factors are likely to influence the difficulty of knowledge transfer: characteristics of the knowledge trans- ferred, of the source, of the recipient, and of the context in which the transfer takes place (e.g., Leonard-Barton, 1990; Teece, 1977; Rogers, 1983). Some researchers place an almost exclu- sive emphasis on the attributes of the knowledge transferred (e.g., Zander and Kogut, 1995; Winter, 1987). Others stress the characteristics of the situation in which the transfer occurs (e.g., Arrow, 1969). However, all four sets of factors can be used together in an eclectic model that allows their relative influence to be measured. Descriptions of the primary variables within each set of factors follow. Characteristics of the knowledge transferred Causal ambiguity Modeling a capability as a production function, Lippman and Rumelt (1982) argued that difficulty in the replication of a capability is most likely to emanate from ambiguity about what the factors of production are and how they interact during production. When the precise reasons for success or failure in replicating a capability in a new setting cannot be determined even ex post, causal ambiguity is present and it is impossible to pro- Exploring Internal Stickiness 31 duce an unambiguous list of the factors of pro- duction, much less measure their marginal contri- bution (Rumelt, 1984; 562). Key to their argument is the notion of irreduc- ible uncertainty. Polanyi (1962: 49) suggested that the undefinable portion of knowledge is embodied in highly tacit human skills. Tacitness could also be a property of collectively held knowledge (Winter, 1987; Kogut and Zander, 1992) and it is often singled out as a central attribute of knowledge with respect to its transfer- ability (Spender, 1993; Nonaka, 1994; Grant, 1996). Causal ambiguity could also result from imperfectly understood idiosyncratic features of the new context in which knowledge is put to use (Tyre and von Hippel, forthcoming; Winter, 1995). Unprovenness Knowledge with a proven record of past useful- ness is less difficult to transfer. Such a record hints of robustness and helps in the process of selecting knowledge for transfer. Without such a record, it is more difficult to induce potential recipients to engage in the transfer (Rogers, 1983) and to legitimize controversial integration efforts (Goodman, Bazerman, and Conlon, 1980; Nelson and Winter, 1982). Characteristics of the source of knowledge Lack of motivation A knowledge source may be reluctant to share crucial knowledge for fear of losing ownership, a position of privilege, superiority; it may resent not being adequately rewarded for sharing hard- won success; or it may be unwilling to devote time and resources to support the transfer. Not perceived as reliable An expert and trustworthy source is more likely than others to influence the behavior of a recipi- ent (e.g., see Perloff, 1993, ch. 6, for a review). When the source unit is not perceived as reliable, is not seen as trustworthy or knowl- edgeable, initiating a transfer from that source will be more difficult and its advice and example are likely to be challenged and resisted (cf. Walton, 1975). Characteristics of the recipient of knowledge Lack of motivation The reluctance of some recipients to accept knowledge from the outside (the 'not invented here' or NIH syndrome) is well documented (e.g., Hayes and Clark, 1985; Katz and Allen, 1982). Lack of motivation may result in foot dragging, passivity, feigned acceptance, hidden sabotage, or outright rejection in the implementation and use of new knowledge (cf. Zaltman, Duncan, and Holbek, 1973). Lack of absorptive capacity Recipients might be unable to exploit outside sources of knowledge; that is, they may lack absorptive capacity (Cohen and Levinthal, 1990: 128). Such capacity is largely a function