MADE TO FIT: HOW PRACTICES VARY AS THEY DIFFUSE
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
姝 Academy of Management Review
2010, Vol. 35, No. 1, 67–92.
MADE TO FIT: HOW PRACTICES VARY AS
THEY DIFFUSE
SHAHZAD M. ANSARI
University of Cambridge
PEER C. FISS
University of Southern California
EDWARD J. ZAJAC
Northwestern University
We extend research on the diffusion of corporate practices by providing a framework
for studying practice variation during diffusion processes. Specifically, we theorize
about how population-level mechanisms of diffusion link with organization-level
mechanisms of implementation that lead to the adaptation of practices. We also
identify technical, cultural, and political elements of fit (or misfit) between diffusing
practices and adopters and analyze how the process of attaining fit across these
elements can trigger different patterns of adaptation.
An extensive body of research on the diffusion across time and space (Greve, 1998; Terlaak &
of practices has significantly enhanced our un- Gong, 2008).
derstanding of “how things—ideas and prac- Furthermore, most of this work has focused on
tices— get from here to there” (Katz, 1999: 145), the diffusion episode of the practice at the inter-
largely by developing parsimonious models organization level—that is, the adoption deci-
that draw on economic (e.g., Banerjee, 1992; sion. Given this interorganizational focus, prior
Bikhchandani, Hirshleifer, & Welch, 1998; models have usually made certain simplifying
Lieberman & Asaba, 2006) and sociological assumptions about the homogeneity of diffusing
mechanisms (e.g., Abrahamson, 1996; Strang & practices across time and space, treating them
Macy, 2001; Strang & Meyer, 1993). In most of this as essentially invariant rather than mutating.
research scholars have typically assumed a This approach, while parsimonious, has led to a
population-level perspective, emphasizing inter- relative neglect of theoretical attention to prac-
organizational conditions. For instance, eco- tice variation at the organization level (Cool,
nomic models tend draw on informational argu- Dierickx, & Szulanski, 1997), with little attention
ments, pointing to a growing level of general to issues of adaptation and internal variety in
information about the value of a practice in af- diffusing practices as they wind their way
fecting diffusion decisions, whereas sociologi- through organizations (e.g., Mamman, 2002;
cal models have tended to use more reputa- O’Mahoney, 2007).
tional arguments that relate to growing Thus, although existing models in diffusion
pressures for social conformity. Taken together, theory have offered considerable insight into
these bodies of literature offer a variety of ra- why practices are initially adopted by an orga-
tional, boundedly rational, and social explana- nization, they typically do not delve deeply into
tions for the adoption and diffusion of practices what happens to such practices during and after
adoption (Wolfe, 1994; Zeitz, Mittal, & McAulay,
We gratefully acknowledge the helpful comments and 1999). We see this as an important omission,
suggestions of former associate editor Pamela Tolbert, three since management practices often cannot be
anonymous reviewers, Paul Adler, Ed Carberry, Bill Cooper, adopted by user organizations as “off-the-shelf”
Marie-Laure Djelic, Henrich Greve, Stefan Jonsson, Mark solutions. Instead, we suggest that diffusing
Kennedy, and participants at the 20th EGOS Colloquium,
Queen’s University, Northwestern University, and the Orga-
practices are likely to evolve during the imple-
nization & Strategy Workshop at the University of Southern mentation process, requiring custom adapta-
California. tion, domestication, and reconfiguration to
67
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright
holder’s express written permission. Users may print, download, or email articles for individual use only.68 Academy of Management Review January make them meaningful and suitable within spe- tices as the degree to which the characteristics cific organizational contexts (Robertson, Swan, of a practice are consistent with the (perceived) & Newell, 1996; Strang & Kim, 2004). needs, objectives, and structure of an adopting To be sure, some scholars have paid attention organization. While this notion of fit is generally to the adaptation of diffusing practices, where seen as self-evident, the way in which different the transfer and diffusion of management prac- types of fit affect the adoption and adaptation tices among different local contexts consist process is much less well understood. We there- of translation, coconstruction, and editing activ- fore suggest that the diffusion process across ities in different cultural and social contexts and time and across adopters should be assessed as may lead to divergence and variability in prac- an issue of dynamic fit between practice and tices that are being adopted, enacted, and adopter and that this fit is influenced by techni- adapted (Johnson & Hagström, 2005; Sahlin- cal, cultural, and political factors. We then posit Andersson, 1996; Westphal, Gulati, & Shortell, that different forms of fit and misfit will trigger 1997; Zilber, 2006). For instance, Zbaracki (1998) different patterns of practice adaptation and examined the relationship between the rhetoric that this understanding will allow us to predict and reality of total quality management (TQM) both the timing (i.e., early versus late) and the in use and showed how the practice was so- form of practice adaptation by an organization. cially constructed and adapted inside different By developing a parsimonious framework of the organizations. Similarly, Frenkel (2005) found adaptation of diffusing practices that places that scientific management and human re- special emphasis on the interaction between sources models imported from the United States characteristics of the diffusing practice and into Israel were reinterpreted by the state, pri- those of the adopter, we aspire to unify and vate employers, and a labor union to be more in reconcile divergent strands of the literature ad- line with prevalent macrocultural discourse. dressing practice diffusion. While such mostly case-oriented studies (e.g., Indeed, we believe our model carries substan- Czarniawska & Sevón, 2005; Johnson & Hag- tive implications for those interested in how ström, 2005; Morris & Lancaster, 2006; Saka, 2004) practices spread. Prior research has to a large offer rich insights into how carriers and hosts extent focused not on variation but on diffusion coconstruct management practices diffusing speed, including identifying and measuring into new settings, they remain quite discon- rates of diffusion and the timing of adoption— nected from the more parsimonious models used that is, early versus late adopters (e.g., Rogers, in large-scale studies of innovation diffusion 1995). The motivation for this emphasis was that (Abrahamson, 2006). Indeed, the dearth of at- policy makers and economists wanted to under- tempts to bridge interorganizational mecha- stand how ”progressive” innovations could nisms of diffusion with intraorganizational im- spread faster in an economy and, implicitly, plementation and adaptation is striking, how ”laggards” could be convinced, thus accel- leading us to propose that an enhanced under- erating the diffusion process (Abrahamson, 1991; standing of the diffusion of corporate practices Rogers & Shoemaker, 1971). The underlying can be achieved by systematically analyzing model of the practice, however, is one of rela- how and why practices are adapted by organi- tively uniform and invariant practices. zations over the course of the diffusion process. We contend, however, that developing a More specifically, we offer a theoretical frame- greater understanding as to when and how work for analyzing how practices vary as they practices change as they diffuse is highly rele- diffuse and are implemented. We begin by pro- vant for several reasons. Two main motivations posing two fundamental dimensions to explain relevant to policy makers are detecting and ex- variation in the ways in which diffusing prac- ploiting variation, on the one hand, and sup- tices are implemented: fidelity and extensive- pressing variation, on the other hand. Regard- ness. ing the first motivation, improving the outcomes Furthermore, we argue that the specific pat- of a diffusion process will frequently require terns of practice adaptation will depend on the knowing where and when enhanced versions of fit between the diffusing practice and the adopt- a practice are likely to appear. In this way the ing organization. Drawing on Nadler and Tush- insights of experimentation can be detected and man’s (1980) work, we define fit of diffusing prac- disseminated to other potential adopters. Alter-
2010 Ansari, Fiss, and Zajac 69
natively, policy makers may be interested in the processes leading to adoption. The first set
assuring consistent and faithful implementation of explanations has its roots in the economic
of a practice, so knowing where variation is literature and builds on the rational actor
likely to emerge is therefore useful in aiding model. It presents arguably the most dominant
attempts to suppress deviation from a preferred perspective in the diffusion of innovation litera-
model. Accordingly, developing a theoretical ture (Rogers, 1995; Sturdy, 2004), conceiving of
framework consisting of technical, cultural, and adopters as rational actors that scan their envi-
political parameters that provides an enhanced ronment and make efficient choices. In keeping
understanding of variation in the diffusion of with Strang and Macy’s (2001) terminology, we
practices, as we endeavor to provide in this refer to these explanations as rational accounts.
study, can assist in localized and targeted The second set of explanations is somewhat
searches for innovative practices and can also more eclectic but overall more closely associ-
allow for more timely interventions, including ated with a sociological perspective and a focus
the possible modification of these parameters on the social embeddedness of actors. These
from policy makers eager to sponsor particular explanations have been variously called “fads
versions of practices deemed more desirable for and fashion perspectives” (Abrahamson, 1991),
business and society. “contagion accounts” (Strang & Macy, 2001), or
We proceed by discussing the theoretical an “institutional perspective” (Jonsson, 2002;
mechanisms proposed in previous approaches Sturdy, 2004). However, none of these labels
to diffusion, and we then introduce our notion of seems ideal, since each tends to exclude other
adaptation (where diffusion meets implementa- significant aspects of a more sociological per-
tion) and its dimensions. Next, we introduce the spective. To recognize this eclecticism, and at
concept of fit as a crucial intermediate mecha- the same time juxtapose it with more economi-
nism for understanding diffusion processes and cally based rational accounts, we refer to this
show how various degrees of fit across multiple second cluster of explanations as social ac-
dimensions result in different degrees of likely counts. We discuss each in more detail below.
adaptation by recipient organizations during
various stages of the diffusion process.1 We con-
Rational Accounts
clude by discussing the implications of our ap-
proach for both the study of practice adaptation Emerging from the field of economics, rational
and diffusion. accounts have an immediate intuitive appeal,
since they focus on the presumed economic ben-
efits that result from the adoption of a practice.
PRIOR MODELS OF PRACTICE ADOPTION
In fact, the connection between cost effective-
The literature on the diffusion of practices ness and the likelihood of diffusion is one of the
among corporations—identified as one of the most widely reported findings in the innovation
key mechanisms in the study of organizations diffusion literature (Rogers, 1995; Strang & Macy,
(e.g., Davis & Marquis, 2005)—is currently char- 2001).
acterized by two sets of explanations regarding Rational accounts tend to come in two ver-
sions. The first, focusing on evolutionary pro-
cesses, suggests that selection forces weed out
1
In conceptualizing adaptation processes, one can con- the weaker performers who fail to adopt an ef-
ceive of four scenarios of adaptation: (1) little or no change in ficient practice (Katz & Shapiro, 1987; Mansfield,
practice or in organization, leading to essentially “as is” 1961). In the second, optimizing version, effective
adoption; (2) change in practice but not in organization,
innovations are adopted by rational decision
leading to adaptation of the practice; (3) change in organi-
zation but not in practice, leading to adaptation of the orga- makers who make the choices that lead to the
nization; and (4) change in both organization and practice, diffusion of beneficent innovations (Chandler,
leading to coevolution or mutual adaptation. Although there 1962; Teece, 1980; Williamson, 1979). In both
will always be some degree of mutual adaptation (Van de forms a key mechanism explaining increasing
Ven, 1986), our intent here is to examine how practices vary
levels of adoption pertains to information cas-
as they diffuse. Thus, we focus on the adaptation of practices
during the diffusion process, rather than on processes of cades (e.g., Banerjee, 1992; Bikhchandani et al.,
organizational change (e.g., Greve & Taylor, 2000; Leonard- 1992, 1998), where adoption processes build mo-
Barton, 1988). mentum as firms use observed behaviors of70 Academy of Management Review January
early adopters, presumably with more accurate allow for the possibility that initial adoption is
information about the practice, to update their driven by economic efficiency rationales. How-
own value expectations regarding a diffusing ever, once a critical mass is reached, efficiency
practice (Terlaak & Gong, 2008). In such models concerns become more and more irrelevant be-
imitation follows from a heuristic of social cause the adoption process is increasingly
proof—that is, firms infer from the actions of driven by bandwagon pressures and legitimacy
other firms what constitutes appropriate actions concerns—that is, processes that are largely de-
to minimize search costs and to avoid the costs coupled from a practice’s technical efficiency
of experimentation (Rao, Greve, & Davis, 2001). (Tolbert & Zucker, 1983). In other words, once a
With greater diffusion more information about threshold is crossed, efficiency concerns are
the utility of a practice reduces its associated replaced by social pressures from outside stake-
uncertainty and, thus, the risk of adoption, holders, leading organizations to adopt prac-
speeding up the diffusion process. However, tices with less consideration of the appropriate-
some rational models also acknowledge that in- ness of the practice for the particular
formation cascades may lead to herding behav- circumstances at hand.
ior, which occurs “when it is optimal for an in- In the strong form, legitimacy arguments hold
dividual, having observed the actions of those that the diffusing practice is at no time techni-
ahead of him, to follow the behavior of the cally efficient—it never employs the optimum
preceding individual without regard to his means for achieving its stated goal, or it may
own information” (Bikhchandani et al., 1992: even be completely ineffective. However, be-
994). Such information cascades may form par- cause of other factors, such as cultural compat-
ticularly fast when early adopters are high- ibility (Soule, 1999) or the normative expecta-
status individuals or are perceived to have tions of outside stakeholders (Abrahamson,
special expertise, leading other firms to imi- 1991; DiMaggio & Powell, 1983), organizations
tate them, even if their private information may still find it advisable to adopt the practice
indicates that adoption is not beneficial (Ban- to increase or maintain their standing in the
erjee, 1992). eyes of their constituency. As a result, these
organizations tend to imitate the models pro-
moted by fashion setters or those used by their
Social Accounts
peers, particularly highly visible and successful
Whereas rational accounts tend to focus on a models.2
growing level of general information about the While providing different rationales for
value of a diffusing innovation, social accounts adoption of practices, both rational and social
tend to emphasize growing levels of pressure accounts of diffusion typically assume a popu-
toward social conformity. Specifically, social ac- lation-level perspective, emphasizing interorgani-
counts tend to assume that organizations fre- zational conditions— either a growing level of
quently imitate other organizations in order to general information about a practice from ra-
appear legitimate and that with increasing in- tional early adopters that can be used to infer its
stitutionalization the adoption of practices is value under uncertainty (e.g., economists’ infor-
therefore often driven by a desire to appear in mation cascades) or growing pressures for
conformance with norms (DiMaggio & Powell, social conformity once enough actors adopt a
1983; Scott, 1995; Sturdy, 2004; Tolbert & Zucker, certain practice (e.g., sociologists’ institutional-
1996). By critiquing the view of adoption as the ization).3 However, a key difference concerns the
result of rational choices, these accounts point durability of behaviors. In rational accounts in-
to the role of group pressures and emphasize the formation cascades can be fragile, and new in-
notion that diffusing practices will frequently be
inefficient or even harmful (Abrahamson, 1991;
2
Strang & Macy, 2001). The assumption of adoption being observable is more
plausible in established, structured industries than it is in
However, efficiency does often enter into so-
emerging industries, where firms may not be as aware of
cial accounts in its functional role of increasing other players because of a lack of shared industry models
the legitimacy of an organization. There are es- and intermediaries (Terlaak & Gong, 2008).
sentially weak and strong forms of this argu- 3
We are grateful to former associate editor Pamela Tol-
ment. In the weak form, legitimacy arguments bert for providing this insight.2010 Ansari, Fiss, and Zajac 71
formation may lead to sudden reversals (Bikh- Zajac, 2006; Green, 2004; Hirsch 1986), or it may
chandani et al., 1992; Lieberman & Asaba, 2006), involve change in the actual implementation of
as was seen in the rise and fall of dotcoms the practice, as when different versions of the
during the Internet bubble. In contrast, social same practice are adopted at different points in
accounts suggest that once a behavior is insti- the diffusion process (Kennedy & Fiss, 2009;
tutionalized, the social order that emerges is Lewis & Seibold, 1993; Westphal et al., 1997). In a
considerably more durable. Although adopters related vein, researchers in the tradition of
have more latitude in adapting the practice dur- Scandinavian institutionalism (e.g., Czarniaw-
ing the preinstitutional stage, increasing insti- ska & Joerges, 1996; Sahlin-Andersson, 1996), ac-
tutionalization and conformity pressures limit tor-network theory (e.g., Callon, 1986; Latour,
that latitude and lead over time to considerably 1986), and social technology transfer (Boyer,
less practice variation (Tolbert & Zucker, 1996). Charron, & Jürgens, 1998; Djelic, 1998; Guillén,
Rational and social accounts of practice diffu- 1994; Zeitlin & Herrigel, 2000) have used the
sion have unquestionably contributed to our un- terms translation (Serres, 1982), editing (Sahlin-
derstanding of the diffusion process, and prior Andersson, 1996), transposition (Boxenbaum &
studies tend to cluster on either side of these two Battilana, 2005), and creolization (Sahlin-
alternative perspectives; rational accounts typ- Andersson & Engwall, 2002) to refer to situations
ically emphasize a technical imperative for where new ideas and practices are adapted to
adoption, and social accounts emphasize a cul- local contexts as they travel during the diffusion
tural imperative for adoption. In line with Hin- process.
ings and Tolbert (2008), we do not see these two As an essential aspect of the implementation
accounts as dichotomous but, rather, as poles on process, adaptation is more likely to be the rule
a continuum, where both can explain diffusion than the exception (Whitten & Collins, 1997), and
under different conditions. it presents a particularly intriguing issue for the
study of diffusion processes in organizational
settings. However, given the lack of a concep-
CONSIDERING PRACTICE ADAPTATION
tual framework for understanding patterns of
As we have noted, the classic diffusion model practice variation across the diffusion process,
is based on invariant practices, where passive adaptation of practices remains a neglected
“accepters” either accept or reject the practice phenomenon. As noted earlier, our position is
(Rogers, 1995: 364). However, a closer consider- that synthesizing insights from studies on pop-
ation of the issue of implementation of the prac- ulation-level diffusion with an original discus-
tice in an organization suggests that few prac- sion of the organization-level implementation of
tices, if any, come out of the diffusion process diffusing practices can lead to greater under-
unchanged (March, 1981; Strang & Soule, 1998). standing of adaptation processes.
We use the term adaptation4 to refer to the pro- To integrate prior models of practice adoption
cess by which an adopter strives to create a with work on practice adaptation and to provide
better fit between an external practice and the a framework for understanding practice adapta-
adopter’s particular needs to increase its ”zone tion across the diffusion process, it is necessary
of acceptance” during implementation (Lewis & to generate the relevant dimensions of practice
Seibold, 1993; Radnor, Feller, & Rogers, 1978). adaptation. To accomplish this we draw on the
This adaptation process may involve change in literature on how knowledge is transmitted and
how a practice is “framed” over time (Fiss & retrieved, which is consistent with the most ba-
sic notion of diffusion as the transmission of
4
ideas (Katz, 1999). Following Yuan et al. (2005),
The diffusion of innovation literature has at times used
the term re-invention to refer to the same process of adapt- Palazzolo, Serb, She, Su, and Contractor (2006),
ing a diffusing innovation (Larsen & Argawalla-Rogers, 1977; and Yuan, Fulk, and Monge (2007), we argue that
Rice & Rogers, 1980; Rogers, 1995). Other terms include reori- adaptation will involve two key dimensions: fi-
entation and variation (Normann, 1971), corruption (Lozeau, delity and extensiveness.
Langley, & Denis, 2002), levels of transfer (Lillrank, 1995),
The first dimension, which we label fidelity,
alteration and optimization (Damanpour & Evan, 1984), re-
configuration (Henderson & Clark, 1990; Meyer & Goes, 1988), relates to whether the adapted practice resem-
emulation with innovation (Westney, 1987), modification bles or deviates in kind from the features of the
(Mamman, 2002), and hybridization (Botti, 1997). previous version of the practice as it is transmit-72 Academy of Management Review January
ted. While Yuan et al. (2007) have used the term tice being implemented—low or high—is closer
accuracy in this regard, we prefer fidelity be- to the notion of scale of implementation. For
cause it does not make assumptions regarding instance, e-business implementation may vary
the normative nature of the prototypical prac- from being selectively implemented to being im-
tice. Fidelity is related to the scope and meaning plemented across the entire span of the organi-
of the practice that is being implemented and zation’s structure and business processes—from
adapted in terms of how ”true” or ”distant” this the procurement department to the field sales
version of the practice is compared to the previ- force to supply chain coordination (Wu et al.,
ous adapted versions of the practice. Thus, if 2003).
late adopters adapt a practice (more or less), it is In short, practices are high fidelity but not
relative to how much the earlier adopters extensive when they are truer to the previous
adapted the practice (more or less) rather than version— but not comprehensively implemented.
relative to some original prototypical version. Practices are extensive but low fidelity if compre-
The notion of a prototype is nonetheless use- hensively implemented— but not true to the previ-
ful to map the terrain of the possible variations ous version.
in an evolving practice over time. Prototypical Illustrating our arguments, Figure 1 shows the
practices may therefore be used to benchmark two dimensions of practice variability, as well
the fidelity of adaptation processes relative to as the patterns of adaptation associated with
the original prototype, as well as relative to them. The top right corner is characterized by
subsequent versions (Lewis & Seibold, 1993).5 high levels of both fidelity and extensiveness.
For instance, in the case of TQM, high-fidelity We call this pattern “full and true” adaptation to
adaptation would include ensuring that a prod- indicate that the adapted practice is being im-
uct or service had higher quality and lower ex- plemented with greater fidelity to its prior ver-
penses associated with it, whereas low-fidelity sion and also in a far-reaching manner. In the
adaptation would emphasize one (but not the top left corner we still find high levels of fidelity
other) improvement. to the prior practice, but the extensiveness of
The second dimension, extensiveness, as- implementation is lower. We call this pattern
sesses whether the degree of practice imple- “low-dosage” adaptation to indicate that the ad-
mentation is greater or lesser than that of the aptation is more aligned with the prior version
previous version of the practice.6 This notion of the practice but presents a more timid effort at
builds on recent research that suggests adopt- implementation in terms of its scope. In the bot-
ing organizations will frequently implement ei- tom right corner we find higher levels of exten-
ther less extensive versions (e.g., Westphal & siveness but lower levels of fidelity. We call this
Zajac, 2001) or more extensive versions (e.g., pattern “tailored” adaptation, indicating that
Hays, 1996; Mooney & Lee, 1999) of a diffusing the adopting organization is using considerable
practice. Extensiveness in adaptation thus indi- resources to implement an extensive version of
cates how far the adapted practice presents far-
reaching or restricted efforts toward implemen- FIGURE 1
tation (Mamman, 2007). Therefore, the concept of Dimensions of Practice Variability and
extensiveness about the ”dosage” of the prac- Adaptation
5
At the high-fidelity end of the continuum, an extreme
case might be the adoption of policies verbatim, including
typographical errors, as documented by Walker (1969). At the
other end of the continuum, adaptations may eventually
reach a point where they bear little resemblance to the
original practice. Here the notion of prototypical features
can again help in introducing threshold values to allow the
researcher to determine whether it may be more useful to
speak of a different rather than an adapted practice.
6
A term related to extensiveness is adoption intensity
(Wu, Mahajan, & Balasubramanian, 2003), also referring to
the degree of implementation of a practice inside an orga-
nization.2010 Ansari, Fiss, and Zajac 73
the practice while also modifying the practice tion’s characteristics.7 However, prior research
substantially, thus implementing a version that has mainly focused on demand-side factors,
is significantly different from its prior version. such as the characteristics of adopters, and has
Finally, the bottom left corner is characterized placed far less emphasis on supply-side factors,
by low levels of both fidelity and extensiveness. such as the characteristics of the diffusing prac-
We call this pattern “distant” adaptation to in- tices. We believe that such an approach can
dicate that adaptation here deviates from the only partially succeed; both demand-side and
prior version, as well as being smaller in scope. supply-side characteristics need to be consid-
To further illustrate these dimensions of prac- ered (Attewell, 1992; Brown, 1981; Cool et al.,
tice adaptation, consider the adaptation of TQM, 1997). Furthermore, rather than considering ei-
which has been the focus of several recent stud- ther demand-side or supply-side factors in iso-
ies (e.g., David & Strang, 2006; Green, Li, & lation, we seek to extend the existing literature
Nohria, 2009; Kennedy & Fiss, 2009; Westphal et by theorizing—at the intersection of supply and
al., 1997; Zbaracki, 1998). The characteristics of demand—about how the characteristics of dif-
TQM include three principles: (1) customer fo- fusing practices interact with the characteristics
cus, (2) continuous improvement, and (3) team- of adopters. This allows us to develop further the
work (Dean & Bowen, 1994). Prototypical imple- concept of compatibility or “goodness of fit” be-
mentation of TQM thus includes practices based tween the attributes of a diffusing item and the
on these principles, such as (1) customer surveys attributes of adopters (Katz, 1999).
and focus groups; (2) flow charts, Pareto analy- Fortunately, the strategy literature and orga-
ses, and statistical process control; and (3) team- nization literature have extensively considered
building methods, such as role clarification and the concept of organizational fit (e.g., Drazin &
group feedback. Variation in the way TQM is Van de Ven, 1985; Venkatraman, 1989). We follow
adapted in an organization can then be as- Nadler and Tushman’s definition of fit as “the
sessed based on the fidelity and extent of TQM degree to which the needs, demands, goals, ob-
implementation relative to this prototypical jectives, and/or structures of one component are
form and prior versions. For instance, low- consistent with the needs, demands, goals, ob-
fidelity adaptation might involve deviation from jectives, and/or structures of another compo-
the core principles by appropriating the practice nent” (1980: 45). While traditional discussions of
for political interests (e.g., Zbaracki, 1998), fit have tended to emphasize the static matching
whereas low-extensiveness adaptation might of organizations to a particular context variable,
involve implementation of TQM practices but in more recent advances have accentuated how fit
less extensive forms (e.g. Westphal et al., 1997). can also be conceptualized dynamically and
In sum, the two dimensions of practice fidelity multidimensionally (Zajac et al., 2000). It is this
and extensiveness provide the foundation for a notion of dynamic, multidimensional fit that we
framework that allows us to connect the rich use here.
work on practice adaptation to overall models of So far we have argued that both demand-side
diffusion patterns. To the extent that adaptation and supply-side attributes matter in practice ad-
signifies differences, fidelity and extensiveness aptation. But how do our arguments speak to the
capture change in kind versus change in degree, dynamic nature of the diffusion process—
respectively. Before moving to make predictions particularly to differences between early and
regarding when we will observe different pat-
terns of adaptation, we first need to consider the
7
reasons for adapting diffusing practices. It is to Fit, of course, is a continuous rather than binary variable
and, thus, will always be measured in degrees. Furthermore,
these issues of fit between the practice and the
we are not assuming that adaptation of a practice will al-
adopting organization that we now turn. ways be triggered by a lack of fit or misfit between a practice
and an organization. Adaptation may also occur within an
organization even when it is not efficient to adapt, either
ADAPTATION AS A RESPONSE TO because of social pressures as organizations observe others’
A LACK OF FIT behaviors or because of coercive pressure imposed by pow-
erful entities. Furthermore, fit may exist even under inac-
A key reason why organizations adapt diffus- tion—a special case that is not our focus here (Zajac, Kraatz,
ing practices is that the characteristics of the & Bresser, 2000). Thus, adaptation (or the lack thereof) may
practice do not fit with the adopter organiza- be characterized by both Type I and Type II errors.74 Academy of Management Review January
later adopters? Who would be more likely to there are also integrated models that draw on
experience misfit during the diffusion process? both rational and social accounts in explaining
Rational accounts point to the importance of diffusion. Probably the most important model
uncertainty and associated mechanisms of that integrates both rational and social argu-
learning (e.g., Banerjee, 1992; Bikhchandani et ments is that of Tolbert and Zucker (1983), who
al., 1998). Here early adopters face greater un- argued that early adopters are mainly con-
certainty regarding the utility and characteris- cerned with the utility of the practice whereas
tics of the practice. Accordingly, misfit is more later adopters are primarily concerned with con-
likely to occur among early adopters, when less formity pressures in a rational to ceremonial
is known about the diffusing practice. Rational shift. However, David and Strang (2006), in their
arguments suggest that early adopters are more study of TQM, suggested a more complex insti-
likely to find misfit during implementation and tutional trajectory, with the practice swinging
to adapt practices, thus leading to more adap- back toward its technical foundations in the mid
tation early on when less is known about the and late 1990s and with its larger institutional
practice. However, over time the uncertainty re- trajectory moving from rational to ceremonial to
garding the utility and characteristics of the rational again. More recently, Kennedy and Fiss
practice decreases as late adopters are able to (2009) offered an integrated diffusion model that
infer the value of the practice from the accumu- suggests that both economic and social motiva-
lated stock of early adopters’ prior decisions tions may, in fact, work in parallel and may both
(Terlaak & Gong, 2008). Accordingly, later adopt- be present among early and later adopters. Note
ers may be more able to avoid practices with a that both integrated models, by combining ra-
low fit for their needs. This also implies that, in tional and social arguments, suggest that misfit
rational accounts, the technical characteristics will occur among both early and later adopters,
of individual adopters (such as size) may be- albeit for different reasons. It would therefore
come better predictors of adoption later in the appear that at least the potential for misfit is
process, whereas social characteristics (such as given across the whole life cycle of the diffusion
susceptibility to conformity pressures) may be- process.
come less useful predictors.
In contrast, social accounts emphasize confor-
FORMS OF FIT AND PATTERNS OF PRACTICE
mity pressures and a desire to look legitimate to
ADAPTATION
outside constituents (e.g., Abrahamson, 1991;
DiMaggio & Powell, 1983; Tolbert & Zucker, 1983). Adaptation in response to a lack of fit presents
In these accounts early adopters have little in- a key issue in the diffusion literature, yet the
centive to adopt practices that do not fit since specific relationship between fit and adaptation
they do not perceive conformity pressures. How- remains largely unexamined. Here we seek to
ever, as conformity pressures mount, later extend the existing literature by theorizing
adopters will be forced to adopt regardless of about how diffusing practices interact with the
their needs, leading to a greater likelihood of characteristics of adopters, resulting in different
misfit among later adopters. In opposition to adaptation patterns. To conceptualize fit we
rational accounts, social accounts thus also sug- draw on Oliver’s (1992) categorization of factors
gest that the technical characteristics of actors influencing organizational practices, and we
become less reliable predictors of adoption later identify three forms of fit that affect adaptation
in the process once an innovation gains legiti- processes: (1) technical fit, (2) cultural fit, and (3)
macy, whereas social characteristics become political fit (see also Sturdy, 2004, for different
better predictors. perspectives on the adoption of management
Both rational and social accounts of diffusion practices). Specifically, we argue that technical,
therefore suggest variation between early and cultural, and political incompatibilities trigger
late adopters due to the interaction between different mechanisms and patterns of adapta-
population-level and organization-level phe- tion on the part of adopting organizations. In-
nomena. However, because of the respective deed, innovative practices tend to vary in the
mechanisms invoked, both accounts have differ- forms of misfit that they typically engender,
ent implications for when misfits of adopted therefore affecting diffusion through corre-
practices are more likely to occur. Furthermore, spondingly different mechanisms for commu-2010 Ansari, Fiss, and Zajac 75
nicating interorganizational variation or con- (Cohen & Levinthal, 1990: 128). For instance, re-
formity. In conceptualizing the three forms of fit, garding the diffusion of the Six Sigma method-
we adopt a boundedly rational perspective ology in the 1990s, one reason GE was far more
(Eisenhardt & Zbaracki, 1992), where adopters successful than Motorola in implementing this
are “cognitive misers” rather than “cognitive practice was that it had already put in place a
dopes” (Rao et al., 2001), where organizations series of its own technologies in quality man-
rely on “cognitive shortcuts” (Tversky & Kahne- agement and human performance management
man, 1974) in making adaptation decisions, and quite compatible with Six Sigma. GE was thus
where adaptation is both a rational and social better positioned than Motorola to generate
process.8 Accordingly, for each of the three forms value from the stringent processes required by
of fit, a poor fit as experienced by an adopter the Six Sigma methodology (Mol & Birkinshaw,
during ongoing assessments and feedbacks (ei- 2008), even though Motorola, not GE, had in-
ther in the face of various crises— e.g., employee vented Six Sigma. Similarly, as the imprinting
resistance and dissatisfaction levels— or with literature has shown, the distinctive stamp of an
evidence of mounting performance problems) organization’s founding environment continues
will make implementation of an unmodified to impact subsequent managerial decisions and
practice more costly and, thus, likely to result in the adoption of organizational practices that are
either adaptation of the diffusing practice or its seen as congruent with inherited organizational
eventual abandonment. Table 1 provides an arrangements (Marquis, 2003; Stinchcombe,
overview of supply-side practice characteristics 1965).
and demand-side characteristics of adopters Demand-side factors may further include in-
and their contexts. traorganization-level factors, such as technolog-
ical background and experience of organization-
al members and executive demographics (e.g.,
Technical Fit
Palmer, Jennings, & Zhou, 1993; Wejnert, 2002;
By technical fit we mean the degree to which Zeitz et al., 1999). Finally, fit may also be affected
the characteristics of a practice are compatible by supraorganization-level factors, including in-
with technologies already in use by potential dustry-level phenomena, such as technological
adopters. On the supply side, practice-level fac- standards and regulatory regimes (Farrell & Sa-
tors relate to the diffusing practice’s technolog- loner, 1985) and maturity levels (e.g., Behrman &
ical foundation and characteristics (Rogers, Wallender, 1976), as well as societal-level phe-
1995; Tornatzky & Klein, 1982). On the demand- nomena, such as the degree of technological
side, organization-level factors affecting techni- advancement, infrastructure, and educational,
cal fit include the recipient organization’s inno- financial, and regulatory institutions (Wejnert,
vativeness (Damanpour, 1991), technological 2002).
base (Adler & Shenhar, 1990), and absorptive
capacity—that is, “the ability of a firm to recog-
Adaptation and Technical Fit
nize the value of new, external information, as-
similate it, and apply it to commercial ends” As we have argued, technical fit depends on
the alignment between practice characteristics
and organizational characteristics. In the case
8
In a preadoption setting firms make their choice under of a low degree of fit, organizations will aim to
uncertainty, based on some expected practice value that is reduce the costs of implementation, regardless
void of any implementation experience. However, postadop- of the original reasons for adoption. Reducing
tion, firms may have higher inferential accuracy based on
technical misfit thus involves the assimilation
information on the specificities of practice value that they
gather through implementation of the practice and/or of practices into existing organizational sys-
through vicarious learning from the postadoption behaviors tems.
of other referent firms (Rao et al., 2001; Terlaak & Gong, 2008). We see the availability of knowledge about a
They may then adapt or even abandon a practice (we are diffusing practice as a key mechanism influenc-
grateful to an anonymous AMR reviewer for this insight).
ing adaptation efforts. Assuming that decision
However, uncertainty may persist even postadoption, and
adaptation—warranted or not—may simply be an imitative makers in organizational settings are generally
response to external social pressures as opposed to a delib- risk averse and that reliable information on the
erate or purposeful activity. effectiveness of a new practice is scarce, the76 Academy of Management Review January
TABLE 1
Supply- and Demand-Side Characteristics by Level of Analysis
Level of Analysis
Supply Side Demand Side
Fit
Characteristics Practice Intraorganizational Organizational Supraorganizational
Technical Technological foundation Individual’s background Organizational absorptive Technological standards
characteristics and characteristics and experience, capacity, technological and regulatory
embodied by the education level, base, innovativeness, regimes, professional
practice technical orientation level of sophistication of bodies, degree of
technologies and systems technological
already in use advancement,
infrastructure,
financial and
regulatory institutions,
educational systems
Cultural Cultural characteristics Beliefs, values, and Organizational Norms, beliefs, and
characteristics such as cultural values preferences about the culture—innovative or values of industry
and meaning appropriateness of closed, values and beliefs associations and
structures embodied by the work practice regional clusters;
the practice macrocultural
discourse; cultural
icons; dominant
institutional logic
Political Normative claims, Interests, relative Formal and informal power Political settlements;
characteristics political “loadings,” power, and agendas structures and rules union agreements;
controversial of organizational within an organization, government
associations embodied members resource dependencies, regulations; types and
by the practice differential positions in character of political
social networks, dominant systems; labor market
coalitions policies; legal
systems; degree of
political freedom;
national policies such
as distributing
concessions and
repressions to various
political, corporate,
and social groups
implementation of a new practice or technology adapted and customized during the diffusion
will frequently proceed cautiously and incre- process (David & Strang, 2006).
mentally (Mooney & Lee, 1999; Rice & Rogers, For instance, Sine, Haveman, and Tolbert
1980). Particularly, if the uncertainty surround- (2005), in their analysis of the independent
ing the practice is high and adopters are unable power sector in the United States, showed that
to reduce misfit and increase the practice’s zone organizations over time adopted newer and risk-
of acceptance, early adopters are likely to avoid ier technologies rather than more established
experimentation and adopt truer or high-fidelity technologies with the development of various
versions of the new practice. However, adapta- types of institutions. Similarly, early adopters
tion efforts can intensify as the practice estab- used just-in-time (JIT) production systems in a
lished by early adopters becomes more elabo- stricter sense—as a way to control the pace of
rate in its specification (Glick & Hays, 1991; production (going from a ”push” to a ”pull” sys-
Hays, 1996), with more details and more ver- tem, thus allowing a drastic reduction in inven-
sions, leading to greater variety and lower fidel- tory). In contrast, later adopters increasingly
ity. This view suggests a process that is the modified JIT and increased its scope, so it even-
reverse of conventional accounts of institution- tually developed into ”big JIT”— “a broad, holis-
alization, with practices becoming increasingly tic, strategic approach for eliminating waste2010 Ansari, Fiss, and Zajac 77
and improving customer service” (Klassen, will implement more extensive ver-
2000: 97). sions of the practice.
In sum, as uncertainty decreases and knowl-
In combination, a lack of technical fit suggests
edge about the practice increases, late adopters a pattern of low-dosage adaptation among early
are exposed to a wider range of possibilities adopters and a pattern of tailored adaptation
regarding the implementation of a practice among later adopters. Figure 2a shows how this
and—with more inferential accuracy—are less
restricted in their adaptation efforts. Accord-
ingly, later adopters are more likely to develop FIGURE 2
increasingly divergent lower-fidelity versions of Patterns of Practice Adaptation
the practice in order to reduce misfit. We restate
the above arguments in the following proposi-
tion.
Proposition 1: When adopters experi-
ence low technical fit between the
practice and the organization, early
adopters will implement higher-fidel-
ity versions whereas later adopters
will implement lower-fidelity versions
of the practice.
At the same time, mechanisms of limited
knowledge and uncertainty reduction also influ-
ence the effect of technical misfit on practice
extensiveness. Since new practices are fre-
quently not well understood at the beginning of
the diffusion process, early adopters are re-
stricted in their ability to reduce misfit and,
thus, are likely to implement not only truer but
also less extensive versions of a given practice.
Later in the diffusion process, when uncertainty
diminishes and there is greater knowledge
about the effectiveness of the practice, we ex-
pect implementation of more extensive versions.
The greater availability of information about the
practice over time therfore allows later adopters
to implement more extensive versions of the
new practice (Glick & Hays, 1991; Hays, 1996;
Mooney & Lee, 1999). The diffusion mechanism
suggested by this process, hence, is one of un-
certainty reduction, where early adopters have
limited knowledge about the innovation and en-
gage in rather limited implementation of a rel-
atively small set of basic features, whereas later
adopters with more information are able to im-
plement increasingly extensive versions.
Proposition 2: When adopters experi-
ence low technical fit between the
practice and the organization, early
adopters will implement less exten-
sive versions whereas later adopters78 Academy of Management Review January
shift from early to later adopters maps onto the individualistic or egalitarian (Bhagat, Kedia,
dimensions of adaptation described earlier. Harveston, & Triandis, 2002) and macrocultural
discourse reflected by best practice awards and
certifications and by celebrity speakers and
Cultural Fit
CEOs— cultural icons like Jack Welch or Steve
By cultural fit we mean the degree to which Jobs (Zeitz et al., 1999).
the characteristics of a diffusing practice are For instance, Hunter (2000) found significant
compatible with the cultural values, beliefs, and differences in the organizational implementa-
practices of potential adopters. New practices tion of such innovative work practices as the
and ideas do not diffuse into a cultural void but, “self-managed team,” based both on the indus-
rather, into a preexisting cultural universe that try sector they operated in and whether they
delineates the roles and responsibilities of its were manufacturing or service establishments.
respective actors and the boundaries of appro- Similarly, since quality circles (QCs) were seen
priate behavior. The concept of cultural fit has as an imported Japanese practice at odds with
informed anthropologically oriented research the U.S. cultural value of individualism and
since the early twentieth century (e.g., Boas, American corporate culture—for instance, the
1925; Lowie, 1914; Spier, 1921), and it was also small role of foremen and cohesive workgroups,
part of the early program regarding the diffu- adversarial management labor relations, and
sion of innovations (Katz, Hamilton, & Levin, the hegemonic position of top management—
1963; Rogers, 1995). However, cultural fit gener- there was little coherent attempt within organi-
ally has played a more peripheral role in stud- zations to promote QCs. Instead, American man-
ies of the diffusion of organizational practices agers and consultants developed a simplified,
(Kedia & Bhagat, 1988: 559). Indeed, the relative context-independent notion of the QC that fit
inattention to the idea of cultural fit led Lopes American notions of “participatory manage-
(1999) to call for more attention to the symbolic ment” (Strang & Kim, 2004; Zeitz et al., 1999).
behavior and repertoire of adopters.9
On the supply side, practice-level factors refer
Adaptation and Cultural Fit
to the cultural characteristics of the diffusing
practice or “cultural object”—specifically, to the In contrast to technical misfit, cultural misfit
meaning structures and cultural values the suggests a somewhat different pattern of adap-
practice embodies (Griswold, 1987). On the de- tation. One strain of research on cultural trans-
mand side, organization-level factors include fer has emphasized the selection process, em-
the values and beliefs prevalent in adopter or- ploying a “rational shopper” metaphor (Whyte,
ganizations. Examples here include corporate 1968). However, as Westney (1987) pointed out,
cultures—values, beliefs, communication styles, this metaphor does not do justice to the amount
mission, and philosophical orientation of the or- of adaptation that frequently takes place. We
ganizations—that impact receptivity to a prac- suggest that adaptation in response to cultural
tice, such as the acceptability of same-sex do- misfit is better addressed by the idea of the
mestic partner benefits (Briscoe & Safford, 2008; “cultural entrepreneur” (DiMaggio, 1992) who
Creed, Scully, & Austin, 2002). At the intraorga- uses culture as a “toolkit” (Swidler, 1986) and
nization level relevant factors refer to individual adapts cultural objects to make them useful in
traits, such as beliefs, values, and preferences relation to local cultural expectations. Particu-
about the appropriateness of the work practice. larly relevant here are adaptations that over-
Supraorganization-level factors include indus- come a cultural misfit by naming and position-
try-level phenomena, such as the industry’s ing the innovation (Hirsch, 1986; Rogers, 1995).
dominant logic (Prahalad & Bettis, 1986) and the While the availability of knowledge is the key
norms, beliefs, and values of industry associa- factor influencing responses to a lack of techni-
tions and regional clusters, as well as society- cal fit, we expect conformity pressures to be the
level phenomena, such as whether cultures are primary mechanism affecting responses to a
lack of cultural fit.
During initial stages of the diffusion process,
9
Hirsch’s (1986) and Hays’ (1996) articles are exceptions in conformity pressures are essentially absent
which the authors paid more attention to cultural fit. since models for conformity have not yet2010 Ansari, Fiss, and Zajac 79
emerged. Accordingly, early adopters have con- formity pressures that come to define the cul-
siderably more latitude to experiment with a tural meaning and acceptable form of a prac-
practice, define it, label it, and adapt it to their tice, misfit will nevertheless affect the extent of
local needs, implementing it in a manner that their practice implementation. Mechanisms of
works for them. However, later in the diffusion conformity pressures suggest a different pattern
process, when established models emerge and regarding the effect of cultural misfit on practice
conformity pressures increase, the ability of the extensiveness. Since early adopters are able to
late adopters to adapt and reduce misfit will be reduce misfit, they are more likely to implement
restricted. This also implies that cultural misfit the practice more extensively. With conformity
is more likely to persist during the latter stages pressures less powerful during initial stages,
of the diffusion process, since later adopters feel there is a reduced need for and benefit from less
restrained in their ability to adapt the practice extensive implementation or decoupling.
and thus implement truer or high-fidelity ver- In contrast, late adopters will be more re-
sions of the practice. Accordingly, once the new stricted in their ability to modify practices and
innovation has been sufficiently modified by the may therefore implement less extensive ver-
cultural entrepreneur to make it acceptable and sions of the practice to reduce the cost of misfit.
it develops a halo of social validation, subse- For instance, Kennedy and Fiss (2009) found that
quent diffusion will largely result in the spread later adopters who aimed to avoid social losses
of a culturally legitimated and institutionalized implemented less extensive versions of TQM.
model (Tolbert & Zucker 1996), leading to con- Similar arguments are suggested by a logic of
straints on the ability of late adopters to modify decoupling,10 where implementation is sym-
the innovation (e.g., Tolbert & Zucker 1983). Be- bolic rather than substantive since organiza-
cause of legitimacy concerns, late adopters tions may engage in ceremonial implementa-
have fewer ”degrees of freedom” to adapt a tion and not integrate the practice within the
practice, even if they experience misfit, and with organization, aiming to show compliance to-
increasing institutionalization one is likely to ward external observers while hiding noncon-
see less variation in practices (Tolbert & Zucker, formity (e.g., Boxenbaum & Jonsson, 2008; Els-
1996). bach & Sutton, 1992; Meyer & Rowan, 1977;
For instance, in their study of the diffusion of Westphal & Zajac, 1994). Later adopters who feel
TQM diffusion, Westphal et al. (1997) argued compelled to adopt a practice because of con-
that early adopters adapted and customized formity pressures will thus be more likely to
TQM practices while later adopters conformed implement less extensive versions of the prac-
to the normative patterns of practices intro- tice. This suggests the following proposition.
duced by these early adopters. Thus, we propose
Proposition 4: When adopters experi-
that the pattern of adaptation in response to
ence low cultural fit between the char-
cultural misfit will lead to considerable adapta-
acteristics of the practice and the or-
tion by early adopters who manipulate the prac-
ganization, early adopters will
tice’s meaning and develop culturally legiti-
implement more extensive versions
mate models, followed by the diffusion of these
whereas later adopters will imple-
models in largely unchanged form by later
ment less extensive versions of the
adopters. Restated as a proposition, we offer the
practice.
following.
These arguments about cultural fit suggest a
Proposition 3: When adopters experi-
pattern that is the inverse of that for technical
ence low cultural fit between the char-
acteristics of the practice and the
organization, early adopters will im- 10
plement lower-fidelity versions Decoupling here refers more to “surface-level” or cere-
monial (less extensive) implementation than to actively “re-
whereas later adopters will imple- working” or modifying the practice to fit with the organiza-
ment higher-fidelity versions of the tion. However, it is possible that the adopted practice being
practice. decoupled during implementation may be qualitatively dif-
ferent from the previous version of the practice (low fidelity)
However, while later adopters will engage in and that decoupling might also entail a change in meaning
less redefinition of the practice because of con- of the practice.You can also read
