ArticlePDF Available

Abstract and Figures

Exploration and exploitation have emerged as the twin concepts underpinning organizational adaptation research, yet some central issues related to them remain ambiguous. We address four related questions here: What do exploration and exploitation mean? Are they two ends of a continuum or orthogonal to each other? How should organizations achieve balance between exploration and exploitation-via ambidexterity or punctuated equilibrium? Finally, must all organizations strive for a balance, or is specialization in exploitation or exploration sometimes sufficient for long-run success? We summarize the contributions of the work in this special research forum and highlight important directions for future research.
Content may be subject to copyright.
THE INTERPLAY BETWEEN EXPLORATION AND
EXPLOITATION
ANIL K. GUPTA
KEN G. SMITH
University of Maryland
CHRISTINA E. SHALLEY
Georgia Institute of Technology
Exploration and exploitation have emerged as the twin concepts underpinning organ-
izational adaptation research, yet some central issues related to them remain ambig-
uous. We address four related questions here: What do exploration and exploitation
mean? Are they two ends of a continuum or orthogonal to each other? How should
organizations achieve balance between exploration and exploitation—via ambidexter-
ity or punctuated equilibrium? Finally, must all organizations strive for a balance, or
is specialization in exploitation or exploration sometimes sufficient for long-run suc-
cess? We summarize the contributions of the work in this special research forum and
highlight important directions for future research.
Since the publication of March’s (1991) pioneer-
ing article, the terms “exploration” and “exploita-
tion” have increasingly come to dominate organi-
zational analyses of technological innovation,
organization design, organizational adaptation, or-
ganizational learning, competitive advantage and,
indeed, organizational survival (e.g., Benner &
Tushman, 2003; Burgelman, 2002; Holmqvist,
2004; Katila & Ahuja, 2002; Lee, Lee, & Lee, 2003;
McGrath, 2001; Sigglekow & Levinthal, 2003). Not-
withstanding the growing reliance of organization-
al research on these twin concepts, an examination
of the literature indicates that the answers con-
tained there to the central questions on this subject
remain incomplete, at times contradictory, and at
best ambiguous. We use the term “central ques-
tions” to refer to the following four issues.
The first issue pertains to definitions and conno-
tations. What do exploration and exploitation re-
ally mean? There appears to be consensus around
the view that exploration refers to learning and
innovation (i.e., the pursuit and acquisition of new
knowledge). However, a similar consensus is lack-
ing on the question of whether exploitation refers
solely to the use of past knowledge or whether it
also refers to the pursuit and acquisition of new
knowledge, albeit of a kind different from that as-
sociated with exploration. For scholars to build a
truly informative and significant body of research
in this area, it is critical that they be clear on defi-
nitional issues. Also, it is important to ensure that
the way exploration and exploitation are empiri-
cally defined is consistent with and appropriate to
their conceptual definitions.
The second issue pertains to orthogonality versus
continuity. Are exploration and exploitation two
ends of a continuum, or two different and orthog-
onal aspects of organizational behavior? In part, the
answer to this question depends on the conceptual
definition of the terms “exploration” and “exploi-
tation.” However, the salience of this question goes
far beyond mere definitions. Theories about the
ease or difficulty with which an organization can
pursue both exploration and exploitation depend
crucially on whether these two tasks are treated as
competing or complementary aspects of organiza-
tional decisions and actions. In addition to theory
development, empirical tests of predictions regard-
ing the impact of “balance” between exploration
and exploitation on organizational performance
would need to be different depending on whether
these two concepts are viewed as mutually anti-
thetical or complementary.
The third issue pertains to ambidexterity versus
punctuated equilibrium. Building on March’s ini-
tial premise that organizational “adaptation re-
quires both exploitation and exploration to achieve
persistent success” (1991: 205), some studies have
concluded that the answer lies in “ambidexterity”
(Benner & Tushman, 2003), whereas others have
concluded that the answer lies in “punctuated
equilibrium” (Burgelman, 2002). Ambidexterity re-
fers to the synchronous pursuit of both exploration
and exploitation via loosely coupled and differen-
tiated subunits or individuals, each of which spe-
cializes in either exploration or exploitation. In
contrast, punctuated equilibrium refers to temporal
rather than organizational differentiation and sug-
Academy of Management Journal
2006, Vol. 49, No. 4, 693–706.
693
gests that cycling through periods of exploration
and exploitation is a more viable approach than a
simultaneous pursuit of the two. As may be clear,
ambidexterity and punctuated equilibrium are
radically different mechanisms. Existing litera-
ture is silent on the questions of whether these
two mechanisms are equally viable, so that an
organization can pick one or the other at will, and
whether exogenous or endogenous contextual
factors should drive the choice between the two
mechanisms.
Finally, the fourth issue pertains to duality ver-
sus specialization. Echoing March’s (1991) argu-
ments about the need for balance between explora-
tion and exploitation, there is near consensus in the
literature that, notwithstanding their radically dif-
ferent dynamics, organizations must learn to excel
at both tasks. We wonder, however, if such a con-
sensus may be somewhat premature and not nec-
essarily logical in all contexts.
In the balance of this paper, we address each of
these four issues and then provide an overview of
the seven papers that comprise this special re-
search forum on managing exploration and exploi-
tation. Finally, we offer a set of research questions
and research questions to guide future research.
DEFINITIONS AND CONNOTATIONS
As noted, the central ambiguity regarding the
definition and implications of exploration and ex-
ploitation lies in whether the two are distinguished
by differences in the type of learning or by the
presence versus the absence of learning. Baum, Li,
and Usher (2000), Benner and Tushman (2002), and
He and Wong (2004) are illustrative of studies that
explicitly embrace the idea that both exploration
and exploitation are associated with learning and
innovation, albeit of different types. Baum, Li, and
Usher suggested that “exploitation refers to learn-
ing gained via local search, experiential refine-
ment, and selection and reuse of existing routines.
Exploration refers to learning gained through pro-
cesses of concerted variation, planned experimen-
tation, and play” (2000: 768). According to Benner
and Tushman, “Exploitative innovations involve
improvements in existing components and build
on the existing technological trajectory, whereas
exploratory innovation involves a shift to a differ-
ent technological trajectory” (2002: 679). Along the
same lines, He and Wong (2004: 483) defined ex-
ploitative innovation as “technological innovation
activities aimed at improving existing product-mar-
ket domains” and exploratory innovation as “tech-
nological innovation aimed at entering new prod-
uct-market domains.” As is evident, in all of these
studies, learning, improvement, and acquisition of
new knowledge are central to both exploitation and
exploration. At the same time, the differences be-
tween the two concepts pertain to whether the new
learning occurs along the same trajectory as the old
one or along an entirely different trajectory.
In contrast, other studies (e.g., Rosenkopf &
Nerkar, 2001; Vassolo, Anand, & Folta, 2004; Ver-
meulen & Barkema, 2001) appear to treat all activ-
ities associated with learning and innovation as
instances of exploration and to reserve the term
“exploitation” for activities in which the central
goal is using past knowledge rather than moving
down any kind of a learning trajectory. This ap-
proach to conceptual interpretation appears ex-
plicit in Rosenkopf and Nerkar’s (2001) study of the
impact of local and nonlocal knowledge search on
the quality of resulting patents. In the case of pat-
ents that were informed exclusively or largely by
local knowledge, they accepted the possibility that
“some readers might consider this a form of exploi-
tation rather than local exploration” (2001: 289).
However, noting that their study focused entirely
on the R&D process and patenting activity, they
preferred to term such cases “the most localized
form of exploration” rather than “exploitation.”
Similarly, in their study of firms’ international ex-
pansion decisions, Vermeulen and Barkema (2001:
459) defined exploration as the “search for new
knowledge” and exploitation as the “ongoing use of
a firm’s knowledge base.” Building on these defi-
nitions, they treated all acquisitions as representing
exploration and all greenfield investments as rep-
resenting exploitation.
Reflecting on these somewhat different ap-
proaches to defining exploration and exploitation,
our conclusion is to build on March’s (1991) logic
and to argue that all activity includes at least some
learning. Even when an organization is attempting
to do nothing more than replicate past actions, it
accumulates experience and goes down the learn-
ing curve, albeit in an incremental manner (Yelle,
1979). Thus, for social systems, there is no such
thing as perfect replication. There is always some
learning, even if it is relatively little and serves
only to reduce variation around the historical
mean. As March noted, “The essence of exploita-
tion is the refinement and extension of existing
competencies, technologies, and paradigms....The
essence of exploration is experimentation with new
alternatives” (1991: 85). To sum up, we would ar-
gue that it is more logical to differentiate between
exploration and exploitation by focusing on the
type or amount of learning rather than on the pres-
ence or absence of learning. Defining these con-
cepts in terms of the presence or absence runs at
694 AugustAcademy of Management Journal
least two kind of risks. One, many activities that, by
March’s (1991) definition, should be viewed as ex-
ploitative would instead be counted and coded as
exploratory. Two, for activities coded as exploit-
ative, researchers may overlook the reliability-en-
hancing learning that results from all human and
organizational attempts to replicate past routines.
In keeping with March (1991), in this discussion
of what exploration and exploitation mean, we
have treated the organization as the unit of analy-
sis. However, variations in the unit of analysis may
well affect the answer to our questions, What is
exploration and What is exploitation? That is,
whether learning differs by type or by presence
versus absence could very well depend on whether
one is focused on the individual level, the team
level, or a more macro organization level. For ex-
ample, an engineer might search and experiment to
discover a new method of producing a product, but
the organization in which he/she is employed
might then exploit this new innovation for profit.
Similarly, it is entirely possible that repetitious
routines of exploitation may not involve much
learning at the individual level (e.g., a machine
operator producing the same widget each day).
However, such an absence of learning is less likely
at the group or organizational level simply because
of variation in skills, knowledge, and experience
across individuals. In other words, at a group or
more macro level, it is more likely that some learn-
ing from experience will take place because of dif-
ferences among individuals. Consequently, what
one individual or organization may view as explor-
atory and experimental learning, another team or
individual may view as exploitative or incremental
learning. The above discussion highlights the need
for researchers to carefully specify a unit of analy-
sis in defining exploration and exploitation.
CONTINUITY VERSUS ORTHOGONALITY
March (1991) appeared very clear in his theori-
zation that, even though both exploration and ex-
ploitation are essential for long-run adaptation, the
two are fundamentally incompatible. March (1991,
1996, 2006) provided several arguments in favor of
this incompatibility. First, exploration and exploi-
tation compete for scarce organizational resources.
Thus, by definition, more resources devoted to ex-
ploitation imply fewer resources left over for explo-
ration, and vice versa. Second, and assuming all
else equal, both types of actions are iteratively self-
reinforcing. Because of the broad dispersion in the
range of possible outcomes, exploration often leads
to failure, which in turn promotes the search for
even newer ideas and thus more exploration,
thereby creating a “failure trap.” In contrast, exploi-
tation often leads to early success, which in turn
reinforces further exploitation along the same tra-
jectory, thereby creating a “success trap.” In short,
exploration often leads to more exploration, and
exploitation to more exploitation. Third, the mind-
sets and organizational routines needed for explo-
ration are radically different from those needed for
exploitation, making the simultaneous pursuit of
both all but impossible. As March noted, “Exploit-
ing interesting ideas often thrives on commitment
more than thoughtfulness, narrowness more than
breadth, cohesiveness more than openness” (1996:
280). To summarize March’s arguments, notwith-
standing the adaptation benefits of both explora-
tion and exploitation, the interplay between the
two occurs in the form of a zero-sum game where
exploration and exploitation compete for scarce re-
sources, attention, and organizational routines; ac-
cordingly, logic dictates that exploration and ex-
ploitation be viewed as two ends of a continuum.
It is all but impossible to dispute March’s logic.
However, it is possible to question some of his key
assumptions. Consider his arguments about the scar-
city of resources. Although it is generally true that
most organizational resources are finite, this need not
be so for all types of resources. Some resources, such
as information and knowledge, may be infinite (Sha-
piro & Varian, 1998). Also, organizations often have
access not only to the resources that they own but
also to resources in their external environments
(Powell, Koput, & Smith-Doerr, 1996). Access to such
external resources may come about either because
these resources constitute public goods (e.g., articles
published in journals) or because the focal organiza-
tion has established strategic alliances with other
stakeholders who privately own or control comple-
mentary resources. Access to external resources con-
siderably eases the constraint imposed on organiza-
tions by the scarcity of internal resources. Katila and
Ahuja’s (2002) conceptualization of exploration and
exploitation provides a nice example of what hap-
pens to these concepts when the relevant resources
do not suffer from the constraint of scarcity. In keep-
ing with March’s (1991) arguments about the benefi-
cial effects of pursuing both exploration and exploi-
tation, Katila and Ahuja (2002) found empirical
support for their prediction that the interaction be-
tween exploration and exploitation will have a posi-
tive impact on new-product development. At the
same time, departing from March’s notion of explo-
ration and exploitation as competing phenomena, Ka-
tila and Ahuja (2002) conceptualized these as orthog-
onal variables. Exploration was operationalized as
search scope (i.e., the propensity to cite different pat-
ents), whereas exploitation was operationalized as
2006 695Gupta, Smith, and Shalley
search depth (i.e., the propensity to cite certain pat-
ents repeatedly). Since the number of patents an or-
ganization may cite is unlimited, and the marginal
cost of accessing the knowledge embedded in one
more patent is highly likely to be modest, it seems
reasonable to conclude that exploitative versus ex-
plorative search (as reflected in patent citation data)
does not suffer from severe resource constraints.
Nerkar’s (2003) study on the antecedents of why
some patents have greater future impact than others
provides an interesting parallel to Katila and Ahuja
(2002). Nerkar looked at the main as well as the in-
teractive effects of exploration and exploitation in the
knowledge search underlying patents. He measured
exploration as the time spread of past knowledge and
exploitation as the recency of past knowledge, and he
left open the possibility that a specific patent might
exhibit high or low levels of both exploration and
exploitation.
Scarcity of resources as well as conflicts over
mind-sets and organizational routines are also non-
issues when a researcher’s interest lies in analyzing
exploration versus exploitation in two different do-
mains that are either loosely connected or con-
nected via standardized/modular interfaces. For
instance, consider a firm such as Cisco. The tech-
nologies and designs embedded in Cisco’s products
suffer from a high rate of obsolescence, making it
imperative that the company pursue a highly ex-
ploratory strategy with respect to technology and
product development. At the same time, even rad-
ically new and different products can be manufac-
tured, sold, and serviced via a preexisting commer-
cialization infrastructure that evolves relatively
slowly (Rangan, 2005). In other words, the inter-
faces between product R&D on the one hand and
manufacturing, sales, and service on the other are
relatively standardized. Note that, in this case, the
resources needed for product R&D are fundamen-
tally different from those needed for complemen-
tary downstream activities. As such, it is easy to
imagine that Cisco could simultaneously engage in
a high degree of exploration in product R&D and a
high rate of exploitation in complementary do-
mains such as manufacturing, sales, and service.
Several recent studies have examined the idea that
organizations operate in multiple domains, not all of
which are tightly coupled via specialized interfaces.
These studies have chosen to treat exploration and
exploitation as simultaneously achievable and thus,
for all practical purposes, orthogonal. Some illustra-
tive examples are Baum, Li, and Usher (2000), Beck-
man, Haunschild, and Phillips (2004), Koza and
Lewin (1998), and Rothaermel (2001). Baum and his
colleagues (2000) viewed an organization’s learning
from its own experience as exploitation and its learn-
ing from others’ experience as exploration; since both
these types of learning are potentially unlimited, they
treated the two as orthogonal. Beckman and her col-
leagues (2004: 259) analyzed interorganizational rela-
tionships and treated “relationships with new part-
ners” as a form of exploration and “additional
relationships with existing partners” as a form of ex-
ploitation; here too, given that the number of inter-
firm relationships has no well-defined limit, the
authors treated exploration and exploitation as or-
thogonal. Koza and Lewin (1998) and Rothaermel
(2001) focused on interfirm alliances and adopted a
similar logic. According to them, any firm can en-
gage in a multiplicity of alliances, and any one of
these alliances can be classified as exploratory or
exploitative.
Answering the question of continuity or orthog-
onality may further depend on the level of analysis.
For example, with division of labor and allocation
of resources, it may be easier for a group, organiza-
tion, or larger system to simultaneously excel at
exploration and exploitation than it is for individ-
uals to do so. As noted, the learning, resources, and
routines necessary for exploration and exploitation
are different. As such, they may be delegated
within a group or organization so that both can be
achieved simultaneously. In this case, management
controls the allocation of decision rights to exploit
or explore. It also may be easier for a group or
organization to switch between exploration and ex-
ploitation when appropriate change routines are in
place and management recognizes the need for
change. For example, Gilson, Mathieu, Shalley, and
Ruddy (2005) found that teams that felt empowered
by their organization to use creative problem solv-
ing as well as standardized routines and proce-
dures had the highest levels of team effectiveness.
In contrast, one can imagine that it would be diffi-
cult for an individual to develop routines to excel
simultaneously at both exploration and exploita-
tion. Further, given the substantial differences in
routines and focus on learning, it may be very dif-
ficult for an individual to even switch between
routines of exploration and exploitation. Amabile
(1996) suggested that individuals who focus on cre-
ativity, exploration, and experimentation are quite
different from those who emphasize appropriate
actions. For example, those focused on creativity
may be intrinsically motivated, whereas individu-
als focused on acting appropriately, especially for
rewards, may be extrinsically motivated. Fiske and
Taylor (1991) further described how well-devel-
oped belief systems resist change and how individ-
uals persevere despite information suggesting that
change is necessary, a conclusion also reached by
Audia, Locke, and Smith (2000).
696 AugustAcademy of Management Journal
The following conclusions summarize our
arguments:
1. The scarcer the resources needed to pursue both
exploration and exploitation, the greater the likeli-
hood that the two will be mutually exclusive—that
is, high values of one will necessarily imply low
values of the other.
2. Within a single domain (i.e., an individual or a
subsystem), exploration and exploitation will gen-
erally be mutually exclusive.
3. Across different and loosely coupled domains
(i.e., individuals or subsystems), exploration and
exploitation will generally be orthogonal, in that
high levels of exploration or exploitation in one
domain may coexist with high levels of exploration
or exploitation in the other domain.
As the above conclusions indicate, we do not
believe that a universal argument can be made in
favor of either continuity or orthogonality. The re-
lationship between exploration and exploitation
depends very much on whether the two compete
for scarce resources and whether or not the analysis
focuses on a single or on multiple domains. Ac-
cordingly, it is important for researchers to ensure
that their chosen premise (i.e., continuity or orthog-
onality) rests on a foundation of logic and theory.
It also is important to note that starting premises
regarding continuity or orthogonality will have di-
rect implications for how researchers test for the
performance implications of pursuing both explo-
ration and exploitation. If the premise is that ex-
ploration and exploitation are two ends of a con-
tinuum (and thus mutually exclusive), then the
correct test for the beneficial effects of balance
would be to test for an inverted U-shaped relation-
ship between degree of exploration (or exploita-
tion) and organizational performance. On the other
hand, if the premise is that exploration and exploi-
tation are orthogonal, then the correct test for the
beneficial effects of balance would be to test for a
positive interaction effect of the two types of learn-
ing on organizational performance. Figure 1 depicts
these arguments in graphical form.
AMBIDEXTERITY VERSUS PUNCTUATED
EQUILIBRIUM
Arguments in favor of the need for both explora-
tion and exploitation are well established and ac-
cepted (Ancona, Goodman, Lawrence, & Tushman,
2001; Benner & Tushman, 2002; Dougherty, 1992;
Eisenhardt & Martin, 2000; Feinberg & Gupta, 2004;
Levinthal & March, 1993; March, 1991, 1996, 2006).
As March noted, “Adaptive systems that engage in
exploration to the exclusion of exploitation are
likely to find that they suffer the costs of experi-
mentation without gaining many of the benefits.
They exhibit too many undeveloped new ideas and
too little distinctive competence. Conversely, sys-
tems that engage in exploitation to the exclusion of
exploration are likely to find themselves trapped in
suboptimal stable equilibria” (1991: 71).
Although near consensus exists on the need for
balance, there is considerably less clarity on how
this balance can be achieved. As mechanisms to
help organizations realize this balance, the two pri-
mary contenders are ambidexterity (Benner &
Tushman, 2003; Burgelman, 1991; Christensen,
1998; Levinthal, 1997; Weick, 1976) and punctu-
ated equilibrium (Burgelman, 2002; Levinthal &
March, 1993; Siggelkow & Levinthal, 2003; Tush-
man & Romanelli, 1985; Vermeulen & Barkema,
2001). Benner and Tushman provided an excellent
articulation of the logic behind ambidexterity:
Ambidextrous organization designs are composed of
highly differentiated but weakly integrated sub-
units. While the exploratory units are small and
FIGURE 1
Testing for the Performance Effects of a Balance
between Exploitation and Exploration
2006 697Gupta, Smith, and Shalley
decentralized, with loose cultures and processes,
the exploitation units are larger and more central-
ized, with tight cultures and processes. Exploratory
units succeed by experimenting— by frequently cre-
ating small wins and losses (Sitkin, 1992). Because
process management tends to drive out experimen-
tation, it must be prevented from migrating into
exploratory units and processes. In contrast, exploi-
tation units that succeed by reducing variability and
maximizing efficiency and control are an ideal lo-
cation for the tight coordination associated with
process management efforts. (2003: 252)
As noted above, ambidexterity is not the only
possible solution to the quest for balance between
exploration and exploitation. Punctuated equilib-
rium, or temporal cycling between long periods of
exploitation and short bursts of exploration, have
been identified as an alternative balancing mecha-
nism that may be both logical and practical. As
Levinthal and March argued, “Less prominent in
the normative literature on strategy and organiza-
tions, but prominent in more descriptive accounts
(Cyert & March, 1992), is the sequential allocation
of attention to divergent goals. While the sequential
allocation of attention is generally viewed as an
outcome of goal conflict and bounded rationality, it
also results in a simplification of experiments in
organizational change” (2003: 98). Following simi-
lar lines, and drawing on his detailed analysis of
Andy Grove’s 1987–98 reign as CEO of Intel Cor-
poration, Burgelman (2002) appeared to suggest
that punctuated equilibrium is a more viable mech-
anism than ambidexterity:
Does optimal long-run adaptation follow a punctu-
ated equilibrium pattern (e.g., Tushman and Ro-
manelli, 1985), perhaps involving a series of dis-
crete periods, each focused on maximally exploiting
the available opportunities, rather than a more con-
tinuous evolutionary process of balancing exploita-
tion of available opportunities at a given time with
preparing the ground for future growth opportuni-
ties?...This study’s findings raise the question of
whether induced and autonomous strategy pro-
cesses are fundamentally at odds with one another
or can be effectively pursued simultaneously. Main-
taining the simultaneity of induced (variation re-
ducing) and autonomous (variation increasing)
strategy processes may involve difficulties similar to
maintaining a balance between exploration and ex-
ploitation in organizational learning (March, 1991).
(Burgelman, 2002: 354)
Given that ambidexterity and punctuated equi-
librium are very different yet both logical and via-
ble ways to achieve balance between exploration
and exploitation, an obvious question surfaces: Are
the two mechanisms equal substitutes, or is the
appropriateness of each mechanism a function of
environmental and organizational context? In ad-
dressing this question, we would argue that the
contingency perspective has considerable merit. If
one is analyzing exploration and exploitation
within a single domain (i.e., an individual OR a
subsystem), and exploration and exploitation are
rightly conceptualized as the mutually exclusive
ends of a continuum, ambidexterity is simply not
an option, and the individual or subsystem must
resort to punctuated equilibrium. In the language of
systems (Henderson & Clark, 1990), this would be
the case when one’s focus is on innovation in sys-
tem-level architecture. By definition, system-level
architecture encompasses an entire system and
constitutes a single domain. Thus, long-term adap-
tation at the level of architecture requires sequen-
tial attention to exploitation and exploration—that
is, punctuated equilibrium. Intel Corporation,
viewed from the perspective of its CEO, fitted this
model (Burgelman, 2002). In contrast, as argued
earlier, if one is analyzing exploration and exploi-
tation in multiple, loosely connected domains, the
two become orthogonal tasks, and it becomes en-
tirely feasible (and perhaps desirable) to pursue
ambidexterity. In the language of systems design,
exploration can be pursued in one module of a
modular system while exploitation is pursued in
another.
We can now summarize the above arguments in
the form of the following conclusions:
1. When analysis is confined to a single domain
(i.e., individual or subsystem) and exploration and
exploitation are conceptualized as two ends of a
common continuum, logic dictates that punctuated
equilibrium be viewed as the appropriate adapta-
tion mechanism for balancing the need for both
exploration and exploitation.
2. When analysis involves action in multiple and
loosely connected domains and exploration and
exploitation are conceptualized as orthogonal,
logic dictates that ambidexterity be viewed as the
appropriate adaptation mechanism for balancing
the need for both exploration and exploitation.
3. Ambidexterity and punctuated equilibrium may
be easier to achieve at an organizational or system
level than at an individual or subsystem level of
analysis.
These conclusions are not as disparate as they may
at first appear to be. Consider a system comprised
of two loosely coupled individuals or subsystems,
A and B. At time t
1
, subsystem A may be pursuing
exploration while B pursues exploitation. At time
t
2
, subsystem A may switch to exploitation, while B
switches to exploration. Thus, within each sub-
system (i.e., each single domain), long-run adapta-
tion occurs via punctuated equilibrium. However,
698 AugustAcademy of Management Journal
across the two domains, long-run adaptation occurs
via ambidexterity. Finally, if the two subsystems
are tightly rather than loosely coupled, then long-
run adaptation even at the level of the entire system
would require punctuated equilibrium and not am-
bidexterity. As Benner and Tushman (2003) noted,
ambidexterity is not a viable option for tightly cou-
pled systems.
DUALITY VERSUS SPECIALIZATION
The analysis we have described so far was built
on an acceptance of March’s (1991) arguments re-
garding the need for every organization to pursue
both exploration and exploitation. We now exam-
ine in more detail whether it might be logical to
predict that, under certain conditions, long-term
survival may be feasible without balance—that is,
by dedicating an organization or system solely to
exploration or solely to exploitation.
We start with the everyday observation that or-
ganizations operate within a broader social system
and as such are interdependent with many other
organizations (Pfeffer & Salancik, 1978; Thompson,
1967). Looking through this lens, it may be possible
to predict that, under certain conditions, the bal-
ance between exploration and exploitation could
be achieved at the level of the broader social system
rather than at the level of individual organizations.
If so, then some organizations may specialize in
exploration, some others in exploitation, while the
balance between the two is achieved via a market
(or quasi-market) interface. Such a scenario is noth-
ing other than the logic of ambidexterity extended
and generalized from the context of single organi-
zations (Benner & Tushman, 2003) to the context of
the broader social system. Building on the logic of
ambidexterity, we would argue that the specializa-
tion strategy is likely to be effective only if the
following conditions are met: (1) The two organi-
zations A and B, where A specializes in exploration
and B in exploitation, control mutually comple-
mentary resources. Such complementarity would
ensure that the output of A’s exploration is not
entirely wasted and that the promising ideas can be
handed over to B for exploitation. Conversely, even
though B focuses solely on exploitation, it has a
constant supply of radically new ideas available
from A. (2) The domain in which organization A
operates is highly dynamic, whereas the domain in
which organization B operates is highly stable. This
variation in the dynamism of environments would
ensure that A faces a persistent need for explora-
tion, whereas B faces a persistent need for exploi-
tation. (3) The degree of mutual cospecialization in
the two sets of resources is low. In the presence of
low cospecialization, a market relationship be-
tween A and B is likely to be a sufficient and stable
means of ensuring that each gets compensated ad-
equately for its contribution (Teece, 1992; William-
son, 1985).
The semiconductor industry, which is experienc-
ing ongoing disaggregation into “fabless” semicon-
ductor companies (which only do product R&D)
and fabrication companies (which only do contract
manufacturing), is a good example of an industry
where the above three conditions appear to be si-
multaneously valid. The complementarity between
semiconductor R&D and semiconductor produc-
tion is obvious. Also, with a relatively modest de-
gree of coordination, most fabrication companies
can manufacture the differing and time-varying
products of fabless companies; thus, cospecializa-
tion in the tangible resources as well as intangible
capabilities of the two sets of organizations tends to
be relatively low.
1
Finally, in the semiconductor
industry, product technologies become obsolete
very rapidly; in contrast, given the extreme capital
intensity of manufacturing resources, the rate of
evolution in production technologies is much
slower. Accordingly, a high degree of ongoing ex-
ploration is far more essential for the pure R&D
companies, whereas a high degree of ongoing ex-
ploitation is far more essential for the pure produc-
tion companies.
It is worth noting that both March (1991) and
Benner and Tushman (2003) signaled the possibil-
ity that, under well-specified conditions, special-
ization rather than duality might be entirely viable.
As March noted, “Finding an appropriate balance
is made particularly difficult by the fact that the
same issues occur at levels of a nested system—at
the individual level, the organizational level, and
the social system level” (1991: 72). In keeping with
this observation, we have argued above that, under
certain conditions, individual organizations may
justifiably focus solely on exploration or exploita-
tion while delegating the task of achieving a bal-
ance between the two to the social system. Along
1
According to field interviews with executives in two
different fabless semiconductor companies headquar-
tered in California, semiconductor manufacturers share
design rules and SPICE models (“SPICE” is “Simulation
Package for Integrated Circuit Emulation”) with the fab-
less companies. As long as the product design of a new
chip conforms to these rules and models, the need for
coordination between a fabless company and a contract
manufacturer is relatively modest. The same facility can
manufacture multiple generations of semiconductor
chips, even when different generations have radically
different product designs.
2006 699Gupta, Smith, and Shalley
similar lines, in concluding their arguments regard-
ing the desirability of ambidextrous organizations,
Benner and Tushman noted: “While our ideas may
be relevant for all types and sizes of firms, they
apply most readily to firms whose strategies in-
clude both exploitative and exploratory innova-
tion....Our propositions are therefore less rele-
vant for firms whose strategies focus solely on
either exploitation or exploratory innovation”
(2003: 252). Although Benner and Tushman (2003)
clearly accepted the possibility of specialized or-
ganizations, they did not articulate the conditions
under which such specialization might be not only
viable but also effective for fostering long-term sur-
vival. We have attempted to do so above.
THE WORK IN THIS
SPECIAL RESEARCH FORUM
In this section, we summarize the papers selected
for this special research forum and connect them,
where appropriate, to the four issues discussed
above. Overall, we received 83 submissions. Of
these, the authors of 15 manuscripts were asked to
revise and resubmit, and 7 manuscripts were ulti-
mately accepted for publication. The 7 articles in
the special forum span all levels of analysis: the
individual operating in a team or an organizational
context (Miller, Zhao, & Calantone, 2006; Taylor &
Greve, 2006), the team (Beckman, 2006; Perretti &
Negro, 2006), the single organization (Siggelkow &
Rivkin, 2006), and the interorganizational context
(Lavie & Rosenkopf, 2006; Wadhwa & Kotha, 2006).
The papers thus range from micro to macro, and we
have arranged them in the issue along this spec-
trum. Table 1 summarizes the key features of each
paper. As is obvious, the papers vary in their treat-
ment of the various issues discussed earlier: conti-
nuity versus orthogonality, ambidexterity versus
punctuated equilibrium, and duality versus spe-
cialization. These variations are entirely consistent
with and reinforce the logic reflected in the above
discussion and conclusions.
Miller, Zhao, and Calantone’s article, “Adding
Interpersonal Learning and Tacit Knowledge to
March’s Exploration-Exploitation Model,” repli-
cates and extends March’s (1991) model simulating
learning within an organization, using an agent-
based simulation. First, these authors extend his
model by considering the role of direct interper-
sonal learning, in addition to learning from an or-
ganizational code. Second, they place individuals
in a location or space. Doing this allows them to
distinguish between distant and local search. Fi-
nally, Miller and colleagues recognize the impor-
tance of tacit knowledge. Tacit knowledge cannot
be transmitted through an organizational code, but
only by individuals sharing their knowledge di-
rectly with others. The first key contribution of
Miller and colleagues’ extension of March’s model
is their recognition of the value of individual learn-
ing and how interpersonal interactions in organiza-
tions are critical for knowledge transfer. The sec-
ond important contribution these authors make is
their offering the first simulation model distin-
guishing tacit and explicit knowledge. A third con-
tribution is that their model captures the spatial
bias that exists in organizations, whereby individ-
uals tend to interact more often with proximate
others. Finally, their model is able to differentiate
between local and distant search as two distinct
aspects of learning. Miller and colleagues’ model
points to the importance of decentralized interper-
sonal learning to overcoming the potential rigidity
of organizations emphasizing exploitation rather
than exploration. Therefore, Miller, Zhao, and
Calantone highlight the fact that some degree of
specialization in exploration or exploitation can
exist at different levels in a system, while the over-
all system exhibits duality.
Taylor and Greve’s article,“Superman or the
Fantastic Four? Knowledge Combination and Ex-
perience in Innovative Teams,” examines
whether the team compositional factors—diver-
sity of knowledge and experience working to-
gether—that lead to variance-enhancing behav-
iors (i.e., exploration) differ from those that lead
to higher mean performance (i.e., exploitation).
They examine these issues by studying teams
involved in the creation and publishing of comic
books (e.g., artists and writers). Taylor and Greve
argue and find that having multiple knowledge
domains leads to the combining of knowledge in
ways that yield innovations. Somewhat surpris-
ingly, they find that similar factors affect both
innovations that lead to extreme success or fail-
ure (as measured by a comic’s commercial value)
and those that lead to high average performance.
Also, individuals were found to be able to com-
bine diverse knowledge more effectively than
teams, which makes sense, given the potential
process losses inherent in team work. A key con-
tribution of this paper is that the researchers look
at the full range of innovations, rather than only
successful innovations. Therefore, they are able to
model factors that can lead to high mean performance
over time, radical innovations, and commercial fail-
ures. This is important, since much of the prior re-
search has not looked at the effect of risk and what
factors lead to failures rather than successes.
700 AugustAcademy of Management Journal
TABLE 1
Summary of the Articles in the Special Research Forum on Managing Exploration and Exploitation
Article
Level of
Analysis Sample/Method
Role of
Exploration/
Exploitation
Definitions and Connotations
of Exploration/Exploitation
Continuity vs.
Orthogonality
Ambidexterity vs.
Punctuated Equilibrium
Duality vs.
Specialization Conclusions
Miller, Zhao, &
Calatone
Individual and
organization
Agent-based
simulation
Dependent variable Differences in rate of learning
(i.e., rapid vs. slow)
Continuous Ambidexterity Specialization at
different levels of
system, duality for
overall system
Direct interpersonal
learning and tacit
knowledge
transmission exist.
Taylor & Greve Individual and
team
Comic books published
1972–96 (archival
and survey data)
Dependent variable Differences in how teams use
knowledge for either
radical or incremental
innovation (i.e., local and
distant search vs. local
search only)
Initially viewed
as continuous,
but results
indicate can
be orthogonal
Ambidexterity Duality possible Similar factors affect both
innovations with
extreme results and
innovations with high
average performance.
Beckman Team and
organization
Young high-technology
firms (interview,
survey, and archival
data)
Dependent variable Differences in level of
learning (i.e., more intense,
distant search vs. local
search only)
Orthogonal Ambidexterity Duality best Founding team prior
company affiliations
affect pursuit of
exploratory and
exploitative strategies.
Perretti &
Negro
Team Hollywood films
produced 1929–58
(archival)
Dependent variable Exploration vs. exploitation
in team design (proportion
of newcomers and
proportion of
new member
combinations)
Continuous Punctuated equilibrium
(implicit assumption)
Specialization:
Exploration only
Decision maker status
and organizational
hierarchy have a U-
shaped relationship
with exploration in
team design.
Siggelkow &
Rivkin
Organization Agent-based
simulation
Independent
variables (local
or distant search
and tight
coupling)
Differences in
interdependencies between
organizational levels can
reverse the effects of
decentralized exploration
Orthogonal:
Focus on
exploration
only
Punctuated
equilibrium: Focus
on exploration only
Specialization:
Exploration only
Exploration can be
decentralized only
when cross-level
interdependencies are
low.
Lavie &
Rosenkopf
Organization U.S. software firms,
1990–2001 (archival:
COMPUSTAT and
Securities Data
Corporation)
Dependent variable Differences in alliance
function, structure, and
attribute across time and
between domains
Continuous Ambidexterity through
alliance type;
punctuated
equilibrium within
domain
Duality between alliance
type; specialization
within domain
Firms can deploy both
ambidextrous alliance
strategies and
punctuated
equilibrium alliance
strategies.
Wadhwa &
Kotha
Interorganiza-
tional
U.S. telecom-
munication
equipment
manufacturers,
during 1989–99
(archival)
Independent
variable (impact
of exploratory
moves on
knowledge
creation)
Corporate venture capital
investments treated as
exploratory moves
Continuous Punctuated equilibrium
(implicit assumption)
Specialization:
Exploration
High involvement
significantly boosts the
knowledge creation
gains from corporate
venture capital
investments.
Second, Taylor and Greve’s findings point to the
value of measuring the career experiences and knowl-
edge of team members, since they found that the right
team composition (for instance, diverse members as
well as depth of experience) can enable an ambidex-
trous approach. Third, their finding that common
factors predict both exploration and exploitation
points to the orthogonality of these approaches for
teams. Finally, they discuss how the dichotomy of
exploration versus exploitation may actually be
driven by differences in goals and expectations for a
task rather than by different team knowledge assets.
Thus, Taylor and Greve’s work highlights the learn-
ing and knowledge exchanges between heteroge-
neous team members that can enhance ambidexterity.
In her article, “The Influence of Founding Team
Company Affiliations on Firm Behavior,” Beckman
proposes that the firm and market experience of
new ventures’ founders (their prior company affil-
iations) shape the type of new firms created as well
as the strategies and structures put into place. She
argues that founding teams with members with
common prior company affiliations are more likely
to have shared beliefs and a common language,
which will contribute to the adoption of routines
for efficiency and incremental improvements in
products and processes (i.e., exploitation). It
should be noted that Beckman views prior common
affiliations as important not because founding team
members have had direct prior work experience
with each other, but because cohesion results from
shared experiences gained through the mere affili-
ation with the same company. On the other hand,
Beckman argues that diverse prior company affili-
ations lead to a variety of perspectives, the discov-
ery of new alternatives, and increased external so-
cial capital, which should lead to innovations (i.e.,
exploration). Finally, she expects founding teams
composed of a mixture of members with both com-
mon and diverse prior company affiliations to pro-
mote organizational ambidexterity and ultimately
greater firm performance. She finds support for
these arguments. A key contribution of this paper is
the message that new firms need to pay attention to
the composite set of experiences that potential
team members have had when constructing found-
ing teams. This view is a nice complement to Tay-
lor and Greve’s (2006) focus on the role of career
experiences and knowledge of team members.
Also, Beckman is able to show that diverse found-
ing teams can be ambidextrous in their use of strat-
egies, an ability that improves firm performance.
Perretti and Negro’s article, “Filling Empty Seats:
How Status and Organizational Hierarchies Affect
Exploration versus Exploitation in Team Design,”
focuses on the introduction of newcomers and the
formation of new member combinations in team
design. Treating exploration and exploitation as
two ends of a continuum, they take the position
that team design can be viewed as more exploratory
when the proportion of newcomers on a team is
high and/or when the proportion of new member
combinations is high. Perretti and Negro’s empiri-
cal setting is the Hollywood film industry during
the period 1929 –58, and their teams are the five-
person groups that contain the key players in the
development of any film: the director, the two lead
actors, the editor, and the director of photography.
They regard team members as new when they are
new to the industry as a whole. In keeping with
Phillips and Zuckerman’s (2001) notion of “mid-
dle-status conformity,” Perretti and Negro find that
team design is most exploratory when it is driven
by either very high status or very low status indi-
viduals. Further, building on Siggelkow and
Levinthal (2003), they also find that team design is
most exploratory within organizational contexts
with either three hierarchical layers or one layer. In
summary, both decision maker status and the ex-
tent of organizational hierarchy have U-shaped re-
lationships with exploration/exploitation choices
in team design.
Siggelkow and Rivkin’s article, “When Explora-
tion Backfires: Unintended Consequences of Mul-
tilevel Organizational Search,” questions the ac-
cepted proposition that decentralization of the
exploration process by bringing it to lower organi-
zational levels enhances exploration and perfor-
mance for an entire organization. In an agent-based
simulation, they find that decentralization can in
fact backfire and diminish exploration as well as
performance for the organization as a whole. How-
ever, they also find that when lower-level units are
loosely connected to one another—for example,
when the level of interdependency between depart-
ments is low—more extensive exploration at lower
levels can increase performance for the organiza-
tion. Siggelkow and Rivkin contribute to our un-
derstanding of how exploration can be achieved in
multilevel organizations by focusing on the ques-
tion of when exploration should be decentralized.
They conclude that decentralization is appropriate
when decisions and departments are modularized.
This paper reinforces our discussion of the appro-
priateness of specialization and certain organiza-
tional design features, such as loosely coupled
structures, for achieving exploration.
Lavie and Rosenkopf’s article, “Balancing Explo-
ration and Exploitation in Alliance Formation,” ex-
plains how equilibrium between exploration and
exploitation is achieved through three forms of al-
liances: function (marketing or R&D), structure (re-
702 AugustAcademy of Management Journal
current partner or new partner), and attribute (sim-
ilar or dissimilar partner). Their focus on different
types of alliances reflects the premise that organi-
zations can pursue exploration and exploitation in
different domains; for example, firms can trade off
exploring in R&D functions versus exploiting in
marketing functions; learning from new partners
versus taking repeat partners; and working with
similar partners versus working with dissimilar
partners. Data from a sample of U.S. software firms
suggested that although path dependencies rein-
force existing patterns of exploration or exploita-
tion within certain domains, software firms do bal-
ance alliance exploration and exploitation both
over time and across domains. Lavie and Rosen-
kopf’s results emphasize the importance of looking
at different types of exploration and exploitation
through a wide lens and using longitudinal data to
study the balancing process. The firms they studied
appeared to be able to deploy both ambidextrous
strategies (exploration and exploitation at the same
time) and punctuated equilibrium strategies
(switching across time). However, if researchers
were to examine a single type of alliance over a
shorter period, specialization rather than balance
might emerge as the more valid description.
Finally, Wadhwa and Kotha’s article, “Knowl-
edge Creation through External Venturing: Evi-
dence from the Telecommunications Equipment
Manufacturing Industry,” examines the factors that
drive technological learning from corporate ven-
ture capital (CVC) investments. These authors treat
exploration and exploitation as two ends of a con-
tinuum and view CVC investments as more explor-
atory when they lead to greater technological learn-
ing on the part of corporate investors. Focusing on
the telecommunications equipment manufacturing
industry, Wadhwa and Kotha find that the number
of CVC investments has an inverted U-shaped rela-
tionship with the extent of technological learning,
suggesting that greater involvement in CVC activity
comes not just with benefits but also with rapidly
rising costs. They also find that the impact of CVC
activity on technological learning is influenced
heavily by whether or not the parent corporation is
actively involved in the investee firm. Involvement
significantly boosts the extent of technological
learning. Most interestingly, when involvement is
high, the relationship between number of CVC in-
vestments and technological learning changes di-
rection and, instead of an inverted U-shaped rela-
tionship, a U-shaped relationship emerges. In
summary, according to Wadhwa and Kotha, tech-
nological learning from CVC investments depends
far more on direct interpersonal interaction be-
tween personnel on both sides than it does on mere
financial investment by larger corporations into
young ventures.
These seven articles significantly contribute to
knowledge of what drives variations in exploration
and exploitation, whether and how a balance be-
tween the two may be achieved, and how variations
in exploration and exploitation impact perfor-
mance. Collectively, these papers also reinforce our
primary conclusions:
(1) Both exploration and exploitation involve
learning, albeit of different degrees and/or types.
(2) Depending on whether one’s focus is on a sin-
gle or multiple domains, exploration and exploita-
tion can be treated as two ends of a continuum or as
orthogonal to each other.
(3) Depending on the context, either ambidexterity
or punctuated equilibrium may serve as the more
appropriate balancing mechanism between explo-
ration and exploitation.
(4) When a multiplicity of subsystems interact
with each other via modular/standardized inter-
faces, the task of balancing exploration and exploi-
tation can be delegated to the higher-level system,
and each subsystem can focus on just exploration
or just exploitation without any major threats to
long-run performance.
TOWARD FUTURE RESEARCH
Building on our review of the literature on explo-
ration versus exploitation and the seven works in-
cluded in this special research forum, the following
emerge as some of the more promising directions
for future research:
First, studies that examine exploration and ex-
ploitation at a micro level are relatively scarce.
Even though our call for papers invited submis-
sions across the full range from the micro (i.e.,
individual) to the macro (i.e., the interorganization-
al) levels, the bulk of the submissions focused on
the more macro levels of analysis. We hope that the
microlevel papers included in this research forum
(Miller, Zhao, & Calantone, and Taylor & Greve)
will add a fillip to research that addresses ques-
tions at the micro level of analysis. For example, at
different stages of the creative or innovative pro-
cess, are different types of people or skills required
to be able to successfully explore or exploit? How
can managers learn to recognize good ideas to ex-
ploit? What happens when others choose to exploit
the ideas of an individual or team? How does pol-
itics come into play in groups in terms of when to
share exploratory ideas and when they may be ex-
ploited by others?
Second, studies spanning multiple levels of anal-
ysis are also relatively scarce. Studies that address
2006 703Gupta, Smith, and Shalley
questions such as the following have the potential
to fill important gaps in scholars’ knowledge base:
What is the individual- and/or team-level origin of
organizational capabilities for exploration and ex-
ploitation? What is the joint effect of intra- and
interfirm networks on exploration and exploita-
tion? How does exploration at one level interact
with exploitation at a lower or higher level? What
are the similarities, differences, and interactions
between an individual’s, a group’s, and an organi-
zation’s creative capacity, diffusive capacity, ab-
sorptive capacity, and risk-taking capacity? Also,
are there parallel processes between levels of anal-
ysis in the manner in which individuals, groups, or
organizations explore and exploit that could shed
light on and inform how these processes should be
managed across levels of analysis?
Third, studies that examine the challenges asso-
ciated with achieving a balance between explora-
tion and exploitation are scarce. In this introduc-
tory article, we highlighted ambidexterity and
punctuated equilibrium as two alternative mecha-
nisms for achieving such a balance. To date, very
few studies have examined the dynamics of each of
these mechanisms in isolation. Further, almost no
attempt has so far been made to compare and con-
trast the feasibility and appropriateness of these
two mechanisms in different contexts.
In conclusion, we began this introductory article
by highlighting the importance of exploration and
exploitation for successful organizational adapta-
tion, technological innovation, organizational
learning, and even organizational survival. We
identified four key research questions related to
exploration and exploitation, including matters of
definition, continuity versus orthogonality, ambi-
dexterity versus punctuated equilibrium, and dual-
ity versus specialization. As our discussion of these
issues and summarization of the seven papers sug-
gests, the twin concepts of exploration and exploi-
tation involve a number of complex processes, vari-
ables, and contingencies. We are hopeful that this
special research forum will advance understanding
of these concepts. Such research is necessary and
important if researchers are to learn how complex
organizational systems can more effectively learn,
adapt, and survive in the long term.
REFERENCES
Amabile, T. M. 1996. Creativity in context (update to
The social psychology of creativity). Boulder: West-
view Press.
Ancona, D. G., Goodman, P. S., Lawrence, B. S., & Tush-
man, M. L. 2001. Time: A new research lens. Acad-
emy of Management Review, 26: 645– 663.
Audia, G., Locke, E. A., & Smith, K. G. 2000. The paradox
of success: An archival and laboratory study of stra-
tegic persistence following radical environmental
change. Academy of Management Journal, 43: 837–
854.
Baum, J. A. C., Li, S. X., & Usher, J. M. 2000. Making the
next move: How experiential and vicarious learning
shape the locations of chains’ acquisitions. Admin-
istrative Science Quarterly, 45: 766 801.
Beckman, C. M. 2006. The influence of founding team
company affiliations on firm behavior. Academy of
Management Journal, this issue.
Beckman, C. M., Haunschild, P. R., & Phillips. D. J. 2004.
Friends or strangers? Firm-specific uncertainty, mar-
ket uncertainty, and network partner selection. Or-
ganization Science, 15: 259 –275.
Benner, M. J., & Tushman, M. L. 2002. Process manage-
ment and technological innovation: A longitudinal
study of the photography and paint industries. Ad-
ministrative Science Quarterly, 47: 676 –706.
Benner, M. J., & Tushman, M. L. 2003. Exploitation,
exploration, and process management: The produc-
tivity dilemma revisited. Academy of Management
Review, 2: 238–256.
Burgelman, R. A. 1991. Intra-organizational ecology of
strategy-making and organizational adaptation. Or-
ganization Science, 2: 239 –262.
Burgelman, R. A. 2002. Strategy as vector and the inertia
of coevolutionary lock-in. Administrative Science
Quarterly, 47: 325–357.
Christensen, C. M. 1998. The innovator’s dilemma. Bos-
ton: Harvard Business School Press.
Cyert, R. M., & March, J. G. 1992. The behavioral theory
of the firm (2nd ed.). Oxford, U.K.: Blackwell.
Dougherty, D. 1992. A practice-centered model of organ-
izational renewal through product innovation. Stra-
tegic Management Journal, 13: 77–92.
Eisenhardt, K. M., & Martin, J. 2000. Dynamic capabili-
ties: What are they? Strategic Management Journal,
21: 1105–1121.
Feinberg, S. E., & Gupta, A. K. 2004. Knowledge spill-
overs and the assignment of R&D responsibilities to
foreign subsidiaries. Strategic Management Jour-
nal, 25: 823–845.
Fiske, S., & Taylor, S. 1991. Social cognition (2nd ed.).
New York: McGraw-Hill.
Gilson, L. L., Mathieu, J. E., Shalley, C. E., & Ruddy, T. M.
2005. Creativity and standardization: Complemen-
tary or conflicting drivers of team effectiveness?
Academy of Management Journal, 48: 521–531.
He, Z.-L., & Wong, P.-K. 2004. Exploration vs. exploita-
tion: An empirical test of the ambidexterity hypoth-
esis. Organization Science, 15: 481– 494.
Henderson, R. M., & Clark, K. B. 1990. Architectural
innovation: The reconfiguration of existing product
704 AugustAcademy of Management Journal
technologies and the failure of established firms.
Administrative Science Quarterly, 35: 9 –30.
Holmqvist, M. 2004. Experiential learning processes of
exploitation and exploration within and between or-
ganizations: An empirical study of product develop-
ment. Organization Science, 15: 70 81.
Katila, R., & Ahuja, G. 2002. Something old, something
new: A longitudinal study of search behavior and
new product introduction. Academy of Manage-
ment Journal, 45: 1183–1194.
Koza, M. P., & Lewin, A. Y. 1998. The co-evolution of
strategic alliances. Organization Science, 9: 255–
264.
Lavie, D., & Rosenkopf, L. 2006. Balancing exploration
and exploitation in alliance formation. Academy of
Management Journal, this issue.
Lee, J., Lee, L., & Lee, H. 2003. Exploration and exploi-
tation in the presence of network externalities. Man-
agement Science, 49: 553–570.
Levinthal, D. A. 1997. Adaptation on rugged landscapes.
Management Science, 43: 377– 415.
Levinthal, D. A., & March, J. G. 1993. The myopia of
learning. Strategic Management Journal, 14 (spe-
cial issue): 95–112.
March, J. G. 1991. Exploration and exploitation in organ-
izational learning. Organization Science, 2: 71–87.
March, J. G. 1996. Continuity and change in theories of
organizational action. Administrative Science
Quarterly, 41: 278–287.
March, J. G. 2006. Rationality, foolishness, and adaptive
intelligence. Strategic Management Journal, 27:
201–214.
McGrath, R. G. 2001. Exploratory learning, innovative
capacity, and managerial oversight. Academy of
Management Journal, 44: 118 –131.
Miller, K. D., Zhao, M., & Calantone, R. 2006. Adding
interpersonal learning and tacit knowledge to
March’s exploration-exploitation model. Academy
of Management Journal, this issue.
Nerkar, A. 2003. Old is gold? The value of temporal
exploration in the creation of new knowledge. Man-
agement Science, 49: 211–229.
Perretti, F., & Negro, G. 2006. Filling empty seats: How
status and organizational hierarchies affect explora-
tion and exploitation in team design. Academy of
Management Journal, this issue.
Pfeffer, J., & Salancik, G. 1978. The external control of
organizations. New York: Harper & Row.
Phillips, D., & Zuckerman, E. W. 2001. Middle-status
conformity: Theoretical restatement and empirical
demonstration in two markets. American Journal of
Sociology, 107: 379429.
Powell, W. W., Koput, K. W., & Smith-Doerr, L. 1996.
Interorganizational collaboration and the locus of
innovation: Networks of learning in biotechnology.
Administrative Science Quarterly, 41: 116 –145.
Rangan, V. K. 2005. Cisco Systems: Managing the go-to-
market evolution (case # 505006). Boston: Harvard
Business School Publishing.
Rosenkopf, L., & Nerkar, A. 2001. Beyond local search:
Boundary-spanning, exploration, and impact in the
optical disk industry. Strategic Management Jour-
nal, 22: 287–306.
Rothaermel, F. T. 2001. Incumbent’s advantage through
exploiting complementary assets via interfirm cooper-
ation. Strategic Management Journal, 22: 687–699.
Shapiro, C., & Varian, H. R. 1998. Information rules.
Boston: Harvard Business School Press.
Siggelkow, N., & Levinthal, D. A. 2003. Temporarily di-
vide to conquer: Centralized, decentralized, and re-
integrated organizational approaches to exploration
and adaptation. Organization Science, 14:
650 669.
Siggelkow, N., & Rivkin, J. 2006. When exploration back-
fires: Unintended consequences of multilevel organ-
izational search. Academy of Management Journal,
this issue.
Sitkin, S. 1992. Learning through failure: The strategy of
small losses. In B. M. Staw & L. L. Cummings (Eds.),
Research in organizational behavior, vol. 14: 232–
266. Greenwich, CT: JAI Press.
Taylor, A., & Greve, H. R. 2006. Superman or the fantastic
four? Knowledge combination and experience in in-
novative teams. Academy of Management Journal,
this issue.
Teece, D. J. 1992. Competition, cooperation, and innova-
tion: Organizational arrangements for regimes of
rapid technological progress. Journal of Economic
Behavior and Organization, 18: 1–25.
Thompson, J. D. 1967. Organizations in action. New
York: McGraw-Hill.
Tushman, M. L., & Romanelli, E. 1985. Organizational
evolution: A metamorphosis model of convergence
and reorientation. In L. L. Cummings & B. M. Staw
(Eds.), Research in organizational behavior, vol. 7:
177–222. Greenwich, CT: JAI Press.
Vassolo, R. S., Anand, J., & Folta, T. 2004. Non-addi-
tivity in portfolios of exploration activities: A real
options-based analysis of equity alliances in bio-
technology. Strategic Management Journal, 25:
1045–1061.
Vermeulen, F., & Barkema, H. 2001. Learning through
acquisitions. Academy of Management Journal, 44:
457–478.
Wadhwa, A., & Kotha, S. 2006. Knowledge creation
through external venturing: Evidence from the tele-
2006 705Gupta, Smith, and Shalley
communications equipment manufacturing indus-
try. Academy of Management Journal, this issue.
Weick, K. E. 1976. Educational organizations as loosely
coupled systems. Administrative Science Quar-
terly, 21: 1–19.
Williamson, O. E. 1985. The economic institutions of
capitalism. New York: Free Press.
Yelle, L. E. 1979. The learning curve: Historical review
and comprehensive survey. Decision Sciences, 10:
302–328.
Anil K. Gupta (agupta@rhsmith.umd.edu) is the Ralph J.
Tyser Professor of Strategy and Organization and the
research director, Dingman Center for Entrepreneurship,
at the Robert H. Smith School of Business, University of
Maryland at College Park. He earned a doctorate in busi-
ness administration from Harvard University. His re-
search interests center around innovation and entrepre-
neurship, globalization, and the creation and leveraging
of knowledge.
Ken G. Smith (kgsmith@rhsmith.umd.edu) is the Dean’s
Chair and Professor of Business Strategy at the Robert H.
Smith School of Business, University of Maryland at Col-
lege Park. He earned a Ph.D. in business policy from the
University of Washington. His research interests include
strategic decision making, competitive dynamics, and the
management of knowledge and knowledge creation.
Christina E. Shalley (christina.shalley@mgt.gatech.edu)
is a professor of organizational behavior and human re-
source management in the College of Management at the
Georgia Institute of Technology. She received her Ph.D.
in business administration from the University of Illinois
at Urbana-Champaign. Her current research interests fo-
cus on investigating the effects of various social and
contextual factors in enhancing or stifling employee cre-
ativity and examining ways to structure jobs and the
work environment to support creative work.
706 AugustAcademy of Management Journal
... Exploiting activity has been associated with alignment, refinement, efficiency, selection, implementation, and mechanistic structures, whereas exploratory activity is linked with search, variation, risk-taking, experimentation, and organic structures in firms (March, 1991). Despite substantial differences separating these activities, the use of the exploration-exploitation paradigm in different fields of study has clearly revealed a contingent relationship between the two, whose influence generally affects business performances (Derbyshire, 2014;Feng et al. 2024;Gupta et al. 2006;Hill and Birkinshaw, 2014). Management of this "ecological interaction" between exploration and exploitation (March, 1991, p. 71) requires expertise in both areas and efficient handling of their interaction in the firm. ...
... The term organizational ambidexterity identifies the ability to dynamically balance exploration and exploitation levels to achieve the ideal equilibrium or combination in the context of both activities (Cao et al. 2009;Gibson and Birkinshaw, 2004;Luger et al. 2018;Roth and Corsi, 2023). This capacity is difficult to cultivate (Gupta et al. 2006;March, 1991), and its effects on performance (Raisch and Birkinshaw, 2008;Raisch et al. 2009) hinge on organizational and environmental factors (Fourné et al. 2019;Stettner and Lavie, 2014;Uotila et al. 2009). ...
... Analysis of the exploration-exploitation balance suffers practical limitations that hinder the advancement of knowledge on this phenomenon. Firstly, the exploration vs. exploitation dichotomy is difficult to define in a general manner (Lavie et al. 2010), and the researcher's discriminating criteria may exert a decisive influence on study findings (Gupta et al. 2006;He and Wong, 2004;O'Reilly and Tushman, 2013). Secondly, the use of the theoretical framework of exploration-exploitation in fields as diverse as knowledge management, innovation, strategic orientation, competitive edge, and the mere survival of businesses has yielded a plethora of contradictory findings on balance and its consequences for firms (Patel et al. 2013;Raisch and Birkinshaw, 2008;Snehvrat et al. 2018;Turner et al. 2013;Wilden et al. 2018). ...
Article
Full-text available
The literature on organizational ambidexterity emphasizes the complex relationship between the balance of exploitation and exploration and its impact on performance. However, the empirical research, while granting special attention to the study of types of ambidexterity and factors conditioning this phenomenon, has largely neglected to analyze balance. Our study is novel in tackling two knowledge gaps in the literature on organizational ambidexterity: the absence of quantitative analyses to pinpoint the optimal exploration-exploitation balance, and the lack of attention paid to the effect of risk on this optimal balance and on organizational ambidexterity in general. In order to address these knowledge gaps, the paper analyzes the technological exploitation-exploration relationship in conditions of uncertainty and the balance that maximizes Value at Risk (VaR) of firm growth. The methodology includes the elaboration of VaR, using the expected value and degree of dispersion in firm growth, and the combination of quantitative methods for hypothesis testing. Results obtained from a sample of 87,911 European firms from all sectors confirm the positive effect on the growth of harmonious efforts in exploration and exploitation, determining technological balance according to an organization's risk-taking. The findings are relevant to theory and decision-making by managers and policymakers.
... According to OLT, exploitation creates reliability in experience through refinement, production, and focused attention, whereas exploration enhances variety in experience through experimentation, trialing, and free association (Levinthal and March 1993, Greve 2007, Holmqvist 2003. 3 An ongoing debate exists regarding the nature and interaction of exploration and exploitation, which are recognized for their different "type or amount of learning rather than on the presence or absence of learning" (Gupta et al. 2006, p. 694). This perspective varies; exploration and exploitation can be treated as either two ends of a continuum or orthogonal to each other, depending on whether the focus is on single or multiple domains (Gupta et al. 2006). Moreover, the debate encompasses the appropriate balancing mechanisms between exploration and exploitation, alternating between ambidexterity and punctuated equilibrium based on the context (Gupta et al. 2006). ...
... This perspective varies; exploration and exploitation can be treated as either two ends of a continuum or orthogonal to each other, depending on whether the focus is on single or multiple domains (Gupta et al. 2006). Moreover, the debate encompasses the appropriate balancing mechanisms between exploration and exploitation, alternating between ambidexterity and punctuated equilibrium based on the context (Gupta et al. 2006). Furthermore, exploration and exploitation are believed to be concurrent and mutually influence one another (Mom et al. 2007). ...
... 112), posing challenges in developing precise measures. Heeding advice from Lavie et al. (2010), we revisited March's (1991) definitions, integrating insights from key studies (e.g., March 1991, Holmqvist 2003, Gupta et al. 2006, Lavie et al. 2010) and feedback from interviews within the manufacturing and retailing industries to delineate supplier learning activities. Specifically, we adapted the measures of exploration from Jansen et al. (2006) and Nijssen et al. (2012) and the measures of exploitation from Azadegan and Dooley (2010) and Jansen et al. (2006). ...
Article
This research examines the impact of contract framing on supplier learning activities within buyer-supplier relationships, an area previously overshadowed by a focus on internal organizational learning mechanisms. Grounded in regulatory focus theory and organizational learning theory, this study assesses how promotion and prevention contract framing influences suppliers’ exploration and exploitation activities. Survey data from 135 buyer-supplier relationship dyads reveals that promotion and prevention framing distinctly affect suppliers’ learning activities. Specifically, during the duration of the contract, (a) compared with prevention contract framing, promotion contract framing has a stronger positive impact on the supplier’s exploration learning activities; and (b) compared with promotion contract framing, prevention contract framing has a stronger negative impact on the supplier’s exploitation learning activities. This study also reveals how the portfolios and clusters of contract framing influence supplier learning activities. Finally, this study reveals that different learning activities result in varied outcomes for the buyer, with exploration boosting continuity and exploitation improving performance. This contribution broadens the understanding of contract framing’s psychological effects on supplier behavior, offering new insights into the dynamics of organizational learning activities and contract design. Funding: This work was supported by National Natural Science Foundation of China [Grant 72072152]; Research Grants Council, University Grants Committee [Grant CityU 11502218]; and Natural Science Foundation of Shandong Province [Grant ZR2023QG003]. Supplemental Material: The online appendix is available at https://doi.org/10.1287/orsc.2022.16526 .
... This allows the algorithm to build an initial surrogate model that estimates the overall behavior of the function. As the optimization progresses, the surrogate model becomes more accurate, and the algorithm shifts towards exploitation (Gupta et al. 2006, Gan et al. 2021. This allows for a more targeted search, potentially leading to a more accurate surrogate model in the neighborhood of the optimal solution and, as a result, faster convergence. ...
Preprint
This tutorial focuses on kriging-based simulation optimization, emphasizing the importance of data efficiency in optimization problems involving expensive simulation models. It discusses how kriging models contribute to developing algorithms that minimize the number of required simulations, particularly in the presence of noisy evaluations. The tutorial compares the performance of kriging-based algorithms against traditional polynomial-based optimization methods using an illustrative example. Additionally, it discusses key extensions of kriging-based algorithms, including multi-objective and constrained optimization, providing insights into their application in complex, real-world settings.
... C reativity and innovation are important sources of competitive advantage for many firms, and such firms often rely on employees to generate creative solutions to important problems (e.g., Amabile 1996;Kachelmeier, Reichert, and Williamson 2008;Manso 2011). Exploration and exploitation represent two distinct types of activities in innovation settings (e.g., March 1991;Gupta, Smith, and Shalley 2006;Manso 2011). Exploration refers to activities that involve searching for new knowledge, experimentation, and risk taking with the goal of discovering new ideas and solutions (March 1991;Manso 2011). ...
Article
Creativity-dependent companies often implement failure disclosure programs to encourage employees to publicly disclose failures encountered in exploration activities with a focus on learning. We examine the effects of failure disclosure on employees’ exploration performance using an experiment in which participants perform a letter-search task by counting the search letters in each question (i.e., exploitation) or identifying an embedded shortcut that applies to all questions (i.e., exploration). We manipulate two factors: (1) whether participants are encouraged to disclose their failed attempts at finding the shortcut to future participants and (2) whether they are evaluated for their task-related skills. We predict and find failure disclosure increases the likelihood of shortcut identification in the presence of evaluation, and this effect is mitigated in the absence of evaluation. Because evaluation is typically present in natural work settings, our theory and results suggest failure disclosure is an effective mechanism to increase employees’ exploration performance.
Article
Full-text available
This paper proposes a novel bio-inspired metaheuristic algorithm, the Chinese pangolin optimizer (CPO), which draws inspiration from the unique hunting behaviors of Chinese pangolins. Chinese pangolins, leveraging their acute sense of smell and innate hunting instincts, can precisely perceive the distance between themselves and their prey, enabling efficient transitions between luring and predation behaviors. The luring behavior consists of two stages: attracting and capturing and moving to feed, while the predation behavior involves three stages: searching and locating, rapid approaching, and digging to feed. These behaviors were successfully simulated through mathematical modeling, and the algorithm’s convergence was systematically analyzed using Markov chain theory, theoretically ensuring the efficiency and reliability of the algorithm in the optimization process. To comprehensively evaluate the performance of the CPO algorithm, 74 standard benchmark functions were utilized, covering unimodal, multimodal, and fixed-dimension multimodal functions, as well as the CEC2017, CEC2019, and CEC2022 test suites. The experimental results and statistical analyses indicate that the CPO algorithm is effective and convergent when addressing complex numerical optimization problems. The CPO algorithm was also successfully applied to three standard engineering design optimization problems and twelve feature selection tasks, further validating its broad applicability and scalability in real-world problems. The experimental results demonstrate that the CPO algorithm outperforms various baseline metaheuristic algorithms and mainstream feature selection methods regarding optimization performance and classification accuracy, fully showcasing its broad applicability and superiority in solving complex real-world problems. Source codes of CPO are publicly available at https://ww2.mathworks.cn/matlabcentral/fileexchange/178109-chinese-pangolin-optimizer.
Article
This article conducts an in‐depth review of the existing literature on network evolution, with an emphasis on understanding how supply chain networks change and evolve. This review has identified a total of 1148 articles in the area of network evolution, comprising 120 articles from business literature and 1028 articles from non‐business literature. Adopting a structured approach, we provide a roadmap for supply chain network evolution (SCNE) to systematically build a multidimensional understanding of SCNE and bridge the gap between theoretical underpinnings and research methods useful to investigate SCNE. We first broaden the scope to identify insights on dynamic network evolution from a wider array of disciplines. Next, we categorize the drivers of SCNE identified within the business literature into exogenous and endogenous factors. We then synthesize theoretical frameworks that lay the groundwork for future research directions proposed in the context of these frameworks. Finally, we outline contemporary methodologies for studying network evolution and evaluate their relevance and application to the supply chain management (SCM) field. Collectively, this roadmap paves the way for a deeper, interdisciplinary understanding of SCNE and outlines a future research agenda, providing a forward‐looking path for supply chain researchers to explore SCNE.
Article
Purpose Embodied intelligent robots are the iconic productivity of the Industry 4.0 era, and their potential to bring about a productivity surge mainly comes from the driving force of robots on innovation rather than efficiency. However, the dynamic impact of robots on the innovation capability of enterprises has not been empirically tested. Design/methodology/approach This study integrates panel vector autoregression and threshold effects to investigate this dynamic relationship by a multi-level analysis based on data of Chinese A-share manufacturing listed enterprises. Findings (1) The short-term momentum of industrial robot applications (IRA) on exploitative innovation (EII) is significant and the long-term momentum on exploratory innovation (ERI) is stronger. (2) EII affected by IRA is the main source of short-term total factor productivity (TFP) growth, while ERI is the driving factor for long-term TFP growth. (3) The impact of IRA on TFP exhibits a double-threshold effect based on ERI and follows a “stepped” incremental pattern. The promoting effect of IRA on TFP will significantly increase only when ERI surpasses certain thresholds. Originality/value Industrial robots accelerate the potential productivity growth in the long term, mainly coming from the augmented contribution of ERI, providing reference and inspiration for enterprises to fully utilize the endogenous growth potential of robots and implement innovation strategies. It also provides forward-looking guidance for organisations to undertake adaptive changes for the forthcoming AI economic revolution.
Article
Social enterprises (SEs), hybrid entities balancing revenue generation and social or environmental goals, often employ bricolage due to resource constraints. Interviews with 37 SE managers unveiled two pivotal bricolage strategies—utilizing SE status–related marketing resources and leveraging available technological resources—as well as how their interplay influences geographical expansion and the contingent roles of stakeholder participation in facilitating their impact. Quantitative studies of 778 U.K. SEs confirm that the predominant facilitator of geographic expansion is the utilization of status-related marketing resources, surpassing the impact of leveraging available technological resources. SEs’ efforts to utilize SE status–related marketing resources should be harmonized with community participation, whereas SEs aiming to leverage available technological resources should align their efforts with employee participation. We also underscore the substitution dynamic between these two bricolage strategies. However, SEs prioritizing employee participation are better positioned to mitigate the challenges arising from this substitution than those emphasizing community participation.
Article
This paper focuses on dynamic capabilities and, more generally, the resource‐based view of the firm. We argue that dynamic capabilities are a set of specific and identifiable processes such as product development, strategic decision making, and alliancing. They are neither vague nor tautological. Although dynamic capabilities are idiosyncratic in their details and path dependent in their emergence, they have significant commonalities across firms (popularly termed ‘best practice’). This suggests that they are more homogeneous, fungible, equifinal, and substitutable than is usually assumed. In moderately dynamic markets, dynamic capabilities resemble the traditional conception of routines. They are detailed, analytic, stable processes with predictable outcomes. In contrast, in high‐velocity markets, they are simple, highly experiential and fragile processes with unpredictable outcomes. Finally, well‐known learning mechanisms guide the evolution of dynamic capabilities. In moderately dynamic markets, the evolutionary emphasis is on variation. In high‐velocity markets, it is on selection. At the level of RBV, we conclude that traditional RBV misidentifies the locus of long‐term competitive advantage in dynamic markets, overemphasizes the strategic logic of leverage, and reaches a boundary condition in high‐velocity markets. Copyright © 2000 John Wiley & Sons, Ltd.
Conference Paper
This paper focuses on dynamic capabilities and, more generally, the resource-based view of the firm. We argue that dynamic capabilities are a set of specific and identifiable processes such as product development, strategic decision making, and alliancing. They are neither vague nor tautological. Although dynamic capabilities are idiosyncratic in their details and path dependent in their emergence, they have significant commonalities across firms (popularly termed 'best practice'). This suggests that they are more homogeneous, fungible, equifinal and substitutable than is usually assumed. In moderately dynamic markets, dynamic capabilities resemble the traditional conception of routines. They are detailed, analytic stable processes with predictable outcomes. In contrast, in high-velocity markets, they are simple, highly experiential and fragile processes with unpredictable outcomes. Finally, well-known learning mechanisms guide the evolution of dynamic capabilities. In moderately dynamic markets, the evolutionary emphasis is on variation. In high-velocity markets, it is on selection. At the level of REV, we conclude that traditional REV misidentifies the locus of long-term competitive advantage in dynamic markers, overemphasizes the strategic logic of leverage, and reaches a boundary condition in high-velocity markets. Copyright (C) 2000 John Wiley & Sons, Ltd.