boumgarden nickerson zenger

Strategic Management JournalStrat. Mgmt. J., 33: 587–610 (2012)

Published online EarlyView in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/smj.1972

Received 18 December 2009; Final revision received 21 July 2011

SAILING INTO THE WIND: EXPLORING THE

RELATIONSHIPS AMONG AMBIDEXTERITY,

VACILLATION, AND ORGANIZATIONAL

PERFORMANCE

PETER BOUMGARDEN,1* JACKSON NICKERSON,2 and TODD R. ZENGER2

1 Hope College, Holland, Michigan, U.S.A.2 Olin Business School, Washington University in St. Louis, St. Louis, Missouri, U.S.A.

While sustainable high performance requires the capacity to simultaneously explore and exploit,the management literature is divided on the most feasible and efficient route toward this end.We review two proposed approaches for achieving simultaneously high levels of explorationand exploitation: organizational ambidexterity and organizational vacillation. To facilitate com-parison, we map these approaches onto a common performance landscape, making precise theempirical question of which delivers superior long run performance. We then analyze canonicalcases from both literatures, examining patterns of decision making and corresponding perfor-mance over time. These cases suggest that vacillation may offer higher long run performance thanambidexterity, while ambidexterity enhances performance on the margin when utilized withinlarger epochs of vacillation. We conclude that ambidexterity and vacillation are complementswith respect to performance, albeit through different mechanisms. Copyright 2012 John Wiley& Sons, Ltd.

INTRODUCTION

Organizational scholars for decades have arguedthat an organization’s long run performancedemands that it simultaneously explore and exploit(Abernathy, 1978; Burns and Stalker, 1961; Cyertand March, 1963; Ghemawat and Ricart, 1993;Lawrence and Lorsch, 1967; O’Reilly and Tush-man, 1997; Thompson, 1967; March, 1991). AsLevinthal and March (1993: 105) succinctly argue,‘the basic problem confronting an organization isto engage in sufficient exploitation to ensure itscurrent viability and, at the same time, devoteenough energy to exploration to ensure future via-bility.’ Implicit in this literature is the assumption

Keywords: organizational design; ambidexterity;exploration-exploitation; vacillation; organization change*Correspondence to: Peter Boumgarden, Department of Man-agement, Hope College, 41 Graves Place, Holland, MI 49523,U.S.A. E-mail: [email protected]

that exploration and exploitation function as com-plements in delivering high levels of organizationalperformance. While performance is increased byexploitation activities, the positive effect ofexploitation on performance is greater as explo-ration increases and vice-versa.1 Consistent withthis assumption, high levels of both explorationand exploitation with some degree of temporalsimultaneity, or what some may call balance, opti-mize the complementary relationship in generatingperformance.

The broad challenge in simultaneously achievingexploration and exploitation is that the organiza-tional design elements that promote explorationare distinct from design elements that promote

1 More formally, complementarily with respect to performanceimplies that ∂P (x)/∂x1∂x2 > 0 where P(x) equals performance,x1 equates to the amount of exploration, and x2 equates to theamount of exploitation. The formal representation assumes Pis concave and satisfies certain technical conditions of concavefunctions (Kannai, 1980).

Copyright 2012 John Wiley & Sons, Ltd.

588 P. Boumgarden, J. Nickerson, and T. R. Zenger

exploitation. Scholars have shown that the orga-nizational designs that promote exploration aredecentralized and organic, while those that pro-mote exploitation are centralized and mechanistic(Burns and Stalker, 1961; Thompson, 1967; Dun-can, 1976; O’Reilly and Tushman, 2008). Indeed,each organizational design demands its own dis-tinct and complementary set of elements regardingstructure, incentives, and culture. Research sug-gests that attempts to design an organization thatboth explores and exploits create inconsistenciesin design elements that diminish the firm’s capac-ity to deliver the desired performance outcome.Indeed, designing an organization that achievesthis simultaneity in exploration and exploitation isconsidered either highly difficult or simply impos-sible (Abernathy, 1978; Christensen, 1997; Cyertand March, 1963; Duncan, 1976; O’Reilly andTushman, 1997, 2004, 2008; Gibson and Birkin-shaw, 2004). Described more precisely, the organi-zational elements that deliver exploitation generatenegative externalities for those that deliver explo-ration and vice versa. The presence of one set oforganizational elements detracts from the effective-ness of the other. The manager seeking to gen-erate both exploration and exploitation, therefore,faces a fundamental paradox. While explorationand exploitation function as complements in gen-erating high performance, the organizational struc-tures and choices that produce them demonstratenegative externalities that undermine the simulta-neous delivery of both. Finding a way to resolvethis fundamental design paradox is essential toachieving high organizational performance.

The organization design literature offers two dis-tinct resolutions to the paradox. One approach,referred to as ‘organizational ambidexterity,’ pro-poses a balance in exploration and exploitationactivities by crafting complex hybrid or dual-structured organizations (O’Reilly and Tushman,2008). The approach calls for one part of the orga-nization to engage in exploration while anotherpart engages in exploitation, with any resultingorganizational inconsistencies overcome throughintegration efforts by top-level managers.Ambidextrous top-level managers, through theirleadership skills, are advised to balance explo-ration and exploitation attempting to achieve astatic equilibrium, selectively integrating andaddressing any negative externalities or incon-sistencies in organizational design (O’Reilly and

Tushman, 2004, 2008; Benner and Tushman, 2003;Raisch et al., 2009).

The alternative approach, referred to as ‘organi-zational vacillation,’ emphasizes dynamicallyachieving high levels of both exploration andexploitation by temporally and sequentially alter-nating between organizational structures that pro-mote either exploration or exploitation, respec-tively (Gulati and Puranam, 2009; Nickerson andZenger, 2002; Siggelkow and Levinthal, 2003).The approach asserts that a structure promotingbalance per se is not necessarily the managerialprescription that follows from the complementar-ity between exploration and exploitation. Instead,this perspective notes that the manager’s task isto optimize long run performance, where perfor-mance is influenced by the levels of explorationand exploitation and not merely their degree ofbalance or simultaneity. By vacillating between(or among) discrete formal organizational modessuch as centralization and decentralization, theorganization may increase dynamically the lev-els of exploration and exploitation beyond thoseachievable through an approach based strictly ona static design choice. The key assumption is thatas structural shifts occur, the levels of explorationand exploitation increase and dissipate with iner-tia (Gulati and Puranam, 2009; Nickerson andZenger, 2002). Hence, the fundamental distinc-tion between these approaches is that with orga-nizational ambidexterity, managers achieve highperformance by deliberately emphasizing a struc-ture that promotes balance in exploration andexploitation, whereas with vacillation, managersachieve high performance by dynamically vacil-lating between structures to achieve high levelsof exploration and exploitation on average, albeitwith inconsistent balance.

In this article, we explore the relationshipsamong ambidexterity, vacillation, and organiza-tional performance. We note at the outset that orga-nizational ambidexterity and organizational vac-illation rely on distinctly different assumptions,which make them theoretically difficult to com-pare. Comparison is further complicated becauseneither theory is entirely clear in its posited rela-tionship to long run economic performance. Toadvance a comparative analysis of the two the-ories, we first define organizational ambidexter-ity and organizational vacillation, along with theirbase assumptions and predictions. Through this

Copyright 2012 John Wiley & Sons, Ltd. Strat. Mgmt. J., 33: 587–610 (2012)

DOI: 10.1002/smj

Ambidexterity, Vacillation, and Organizational Performance 589

exercise, we seek to generate a common theoret-ical language. We then make explicit the theo-retical relationships among exploration, exploita-tion, and economic performance by developing alandscape that maps performance as a functionof these dual capabilities. Next, we overlay bothframeworks onto this shared performance land-scape to directly compare their hypothesized per-formance trajectories over time. By mapping ontoa common performance landscape, we make clearthe empirical question regarding the relationshipbetween each theory and long run performance.The key empirical question is whether organi-zational ambidexterity maps onto a performancetrajectory that is superior to or inferior to theperformance trajectory delivered by organizationalvacillation. The question, therefore, is not whichapproach delivers balanced exploitation and explo-ration, but which approach delivers an abundanceof both. Our empirical question is then initiallyinvestigated by turning to two longitudinal canon-ical case studies.

The first case study examines in detail 25 yearsof Hewlett-Packard’s organizational history, a briefsynopsis of which Nickerson and Zenger (2002)used as the archetype to illustrate vacillation.The second case study examines with new detailsapproximately 20 years of USA Today’s effortsto explore and exploit the Internet space throughwhat ultimately became USAToday.com, a canon-ical illustration that O’Reilly and Tushman (2004)use to describe organizational ambidexterity. Thedetails of these case studies offer new insightsinto the currency of the ambidexterity and vacil-lation theories. For instance, the Hewlett-Packardcase illustrates that management vacillated aboutevery four to six years between organizationalstructures focused on generating either explorationor exploitation, but within these epochs Hewlett-Packard experienced periods of both balance inexploration and exploitation. The 20-year assess-ment of USA Today’s efforts to generate an onlinebusiness also suggests period of organizationalambidexterity and a similar pattern of vacillationbetween decentralization and integration. In bothsettings, senior management’s target was high eco-nomic performance. Moreover, with every organi-zational change, managers expressed that a balanceof exploration and exploitation was sought. Yet, inevery instance as managers pursued high perfor-mance, they consistently compromised on balancethrough a structural change that in the short run (or

medium term) aggressively promoted either explo-ration or exploitation, respectively.

In this article, we seek to theoretically explainand illustrate the comparative functionality oforganizational ambidexterity and vacillation. Inaddition, we attempt to compare their respectiverelationships as strategic paths to economic perfor-mance. While our case studies cannot statisticallyrespond to our empirical question, they nonethe-less provide suggestive evidence of the importantrole organizational vacillation and organizationalambidexterity play in achieving high performanceby delivering an abundance of exploration andexploitation, while also offering new evidence fortheory building. Importantly, we find evidence inour case studies that brief episodes of organiza-tional ambidexterity are delivered by and embed-ded within organizational vacillation. Ambidex-trous leadership, therefore, plays an important rolefor generating economic performance within broadpatterns of organizational vacillation. Our insightprovides the foundation for a theoretical recon-ciliation between organizational ambidexterity andorganizational vacillation.

THEORY

The capacity to both explore or ‘search for new,useful adaptations,’ and exploit through ‘the useand propagation of known adaptations’ (Fang, Lee,and Schilling, 2010: 626) is vital to the survivaland performance of organizations (Levinthal andMarch, 1993). Exploration and exploitation func-tion as complements in generating organizationalperformance. Thus, a capacity to exploit innova-tions is more valuable in the presence of a largercapacity to generate innovations, just as a capac-ity to generate innovations is more valuable inthe presence of a larger capacity to exploit them.Moreover, there may be particular benefit to simul-taneously generating high levels of exploration andexploitation, rather than generating exploration andexploitation asynchronously.

While achieving high levels of both of theseactivities concurrently is the aim of many orga-nizations, the literature suggests that this out-come is not easily achieved (Abernathy 1978;Christensen 1997; Cyert and March, 1963; Dun-can, 1976; O’Reilly and Tushman, 1997, 2004,2008; Gibson and Birkinshaw, 2004). The diffi-culty arises because the organizational structures


DOI: 10.1002/smj


that promote exploration are generally regardedas distinct from those that generate exploitation(e.g., Nickerson and Zenger, 2002). Specifically,organizational structures that promote explorationinvolve a distinctly different set of complementarydesign elements than organizational structures thatpromote exploitation. Efforts to mix these com-peting bundles of design choices may yield inef-fective organizational design. Figure 1 provides asimple representation of this tension. While explo-ration and exploitation operate as positive comple-ments in generating organizational performance,negative externalities generally define the relation-ship between the complementary set of designelements that generate exploration and the com-plementary set of design elements that generateexploitation. A choice to design solely for explo-ration (or exploitation) ignores the positive com-plementary that accompanies simultaneous successin generating both, but designing to generate bothconfronts the negative externalities associated withmixing the design elements in a way that targetsboth.

Consistent with Raisch et al. (2009), we high-light two distinct approaches to resolving this con-tradiction and developing such dual capability: astatic approach in which firms adopt ambidextrousorganizational structures that balance the pursuit ofexploration and exploitations (O’Reilly and Tush-man, 2008; Raisch & Birkinshaw, 2008) and adynamic approach in which firms sequence theadoption of structures that target either explorationor exploitation (Nickerson and Zenger, 2002;Siggelkow and Levinthal, 2003; Gulati and

Figure 1. The ambidexterity design dilemma

Puranam, 2009). We will develop theoretical argu-ments for each, noting in particular the tensionsbetween designing to exploit the performance com-plementarities between exploration and exploita-tion and designing to maximize complementaritiesamong organizational design elements.

Organizational ambidexterity

The static approach to organizational ambidexter-ity involves the structural separation of explorationand exploitation activities into distinct units, adual structure coupled with a set of leadershipdirectives focused on integrating across this dis-crete structuring (O’Reilly and Tushman, 2008).For example, managers of an ambidextrous orga-nization orient some units toward exploration withsignificant autonomy (i.e., a new venture division)and other units toward exploitation (i.e., divisionsfocused on current product refinement and mar-keting) with a focus on coordination, resourcesharing, and centralized control. Integration acrossthese groups takes place when a senior leadershipgroup develops and reinforces ‘a common strategicintent, an overarching set of values, and targetedstructural linking mechanisms to leverage sharedassets’ (O’Reilly and Tushman, 2008: 195). Recentwork suggests that middle management may alsoplay an important role in facilitating this integra-tion (Taylor and Helfat, 2009).

The argument for organizational ambidexteritybegins by assuming a structural tension betweendesigning for exploration and designing forexploitation, but then highlighting a top manage-ment capability for resolving incompatible orga-nizational structuring through leadership. Echo-ing the arguments for loosely coupled systems(Weick, 1976), researchers argue that overly dis-ruptive technologies generated through explorationmust be segmented from the rest of the organi-zation or they run the risk of undermining themore exploitative processes of the organization(Christensen, 1997). The complementary integra-tion behaviors are necessary because high per-formance requires capabilities in exploration andexploitation to be intertwined across the entireorganization, rather than focused solely with a spe-cific subunit. As Benner and Tushman (2003: 247)argue, ambidextrous organizations must ‘build inboth tight and loose coupling simultaneously,’ as‘integration and differentiation are complemen-tary, not alternative, mechanisms for achieving


DOI: 10.1002/smj


organizational effectiveness’ (Raisch et al., 2009:687). Integration across these segmented systemstakes place as management uses a balanced set ofincentives to effectively manage the inconsistentalignments (Smith and Tushman, 2005). The inte-gration serves to ‘synchronize the team’s socialand task processes, including the quality of infor-mation exchange, collaborative behavior, and jointdecision making,’ thus helping ‘its members dealwith the contradictory knowledge processes thatunderpin the attainment of an exploitative andexploratory orientation, such that greater integra-tion enhances the likelihood of jointly pursuingboth’ (Lubatkin et al., 2006: 647). Researchersargue that organizational ambidexterity results inan organization capable of navigating ‘a balancebetween the need to be small and large, central-ized and decentralized, and focused both on theshort term and long term simultaneously’ (Bennerand Tushman, 2003: 248).

While much of the literature on ambidexterityexamines large organizations with the capacity tostructurally separate exploring units from exploit-ing units, the deeper organizational challenge is tosimultaneously pursue exploration and exploitationat all levels of the organization, including levelswhere structural separation seems quite infeasi-ble. Indeed, as March (1991: 72) correctly notes,the task of ‘finding an appropriate balance ismade particularly difficult by the fact that thesame issues occur at levels of a nested system.’Thus, entire multibusiness organizations, singledivisions, departments, work teams, and even indi-viduals all face this need for both exploration andexploitation. At the corporate and business unitlevel, however, researchers advocating the staticmodel of organizational ambidexterity highlightthe importance of top management team leadershipand a separation of distinct units that explore whileothers exploit. The approach arguably providesa stable balance of exploration and exploitationactivities over time. While inconsistencies mayarise among design elements, scholars claim that atop management team capable of integrating thesecompeting goals will minimize the negative con-sequences of the tension. Organizational ambidex-terity emphasizes the pursuit of balance betweenexploration and exploitation, consistent with thepresence of complementarity between them andimplies that any negative externalities that accom-pany inconsistency in design elements are mini-mal and can be overcome, or at least mitigated,

through effective leadership and nuanced design.Successfully pursing balance between explorationand exploitation in essence implies the ability tomaintain a more or less static equilibrium.

Organizational vacillation

Organizational vacillation is a dynamic approachto achieving high performance through simultane-ously high levels of exploration and exploitation.The approach advocates modulating between astructural orientation focused on exploration and astructural orientation focused on exploitation. Thetheoretical case for this dynamic approach hingeson the presence of inertia in the informal out-comes from formal shifts in organizational designand on the capacity of an internally consistent andmore focused design structure to generate elevatedlevels of either exploration or exploitation (Nicker-son and Zenger, 2002). Complementarity in designelements creates a form of discreteness in organi-zational design (Mintzberg, 1979; Milgrom, Qian,and Roberts, 1991; Williamson, 1991). While com-plementarity in design elements can certainly beviolated, doing so undermines the design’s capac-ity to achieve high levels of either exploration orexploitation respectively.

Nickerson and Zenger (2002), however, arguethat while complementarities in design encour-age discreteness in the choice of formal organiza-tional structure rather than an ambidextrous mixingof elements, the informal organization, governedby significant inertia, responds with much greatercontinuity. Thus, while the formal structure or‘the normative system designed by management’(Scott, 1981: 82) can be quite abruptly shifted,the degree of actual exploration or exploitationproduced reflects the underlying informal organi-zation—the routines, decision-making processes,and knowledge flows within the organization,as well as the general behaviors, decisions, andactions of individuals within the organization. Theinformal behaviors, processes, communication pat-terns, and routines adjust more continuously inresponse to structural shifts. Thus, formal struc-tural choices influence the shape of the informalorganization (Stevenson, 1990; Shrader, Lincoln,and Hoffman, 1989; Kadushin and Brimm, 1990),albeit more gradually with the pace contingent onthe magnitude of organizational inertia present.


DOI: 10.1002/smj


By choosing a structure characterized by decen-tralization and autonomy, the organization gen-erates knowledge flows, communication patterns,and decision-making routines that promote highlevels of exploration; then, in response to struc-tural change toward greater centralization and inte-gration, these informal organization elements shift(Nickerson and Zenger, 2002). Repeated mod-ulation provides the ‘functionality to temporar-ily achieve intermediate levels’ (Nickerson andZenger, 2002: 560) of exploration and exploitationand, thus, produces brief periods of dual capabilityand a definable high performance trajectory.

In sum, in advocating for organizational vacil-lation, Nickerson and Zenger (2002) suggest thatthis structuring provides relatively high amountsof both exploration and exploitation, albeit in agradually shifting mix. While organizational vac-illation may not produce the stable balance ofexploration and exploitation that is the target oforganizational ambidexterity, the key question iswhether an effectively timed modulation results inan informal organization that produces higher andmore abundant levels of exploration and exploita-tion on average, although in a varying mix.2 Thus,the basic logic of vacillation is that an emphasis onstatic balance in exploration and exploitation com-promises the levels of each that are attained andthat higher performance accompanies the higherlevels of exploration and exploitation achievedthrough a dynamic approach of organizational vac-illation.

MAPPING A COMMON THEORETICALFRAME FOR COMPARISON

While both organizational ambidexterity and orga-nizational vacillation posit a relationship betweenstructural choices and corresponding levels ofexploration and exploitation, comparison is com-plicated in that neither theory is clear in explic-itly defining the relationships among exploration,exploitation, and economic performance. Thus, tofacilitate theory comparison, we delineate these

2 While we don’t address the ideal cadence of change, ourargument shares much in common with Probst and Raisch(2005: 101) who suggest that to remain ‘balanced,’ firms mustidentify ‘early warning signs’ that signal the need for a ‘coursecorrection’ or ‘counter measures.’ As they note, one of the keychallenges for senior leadership is determining when ‘the timefor a course correction has come’ and then determining thecorrect ‘counter measures’ (Probst and Raisch, 2005: 100).

relationships more clearly on a common perfor-mance landscape. We first identify how each theorymaps onto the common framework. Then we makemore theoretically precise the comparative empir-ical question that discriminates between these twoperspectives.

Exploration, exploitation, and performance

To map the relationships among exploration,exploitation, and performance, we articulate threefoundational assumptions. First, we assume that,ignoring for the moment the costs of organization,short run organizational performance (conceptu-alized as an instantaneous rate of expected eco-nomic profitability3) increases with correspondingincreases in the level of exploration or exploita-tion activity within the organization. Long runperformance, therefore, is defined by integratinginstantaneous economic performance over time.For simplicity, we conceptualize two orthogonalvariables—one for the level of exploration and onefor the level of exploitation, where the level repre-sents flow of investments into each type of activity.Second, we assume that exploration and exploita-tion are complementary in generating performance,where complementarity is defined as the propertythat doing more of one activity raises the marginalperformance return of the other (see Milgrom andRoberts, 1990).4 Third, we assume that, consis-tent with standard neoclassical economic assump-tions of convexity, the marginal increase in perfor-mance benefits from increased exploration as wellas from exploitation are positive but with diminish-ing returns to scale.5 Our assumptions are gener-ally consistent with empirical and theoretical workin the ambidexterity and learning literatures thatfinds, for example, a positive relationship betweenthe interaction of exploration and exploitation andsales growth (He and Wong, 2004). The assump-tions also are consistent with the underlying logic

3 The temporal realization of profit differs for exploration andexploitation. We assume that expected economic profitabilityoffers a theoretical construct that accounts for the future pathsof profits generated from past and current levels of investmentsin exploration and exploitation, but does not incoporate returnsfrom potential future investments.4 Milgrom and Roberts (1990) define complementarity in mathe-matical terms for smooth functions as ∂π 2/∂x1∂x2 ≥ 0 where xi

for all i represent the levels of activities conducted to produceresulting profits.5 Mathematically, these assumptions are ∂π/∂x1 ≥ 0; ∂π 2/∂x1

2≤ 0; ∂π/∂x2 ≥ 0; ∂π 2/∂x12

2≤ 0.


DOI: 10.1002/smj


Figure 2. Performance landscape

of vacillation (Nickerson and Zenger, 2002). Math-ematically from these assumptions, we infer thatthe cross-partial derivative of performance withrespect to exploration and exploitation within therelevant range is positive; though, we anticipatethat the overarching performance landscape is con-cave, implying diminishing returns to scale.6

Given our assumptions, Figure 2 provides athree-dimensional representation of the relation-ships among the three variables that satisfies ourassumptions. Specifically, exploration andexploitation are located along the x- and y-axes,respectively, and performance, as expected eco-nomic profitability (a function of such explorationand exploitation combinations), is positioned ver-tically on the z-axis.7

Performance trajectories of ambidexterity and

vacillation

Having defined a performance landscape absentthe costs of organization, we now explore howeach theory maps onto the performance landscapeonce we account for the costs of organization (see

6 Although our model is not formally analytic in nature, we antic-ipate in our formulation that a more formal approach will includeassumptions about smoothness and other desirable mathematicalproperties. For instance, a concave quadratic function within aspecified relevant range would satisfy our assumed constraints.7 In this framework, performance is conceptualized as instanta-neous in that it reflects the short-run performance of an orga-nization at any given point in time. To understand performancein the long run, it is necessary to observe movement along thisperformance landscape over time.

Figure 1). Two classes of organizational costs arerelevant for our analysis. The first class of costs isassociated with the cost of organizing each com-plementary set of design elements used to supporteither exploration or exploitation. For instance, itis commonly argued that a centralized organiza-tional structure is needed to facilitate exploita-tion. In contrast, decentralized organizations gen-erate exploration. Each one of these organizationalstructures requires some setup costs to configurethe set of design elements, as well as administra-tive costs to maintain and operate them. For pur-poses of simplifying exposition, we momentarilyset aside setup costs as well as any administrativecosts to maintain and operate purely complemen-tary configurations of design elements, such aseither centralization or decentralization. Instead,we focus on only those costs associated with thenegative externality from mixing competing designelements with the scope of these costs measuringthe magnitude of negative externalities. As we willdescribe, the magnitude of such costs contributesignificantly in differentiating the empirical pre-dictions of ambidexterity from those of vacillation.We now turn to describing the magnitude of thispotential negative externality and how it influencesthe performance predicted from each theory.

The argument for organizational ambidexterityachieving high performance hinges on the assump-tion that the negative externalities that accompanyorganizational design choices needed to deliverexploration and exploitation are moderate or can,through skillful management, be minimized. Yet,


DOI: 10.1002/smj


Figure 3. Performance comparison of ambidexterity and vacillation strategies

research suggests the negative externalities may besubstantial. For instance, as Martens, Debackere,and Van Looy (2005: 210) argue, ‘the idea of. . .ambidextrous organizations is diametricallyopposed to the notion of internal consistencythat has dominated the literature on organizationaldesign over the last decades.’ Theorists in thestructural configuration literature argue that ‘effec-tive organizations achieve an internal consistencyamong their design parameters, in effect, a struc-tural configuration’ (Mintzberg, 1979: 300), andthey empirically demonstrate that organizationaltraits are not randomly distributed within popu-lations of organizations, but rather appear in clus-ters (Miller and Friesen, 1980; Doty et al., 1993).Therefore, adopting an organizational form thatdeviates from these clusters results in lower per-formance and a diminished probability of survival(Doty, Glick, and Huber, 1993; Mintzberg, 1979).

Economists have also emphasized the funda-mental complementarity among design elementsin discretely different organizational forms. Forexample, Williamson (1991) describes alterna-tive governance forms as characterized by a syn-drome of mutually supporting attributes and viewsattempts to craft organizational forms that vio-late these complementarities as problematic andunlikely to survive. Organizational theorists, usingpunctuated equilibrium models, have emphasizedconcepts of complementarity to describe patterns

of organizational change that result in rather dis-crete organizational choices (Gersick, 1991; Tush-man and Romanelli, 1985). The discontinuousnature of both change and organizational choicesreflects the ‘deep structure’ that governs these sta-ble forms (Gersick, 1991) and the resulting prob-lems of structuring without such consistency.

Given the objective of high performance throughorganizational ambidexterity, we expect anambidextrous organization is structured and led toproduce an approximate balance of exploration andexploitation. Such a balance will fall somewhereon the performance landscape where explorationand exploitation are relatively equivalent—twoexamples being Points One and Two in Figure 3.However, all points along a vector from the ori-gin through Points One and Two are positions ofbalance. The critical question concerns the per-formance level that organizational ambidexterityproduces. In other words, does the performancethat can be delivered by ambidexterity lie closerto Point One or Point Two? The greater the neg-ative externality (organizational cost) delivered byinconsistent design elements, the lower the actualexploration and exploitation delivered by the orga-nizational structure. Put differently, the greater thenegative externality from organizational ambidex-terity, the lower and more distant the level ofperformance achieved will be from Point One. Ifthe negative externality is relatively small, which


DOI: 10.1002/smj


ambidexterity scholars implicitly assume, then anambidextrous organization generates abundant lev-els of exploration and exploitation correspond-ing to high performance indicated by Point One.In contrast, if the negative externality deliveredfrom inconsistent design elements is great, thenthe actual levels of exploration and exploitationare much less than the ideal and project onto theperformance landscape at a much lower level ofperformance—symbolized by the location of PointTwo or even lower.

While the debate over the merits of structuralambidexterity involves a discussion about whereperformance lies between Points One and Two inFigure 2, organizational vacillation maps onto theperformance landscape in a very different way.For instance, instead of balancing exploration andexploitation on the performance landscape, vacil-lation in formal structure leads to a dynamic pathon the landscape that traverses from one side ofthe landscape to the other as initiated by discretechanges in the formal organizational structure.To understand the path delivered by vacillation,assume an initial location on the performance land-scape in which an organization is centralized deliv-ering high levels of exploitation and comparativelylower exploration. Following the vacillation logic,managers desiring more exploration restructure theformal elements of the organization’s design todecentralize. The structural shift quickly increasesthe level of exploration and, due to the embed-ded nature of culture, routines, and communica-tion patterns, may allow for temporary retention ofthe cross-unit coordination developed in the previ-ous structuring. In periods between formal struc-tural changes, the informal organization simultane-ously reflects both the current structure promotingexploitation (exploration) and the prior structurepromoting exploration (exploitation). The simul-taneous increase in the level of exploration anda slowly diminishing level of exploitation allowsorganizational vacillation to generate a higher levelof performance in the short run than a static choiceto adopt one or the other.

Eventually, as the autonomy of the decentral-ized units gradually unwinds the cross-unit coor-dination, the organization repositions to a highlevel of exploration with a low level of exploita-tion. As the temporally achieved dual capabil-ity of exploration and exploitation depletes overtime as “mechanistic’ firms. . . become still moremechanistic over time. . . and firms moving in an

‘organic’ direction become still more organic later’(Miller and Friesen, 1980: 592; c.f. Nickersonand Zenger, 2002), management again restructuresand reverses the dynamic performance trajectory.Vacillation particularly improves performance overa static choice if there is an asymmetry in thepace at which exploration and exploitation respondto shifts in the formal structure; in particular ifa structural shift rapidly accelerates the forma-tion of one performance dimension, while inertiaslows the decay of the other. Thus, if a central-ized firm, which is effectively exploiting, suddenlydecentralizes and, in response, exploitation slowlydecays while exploration quickly accelerates, com-plementarity in exploration and exploitation willelevate performance through switching. Similarly,if the same firm that is now decentralized (andeffectively exploring) suddenly centralizes and, inresponse, exploration slowly decays while explo-ration quickly accelerates, then switching againwill elevate performance. The vacillating back andforth scribes the performance trajectory that movesacross the landscape. The difference between highand low performance vacillation is shown by thedifferent performance locations of the two dual-arrowed lines in Figure 2 (Path Three); specifi-cally, low performance vacillation is shown bythe line closer to the origin and higher perfor-mance movements by the higher-arced line furtherfrom this point. The trajectories of these arcs aredetermined by the inertia present, i.e., how fastorganizational attributes both accelerate and decayin response to shifts in structure.

With both frameworks theoretically mapped ontoa common performance landscape, we can nowformulate the fundamental question of whetherambidexterity is located at a higher level on thelandscape compared to the performance region thatvacillation generates or vice versa. With referenceto Figure 3, the question is whether ambidexterityachieves performance at Point One, which yieldsperformance comparatively greater than that deliv-ered by vacillation along Path Three, or whetherambidexterity achieves performance at Point Two,which yields performance comparatively lowerthan that delivered by vacillation along Path Three.Returning to the logic summarized in Figure 1,the question is whether static design decisionstargeting balance yield a greater abundance ofexploration and exploitation than do design deci-sions that emphasize dynamic vacillation between


DOI: 10.1002/smj


complementary design choices that target explo-ration and exploitation.

CASE EVIDENCE

To examine our fundamental question, we focuson an in-depth case analysis of two familiar illus-trations in this literature. We hope a focus onfamiliar examples that develop a commonly under-stood set of facts will permit more robust dis-cussion and debate about the source of high per-formance through exploration and exploitation, inmuch the same way that the GM-Fisher Body casehas provided the backdrop for extended theoret-ical discussions and debates in the literature onthe topic of vertical integration (see Klein, Craw-ford and Alchian, 1978; Casadesus-Masanell andSpulber, 2000; Coase, 2000; Freeland, 2000; Klein,2000). We first focus our attention on the Hewlett-Packard illustration, which was briefly detailed byNickerson and Zenger (2002) in developing theefficiency of organizational vacillation as an expla-nation for organizational performance. In adding tothis prior analysis, we draw on a much wider arrayof data sources to develop a more complete historyof HP’s evolution over a longer time period of25 years. Moreover, lest we be accused of merelyselecting illustrations that highlight the virtue ofvacillation over structural ambidexterity, our sec-ond case study focuses on USAToday.com, a casefrequently used to illustrate the benefits of orga-nizational ambidexterity (see O’Reilly and Tush-man, 2004, 2008; Tushman et al., 2010). Further,to address March’s (1991) argument that the explo-ration and exploitation tension operates at multiplelevels of analysis, our case illustrations are at twodifferent levels: the firm or multibusiness level andthe business unit level.

We approach our task by examining patternsof managerial decision making over time. Ourmethodological approach is similar to previous useof case studies for theory testing (e.g. Anderson,1983; Pinnfield, 1986), though our time horizon islong. We assume managers are boundedly rationalin their choice of organizational design (Simon,1955) and are selecting structures over time thatrespond to inefficiencies generated by the currentchoices (or changes in environment). If organiza-tional vacillation yields high average performancewhile organizational ambidexterity yields lowerperformance, managers should pursue a pattern of

modulation over time. If, however, organizationalvacillation yields low performance and organiza-tional ambidexterity high performance, manage-ment should pursue organizational ambidexterityover extended periods of time. The case studiesprovide data relevant to understanding when andhow firms develop capability in both explorationand exploitation, thereby serving the purpose oftheory elaboration (Lee, Mitchell, and Sablynski,1999).

Hewlett-Packard

We examine Hewlett-Packard over a 25-yearperiod—a time during which HP became thelargest information technology company in theworld. Our analysis of HP is based on a varietyof data sources, including news articles identi-fied through Lexis-Nexis, case studies developedat Washington University (Zillmer and Zenger,2002a, 2002b, 2002 c, 2002d), Stanford University(Burgelman and Meza, 2000, 2004a, 2004b), andHarvard University (King, 1984), annual reports,and 25 years of securities analyst reports (1,026reports) on Hewlett-Packard from Bears Stearns,Credit Suisse, Deutche Bank, PaineWebber, Pru-dential, and Smith Barney. We also analyzed tran-scripts from personal interviews with former CEOJohn Young conducted by two of the authors in1998, as well as a memoir Carly Fiorina (2006)wrote about her experiences at HP. Using thesedata sources, we constructed a focused historyof HP which we have organized into sectionsdemarcated by significant shifts in organizationalstructure.

A decentralized HP (pre-1982)

By the early 1980s, Hewlett-Packard had firmlyestablished itself as one of the world’s most inno-vative corporations—a company remarkable in itscapacity to explore new technology and exploitwhat it discovers. A 1983 Fortune survey of thetop 6,000 organizations ranked HP third in innova-tiveness. Over several decades, HP had developeda leading position in test and measurement equip-ment, designing and producing largely stand-aloneproducts. Much of HP’s success in innovation wascredited to a structure of decentralization, with 45small, autonomous divisions crafted around spe-cific products. Each division maintained controlover marketing, production, and product design.


DOI: 10.1002/smj


The decentralized structure was accompanied bya culture of cooperation and autonomy identifiedas the ‘HP Way.’ However, in recent years, HPhad become an important player in the comput-ing industry which, for the first time in its history,demanded a significantly higher level of integra-tion across divisional products and services.

A move toward centralization (1982 to 1988)

Despite HP’s remarkable historic success in inno-vation, by the early-1980s, both external observersand company management noted significant prob-lems. CEO Young stated the biggest challenge atHP was finding a way to balance the organizationby ‘orchestrat(ing) the divisions and provid(ing)a strategic glue and direction for the computereffort, while keeping the work units small’ (Busi-

nessWeek, 1982: 73). Yet, as HP’s product port-folio increasingly drifted away from instrumentsand into computing and software, customers beganto complain about compatibility across HP prod-ucts and increasingly demanded integrated solu-tions rather than stand-alone components. Analystscritiqued HP for having poor ‘integration of instru-ments and computer sales efforts’ (Muratore, 1984:29). For the first time in its history, HP facedstrong pressure to coordinate designs, products,and marketing efforts across divisions—all keysteps in efficiently exploiting existing technolog-ical strength.

Starting in 1983, HP made a series of movesto remedy these problems and thereby achievegreater balance. These structural shifts consoli-dated product divisions around shared customerlines. In late 1983, HP began organizing disparatecomputer groups under a common umbrella toimprove the coordination and efficiency of market-ing, manufacturing, and engineering. By July 1984,HP reorganized the, ‘previous product-group-basedorganization into four major sectors, includingone devoted entirely to marketing and selling’(Hewlett-Packard, 1984: 5) to provide integrationin both sales and product development efforts.In addition, HP stripped away up to 40 percentof each division’s marketing dollars and placedthese in a centralized marketing group that over-saw strategic marketing for the entire ComputerGroup. Consolidation continued in January 1985,with the grouping of three separate units into a sin-gle Manufacturing, Medical and Analytical Group.At the conclusion of this centralization phase, the

historic, decentralized HP had been dramaticallytransformed. Autonomy within the product divi-sions had been significantly curtailed. Product divi-sions had been collapsed and most other divisionswere clustered under larger structural umbrellas.Sales and marketing had been significantly central-ized and organized by customer categories ratherthan product divisions.

The reorganization yielded positive performancebenefits over the next several years. HP becamemore skilled in delivering integrative customersolutions as evidenced by products that bridgedhistoric computer and instrument divides. Man-agers agreed on common standards and platforms,such as RISC technology and UNIX, around whichproduct and software were built, producing a moreunified and integrated product line. Analyst com-ments reflected the exploitation benefits of cen-tralization, noting an increased ability to see cus-tomers’ problems across products and market HP’soverall capabilities. Surprisingly, given HP’s longtradition of autonomy and decentralization, therewas very little concern expressed by securities ana-lysts, the business press, or management itself thatthe shift toward centralization might undermineHP’s innovativeness. Instead, the sentiment wasthat HP had gained greater balance, as reflected ina 1984 Prudential analyst report that said:

‘The independence and limited perspective of

the individual divisions (before reorganization),

however, has tended to produce products not

integrated from either a product or a market

perspective. The reorganization regroups these

divisions, but it does not change the basic con-

cept of independent units’ (Muratore, 1984: 30).

Finally, the structural shift seemed to initiallyhave a positive impact on HP’s operating perfor-mance, infusing, in particular, a greater balance ofemphasis on exploration and exploitation. Duringthe next several years, HP’s stock price outper-formed the S&P 500.

A return to decentralization (1988 to 1994)

Despite what were likely intentions of sustainedbalance, by the late-1980s, HP had becomemarkedly more bureaucratic, with innovation sig-nificantly curtailed. HP’s share price had alsoslumped. Analysts critiqued HP for its lack ofinnovation and lateness to market, particularly in


DOI: 10.1002/smj


contrast to smaller rivals such as Sun Microsys-tems. Analysts credited delays in the release ofboth the Model 950 microcomputer and the Spec-trum Series to the growing bureaucracy. The Wall

Street Journal characterized the organizationalenvironment as one in which ‘even the simplestdecision was sent up the management chain, some-times all the way to Mr. Young’ (Yoder, 1991:1). By the late-1980s, the founders, David Packardand William Hewlett, along with CEO Young wereconvinced the centralized structure was constrain-ing innovation.

Beginning in 1988 and then abruptly in 1990,HP took aggressive steps to increase innova-tion by granting divisions greater autonomy. InApril 1988, HP separated the Computer Periph-erals Group into three distinctly separate productgroups for Publishing Products, Hardcopy Tech-nology, and Mass Storage. HP split the Printer andDisk Drive groups away from Computer ProductsGroup in 1990, to give the computer group morefreedom to concentrate on microcomputers andworkstations, a movement back toward a decen-tralized product structure. In an attempt to furtherenhance innovation through autonomy, divisionalheads, such as Lew Platt, were moved away fromcorporate headquarters and given significant auton-omy to work on major projects without havingto consult CEO Young. In 1991, HP did mergeseveral divisions under a Personal ComputationBusiness (PCB), but the new division gained sig-nificant autonomy from corporate headquarters. In1992, newly promoted CEO Lew Platt further splitout the Computer Systems Organization, eliminat-ing a layer of management and breaking up theoperations of its networked systems group intofour independent units oriented around products,all with autonomy to make their own strategic andmarketing decisions. Thus, by 1992, HP movedits formal structure significantly back toward itsdecentralized roots in an effort to enhance explo-ration.

Analysts noted improved performance in HPdivisions, expressing new confidence in HP’scapacity to generate innovative products and rev-enue growth. By June 1991, Smith Barney ana-lysts were lauding HP for ‘the removal of layersof decision makers and the exodus of ‘manage-ment by committee,” and its return to, ‘its moreentrepreneurial roots” (Wang, 1991a: 5). Analystsand HP executives also expressed a clear senti-ment that HP was balanced again by being both

integrated and innovative, able to both diversifyinto new business and maintain cost control (Wang,1991b). Prudential analysts commented on HP’snewfound balance in exploration and exploitationand its capacity to deliver earnings growth throughboth revenue growth led by products and signifi-cant cost controls (Conigliaro, 1991). The 1992Annual Report also spoke of an improved abil-ity to ‘balance responsiveness with the steadfastpursuit of excellence’ (1992: 15). Favorable com-ments by analysts continued into 1993 and 1994;although, the focus was less on balance and moreon renewed innovation through decentralization.Analysts at Smith Barney felt the previous moveto decentralize had transformed HP into:

‘A significantly more nimble organization and

the improvement or increase in new products

as a percent of total received. These attributes

are consistent with the company’s history but

have perhaps been raised to a new level of

proficiency’ (Wang, 1995: 5).

1995 to 1998: movement back toward

centralization

The greater balance between exploration andexploitation was not long lived. By 1995, theemphasis again shifted toward exploitation throughgreater coordination across businesses. Analystshighlighted the need for integrated customer solu-tions due to the convergence of multiple computingsystems into a single information technology envi-ronment (Neff, Wu, and Bean, 1999). Customers,popular press, and analysts alike criticized HP for‘functioning almost like separate computer compa-nies under the same brand name’ (Vijayan, 1998).

In a series of moves starting in 1995, HP for-mally adopted a more centralized structure. HP’sleadership felt this centralization effort came atthe right time to pull together the benefits ofautonomous divisions and their resulting techno-logical expertise. In August 1995, HP broughttogether the Computer Products Organization, theComputer Systems Organization, and the World-wide Customer Support Operations into a collab-orative Computer Organization Group. The 1995Annual Report articulated clearly that the aim ofthis centralization was to find balance betweeninnovation and integration:

‘By unifying our computer activities, we can

leverage strengths and deliver the integrated


DOI: 10.1002/smj


solutions customers are looking for while main-

taining the benefits of focused businesses’

(Hewlett-Packard, 1995: 3).

By 1997, despite increased centralization, ana-lysts still highlighted HP’s capacity for effectiveexploration through rapid entry into new markets(Jones, Jones, and McBride, 1997).

In 1997, HP further centralized by reorganizingthe sales force around customer groups, rather thanproducts, with the aim of creating a sales forcewith broad product knowledge and the capacityto create solutions. In early 1998, HP merged theTest and Measurement Group and MeasurementSystems Organization into a combined Measure-ment Organization, a move that pulled togetherall product groups associated with HP Electronics,Test and Measurement, Medical Products, Chem-ical Analysis, and Components Groups. To facil-itate additional coordination, Lew Platt created aformal executive committee to address company-wide strategic issues including business evaluation,portfolio management, structure and governance,and management and employee practices. Late in1998, HP combined the Enterprise Systems andthe Software and Services Group to redevelop theUNIX solutions business and allow for the pre-sentation of ‘one face to the customer’ (Vijayan,1998). In the 1998 Annual Report, HP execu-tives claimed a degree of balance, suggesting thatthey now operated a ‘faster, more competitivecompany, with an improved product and servicesoffering, greater ability to deliver solutions, anda more focused organizational structure’ (Hewlett-Packard, 1998: 2)

HP divides and decentralizes again (1998 to

1999)

Despite positive exploitation benefits from central-ization, signs of significant imbalance and per-formance decline appeared by 1999. Analystsagain labeled HP as ‘big, bloated, and bureau-cratic’ (Lambeth, 1999). The integrated divisionsand bureaucratic decision-making process shelvedgood ideas across the organization. HP began toexperience poor performance in test and mea-surement and began losing market share in otherdivisions to competitors pushing more innovativeproducts. HP began to be viewed as lagging ininnovation, and its stock price dropped in late 1998as it struggled to regain market share and adjust to

the falling prices in the semiconductor, PC, andprinter industries. News media and analysts alsosuggested that the individual business units, espe-cially Test and Measurement, were unable to getthe attention that they needed in the integratedstructure.

From late-1998 through 1999, Platt made aseries of decisions to increase innovation throughgreater autonomy of the divisions. Moreover, inMarch 1999, to further increase autonomy, HPspun off its testing, medical, and chemical productsinto the new company, Agilent. By April 1999,HP was segmented into four distinct businessunits (Enterprise Computing Solutions, ComputerProducts, Inkjet Imaging Solutions, and LaserJetImaging Systems) oriented primarily around prod-ucts, each with a CEO and the latitude to for-mulate business and partner strategies independentof one another and corporate approval. The 1999Annual Report suggested HP’s desire was for thenew structure to encourage flexibility and inno-vation. Moreover, language in the annual reporthad shifted away from integration and customersolutions—key elements of a focus on exploita-tion—to a renewed emphasis on flexibility andinnovation.

The move to spin-out the Test and Measure-ment division was met with acclaim by analystswho cited the minimal attention HP devoted tothe division. Analysts at Bears Stearns lauded theannounced spin-off as a way to give the divisionincreased attention (Neff et al., 1999). Analysts atDeutche Bank felt the move to spin out Agilentwould ‘provide increased visibility into Hewlett-Packard’s core business’ (Rueppel et al., 1999:1). The move toward product divisions was alsoviewed positively by analysts as a device thatallowed HP to capture intellectual property moreeffectively than the previous organizational struc-ture. By late 1999, the stock was again outperform-ing the S&P 500.

Fiorina’s march toward centralization (1999 to

2005)

When new CEO Carly Fiorina arrived in 1999,analysts were already demanding significant reor-ganization, emphasizing coordination and exploita-tion across divisions. Fiorina’s assessment of HPwas an organization overrun with internal compe-tition and duplicated effort—and an organizationwith which it was difficult for customers to do


DOI: 10.1002/smj


business. Analysts at Prudential pointed to HP’sgrowth in size causing ‘inefficiencies across theorganization’ (Alexy and Park, 2000: 6). They notethat HP’s ‘lack of focus had created opportuni-ties for its competitors’ (Alexy and Park, 2000: 6).Internally, managers feared that HP had missedsignificant opportunities in the Internet.

Starting in 1999, Fiorina made a series of signif-icant moves toward centralization. She saw thesemoves as an effort to rather explicitly undo or‘offset the patterns of the previous organization’(Fiorina, 2006: 191). Regarding the benefits anddrawbacks of the previous structure, Fiorina sug-gested to shareholders in 2000 in rather explicitterms that she was trying to balance the organi-zation—to address exploitation objectives that thedecentralized structure poorly achieved:

‘Our highly decentralized structure has enabled

us to move quickly. In an Internet age, it’s clear

that we must maintain this speed. At the same

time, our value to customers lies not only in indi-

vidual products, but in delivering total solutions.

To achieve this, we are focusing on strategic

opportunities that fall between our traditional

businesses, or cross the lines of one of more of

our business’ (Hewlett-Packard, 2000: 11).

HP adopted a ‘front-back’ organizational struc-ture with two customer-facing sales organiza-tions and two focused on production and productdevelopment. In the 2000 Annual Report, Fiorinaemphasized the need to ‘quickly. . . turn inven-tive ideas into world-class technology solutions’(Hewlett-Packard, 2000: 11), which she felt wasachievable only through greater centralization. In2001, HP completed a dramatic acquisition ofCompaq followed by its direct integration into thiscentralized structure. In May 2004, HP created acomplete Customer Sales Organization to furtherintegrate the sales and marketing efforts across theentire organization rather than by specific productgroups.

Analysts praised Fiorina for initiating muchneeded changes and for attempting to integratethe highly decentralized organization. They notedHP’s new and balanced capacity to sustain ‘earn-ings growth while re-engineering and streamliningoperations’ (Young, 1999: 2). New products andcustomer solutions and strategies demonstrated thebenefits of HP’s new integration. For instance, HPunveiled a digital entertainment strategy that pulled

together divergent strengths across the firm. HPadditionally realized significant savings in infras-tructure costs from the shift toward centralization.

Hurd decentralizes (2005 to 2008)

Despite benefits from centralization, within a fewyears analysts were commenting on HP’s prob-lems with innovation, noting that HP was failingto ‘grow at the same rate as its more narrowlyfocused competitors’ (Altherr and Haneman, 2000:3). In 2004, interim CEO Robert Wayman alsorecognized the need to shift from a focus on ‘con-solidation, integration, and cost cutting’ to a focuson ‘accelerating profitable growth, driving lever-age across HP’s product portfolio and extendingleadership into new categories’ (Hewlett-Packard2004: 1). Moreover, the media began to call foreither greater decentralization or the actual breakupof HP as a solution, including either the sale of thePC or the printers business. HP had become slowand bureaucratic and insiders to the HP situationfelt Fiorina’s centralization considerably slowedorganizational processes that were central to explo-ration. HP’s board began pushing for decentraliza-tion efforts. However, despite HP’s prior history ofstructural change by each of the prior two CEOs,Fiorina actively resisted board pressure to change;she insisted on maintaining the current centralizedstate (Tam, 2005).

At a board meeting in mid-January 2005, HPboard members decided on a management reor-ganization plan despite Fiorina’s objections. Notsurprisingly, she was replaced in early Februarywith new CEO, Mark Hurd. The 2005 AnnualReport addressed the bureaucracy that Hurd dis-covered upon taking the helm:

‘In a few cases there were nine layers of man-

agement between the CEO and a customer. . .

Some business divisions had less than 30 per-

cent of their budgets directly under their control

because of the way costs were allocated. When

this kind of organizational design is applied to a

company of HP’s scale, it represents the under-

pinnings of slow decision-making and confusion

in terms of accountability’ (Hewlett-Packard,2005: 2).

Almost immediately, CEO Mark Hurd stream-lined staff and significantly decentralized to rekin-dle innovation. In July 2005, Hurd dismantled the


DOI: 10.1002/smj


Customer Solutions Group and divided respon-sibility to each of the specific product groups,putting product groups in charge of their own mar-keting, product development, and sales.

As a result, HP’s profit and innovationrebounded. Analysts applauded the shift back todecentralization, suggesting the new structure‘reinforces its long-standing philosophy on thestrength of the HP portfolio’ (McCullough, 2005:1). Moreover, they again spoke of regained bal-ance, in which HP had strengthened ‘its abil-ity to balance growth and profitability’ (Sempleand Kelleher, 2006: 1). Disbanding the central-ized sales and marketing organization, the Cus-tomer Solutions Organization, has been viewed asinstrumental in HP’s resurgence in the PC market,including regaining the No. 1 market share fromDell.

Assessment

Over 25 years, HP pursued a rather remarkablepattern of reorganizations, each structural changeseemingly focused on correcting the outcomes ofthe past structure. When HP’s autonomous divi-sions produced excessive redundancy and incon-sistency, management formally reorganized, cen-tralizing key cross-divisional activities and collect-ing divisions under larger organizational umbrel-las that could fuel integration and consistency aswell as lower costs. Such efforts were directlyfocused on more effectively exploiting existingbusinesses and products—reducing inefficiencies,eliminating redundancies, and combining and inte-grating products to satisfy customers. These formalorganizational structures rapidly delivered consis-tency, integration, and solution selling. However,they also generated bureaucracy that increasinglyattenuated and delayed innovativeness and explo-ration. As these bureaucratic costs increased withtime, management sought increased exploration byformally restructuring. Divisions were separatedand centralized activities were redistributed backinto autonomous product divisions. The divisionsquickly responded with increased exploration ush-ering in a period of greater balance between explo-ration and exploitation, but eventually undermin-ing the level of exploitation again.

While HP leadership frequently spoke of seekingto balance or rebalance the organization, essen-tially expressing a desire for ambidexterity, itnonetheless frequently engaged in structural

change—change that shifted the organizationtoward either greater exploration or greaterexploitation. HP’s inability to achieve static bal-ance with high levels of both exploration andexploitation led it to vacillate. Management cen-tralized when too little exploitation (in terms ofcoordination and integration) was identified. As aresult, the firm enjoyed episodes where the for-mal structure delivered exploitation and the inertialinformal structure delivered exploration. Managersdecentralized when too little exploration (in termsof flexibility and autonomy) was realized. Thatsaid, we also found evidence that senior leadershipattempted to balance within each epoch, whichmay have elevated performance. Nonetheless, suchbalancing efforts eventually were insufficient and,repeatedly, an organizational change led to a newepoch. While HP’s stock price performance driftedup and down in these 25 years, the stock signif-icantly outperformed all market indices over thistime period and HP emerged as the world’s largestglobal IT company. Note also that this case illus-tration highlights an important boundary conditionfor both perspectives. Prior to the early 1980s,decentralization and the primary pursuit of explo-ration was a stable equilibrium for HP. It was therapid expansion into computing that elevated theneed for integration across divisions that triggeredefforts to achieve exploration and exploitation atthe corporate level.

USA Today Online

USA Today’s Internet business emerged during a15-year time period—a time when growth of theInternet caused a fundamental shift in the formatthrough which news is delivered. Our discussionof this case is based on several sources. For theearly periods, we draw heavily from existing casestudies and academic articles featuring this episode(O’Reilly and Tushman, 2004, 2008; Tushman,Roberts, and Kiron, 2001a, 2001b). We augmentthe study with data collected from a variety ofsecondary news reports, from annual reports ofUSA Today’s parent company, Gannett Company,Inc., and from an interview with a USA Today

executive. Using these data sources, we constructa focused history of USA Today, highlighting itssuccessful efforts to explore the online business,while in the process exploit the existing printbusiness. Thus, while our discussion of HP focusedon its aggregate, corporate-level organization, this


DOI: 10.1002/smj


case highlights the efforts to both explore andexploit within a focused business segment: onlinenews.

Pre-1995 integration

USA Today (USAT) launched in 1982 with the goalof creating a national daily newspaper. With a keystrategy of providing the newspaper free of chargeat hotels around the country, the newspaper experi-enced 10 years of financial losses before the paperbegan earning profits. Tom Curley, the young man-ager who coordinated the original market researchthat led to the launching of the newspaper, becamepresident in 1986, with responsibility for circula-tion. In 1991, he added the title of publisher andassumed responsibility for the newsroom in 1994.In 1995, he expanded his responsibilities to includeUSAT’s international edition and Internet business.

Efforts to explore Internet expansion opportu-nities began within four years of the launch ofUSA Today. Beginning in 1986, USAT launchedseveral projects that experimented with a bulletinboard service that allowed users to play chessand peruse an online sports center. These projectsevolved into a small organization embedded withinUSAT called the USA TODAY Information Net-work. While these online projects exploited someof USAT’s existing sports content and providedsome valuable Internet experience, by the mid-1990s nothing in the Internet realm had emergedinto a substantial source of revenue growth forUSAT.

A stand-alone division (1995 to 2000)

By 1995, global activity on the Internet wasexploding. USAT recognized that its own effortsto explore in this space were failing or insuffi-cient. Therefore, they closed down the Informa-tion Network in March 1995 and launched a newindependent division, USAToday.com (Online), inApril 1995. In forming this new division, Curleymade a substantial investment and granted exten-sive autonomy to its general manager, LorraineCichowski, to run the business autonomously as a‘stand-alone’ operation entirely separate from theUSAT newsroom operation and culture (Tushmanet al., 2001a).

The stand-alone structure of the Online divisionled to dramatic experimentation, rapid innovation,and a business model that increasingly diverged

from the print business. The original Online modelwas positioned as a dial-up online newspaperproviding up-to-the-minute sports statistics. Cus-tomers were charged a subscription of $12.95 permonth for unlimited access to the online newspaperand three hours of access to the broader Inter-net. However, this model was quickly scrappedin favor of an advertising-based revenue model.A month later, Online introduced another majorchange by shifting from a daily published edition,consistent with the print business format, to a for-mat of continuous updating and breaking news.This strategy shift meant Cichowski had to relymostly on wire services for breaking news. Theformat also meant vastly different work habitsand culture as compared to the USAT print news-room. Cichowski also chose to hire staff with adistinctly different and younger demographic pro-file than those already within the print business.Moreover, incentives and the culture were strik-ingly different from the USAT print newsroom, andthey were shaped to support the continuous updateformat. Furthermore, decision processes and struc-tures discouraged coordination between USAT andOnline. The two newsrooms were physically sepa-rated on different floors. The autonomous divisionbegan to see success with revenue increasing at adouble-digit pace in 1996 and 1997. The organiza-tion turned a profit in the fall of 1998, becomingthe first newspaper-owned online news service toachieve this distinction.

Online’s autonomy and success came with apredictable cost, however. By 1998, a growingconflict emerged between Online and the printnewsroom. Online found news from the USATnewsroom to be increasingly useless, as the writ-ers worked to the one daily deadline for the printnewspaper. As a result, by 1998 only 20 percentof Online’s news came from the print newsroom.Conflict also arose since the accuracy of Onlinestories occasionally slipped below USAT stan-dards, thus undermining what USAT saw as theirkey brand attributes: fairness, accuracy, and trust.Conflict also arose over access to news sourcesand ownership of stories. Print reporters were con-cerned that news sources might stop returningcalls if they had already returned calls to Online’sreporters. Even more pressing was the concern thatthe Online business essentially leaked key storiesto their print competitors, who could follow theOnline Web site and with minor source check-ing, and then include similar stories in their print


DOI: 10.1002/smj


news publications. All in all, USAT print person-nel were concerned that online media would soonput them out of business. The sense of competitionbetween the two autonomous units was summed upby an Online staffer who said ‘there was almosta sense of ‘watch what we can do without you(the paper).’ We felt like the paper people thoughtwe were a joke, so this didn’t help matters much’(Tushman et al., 2001a: 10). In concert with dete-rioration of relationships between the two news-rooms, Cichowski’s relationship with Curley hadalso deteriorated by 1998 because of the tensionsbetween the two divergent arms of USAT. In partto resolve the growing issues, in 2000 Cichowskipushed for further autonomy by spinning off theonline business.

2000 to 2005: selective integration

Curley, however, perceived a larger opportunityin moving the opposite direction—a direction thatwould exploit significant synergies across the busi-ness units by promoting the integration of the printand Online groups in which he termed a ‘network’strategy. With this strategy, he planned to leveragenews gathering/editorial capabilities across onlineand print newsrooms, as well as Gannett’s networkof television stations. His initial efforts to promotethis strategy involved management changes, butthese efforts proved insufficient and in spring of2000, Online continued to operate as an indepen-dent business unit.

With the view that ‘separateness equals death’(Tushman et al., 2001a: 13), Curley replacedCichowski and 40 percent of the senior manage-ment team. While the Online newsroom remainedphysically separated from the print newsroom,Curley insisted on weekly editorial meetings thatforced both staffs to interact and integrate. He alsoshifted senior team incentives so that bonus incen-tives were based on the combined performanceof Online and print growth. Curley additionallylaunched USA LIVE, which offered news throughGannett-owned television stations.

Very quickly, the online and print newsrooms,along with television, began to share stories andbetter coordinate. Jeff Weber, Cichowski’s replace-ment, also took steps to shift the brand iden-tity of Online closer to the newspaper’s brandimage. He met twice a month with the com-bined organization’s publisher and editor to ‘figureout the guidelines for how the newspaper and

Online should play in the same sandbox’ (Tushmanet al., 2001b: 2). Online editors started attendingdaily newsroom’s meetings. The immediate resultswere impressive. The average unique monthly vis-itors increased from 300,000 a month under toCichowski to 400,000 under Weber, with visitorsspending more than 18 minutes per visit versus 11minutes previously. However, while the numberof visitors grew, USAToday.com nonetheless lostmarket share.

Post-2005: complete integration

The shift toward closer integration of Onlineand print continued over the years from 2000to 2005, culminating with Jeff Weber announc-ing in December 2005 that the print and onlinenewsrooms would structurally merge. In the pro-cess, any remaining autonomy between the twogroups was eliminated. In announcing the shift,Craig Moon, president and publisher of USA Today

stated that ‘what I’m basically here to tell youtoday is that there no longer is going to be a dot-com newsroom. There no longer is going to bea print newsroom’ (USA Today, 2005). By 2005,the opportunities to exploit cost savings throughintegration simply overwhelmed any gains fromseparation. USAT print and online had become asingle integrated unit with no separate accountingsystems or structures of any form.

Assessment

Unlike the focus on an entire organization inthe case of HP, our focus on USA Today high-lights how a specific business unit was struc-turally governed over time. Over a 19-year period,we see a pattern of a new business beginningfully integrated, then radically becoming decen-tralized and autonomous to facilitate exploration,then selectively reintegrated, only to fully rein-tegrate as efforts shifted to exploitation. Prior to1995, USAT’s attempts to develop an Internet busi-ness were simply embedded within USAT. Theseefforts were rather limited in their success untila highly autonomous USAToday.com was formedunder Cichowski and made independent from anequally autonomous USAT. With this decentral-ized, autonomous structure, exploration flourished.Yet predictably, a lack of cross-unit coordinationwas problematic, prompting pressure to restruc-ture. We see evidence of this pressure in the tenor


DOI: 10.1002/smj


of interactions between Online and USAT from1995 to 2000. After separating USA Today Onlinein 1995, there was an extended period of timein which the business seemed to both simultane-ously explore new innovations and exploit exist-ing resources in the print business. For example,Online experimented with a sports portal offeringup-to-the-minute sports statistics and other contentfrom the newsroom for an individual subscriptionfee. However, when this approach proved insuf-ficient, Online quickly moved to exploring otheroptions irrespective of the effect this new strat-egy had on USAT—options that were much lessexploitative of the print business resources. Suchexploration led to increasing frictions between thetwo autonomous units in addition to redundantnews generation efforts.

As the cost of this failure to coordinate acrossbusinesses increased, consistent with organiza-tional vacillation, management began selectivelyrestructuring the formal organization to gain thegreater coordination and integration, while stillmaintaining USAT Online as a separate business.In theory, a formal reorganization would shiftgoals, reporting relationships, decision rights, andincentives in a consistent way. We see just thisin 2000 as Curley’s structural changes involvedreplacing recalcitrant staff, directing the new staffto hold weekly meetings, and introducing an incen-tive structure based on group performance insteadof individual performance. These selective effortsat integration are consistent with the organiza-tional ambidexterity argument and appear to haveenhanced their capacity to exploit across thesetwo distinct business units. However, eventuallyin 2005, management determined that this selec-tive integration was insufficient and the Onlineand print newsrooms became fully integrated (USA

Today Staff, 2005). We view this pattern of inte-gration, separation, and reintegration as suggestiveof organizational vacillation.

It is clear from Curley’s own assessment thatthe pattern of first exploration through autonomyand structural separation, followed by a focuson exploitation and economizing through activeintegration, was critical to his success in achievingambidexterity. In reflecting on the business, Curleycommented:

‘Originally, when Online launched, they did

need some space, some separateness to figure

out who they were. I’m not sure that they would

have embraced breaking news as early and as

fully as they did if they’d been fully integrated

with the paper. And, you can’t just blame them

for this separateness—the first four years when

they were up and running, people in the news-

room didn’t even have a PC on their desk to

access the Web. But, by 1999, it was pretty clear

that greater integration was required’ (Tushmanet al., 2001b: 3).

Note the clear sentiment in Curley’s commentsthat decentralization and autonomy were essentialto their initial success in innovation and explo-ration. When Curley formally reorganized in 2000,Online and USAT shifted their focus to exploit-ing synergistic opportunities. The important obser-vation is that achieving exploration demanded astructural change, as did achieving exploitation.Thus, while USAT certainly pushed to organize forbalance with what appears to be some extendedsuccess, the performance of the online businessin exploring and exploiting was also significantlydriven by the dynamics of structural change towardboth integration and separation.

In fact, USAT’s use of autonomous units to cre-ate space for an exploration outside of traditionalcompetencies was nothing new for the organiza-tion. In the past, USAT had used a similar strategywith several ventures, including their telemarket-ing group, USA Today TV, and Sports Weekly.With all these projects, management granted acertain level of autonomy to facilitate its devel-opment. Then, depending on the success of thisproject and whether or not it had developed syn-ergies with the larger organization, these projectswere reintegrated with USAT, spun out, or simplyshut down.

DISCUSSION

In these two cases, we find support for patternsof structural vacillation playing a vital role in cor-porate efforts to develop exploration and exploita-tion, together resulting in high long run perfor-mance. Specifically, we find that in both HP andUSA Today, efforts to structurally segment theorganization and build in cross-unit integrationthrough leadership initiatives were not sustain-able over time, leading us to infer that organiza-tional ambidexterity was insufficient in the longrun at delivering high performance. We also find


DOI: 10.1002/smj


that repeated structural modulation correspondedto periods of capability in both exploration andexploitation and overall organizational success.While two case studies are insufficient to providestrong empirical support, each illustrates the roleof organizational vacillation in delivering high per-formance over the long run through the provisionof both exploration and exploitation, even thoughmuch of the time the levels of each were unbal-anced. At HP, the observed exploration, exploita-tion, and superior organizational performance overa 25-year period was achieved through a pat-tern of organizational vacillation. Note that thesestructural shifts targeted increased balance, butachieved it only temporarily with epochs of struc-tural change. In the case of USA Today, while wefind evidence that managers attempted to success-fully implement an organizationally ambidextrousapproach in 2000, an extension of the time horizonboth before and after 2000 reveals that the attemptwas nonetheless embedded in a pattern of orga-nizational vacillation in which online efforts werefirst centralized within USAT, then decentralized,and subsequently recentralized.

While our case studies highlight vacillation’srole in generating high long run performance, wealso find evidence that ambidexterity as a leader-ship initiative yields performance benefits withinthe epochs of organizational vacillation. Initiativesto build coordination across a segmented organi-zation with cross-unit teams, linking mechanisms,and a unifying culture and team processes—facetsO’Reilly and Tushman (2008) describe in their def-inition of ambidexterity—appear to be beneficialto performance at least in the short run in the midstof transitions from one organizational structure toanother. Our finding is consistent with the cross-sectional empirical results that are suggestive of apositive impact of organizational ambidexterity onvarious performance metrics (He and Wong, 2004;O’Reilly and Tushman, 2008), while additionallysuggesting that research with longer time horizonsis essential to evaluate vacillation’s role in shapingthese outcome over time.

Our discoveries from the two case studies aresuggestive of a theoretical reconciliation betweenthe organizational ambidexterity and organizationalvacillation perspectives. In essence, our case stud-ies indicate that vacillation, within the boundaryconditions described below, can yield higher longrun performance. Yet, those managers who engagein organizational ambidexterity within each epoch

of vacillation can deliver even higher long runperformance as their efforts attempt to extend theperiod of balance as well as increase the amountof exploration and exploitation in the short run.Therefore, we conclude that the two perspectivespresent some capacity for integration. For instance,we believe the theory of organizational vacilla-tion can be extended to acknowledge that effort byleaders to achieve ambidexterity can shift upwardthe organizational vacillation performance pathdescribed in Figure 3. That said, more theory andempirical work is needed to understand the precisepath of the curve and how these efforts impact suchfactors like the frequency of vacillation. Advanc-ing this combined theory not only provides a wayto reconcile the two perspectives, but also pro-vides value by building a more comprehensive andcumulative theory of organization that can benefitleaders.

As noted earlier, our case studies reflect differingunits of analysis. The HP illustration employs thecorporation as the unit of analysis, whereas theUSA Today illustration occurred at the businessunit level. While the two units of analysis differ,both units illustrate vacillation between discretestructural alternatives due to the facilitating roleof organizational inertia and illustrate managersattempting to manage ambidextrously. That said,different units of analysis might yield differencesin the ideal rate of vacillation and the natureof structural choices. We also envision that thechallenges of ambidexterity may differ because ofthe differing authority structures found in corporateversus business unit levels of analysis.

The intensity of these differences may reflectdifferences in the intensity of the selection envi-ronment (i.e., the pressure to perform). Williamson(1985) draws this distinction and suggests that dif-ferences in selection environments serve to influ-ence the time span by which certain organizationalprocesses take place. The selection environmentsbetween Hewlett-Packard, a publicly traded com-pany, and USA Today, a business unit within thelarger Gannett Corporation, are likely to differ.The leaders at Hewlett-Packard arguably operatedwithin a relatively strong selection environment asa result of its public ownership. Failure to performin the short run quickly aggregates pressure forchange from external shareholders and increasesthe likelihood of strategic or structural adjustment.By contrast, USA Today is a subunit of the larger


DOI: 10.1002/smj


Gannett Corporation. We suggest that embedded-ness within a larger organization may, at times,insulate the business unit from selection pressures.The notion that selection pressures may vary bythe unit of analysis represents an opportunity toexpand and refine vacillation theory.

The primary alternative explanation for finiteepisodes of organizational ambidexterity and orga-nizational vacillation is that the environment ischanging. If the environment is changing, thencontingency theory would predict that so tooshould an organization’s structure. Over the longtime horizons studied, the environment wasundoubtedly changing in both cases. However, fora contingency explanation to explain the observedphenomena, the environment would need to changeso as to exogenously vacillate in its need for explo-ration and exploitation in a pattern consistent withthe observed patterns in structural change. Thisseems quite unlikely in either illustration. Overthe 25 years of the study, the computer indus-try tended to move in the direction of proprietarysystems to open standards, from mainframes tonetworks of microcomputers, from large corporatecustomers to a diversity of business and consumercustomers. These trends did not reverse. Similarly,trends in the news industry have been unidirec-tionally driven by the Internet and related tech-nologies, which likewise have not reversed. Bothenvironments seem to have consistently demandedexploration and exploitation.

Both theories of organizational ambidexterityand organizational vacillation are valid within a setof boundary conditions. Key conditions are high-lighted here and build on those identified by Nick-erson and Zenger (2002) and Gulati and Puranam(2009). For instance, both theories rely on man-agers desiring multiple performance dimensions,each of which is best achieved by a rather discreteorganizational form. However, in those environ-ments where only exploitation is required, man-agers can simply centralize or otherwise structureto exploit with little concern for generating bal-ance through either vacillation or ambidexterity.A second and related boundary condition is thatboth theories assume that complementarities existamong organizational design elements. If this werenot the case, exploration and exploitation could beorganized quite separately in a dual structure with-out any conflict or design tension. Both theoriesalso rely on the condition of a selection environ-ment that is not so hypercompetitive that firms

pursuing exploration to any degree are selectedout. In this type of extreme selection environment,the benefits of vacillation would not have time toaccrue. Moreover, for ambidexterity, any reduc-tion in exploitation caused by pursuing explorationwould result in the firm being selected out. Thetheories do differ in one important set of boundaryconditions. Ambidexterity suggests that the bene-fits of crafting somewhat conflicting organizationalstructures exceed the costs of doing so. In contrast,organizational vacillation assumes that the iner-tia in the informal organization causes explorationand exploitation to differentially wax and wane inresponse to structural shifts and, as a result, gener-ates benefits from vacillation that over time exceedthe cost of structural change. We assume all ofthese boundary conditions hold in our case studies.

Sailing into the wind

To help convey the findings of our research tomanagers, we believe a metaphor may be partic-ularly helpful. The challenge of achieving highlevels of both exploration and exploitation par-allels the challenge sailors face in attempting tosail into the wind. The sailor knows that layinga course directly into the wind not only slowsprogress, but ensures regress. In contrast, skill-fully configuring the boat’s mainsail, foresail, andrudder to set a course 40 degrees off wind can gen-erate tremendous speed. While traveling on this‘close-hauled’ course can maintain a fast speed,the boat is not sailing directly toward the desireddestination. Indeed, sailing for too long on thiscourse takes the boat far from its desired destina-tion. Hence, the sailor comes about, reconfiguresthe sails and rudder, and sets a course 40 degreesoff wind. While each course correction or ‘tack’imposes a loss in forward momentum, the skillfulsailors masters these reconfigurations so as to min-imize momentum loss and enable the boat to sail,on average, into the wind and achieve the objectivefaster than staying on one course for an extendingperiod of time.

In much the same way, by adopting a coherentbundle of organizational attributes, managementgenerates momentum toward increasedexploration. By adopting a distinctly different setof attributes, the organization generates momen-tum toward increased exploitation. An effort tocraft an organization that generates both in per-fect balance may simply stifle both. Inconsistent


DOI: 10.1002/smj


choices compromise the capacity of the organi-zation for movement of any type, while consis-tent choices can accelerate the pace toward eitherincreased exploration or exploitation. By vacillat-ing between consistent sets of choices and allowingthe informal organization to respond accordingly,exploitation and exploration are generated at highlevels with a varying degree of balance, dependingon time from the previous change. Nonetheless,each tack or shift in structure imposes costs onthe organization, including lost momentum thatcurtails performance, much like the overzealouscaptain that loses momentum from tacking ineffi-ciently or too often. Ambidextrous managers maybe capable of tacking with greater efficiency ormetaphorically maintaining a course fewer degreesoff wind, thereby improving organizational perfor-mance, at least on the margin. Once efficiency isachieved, a key task of the manager is to identifywhen changes in formal organization are requiredand what these changes entail. Over time, thesechoices essentially define the cadence of changeor the pace of vacillation.

Limitations and future research directions

This article is not without weaknesses. Whileour qualitative analysis of historically wide anddeep canonical cases provides the benefits of rich-ness critical to understanding the mechanisms thatdeliver ambidexterity, the method does not pro-vide for conclusive statistical assessment of thecompeting mechanisms. The limitation of archivalresearch is also visible in this project. Our data islimited by its sources, as well as self-presentationand retrospective biases. Particular care must beexercised in the interpretation of annual reportsthat are biased in presentation. Our use of analystreports, interviews, and secondary news sourcesprovides us a way around some of these prob-lems inherent in using annual reports, though thesesources are subject to limited information. Finally,even though our archival work is extensive andfrom a multitude of perspectives, we still are notable to completely and precisely describe the infor-mal organization as it follows shifts in the for-mal organization. The analysis of USA Today wasadditionally limited by the fact that it is not apublic company in and of itself, so intensive anal-ysis of management’s action was largely limited tomaterial from or about the parent company and/orsecondary news sources and academic literature.

Though the observed patterns in these two casesappear to suggest that organizational vacillationplayed a role in the development of organizationalambidexterity, other factors not accessible to usmay have influenced the organizational dynam-ics. Thus, our findings are tentative and we callfor more comprehensive and large-scale empiricalresearch.

We recommend that future research on organi-zational ambidexterity take a temporally rich per-spective on the role of strategic and organizationaldecisions. We believe organizational research isbetter informed by moving away from snapshotsof organizational strategies and, instead, explor-ing dynamics and histories. Much like the litera-ture on emergent strategies (Mintzberg, 1978), weargue that strategy (whether deliberate or emer-gent) is most clearly visible through a tempo-rally rich frame. Large-scale empirical examina-tion of the phenomenon is the next research step.Our case studies illustrate the challenges involvedin developing such a data set. Information onstructural shifts in organizations is not presentlycollected in widely available databases. Vacilla-tion is episodic, and identifying episodes requiresdetailed case studies. That said, these cases studiesshould provide guidance with respect to construct-ing case studies for other firms, as well as to codingepisodes of change. Our hope is that details foreach firm provided herein will speed coding effortsfor other firms, thereby enabling large-scale empir-ical research on vacillation and ambidexterity.

Our findings provide important implications formanagement. We argue that managers who arewilling to pursue dual capability dynamically byskillfully vacillating among bundles of comple-mentary elements are best able to achieve sus-tained high performance over time. The leaders ofambidextrous organizations must be able to under-stand which formal structural decisions facilitatethe achievement of an ambidextrous informal orga-nization and when to employ them. In this way, theambidextrous leader is much like the captain sail-ing into the wind—one who progresses toward thedestination by skillful tacking between positions.

REFERENCES

Abernathy WJ. 1978. The Productivity Dilemma: Road-block to Innovation in the Automobile Industry . JohnsHopkins University Press: Baltimore, MD.


DOI: 10.1002/smj


Alexy K, Park C. 2000. Hewlett-Packard Company,company report. Prudential Securities, Inc. 21 March.

Altherr MR, Haneman N. 2000. Hewlett-Packard Com-pany, company report. Credit Suisse First Boston Cor-poration. 15 December.

Anderson PA. 1983. Decision making by objectionand the Cuban missile crisis. Administrative ScienceQuarterly 28: 201–222.

Benner MJ, Tushman ML. 2003. Exploitation, explo-ration, and process management: the productivitydilemma revisited. Academy of Management Review28: 238–256.

Bradach J. 1997. Using the plural form in themanagement of restaurant chains. AdministrativeScience Quarterly 42: 276–303.

Burgelman RA, Meza PE. 2000. The new HP way.Stanford Graduate School of Business CaseSM-72.Stanford University.

Burgelman RA, Meza PE. 2004a. The new HP in2004 (A): Leading strategic integration. StanfordGraduate School of Business Case SM-125A. StanfordUniversity.

Burgelman RA, Meza PE. 2004b. The new HP in2004 (B): Winning in the core business. StanfordGraduate School of Business Case SM-125B. StanfordUniversity.

Burns T, Stalker GM. 1961. The Management ofInnovation . Tavistock: London, U.K.

BusinessWeek. 1982. Can John Young redesign Hewlett-Packard? 6 December.

Casadesus-Masanell R, Spulber DF. 2000. The fable ofFisher Body. Journal of Law and Economics 43:67–104.

Christensen CM. 1997. The Innovator’s Dilemma: WhenNew Technologies Cause Great Firms to Fail . HarvardBusiness School Press: Boston, MA.

Coase RH. 2000. The acquisition of Fisher Body byGeneral Motors. Journal of Law and Economics 43:15–32.

Conigliaro L. 1991. Hewlett-Packard Company, companyreport. Prudential Securities, Inc. 24 May.

Cyert RM, March JG. 1963. Behavioral Theory of theFirm . Blackwell Business: Cambridge, MA.

Doty DH, Glick WH, Huber GP. 1993. Fit, equifinality,and organizational effectiveness: a test of twoconfigurational theories. Academy of ManagementReview 36: 1196–1250.

Duncan R. 1976. The ambidextrous organization: design-ing dual structures for innovation. In The Managementof Organization (Vol. 1), Killman RH, Pondy LR,Sleven D (eds). North Holland: New York; 167–188.

Fang C, Lee J, Schilling MA. 2010. Balancing explo-ration and exploitation through structural design: theisolation of subgroups and organizational learning.Organization Science 21(3): 625–642.

Fiorina C. 2006. Tough Choices: A Memoir . PenguinGroup: New York.

Freeland RF. 2000. Creating holdup through verticalintegration: Fisher Body revisited. Journal of Law andEconomics 43(1): 33–66.

Gersick CJG. 1991. Revolutionary change theories: amulti-level explanation of the punctuated equilibrium

paradigm. Academy of Management Review 16:30–36.

Ghemawat P, Ricart JE. 1993. The organizational tensionbetween static and dynamic efficiency. StrategicManagement Journal, Winter Special Issue14: 59–73.

Gibson CB, Birkinshaw J. 2004. The antecedents,consequences, and mediating role of organizationalambidexterity. Academy of Management Journal 47:209–226.

Gulati R, Puranam P. 2009. Renewal through reorganiza-tion: the value of inconsistencies between formal andinformal organization. Organization Science 20(2):422–440.

He ZL, Wong PK. 2004. Exploration vs. exploitation:an empirical test of the ambidexterity hypothesis.Organization Science 15: 481–494.

Hewlett-Packard. 1984. Annual Report. ThompsonResearch. Available at: http://research.thomsonib.com/(accessed 25 July 2011).








Jones JJ, Jones C, McBride C. 1997. Modest fourthquarter EPS shortfall; continued growth opportunitiesin 1998. Salomon Smith Barney. 17 December.

Kadushin C, Brimm M. 1990. Why networking fails:double binds and the limitations of shadow networks.Paper presented at the International Social NetworkConference, San Diego, CA.

Kannai Y. 1980. The AELP definition of complemen-tarity and least concave utility functions. Journal ofEconomic Theory 22: 115–117.

King R. 1984. Hewlett-Packard: challenging the entre-preneurial culture. Harvard Business School Case 9-384-035. Harvard University.

Klein B. 2000. Fisher-General Motors and the nature ofthe firm. Journal of Law and Economics 42: 105–132.

Klein B, Crawford RG, Alchian AA. 1978. Verticalintegration, appropriable rents, and the competitivecontracting process. Journal of Law and Economics21: 297–326.

Lambeth J. 1999. Hewlett-Packard aims to divideand conquer. Computing. 11 March. Available at:http://www.computing.co.uk/ctg/news/1819839/


DOI: 10.1002/smj


hewlett-packard-aims-divide-conquer (accessed 13February 2012).

Lawrence PR, Lorsch JW. 1967. Organization and Envi-

ronment: Managing Differentiation and Integration .Harvard Business School Press: Boston, MA.

Lee TW, Mitchell TR, Sablynski CJ. 1999. Qualitativeresearch in organizational and vocational psychology.Journal of Vocational Behavior 55: 161–187.

Levinthal DA, March JG. 1993. The myopia of learning.Strategic Management Journal , Winter Special Issue14: 95–112.

Lubatkin MH, Simsek Z, Ling Y, Veiga JF. 2006.Ambidexterity and performance in small- to medium-sized firms: the pivotal role of top management teambehavioral integration. Journal of Management 32:646–672.

March JG. 1991. Exploration and exploitation inorganizational learning. Organization Science 2:71–87.

Martens T, Debackere K, Van Looy B. 2005. Organizingfor continuous innovation: on the sustainability of theambidextrous organization. Creativity and Innovation

Management 14: 208–221.McCullough A. 2005. Hewlett-Packard, laying down the

new law. Credit Suisse First Boston Corporation. 13December.

Milgrom P, Qian Y, Roberts J. 1991. Complementarities,momentum, and the evolution of modern manufactur-ing. American Economics Review 81: 84–88.

Milgrom P, Roberts J. 1990. Tie economics of modernmanufacturing: technology, strategy, and organization.American Economic Review 80: 511–528.

Miller D, Friesen PH. 1980. Momentum and revolutionin organizational adaptation. Academy of Management

Journal 23: 591–614.Mintzberg H. 1978. Patterns in strategy formation.

Management Science 24: 934–948.Mintzberg H. 1979. The Structuring of Organizations .

Prentice Hall: Englewood Cliffs, NJ.Muratore CL. 1984. Hewlett-Packard Company, com-

pany report. Prudential-Bache Securities. 20 Septem-ber.

Neff AJ, Wu S, Bean WB. 1999. Hewlett-Packard, themetamorphosis continues. Bear Stearns. 25 May.

Nickerson JA, Zenger TR. 2002. Being efficientlyfickle: a dynamic theory of organizational choice.Organization Science 13: 547–566.

O’Reilly CA, Tushman ML. 1997. Winning Through

Innovation: A Practical Guide to Leading Organiza-

tional Change and Renewal . Harvard Business SchoolPress: Boston, MA.

O’Reilly CA, Tushman ML. 2004. The ambidextrousorganization. Harvard Business Review 82: 74–81.

O’Reilly CA, Tushman ML. 2008. Ambidexterity asa dynamic capability: resolving the innovator’sdilemma. Research in Organizational Behavior 28:185–206.

Pinnfield LT. 1986. A field examination of perspectiveson organizational decision making. Administrative

Science Quarterly 31: 365–388.

Probst G, Raisch S. 2005. Organizational crisis: the logicof failure. Academy of Management Executive 19:90–105.

Raisch S, Birkinshaw J. 2008. Organizational ambidex-terity: antecedents, outcomes, and moderators. Journalof Management 34(3): 375–409.

Raisch S, Birkinshaw J, Probst G, Tushman ML. 2009.Organizational ambidexterity: balancing exploitationand exploration for sustained performance. Organiza-tion Science 20(4): 685–695.

Rueppel PC, Carboy ME, Raber E. 1999. Hewlett-Packard Company, company report. Deutsche Bank.24 August.

Scott WR. 1981. Organizations: Rational, Natural, andOpen Systems . Prentice Hall: Englewood Cliffs, NJ.

Semple R, Kelleher C. 2006. Hewlett-Packard, more gasin the tank? Credit Suisse First Boston Corporation. 8November.

Shrader CB, Lincoln JR, Hoffman AN. 1989. Thenetwork structures of organizations: effects oftask contingencies and distributional form. HumanRelations 42: 43–67.

Siggelkow N, Levinthal DA. 2003. Temporarily divide toconquer: centralized, decentralized, and reintegratedorganizational approaches to exploration and adapta-tion. Organization Science 14: 650–669.

Simon HA. 1955. A behavioral model of rational choice.Quarterly Journal of Economics 69: 99–118.

Smith WK, Tushman ML. 2005. Managing strategic con-tradictions: a top management model for manag-ing innovation streams. Organization Science 16:522–536.

Stevenson W. 1990. Formal structure and networksof interaction with organizations. Social ScienceResearch 19: 113–131.

Taylor A, Helfat CE. 2009. Organizational linkagesfor surviving technological change: complementaryassets, middle management, and ambidexterity.Organization Science 20(4): 718–739.

Thompson JD. 1967. Organizations in Action . McGraw-Hill: New York.

Tushman M, Smith WK, Wood RC, Westerman G,O’Reilly C. 2010. Organizational designs and inno-vation streams. Industrial and Corporate Change 19:1331–1336.

Tushman ML, Roberts MJ, Kiron D. 2001a. USA Today :Pursuing the Network Strategy (A), 9-402-010,Harvard Business School Press: Boston, MA.

Tushman ML, Roberts MJ, Kiron D. 2001b. USA Today :Pursuing the Network Strategy (B), 9-402-010,Harvard Business School Press: Boston, MA.

Tushman ML, Romanelli E. 1985. Organizational evo-lution: a metamorphosis model of convergence andreorientation. Research in Organizational Behavior 7:171–222.

USA Today. 2005. USA Today merging print, onlinenewsrooms. 12 December. Available at: http://www.usatoday.com/money/media/2005-12-12-usat-newsrooms x.htm (accessed 13 February 2012).

Vijayan J. 1998. HP reorgs to clarify message. Com-puterworld. 26 October. Available at: http://www.accessmylibrary.com/article-1G1-53134684/hp-


DOI: 10.1002/smj


reorgs-clarify-message.html (accessed 13 February2012).

Wang SF. 1991a. Hewlett-Packard Company, companyreport. Smith Barney, Harris Upham and Co. 7 March.

Wang SF. 1991b. Hewlett-Packard Company, companyreport. Smith Barney, Harris Upham and Co. 3 June.

Wang SF. 1995. Hewlett-Packard Company, companyreport. Smith Barney. 19 January.

Weick KE. 1976. Educational organizations as looselycoupled systems. Administrative Science Quarterly 21:1–19.

Williamson OE. 1985. The Economic Institutions ofCapitalism . Free Press: New York.

Williamson OE. 1991. Comparative economic organiza-tion: the analysis of discrete structural alternatives.Administrative Science Quarterly 36: 269–296.

Yoder SK. 1991. A 1990 reorganization at Hewlett-Packard already is paying off. Wall Street Journal .22 July.

Young D. 1999. Hewlett-Packard research note. PaineWebber. 1 December.

Zillmer L, Zenger TR. 2002a. Hewlett-Packard (A): Thesearch for an efficient organizational culture. OlinSchool of Business, Washington University in StLouis.

Zillmer L, Zenger TR. 2002b. Hewlett-Packard (B): Thesearch for an efficient organizational culture. OlinSchool of Business, Washington University in StLouis.

Zillmer L, Zenger TR. 2002c. Hewlett-Packard (C): Thesearch for an efficient organizational culture. OlinSchool of Business, Washington University in StLouis.

Zillmer L, Zenger TR. 2002d. Hewlett-Packard (D): Anew captain. Olin School of Business, WashingtonUniversity in St Louis.


DOI: 10.1002/smj

boumgarden nickerson zenger

Documents