Jawn Lam

Ambidextrous organization

Organizational StructuresProfessional, project, and knowledge forms13 min read

Units that exploit the existing business and units that explore new ones are kept structurally separate but joined under one senior team.

What it is

An ambidextrous organization runs two kinds of unit under one senior team: exploiting units refine the existing business, and exploring units build new technologies, products, or markets. Each has its own processes, structures, and cultures; the senior team integrates them through a shared strategic intent, resource allocation, and targeted links (O'Reilly and Tushman 2004).

The ambidextrous organization keeps exploring apart from exploiting and joins them at the top

Select a unit or a letter to read it; the key above the drawing names the lines and letters used.

Reports toAdjust to each other directlyDDirect supervisionNStandardization of normsDifferentiation contested: one unit can explore and exploit at once, with no separate unit (Gibson and Birkinshaw 2004).

Targeted linksSenior teamSeparate unitsSenior teamExploitingunitExploitingunitExploringunitBy purpose: exploitBy purpose: exploreDN
Targeted linksSenior teamSeparate unitsSeniorExploitExploitExploreBy purpose: exploitBy purpose: exploreDN
Select a unit or a letter to read it. D, direct supervision: One person takes responsibility for the work of others, giving instructions and monitoring their actions. N, standardization of norms: Shared beliefs and values, usually across the whole organization, guide the work.Senior team. Layer: Senior team. Reports to no one. Integrates the units and allocates resources through a shared strategic intent and targeted links (O'Reilly and Tushman 2004). In a survey of 230 firms, separating the units worked through senior team social integration and cross-functional interfaces (Jansen and colleagues 2009).Exploiting unit. Layer: Separate units. Reports to Senior team. By purpose: exploit. Refines the existing business. That its own head holds delegated power is our judgment, and that its processes are formal is our reading, since no text read gives the level.Exploiting unit. Layer: Separate units. Reports to Senior team. By purpose: exploit. Also part of the existing business; it holds its boundaries (our judgment).Exploring unit. Layer: Separate units. Reports to Senior team. By purpose: explore. Builds new technologies, products, or markets with its own processes, structures, and culture (O'Reilly and Tushman 2004); that its processes are loose is our reading.D, direct supervision. One person takes responsibility for the work of others, giving instructions and monitoring their actions.N, standardization of norms. Shared beliefs and values, usually across the whole organization, guide the work.
The structure in words
  1. Senior team
    • Senior team, reports to no one, coordinated by direct supervision. Integrates the units and allocates resources through a shared strategic intent and targeted links (O'Reilly and Tushman 2004). In a survey of 230 firms, separating the units worked through senior team social integration and cross-functional interfaces (Jansen and colleagues 2009).
  2. Separate units
    • Exploiting unit, reports to Senior team, grouped by purpose: exploit. Refines the existing business. That its own head holds delegated power is our judgment, and that its processes are formal is our reading, since no text read gives the level.
    • Exploiting unit, reports to Senior team, grouped by purpose: exploit. Also part of the existing business; it holds its boundaries (our judgment).
    • Exploring unit, reports to Senior team, grouped by purpose: explore. Builds new technologies, products, or markets with its own processes, structures, and culture (O'Reilly and Tushman 2004); that its processes are loose is our reading.
  • Exploiting unit and Exploring unit: adjust to each other directly (Targeted links).
  • Across the whole organization: standardization of norms.
  • Differentiation contested: one unit can explore and exploit at once, with no separate unit (Gibson and Birkinshaw 2004).

Units that exploit the existing business and a unit that explores stand apart, each with its own processes, structures, and cultures. The senior team joins them through a shared strategic intent, resource allocation, and targeted links.

Source: O'Reilly and Tushman 2004, a practitioner's report of a multi-unit comparison; Jansen and colleagues 2009, a two-wave survey of 230 firms; Gibson and Birkinshaw 2004, a survey of 4,195 people in 41 business units.

  • Grouping: product or purpose, our judgment: structurally independent units (O'Reilly and Tushman 2004).
  • Specialization: medium, our judgment: each unit holds the functions its mode of learning needs.
  • Formalization: spans low and high, by unit, our judgment: exploratory units get their own processes, but no text read gives the levels (O'Reilly and Tushman 2004).
  • Centralization and layers: mixed centralization and moderate layers, our judgment; no source read sets the depth.
  • Decision rights: limited vertical decentralization, Type C in Mintzberg's (1980) synthesis, propositions without a new test; our judgment: each unit's head holds delegated power.
  • Coordinating mechanism: spans direct supervision and standardization of norms, our judgment: the senior team and a shared intent.
  • Lateral mechanisms: task forces and teams, on evidence: in a two-wave survey of 230 firms, cross-functional interfaces carried part of separation's effect (Jansen and colleagues 2009).
  • Differentiation against integration: contested. High differentiation with heavy integration (Jansen and colleagues 2009), or low to moderate, since one unit can be ambidextrous alone (Gibson and Birkinshaw 2004, a survey of 4,195 people in 41 business units) and differentiation did not condition the payoff (Jansen, Simsek, and Cao 2012, a quantitative study of 285 units).
  • Mode of exchange: bureaucracy, our judgment.
  • Boundary permeability: selective, our judgment, from no text read: exploring units opened to new markets and technologies, exploiting units holding their boundaries.

Where the ambidextrous organization sits on the eleven dimensions

The filled segments are the form's levels; select a row for its reason and sources.

Select a dimension to read why the form sits there, and on what basis.Basis of grouping: Product or purpose. Basis: Judgment. Structural ambidexterity organizes exploratory efforts as structurally independent units, each with its own processes, structures, and cultures, integrated at the senior management level; contextual ambidexterity achieves both inside one business unit with no separate explore unit, and whether to separate or integrate is an unresolved tension; either way the business unit is the basis, which is judgment. Sources: O'Reilly and Tushman 2004; Gibson and Birkinshaw 2004; Raisch and colleagues 2009.Specialization: Medium. Basis: Judgment. Each unit holds the functions its own mode of learning needs; judgment. Sources: O'Reilly and Tushman 2004.Standardization and formalization: Low and high. Basis: Judgment. Exploratory units are kept separate from exploitative ones and allowed their own processes, structures, and cultures, the dual structures Duncan proposed, with some units focused on alignment and others on adaptation; that the exploitative units run formal processes and the exploratory ones loose processes, so that the form spans the scale by unit, is our reading, since no text read gives the levels. Sources: O'Reilly and Tushman 2004; Duncan 1976.Centralization: Mixed. Basis: Judgment. The senior team integrates and allocates, while each unit runs its own work; judgment. Sources: O'Reilly and Tushman 2004.Layers and span of control: Moderate. Basis: Judgment. Nothing in the sources read sets the depth; judgment. Sources: O'Reilly and Tushman 2004.Decision rights and the locus of authority: Limited vertical decentralization. Basis: Judgment. Each unit has its own head with delegated power, joined only at the top, as in Type C; judgment. Sources: O'Reilly and Tushman 2004; Mintzberg 1980.Prime coordinating mechanism: Direct supervision and standardization of norms. Basis: Judgment. Integration runs through the senior team and a shared strategic intent; in one study the effect of structural separation worked through senior team social integration and cross-functional interfaces. Sources: O'Reilly and Tushman 2004; Jansen and colleagues 2009.Lateral mechanisms: Task forces and teams. Basis: Evidence. Formal cross-functional interfaces carried part of the effect of structural differentiation on ambidexterity. Sources: Jansen and colleagues 2009.Differentiation against integration: High differentiation, heavy integration. Basis: Evidence. Structural differentiation of exploit and explore units works through senior team integration and cross-functional interfaces. Sources: Jansen and colleagues 2009; Tushman and O'Reilly 1996; Gibson and Birkinshaw 2004; Jansen and colleagues 2012. View 2: Low differentiation, light integration to moderate differentiation, some integrating roles. Contextual ambidexterity achieves alignment and adaptability at once inside one business unit, without differentiated subunits, and a later study of 285 units found structural differentiation did not condition the link between unit ambidexterity and performance. Sources: Gibson and Birkinshaw 2004; Jansen and colleagues 2012.Mode of exchange and control: Bureaucracy. Basis: Judgment. Inside one hierarchy under one senior team; judgment. Sources: O'Reilly and Tushman 2004.Boundary permeability: Selective. Basis: Judgment. The explore units are opened to new markets and technologies, while the exploit units hold their boundaries; judgment. Sources: Tushman and O'Reilly 1996; Ashkenas and colleagues 1995.

Two of the eleven levels rest on evidence, none on theory, and nine on judgment. One row is contested: Differentiation against integration.

Source: O'Reilly and Tushman 2004, a practitioner's report of a multi-unit comparison; Gibson and Birkinshaw 2004, a survey of 4,195 people in 41 business units; Raisch and colleagues 2009; Duncan 1976; Mintzberg 1980, a synthesis of earlier research, offering propositions without a new test; Jansen and colleagues 2009, a two-wave survey of 230 firms; Tushman and O'Reilly 1996; Jansen and colleagues 2012, a quantitative study of 285 units; Ashkenas and colleagues 1995.

Where it came from

By Gibson and Birkinshaw's (2004) account, Robert Duncan first used the term, in a 1976 chapter, unread here, proposing dual structures for innovation. James March (1991), in formal models with no field data, named the tension: adaptive processes refine exploitation faster than exploration.

The schools of organization design and the forms' origins, 1900 to now

Each row is a school's era; each square under them is a form at its first year. Select a row or a square to read it; on a form's page, its own square and span are marked, with the schools that treat it.

1900192019401960198020002020Bureaucracy and its criticsClassical administrationHuman relations and System 4Transaction cost economicsCarnegie schoolSocio-technical systemsLabor-managed firmModularity and mirroringStrategy and structureContingency theoryFrench bureaucratic schoolJapanese-firm literatureScandinavian and DutchConfiguration theoryInformation processingLabor processStrategic choiceGerman codeterminationOrganizational economicsInstitutional theoryPopulation ecologyResource dependenceBusiness groupsScandinavian institutionalSocietal effectNetwork perspectiveComputational designNew-forms debateAlgorithmic managementForms by first yearAmbidextrous organization, 1976 to 2004Now

The forms by first year

1900192019401960198020002020Ambidextrous organization, 1976 to 2004Now

Schools that treat this form: the Carnegie school and the attention-based view; Computational organization design and search.

Select a school's row or a form's square to read it.Select a form's square to read it; open a school below.Bureaucracy and its critics1900s to 1950s, revived from the 1990sWhat it claimsWeber held that rational-legal authority exercised through an office hierarchy, written rules, trained officials, and promotion by qualification is technically superior in precision, speed, and reliability. Critics studied its unintended consequences, Michels's iron law sharpest.What it has shownMerton (1940) argued, with no test, that rules can turn from means into ends; case studies followed. The Aston group concluded that the type's dimensions vary independently, and Wagner's meta-analysis (2021) found that many of their interrelationships have weakened.Who disputes itHuman relations and System 4 sets the costs of impersonal rules to motivation and cooperation against the technical superiority Weber stresses.Bureaucracy and its critics in The Schools of Organization DesignClassical administration1910s to 1940s, and in German to the 1970sWhat it claimsThere are general principles of administration, among them division of work, unity of command, a scalar chain of authority, and a limited span of control, and a structure built to them will be efficient in most settings. Taylor added that planners should set work methods.What it has shownThe school offered reasoning and experience, not tests. Woodward (1965) found that the spans and levels of the more successful firms in her sample differed by production system, and Simon (1946) showed that the principles come in pairs that point in opposite directions.Who disputes itStructural contingency theory found that the structure of successful firms varies with technology and environment, where classical writers treat their principles as general.Classical administration in The Schools of Organization DesignHuman relations and System 41920s to 1960sWhat it claimsOutput depends on the social relations of work, the norms of informal groups, and the style of supervision as much as on formal design. Likert's System 4, overlapping work groups joined by linking-pin managers under participative management, outperforms authoritative systems.What it has shownThe founding evidence is the Hawthorne studies, which later reanalyses dispute (Carey 1967; Levitt and List 2011). Likert's support came largely from his own institute's surveys. Wagner (1994) found participation's effects on performance and satisfaction significant but small.Who disputes itLabor process and critical theory reads human relations as a technique of managerial control rather than a change in the distribution of power.Human relations and System 4 in The Schools of Organization DesignTransaction cost economics and the M-form hypothesis1937, then the 1970s to the presentWhat it claimsFirms exist, and are organized as they are, to economize on transaction costs. Hierarchy replaces the market when assets are specific, uncertainty is high, and exchange is frequent, and the multidivisional form should outperform the functional form in large, diversified firms.What it has shownArmour and Teece (1978) found, in United States petroleum firms from 1955 to 1973, a positive relationship between the multidivisional form and profitability while the form was spreading, but not once an equilibrium in organizational form was reached.Who disputes itThe network perspective on organizations: Granovetter argued that the theory undersocializes economic action, which is embedded in networks of personal relations that also produce trust and order.Transaction cost economics and the M-form hypothesis in The Schools of Organization DesignThe Carnegie school and the attention-based view1940s to the presentWhat it claimsPeople in organizations are boundedly rational, so structure matters because it sets the premises of their decisions and channels their attention: who sees which problems, which goals they pursue, and which routines they follow. Ocasio makes structure distribute attention.What it has shownMarch and Simon (1958) drew propositions from the existing literature, and Cyert and March (1963) added process models and simulations. Two studies show hierarchy shaping what is sent up as well as what is chosen (Reitzig and Maciejovsky 2015; Keum and See 2017).Who disputes itOrganizational economics: decision rights, authority, and complementarities models a principal and agents with defined objectives; the Carnegie school, firms as coalitions with unresolved goal conflict.The Carnegie school and the attention-based view in The Schools of Organization DesignSocio-technical systems1950s to 1980sWhat it claimsAn organization is a social system and a technical system that must be designed together. The same technology admits more than one way of organizing work, and groups that own a whole task and control its variances at the source perform better and are more satisfied.What it has shownTrist and Bamforth (1951) found that longwall mechanization broke up self-regulating work groups. A review of 134 experiments found them extremely successful overall, and Stewart (2006) linked team autonomy to performance; studies of socialist brigades and Volkswagen qualify it.Who disputes itLabor process and critical theory: Kelly argued that labor intensification, not joint optimization, accounts for what socio-technical redesign achieved, and that the part of pay incentives was seriously underestimated.Socio-technical systems in The Schools of Organization DesignThe economics of the labor-managed firm and of enterprise ownership1958 to the presentWhat it claimsA firm run by its workers behaves differently from one run for investors: Ward modeled the Yugoslav-style firm as maximizing income per worker. Hansmann argued that an ownership form's success depends on the balance between the costs of market contracting and of ownership.What it has shownBurdín (2014) found a 29 percent lower hazard of dissolution for Uruguayan worker-managed firms than for conventional ones, and Fakhfakh, Pérotin, and Gago (2012) found French labor-managed firms as productive or more so in every industry studied.Who disputes itOrganizational economics (Jensen and Meckling) contrasts the labor-managed firm, in which tradable residual claims are legally prohibited, with cooperatives and partnerships that arise out of voluntary contracting.The economics of the labor-managed firm and of enterprise ownership in The Schools of Organization DesignModularity and the mirroring hypothesis1960s to the presentWhat it claimsComplex systems are built of nearly decomposable parts, and an organization's structure tends to mirror the architecture of what it makes: modular products allow loosely coupled organizations, and tightly coupled products need tightly coupled teams. Conway's law runs in reverse.What it has shownColfer and Baldwin (2016) review the empirical studies and report both support and principled exceptions. Direct tests point both ways: MacCormack, Baldwin, and Rusnak (2012) found open source products more modular than their commercial pairs; dynamic components break the mirror.Who disputes itNo school in the record disputes it head on; it answers information processing, which meets the coordination load inside the organization; mirroring writers locate it in the product's architecture.Modularity and the mirroring hypothesis in The Schools of Organization DesignStrategy and structure1960s to 1980sWhat it claimsStructure follows strategy: firms that diversify by product or geography outgrow the centralized functional structure and, after a lag and usually a crisis, adopt divisions. Chandler drew this from the histories of DuPont, General Motors, Standard Oil of New Jersey, and Sears.What it has shownChandler (1962) is comparative business history of four firms set against a wider survey. Rumelt (1974) related diversification and structure to performance, and Egelhoff (1988), with 34 multinationals, found some of Stopford and Wells's relationships supported and others not.Who disputes itInstitutional theory: in Fligstein's model divisions spread with a product-related or unrelated strategy, presidents from sales or finance, and other firms in the industry changing; strategy was one of three.Strategy and structure in The Schools of Organization DesignStructural contingency theory1960s to the presentWhat it claimsThere is no one best structure. Performance depends on fit between structure and contingencies such as environment, technology, size, and strategy, on an open-system view of the organization. Drazin and Van de Ven separate fit as selection, interaction, and systems.What it has shownBurns and Stalker, Woodward, and the Aston group related structure to context; Drazin and Van de Ven (1985) supported selection and systems fit only. Pennings (1975) found that fit did not explain effectiveness; Schoonhoven (1981) found traditional hypotheses unsupported.Who disputes itStrategic choice: Child argued that contingency theory explains structure by the constraints of context and leaves out the political process by which power-holders choose strategic action, structural forms included.Structural contingency theory in The Schools of Organization DesignThe French bureaucratic school (strategic analysis)1960s to 1980sWhat it claimsA bureaucracy is a system of power relations organized around zones of uncertainty. Impersonal rules and centralization protect members from depending on one another, those who control uncertainties hold power, and, unable to correct its errors, it changes only through crisis.What it has shownLe phénomène bureaucratique (1963; English 1964) rests on two case studies, a clerical agency and a state industrial monopoly. L'acteur et le système (1977; English 1980) is theoretical.Who disputes itSocietal effect and comparative business systems: D'Iribarne argues that national logics of conduct, such as a French logic of honor, explain differences that Crozier attributed to power games.The French bureaucratic school (strategic analysis) in The Schools of Organization DesignThe Japanese-firm literature1960s to the presentWhat it claimsThe large Japanese firm coordinates operating decisions through horizontal information sharing among semi-autonomous shops and uses its ranking hierarchy mainly as an incentive scheme (Aoki). It creates knowledge through middle-up-down management (Nonaka and Takeuchi).What it has shownAoki (1986) is a formal model. Lincoln, Hanada, and McBride (1986) compared 51 Japanese and 55 United States manufacturing plants, and Nonaka and Takeuchi (1995) rest on case studies, chiefly of Japanese firms. Fruin (1994) found the enterprise system defined at three levels.Who disputes itLabor process and critical theory reads shop-floor delegation as control: Sewell and Wilkinson argue that just-in-time and total quality regimes delegate tactical responsibility while centralizing strategic control.The Japanese-firm literature in The Schools of Organization DesignThe Scandinavian and Dutch socio-technical schools1960s to 2000sWhat it claimsThe Scandinavian program sought democracy at work through field experiments with semi-autonomous groups, then through democratic dialogue. The Dutch school's design theory turns complex organizations with simple jobs into simple organizations with complex jobs.What it has shownEmery and Thorsrud reported field experiments in Norwegian industry; Gustavsen (1992) records that they spread less than hoped, his reason for turning to dialogue. Berggren studied work organization in the Swedish auto industry. The Dutch work rests on design cases and argument.Who disputes itThe Japanese-firm literature: defenders of the Toyota system hold that standardized work with employee participation learns faster; Berggren reports Swedish long-cycle plants matching a line plant on productivity.The Scandinavian and Dutch socio-technical schools in The Schools of Organization DesignConfiguration theory1970s to the presentWhat it claimsStructures cluster into a small number of internally coherent configurations whose elements reinforce one another, so organizations move between a few types in large jumps, not small steps. Mintzberg's five configurations each rest on one coordinating mechanism.What it has shownDoty, Glick, and Huber (1993) found that fit to Miles and Snow's ideal types predicted effectiveness while fit to Mintzberg's did not. A meta-analysis of 40 tests (Ketchen and colleagues 1997) found that configuration partly explains performance, more so with broad definitions.Who disputes itStructural contingency theory: Donaldson counters that most organizations sit at intermediate points on a continuum of fits, not in a few configurations, and that studies over time find mostly incremental change.Configuration theory in The Schools of Organization DesignInformation processing1970s to the presentWhat it claimsThe more uncertain the task, the more information must be processed during its execution. An organization either reduces that need, through slack resources or self-contained tasks, or raises its capacity, through vertical information systems or lateral relations up to the matrix.What it has shownGalbraith (1973, 1974) built the framework from cases and from Simon's theory of bounded rationality, and Tushman and Nadler (1978) set it out as an integrating concept for design. Daft and Lengel (1986) argued that structure must reduce equivocality as well as supply data.Who disputes itThe Carnegie school and the attention-based view: Ocasio holds that firm behavior results from how the firm channels and distributes attention, so structure governs where attention goes, not only how much is processed.Information processing in The Schools of Organization DesignLabor process and critical theory1970s to the presentWhat it claimsStructure is above all a system of control over labor. Braverman argued that management deskills work by separating conception from execution; Edwards traced a sequence from simple control by the boss to technical control through machines to bureaucratic control through rules.What it has shownHistorical and theoretical (Braverman 1974; Edwards 1979). Delbridge (1995), observing a Japanese-owned plant in England, found that just-in-time and total quality regimes combined the control of labor with management's economic goals more completely than piece rates had.Who disputes itBureaucracy and its critics: Adler argues that labor process theory ignores the progressive socialization of the labor process, which drives long-term skill upgrading, and disputes deskilling as the trend.Labor process and critical theory in The Schools of Organization DesignStrategic choice1970s to the presentWhat it claimsStructure is the outcome of choices by those with power in the organization, who decide on strategic action, including the structural form, the manipulation of environmental features, and the performance standards to be met. Context constrains but does not determine structure.What it has shownLargely conceptual: Child (1972) built the argument from gaps in contingency findings, and the empirical case rests mostly on the modest share of structural variance contingencies explain. A 2026 review of 42 studies finds upper echelons theory their main theoretical anchor.Who disputes itStructural contingency theory: Donaldson replies that organizations mostly adopt the structures their contingencies dictate, to avoid the loss of performance misfit brings, so choice over structure is at most limited.Strategic choice in The Schools of Organization DesignThe German codetermination literature1970s to the presentWhat it claimsWorks councils and employee board seats form a structural layer through which coordination can become negotiation between management and employees. Defenders argue that it supports long-term employment, training, and cooperation; critics, that it weakens shareholder control.What it has shownJäger, Schoefer, and Heining (2021) found board-level codetermination left wages unchanged and, if anything, raised capital formation. Jirjahn (2011) reports studies pointing to positive works council effects; a meta-analysis found codetermination laws tied to lower productivity.Who disputes itOrganizational economics (Gorton and Schmid) found that firms with equal employee and shareholder representation did worse for their shareholders than firms with one-third employee representation.The German codetermination literature in The Schools of Organization DesignOrganizational economics: decision rights, authority, and complementarities1976 to the presentWhat it claimsDesign is the joint choice of who holds which decision rights, how performance is measured, and how it is rewarded. Rights should be placed where the specific knowledge lies, formal authority differs from real authority, and practices are complements that pay off in clusters.What it has shownIchniowski, Shaw, and Prennushi (1997) found, on 36 steel finishing lines, that lines using a set of innovative work practices achieved substantially higher productivity, consistent with complementarities. Bloom and colleagues found more decentralization where trust was higher.Who disputes itGhoshal argued that management research has shaped managers' worldview with ideologically inspired amoral theories, built on a scientific model that excludes human intentionality and so denies moral considerations.Organizational economics: decision rights, authority, and complementarities in The Schools of Organization DesignInstitutional theory1977 to the presentWhat it claimsFormal structures are often adopted because they are legitimate in the organization's field, as rationalized myths, and then decoupled from the work done. Organizations in a field grow alike through coercive, mimetic, and normative isomorphism, whether or not it is efficient.What it has shownTolbert and Zucker (1983) found that city characteristics predicted early but not later adoption of civil service reform, which they read as institutionalization. Palmer, Jennings, and Zhou (1993) found support for institutional and economic accounts of late M-form adoption.Who disputes itStructural contingency theory: Kraatz and Zajac found little support for neoinstitutional predictions in 631 liberal arts colleges, which grew less alike; the change they studied was curricular, not structural.Institutional theory in The Schools of Organization DesignPopulation ecology1977 to the presentWhat it claimsOrganizations face strong structural inertia, because selection favors the reliability and accountability of stable structures. Change in a population's forms comes mainly from founding and failure, not adaptation, and changing core features resets the liability of newness.What it has shownAmburgey, Kelly, and Barnett (1993) found that the effects of change on the hazard of failure vary by type of change and over time, and that rates of change depend on how often and how recently an organization has changed. Carroll and Hannan (2000) set out the program's methods.Who disputes itFor structural contingency theory, Donaldson argues that ecology understates managers' capacity to adapt structure.Population ecology in The Schools of Organization DesignResource dependence1978 to the presentWhat it claimsOrganizations depend on outside parties for critical resources, and structure follows the management of that dependence: units that cope with critical dependencies gain power, and boundaries are redrawn through cooptation, joint ventures, and mergers to absorb constraint.What it has shownPfeffer and Salancik (1978) assembled studies of boards, mergers, and joint ventures. A meta-analysis of 157 tests (Drees and Heugens 2013) corroborated the main prediction: organizations meet resource dependencies with interorganizational arrangements that raise their autonomy.Who disputes itTransaction cost economics and the M-form hypothesis explains boundaries by efficiency; resource dependence, by power and constraint.Resource dependence in The Schools of Organization DesignThe business-groups literature1978 to the presentWhat it claimsGroups of legally independent firms bound by ownership and social ties are a durable form of their own, neither one firm nor a market. Where institutions function poorly, groups can benefit member firms and can destroy value; they are sometimes paragons, sometimes parasites.What it has shownKhanna and Palepu (2000) found that affiliates of the most diversified Indian groups outperformed unaffiliated firms. A meta-analysis of 141 studies (Carney and colleagues 2011) found that affiliation lowers performance in general, less so where institutions are underdeveloped.Who disputes itOrganizational economics: agency writers read pyramids and cross-holdings as devices that separate control from ownership; Almeida and Wolfenzon offer a financing rationale for pyramids instead.The business-groups literature in The Schools of Organization DesignScandinavian institutionalism1980s to the presentWhat it claimsIdeas of organizing do not diffuse unchanged; they travel and are translated as they are adopted, so a form taken up in one place is not the same form elsewhere. Organizations can also be built of organizations (meta-organizations), which behave differently.What it has shownA discussion paper by Czarniawska and Joerges (1995) contrasts the diffusion of ideas with their translation, drawing on examples such as Swedish municipal reform; the work on meta-organizations (Ahrne and Brunsson 2008) is conceptual.Who disputes itInstitutional theory can read the same processes, seen from a translocal view, as institutional isomorphism, though translation holds that each adopter modifies an idea, so an institution re-created elsewhere is new.Scandinavian institutionalism in The Schools of Organization DesignSocietal effect and comparative business systems1980s to the presentWhat it claimsFirms matched on size, technology, and industry differ systematically by country, because how a society educates, trains, recruits, and promotes its workforce, and its industrial relations, shape how work is divided and supervised, so structure is partly a national product.What it has shownMaurice, Sorge, and Warner (1980) compared closely matched manufacturing units in France, West Germany, and Great Britain and found more pervasive differences in their organization than usually recognized. Whitley (1999) set out a comparative account of business systems.Who disputes itStructural contingency theory stresses culture-free elements of structure; matched comparisons find nationally different shapes of organization.Societal effect and comparative business systems in The Schools of Organization DesignThe network perspective on organizations1985 to the presentWhat it claimsOrganizations and their members are nodes joined by ties, and the pattern of ties shapes what they can do. Economic action is embedded in networks of personal relations (Granovetter 1985), and network forms are a distinct form, not a hybrid of market and hierarchy (Powell 1990).What it has shownPodolny and Page (1998) review the research and find that network forms foster learning, confer status and legitimacy, bring economic benefits, help manage resource dependencies, and give employees considerable autonomy, and they ask for attention to networks' dysfunctions.Who disputes itTransaction cost economics: Williamson treats the hybrid as a third generic form between market and hierarchy; for Powell, network forms are a distinct form, not a mix of the two.The network perspective on organizations in The Schools of Organization DesignComputational organization design and search1990s to the presentWhat it claimsDesign choices interact, so design space is a rugged landscape with many local peaks, and structure affects how well an organization searches it. Decentralizing interdependent decisions can trap units on poor peaks; temporary decentralization, then reintegration, can help.What it has shownMostly simulation, with one field test: Csaszar (2012), with more than 150,000 stock-picking decisions by 609 mutual funds, found that decision-making structure affects the number of initiatives pursued and the omission and commission errors made.Who disputes itNo school in the record disputes it head on; it disputes structural contingency theory, whose bivariate fit, its simulations show, cannot capture how the best design depends on interactions among choices.Computational organization design and search in The Schools of Organization DesignThe new-forms debate1990s to the presentWhat it claimsOrganizations must solve four universal problems (dividing tasks, allocating them, providing rewards, and providing information), and a form is new when it solves them in new ways. Self-managing organizations radically decentralize authority; skeptics hold hierarchy reappears.What it has shownFoss (2003) traced the rise and partial reversal of Oticon's spaghetti organization to the problem of selective intervention. A meta-analysis of 54 studies of 13,914 teams (Greer and colleagues 2018) found that hierarchy within teams on net slightly lowers team performance.Who disputes itBureaucracy and its critics: removing formal structure leaves informal elites whose power is unaccountable (Freeman), and democratic and network organizations are much less hierarchy-free than their labels suggest.The new-forms debate in The Schools of Organization DesignThe algorithmic-management literature2015 to the presentWhat it claimsSoftware does much of the directing, evaluating, and disciplining that managers and rules once did. Kellogg, Valentine, and Christin (2020) call algorithmic control a new contested terrain; Puranam (2021) treats dividing decisions between people and machines as a design problem.What it has shownMostly qualitative field studies of gig work. Wood and colleagues (2019), with 107 interviews in six countries and 679 survey respondents, found that algorithmic management tends to give remote gig workers flexibility and autonomy, but also low pay, isolation, and overwork.Who disputes itLabor process and critical theory reads control at work as contested terrain; Kellogg, Valentine, and Christin take up that view but argue that algorithmic control is distinct from earlier forms.The algorithmic-management literature in The Schools of Organization DesignFunctional structureOrigin: 1911 to 1937.Family: Classical hierarchies.Line and staffOrigin: 1916 to 1943.Family: Classical hierarchies.Multidivisional (M-form) by productOrigin: 1920s practice; named and analyzed 1962 to 1975.Family: Divisionalized forms.Machine bureaucracyOrigin: 1922 (Weber); 1979 (Mintzberg).Family: Classical hierarchies.Divisional by customer or market segmentOrigin: 1937; 2005.Family: Divisionalized forms.Matrix organizationOrigin: 1950s to 1960s practice; analyzed 1971 to 1977.Family: Lateral and multidimensional forms.Co-determined organizationOrigin: 1951 to 1976 (the laws); 1984 to 2022 (the main studies).Family: Forms from other traditions.Socio-technical autonomous work groupsOrigin: 1951 to 1965.Family: Team-based and self-managing forms.Stratified hierarchy (requisite organization)Origin: 1951 to 1989.Family: Classical hierarchies.Cooperative organizationOrigin: 1958 to 1999.Family: Forms from other traditions.Linking-pin organization (System 4)Origin: 1961 to 1967.Family: Team-based and self-managing forms.Divisional by regionOrigin: 1962 to 1972.Family: Divisionalized forms.Adhocracy (innovative organization)Origin: 1968 to 1979.Family: Professional, project, and knowledge forms.J-form (the Japanese firm)Origin: 1968 to 1990.Family: Forms from other traditions.Professional bureaucracyOrigin: 1968 to 1979.Family: Professional, project, and knowledge forms.Holding company (H-form)Origin: 1972 to 1975; 1987.Family: Divisionalized forms.Ambidextrous organizationOrigin: 1976 to 2004.Family: Professional, project, and knowledge forms.Simple structureOrigin: 1979.Family: Classical hierarchies.Hybrid structureOrigin: 1980 to the present.Family: Lateral and multidimensional forms.Missionary (ideological) organizationOrigin: 1980 to 1989.Family: Ideological forms.Network, modular, and virtual organizationOrigin: 1986 to 1990.Family: Network and market-like forms.Transnational and multidimensional organizationOrigin: 1987 to 1989.Family: Lateral and multidimensional forms.Business group (keiretsu, chaebol, family house)Origin: 1988 to 2007.Family: Forms from other traditions.Shamrock organizationOrigin: 1988 to 1989.Family: Network and market-like forms.Federal organizationOrigin: 1989 to 1992.Family: Network and market-like forms.Self-managing organization (less-hierarchical forms)Origin: 1989 to 2017.Family: Team-based and self-managing forms.Boundaryless organizationOrigin: 1990 to 1995.Family: Network and market-like forms.Front-back organizationOrigin: 1990s to 2007.Family: Lateral and multidimensional forms.Professional partnership (P2 form)Origin: 1990.Family: Professional, project, and knowledge forms.The corporate center and shared servicesOrigin: 1991 to 2000.Family: Divisionalized forms.Internal market organizationOrigin: 1993 to 1994.Family: Network and market-like forms.Process (horizontal) organizationOrigin: 1993 to 1999.Family: Lateral and multidimensional forms.Cellular organizationOrigin: 1994 to 1997.Family: Network and market-like forms.Hypertext organizationOrigin: 1994 to 1995.Family: Forms from other traditions.Team-based organizationOrigin: 1995.Family: Team-based and self-managing forms.Project-based organizationOrigin: 2000 to 2006.Family: Professional, project, and knowledge forms.Agile at scaleOrigin: 2001 to 2016.Family: Contemporary and emerging forms.Distributed (all-remote) organizationOrigin: 2002 to 2021.Family: Contemporary and emerging forms.Open-source and community-based organizationOrigin: 2002 to 2007.Family: Contemporary and emerging forms.Algorithmically managed organizationOrigin: 2015 to 2020.Family: Contemporary and emerging forms.Team of teamsOrigin: 2015.Family: Team-based and self-managing forms.

The schools by first year

  1. Bureaucracy and its critics1900s to 1950s, revived from the 1990sWhat it claimsWeber held that rational-legal authority exercised through an office hierarchy, written rules, trained officials, and promotion by qualification is technically superior in precision, speed, and reliability. Critics studied its unintended consequences, Michels's iron law sharpest.What it has shownMerton (1940) argued, with no test, that rules can turn from means into ends; case studies followed. The Aston group concluded that the type's dimensions vary independently, and Wagner's meta-analysis (2021) found that many of their interrelationships have weakened.Who disputes itHuman relations and System 4 sets the costs of impersonal rules to motivation and cooperation against the technical superiority Weber stresses.
  2. Classical administration1910s to 1940s, and in German to the 1970sWhat it claimsThere are general principles of administration, among them division of work, unity of command, a scalar chain of authority, and a limited span of control, and a structure built to them will be efficient in most settings. Taylor added that planners should set work methods.What it has shownThe school offered reasoning and experience, not tests. Woodward (1965) found that the spans and levels of the more successful firms in her sample differed by production system, and Simon (1946) showed that the principles come in pairs that point in opposite directions.Who disputes itStructural contingency theory found that the structure of successful firms varies with technology and environment, where classical writers treat their principles as general.
  3. Human relations and System 41920s to 1960sWhat it claimsOutput depends on the social relations of work, the norms of informal groups, and the style of supervision as much as on formal design. Likert's System 4, overlapping work groups joined by linking-pin managers under participative management, outperforms authoritative systems.What it has shownThe founding evidence is the Hawthorne studies, which later reanalyses dispute (Carey 1967; Levitt and List 2011). Likert's support came largely from his own institute's surveys. Wagner (1994) found participation's effects on performance and satisfaction significant but small.Who disputes itLabor process and critical theory reads human relations as a technique of managerial control rather than a change in the distribution of power.
  4. Transaction cost economics and the M-form hypothesis1937, then the 1970s to the presentWhat it claimsFirms exist, and are organized as they are, to economize on transaction costs. Hierarchy replaces the market when assets are specific, uncertainty is high, and exchange is frequent, and the multidivisional form should outperform the functional form in large, diversified firms.What it has shownArmour and Teece (1978) found, in United States petroleum firms from 1955 to 1973, a positive relationship between the multidivisional form and profitability while the form was spreading, but not once an equilibrium in organizational form was reached.Who disputes itThe network perspective on organizations: Granovetter argued that the theory undersocializes economic action, which is embedded in networks of personal relations that also produce trust and order.
  5. The Carnegie school and the attention-based view1940s to the presentWhat it claimsPeople in organizations are boundedly rational, so structure matters because it sets the premises of their decisions and channels their attention: who sees which problems, which goals they pursue, and which routines they follow. Ocasio makes structure distribute attention.What it has shownMarch and Simon (1958) drew propositions from the existing literature, and Cyert and March (1963) added process models and simulations. Two studies show hierarchy shaping what is sent up as well as what is chosen (Reitzig and Maciejovsky 2015; Keum and See 2017).Who disputes itOrganizational economics: decision rights, authority, and complementarities models a principal and agents with defined objectives; the Carnegie school, firms as coalitions with unresolved goal conflict.
  6. Socio-technical systems1950s to 1980sWhat it claimsAn organization is a social system and a technical system that must be designed together. The same technology admits more than one way of organizing work, and groups that own a whole task and control its variances at the source perform better and are more satisfied.What it has shownTrist and Bamforth (1951) found that longwall mechanization broke up self-regulating work groups. A review of 134 experiments found them extremely successful overall, and Stewart (2006) linked team autonomy to performance; studies of socialist brigades and Volkswagen qualify it.Who disputes itLabor process and critical theory: Kelly argued that labor intensification, not joint optimization, accounts for what socio-technical redesign achieved, and that the part of pay incentives was seriously underestimated.
  7. The economics of the labor-managed firm and of enterprise ownership1958 to the presentWhat it claimsA firm run by its workers behaves differently from one run for investors: Ward modeled the Yugoslav-style firm as maximizing income per worker. Hansmann argued that an ownership form's success depends on the balance between the costs of market contracting and of ownership.What it has shownBurdín (2014) found a 29 percent lower hazard of dissolution for Uruguayan worker-managed firms than for conventional ones, and Fakhfakh, Pérotin, and Gago (2012) found French labor-managed firms as productive or more so in every industry studied.Who disputes itOrganizational economics (Jensen and Meckling) contrasts the labor-managed firm, in which tradable residual claims are legally prohibited, with cooperatives and partnerships that arise out of voluntary contracting.
  8. Modularity and the mirroring hypothesis1960s to the presentWhat it claimsComplex systems are built of nearly decomposable parts, and an organization's structure tends to mirror the architecture of what it makes: modular products allow loosely coupled organizations, and tightly coupled products need tightly coupled teams. Conway's law runs in reverse.What it has shownColfer and Baldwin (2016) review the empirical studies and report both support and principled exceptions. Direct tests point both ways: MacCormack, Baldwin, and Rusnak (2012) found open source products more modular than their commercial pairs; dynamic components break the mirror.Who disputes itNo school in the record disputes it head on; it answers information processing, which meets the coordination load inside the organization; mirroring writers locate it in the product's architecture.
  9. Strategy and structure1960s to 1980sWhat it claimsStructure follows strategy: firms that diversify by product or geography outgrow the centralized functional structure and, after a lag and usually a crisis, adopt divisions. Chandler drew this from the histories of DuPont, General Motors, Standard Oil of New Jersey, and Sears.What it has shownChandler (1962) is comparative business history of four firms set against a wider survey. Rumelt (1974) related diversification and structure to performance, and Egelhoff (1988), with 34 multinationals, found some of Stopford and Wells's relationships supported and others not.Who disputes itInstitutional theory: in Fligstein's model divisions spread with a product-related or unrelated strategy, presidents from sales or finance, and other firms in the industry changing; strategy was one of three.
  10. Structural contingency theory1960s to the presentWhat it claimsThere is no one best structure. Performance depends on fit between structure and contingencies such as environment, technology, size, and strategy, on an open-system view of the organization. Drazin and Van de Ven separate fit as selection, interaction, and systems.What it has shownBurns and Stalker, Woodward, and the Aston group related structure to context; Drazin and Van de Ven (1985) supported selection and systems fit only. Pennings (1975) found that fit did not explain effectiveness; Schoonhoven (1981) found traditional hypotheses unsupported.Who disputes itStrategic choice: Child argued that contingency theory explains structure by the constraints of context and leaves out the political process by which power-holders choose strategic action, structural forms included.
  11. The French bureaucratic school (strategic analysis)1960s to 1980sWhat it claimsA bureaucracy is a system of power relations organized around zones of uncertainty. Impersonal rules and centralization protect members from depending on one another, those who control uncertainties hold power, and, unable to correct its errors, it changes only through crisis.What it has shownLe phénomène bureaucratique (1963; English 1964) rests on two case studies, a clerical agency and a state industrial monopoly. L'acteur et le système (1977; English 1980) is theoretical.Who disputes itSocietal effect and comparative business systems: D'Iribarne argues that national logics of conduct, such as a French logic of honor, explain differences that Crozier attributed to power games.
  12. The Japanese-firm literature1960s to the presentWhat it claimsThe large Japanese firm coordinates operating decisions through horizontal information sharing among semi-autonomous shops and uses its ranking hierarchy mainly as an incentive scheme (Aoki). It creates knowledge through middle-up-down management (Nonaka and Takeuchi).What it has shownAoki (1986) is a formal model. Lincoln, Hanada, and McBride (1986) compared 51 Japanese and 55 United States manufacturing plants, and Nonaka and Takeuchi (1995) rest on case studies, chiefly of Japanese firms. Fruin (1994) found the enterprise system defined at three levels.Who disputes itLabor process and critical theory reads shop-floor delegation as control: Sewell and Wilkinson argue that just-in-time and total quality regimes delegate tactical responsibility while centralizing strategic control.
  13. The Scandinavian and Dutch socio-technical schools1960s to 2000sWhat it claimsThe Scandinavian program sought democracy at work through field experiments with semi-autonomous groups, then through democratic dialogue. The Dutch school's design theory turns complex organizations with simple jobs into simple organizations with complex jobs.What it has shownEmery and Thorsrud reported field experiments in Norwegian industry; Gustavsen (1992) records that they spread less than hoped, his reason for turning to dialogue. Berggren studied work organization in the Swedish auto industry. The Dutch work rests on design cases and argument.Who disputes itThe Japanese-firm literature: defenders of the Toyota system hold that standardized work with employee participation learns faster; Berggren reports Swedish long-cycle plants matching a line plant on productivity.
  14. Configuration theory1970s to the presentWhat it claimsStructures cluster into a small number of internally coherent configurations whose elements reinforce one another, so organizations move between a few types in large jumps, not small steps. Mintzberg's five configurations each rest on one coordinating mechanism.What it has shownDoty, Glick, and Huber (1993) found that fit to Miles and Snow's ideal types predicted effectiveness while fit to Mintzberg's did not. A meta-analysis of 40 tests (Ketchen and colleagues 1997) found that configuration partly explains performance, more so with broad definitions.Who disputes itStructural contingency theory: Donaldson counters that most organizations sit at intermediate points on a continuum of fits, not in a few configurations, and that studies over time find mostly incremental change.
  15. Information processing1970s to the presentWhat it claimsThe more uncertain the task, the more information must be processed during its execution. An organization either reduces that need, through slack resources or self-contained tasks, or raises its capacity, through vertical information systems or lateral relations up to the matrix.What it has shownGalbraith (1973, 1974) built the framework from cases and from Simon's theory of bounded rationality, and Tushman and Nadler (1978) set it out as an integrating concept for design. Daft and Lengel (1986) argued that structure must reduce equivocality as well as supply data.Who disputes itThe Carnegie school and the attention-based view: Ocasio holds that firm behavior results from how the firm channels and distributes attention, so structure governs where attention goes, not only how much is processed.
  16. Labor process and critical theory1970s to the presentWhat it claimsStructure is above all a system of control over labor. Braverman argued that management deskills work by separating conception from execution; Edwards traced a sequence from simple control by the boss to technical control through machines to bureaucratic control through rules.What it has shownHistorical and theoretical (Braverman 1974; Edwards 1979). Delbridge (1995), observing a Japanese-owned plant in England, found that just-in-time and total quality regimes combined the control of labor with management's economic goals more completely than piece rates had.Who disputes itBureaucracy and its critics: Adler argues that labor process theory ignores the progressive socialization of the labor process, which drives long-term skill upgrading, and disputes deskilling as the trend.
  17. Strategic choice1970s to the presentWhat it claimsStructure is the outcome of choices by those with power in the organization, who decide on strategic action, including the structural form, the manipulation of environmental features, and the performance standards to be met. Context constrains but does not determine structure.What it has shownLargely conceptual: Child (1972) built the argument from gaps in contingency findings, and the empirical case rests mostly on the modest share of structural variance contingencies explain. A 2026 review of 42 studies finds upper echelons theory their main theoretical anchor.Who disputes itStructural contingency theory: Donaldson replies that organizations mostly adopt the structures their contingencies dictate, to avoid the loss of performance misfit brings, so choice over structure is at most limited.
  18. The German codetermination literature1970s to the presentWhat it claimsWorks councils and employee board seats form a structural layer through which coordination can become negotiation between management and employees. Defenders argue that it supports long-term employment, training, and cooperation; critics, that it weakens shareholder control.What it has shownJäger, Schoefer, and Heining (2021) found board-level codetermination left wages unchanged and, if anything, raised capital formation. Jirjahn (2011) reports studies pointing to positive works council effects; a meta-analysis found codetermination laws tied to lower productivity.Who disputes itOrganizational economics (Gorton and Schmid) found that firms with equal employee and shareholder representation did worse for their shareholders than firms with one-third employee representation.
  19. Organizational economics: decision rights, authority, and complementarities1976 to the presentWhat it claimsDesign is the joint choice of who holds which decision rights, how performance is measured, and how it is rewarded. Rights should be placed where the specific knowledge lies, formal authority differs from real authority, and practices are complements that pay off in clusters.What it has shownIchniowski, Shaw, and Prennushi (1997) found, on 36 steel finishing lines, that lines using a set of innovative work practices achieved substantially higher productivity, consistent with complementarities. Bloom and colleagues found more decentralization where trust was higher.Who disputes itGhoshal argued that management research has shaped managers' worldview with ideologically inspired amoral theories, built on a scientific model that excludes human intentionality and so denies moral considerations.
  20. Institutional theory1977 to the presentWhat it claimsFormal structures are often adopted because they are legitimate in the organization's field, as rationalized myths, and then decoupled from the work done. Organizations in a field grow alike through coercive, mimetic, and normative isomorphism, whether or not it is efficient.What it has shownTolbert and Zucker (1983) found that city characteristics predicted early but not later adoption of civil service reform, which they read as institutionalization. Palmer, Jennings, and Zhou (1993) found support for institutional and economic accounts of late M-form adoption.Who disputes itStructural contingency theory: Kraatz and Zajac found little support for neoinstitutional predictions in 631 liberal arts colleges, which grew less alike; the change they studied was curricular, not structural.
  21. Population ecology1977 to the presentWhat it claimsOrganizations face strong structural inertia, because selection favors the reliability and accountability of stable structures. Change in a population's forms comes mainly from founding and failure, not adaptation, and changing core features resets the liability of newness.What it has shownAmburgey, Kelly, and Barnett (1993) found that the effects of change on the hazard of failure vary by type of change and over time, and that rates of change depend on how often and how recently an organization has changed. Carroll and Hannan (2000) set out the program's methods.Who disputes itFor structural contingency theory, Donaldson argues that ecology understates managers' capacity to adapt structure.
  22. Resource dependence1978 to the presentWhat it claimsOrganizations depend on outside parties for critical resources, and structure follows the management of that dependence: units that cope with critical dependencies gain power, and boundaries are redrawn through cooptation, joint ventures, and mergers to absorb constraint.What it has shownPfeffer and Salancik (1978) assembled studies of boards, mergers, and joint ventures. A meta-analysis of 157 tests (Drees and Heugens 2013) corroborated the main prediction: organizations meet resource dependencies with interorganizational arrangements that raise their autonomy.Who disputes itTransaction cost economics and the M-form hypothesis explains boundaries by efficiency; resource dependence, by power and constraint.
  23. The business-groups literature1978 to the presentWhat it claimsGroups of legally independent firms bound by ownership and social ties are a durable form of their own, neither one firm nor a market. Where institutions function poorly, groups can benefit member firms and can destroy value; they are sometimes paragons, sometimes parasites.What it has shownKhanna and Palepu (2000) found that affiliates of the most diversified Indian groups outperformed unaffiliated firms. A meta-analysis of 141 studies (Carney and colleagues 2011) found that affiliation lowers performance in general, less so where institutions are underdeveloped.Who disputes itOrganizational economics: agency writers read pyramids and cross-holdings as devices that separate control from ownership; Almeida and Wolfenzon offer a financing rationale for pyramids instead.
  24. Scandinavian institutionalism1980s to the presentWhat it claimsIdeas of organizing do not diffuse unchanged; they travel and are translated as they are adopted, so a form taken up in one place is not the same form elsewhere. Organizations can also be built of organizations (meta-organizations), which behave differently.What it has shownA discussion paper by Czarniawska and Joerges (1995) contrasts the diffusion of ideas with their translation, drawing on examples such as Swedish municipal reform; the work on meta-organizations (Ahrne and Brunsson 2008) is conceptual.Who disputes itInstitutional theory can read the same processes, seen from a translocal view, as institutional isomorphism, though translation holds that each adopter modifies an idea, so an institution re-created elsewhere is new.
  25. Societal effect and comparative business systems1980s to the presentWhat it claimsFirms matched on size, technology, and industry differ systematically by country, because how a society educates, trains, recruits, and promotes its workforce, and its industrial relations, shape how work is divided and supervised, so structure is partly a national product.What it has shownMaurice, Sorge, and Warner (1980) compared closely matched manufacturing units in France, West Germany, and Great Britain and found more pervasive differences in their organization than usually recognized. Whitley (1999) set out a comparative account of business systems.Who disputes itStructural contingency theory stresses culture-free elements of structure; matched comparisons find nationally different shapes of organization.
  26. The network perspective on organizations1985 to the presentWhat it claimsOrganizations and their members are nodes joined by ties, and the pattern of ties shapes what they can do. Economic action is embedded in networks of personal relations (Granovetter 1985), and network forms are a distinct form, not a hybrid of market and hierarchy (Powell 1990).What it has shownPodolny and Page (1998) review the research and find that network forms foster learning, confer status and legitimacy, bring economic benefits, help manage resource dependencies, and give employees considerable autonomy, and they ask for attention to networks' dysfunctions.Who disputes itTransaction cost economics: Williamson treats the hybrid as a third generic form between market and hierarchy; for Powell, network forms are a distinct form, not a mix of the two.
  27. Computational organization design and search1990s to the presentWhat it claimsDesign choices interact, so design space is a rugged landscape with many local peaks, and structure affects how well an organization searches it. Decentralizing interdependent decisions can trap units on poor peaks; temporary decentralization, then reintegration, can help.What it has shownMostly simulation, with one field test: Csaszar (2012), with more than 150,000 stock-picking decisions by 609 mutual funds, found that decision-making structure affects the number of initiatives pursued and the omission and commission errors made.Who disputes itNo school in the record disputes it head on; it disputes structural contingency theory, whose bivariate fit, its simulations show, cannot capture how the best design depends on interactions among choices.
  28. The new-forms debate1990s to the presentWhat it claimsOrganizations must solve four universal problems (dividing tasks, allocating them, providing rewards, and providing information), and a form is new when it solves them in new ways. Self-managing organizations radically decentralize authority; skeptics hold hierarchy reappears.What it has shownFoss (2003) traced the rise and partial reversal of Oticon's spaghetti organization to the problem of selective intervention. A meta-analysis of 54 studies of 13,914 teams (Greer and colleagues 2018) found that hierarchy within teams on net slightly lowers team performance.Who disputes itBureaucracy and its critics: removing formal structure leaves informal elites whose power is unaccountable (Freeman), and democratic and network organizations are much less hierarchy-free than their labels suggest.
  29. The algorithmic-management literature2015 to the presentWhat it claimsSoftware does much of the directing, evaluating, and disciplining that managers and rules once did. Kellogg, Valentine, and Christin (2020) call algorithmic control a new contested terrain; Puranam (2021) treats dividing decisions between people and machines as a design problem.What it has shownMostly qualitative field studies of gig work. Wood and colleagues (2019), with 107 interviews in six countries and 679 survey respondents, found that algorithmic management tends to give remote gig workers flexibility and autonomy, but also low pay, isolation, and overwork.Who disputes itLabor process and critical theory reads control at work as contested terrain; Kellogg, Valentine, and Christin take up that view but argue that algorithmic control is distinct from earlier forms.

The forms by first year

  1. Functional structure: 1911 to 1937
  2. Line and staff: 1916 to 1943
  3. Multidivisional (M-form) by product: 1920s practice; named and analyzed 1962 to 1975
  4. Machine bureaucracy: 1922 (Weber); 1979 (Mintzberg)
  5. Divisional by customer or market segment: 1937; 2005
  6. Matrix organization: 1950s to 1960s practice; analyzed 1971 to 1977
  7. Co-determined organization: 1951 to 1976 (the laws); 1984 to 2022 (the main studies)
  8. Socio-technical autonomous work groups: 1951 to 1965
  9. Stratified hierarchy (requisite organization): 1951 to 1989
  10. Cooperative organization: 1958 to 1999
  11. Linking-pin organization (System 4): 1961 to 1967
  12. Divisional by region: 1962 to 1972
  13. Adhocracy (innovative organization): 1968 to 1979
  14. J-form (the Japanese firm): 1968 to 1990
  15. Professional bureaucracy: 1968 to 1979
  16. Holding company (H-form): 1972 to 1975; 1987
  17. Ambidextrous organization: 1976 to 2004
  18. Simple structure: 1979
  19. Hybrid structure: 1980 to the present
  20. Missionary (ideological) organization: 1980 to 1989
  21. Network, modular, and virtual organization: 1986 to 1990
  22. Transnational and multidimensional organization: 1987 to 1989
  23. Business group (keiretsu, chaebol, family house): 1988 to 2007
  24. Shamrock organization: 1988 to 1989
  25. Federal organization: 1989 to 1992
  26. Self-managing organization (less-hierarchical forms): 1989 to 2017
  27. Boundaryless organization: 1990 to 1995
  28. Front-back organization: 1990s to 2007
  29. Professional partnership (P2 form): 1990
  30. The corporate center and shared services: 1991 to 2000
  31. Internal market organization: 1993 to 1994
  32. Process (horizontal) organization: 1993 to 1999
  33. Cellular organization: 1994 to 1997
  34. Hypertext organization: 1994 to 1995
  35. Team-based organization: 1995
  36. Project-based organization: 2000 to 2006
  37. Agile at scale: 2001 to 2016
  38. Distributed (all-remote) organization: 2002 to 2021
  39. Open-source and community-based organization: 2002 to 2007
  40. Algorithmically managed organization: 2015 to 2020
  41. Team of teams: 2015

Twenty-three of the 29 schools run to the present, and 16 of those began in the 1970s or later. The forms arrive in waves: 14 of the 41 first appear between 1985 and 1994.

Source: Fayol 1949; Weber 1978; Mayo 1933; Trist and Bamforth 1951, an interview-based field study in one industry; Burns and Stalker 1961, a field study of twenty firms; Mintzberg 1979; Child 1972a; Coase 1937; Meyer and Rowan 1977; Hannan and Freeman 1977; Galbraith 1973; Jensen and Meckling 1976; Crozier 1963; Emery and Thorsrud 1969; Yoshino 1968; Streeck 1984a; Kellogg and colleagues 2020, a review of earlier research through one theoretical lens, with no new data; Puranam and colleagues 2014, a conceptual argument; Chandler 1962, several independent case histories by a historian; Pfeffer and Salancik 1978; Simon 1946; Maurice and colleagues 1980, a comparison of matched factories in three countries; Levinthal 1997; Simon 1962; Braverman 1974; Hamilton and Biggart 1988, a comparison of three economies by scholars; Ward 1958; Czarniawska and Joerges 1995; Granovetter 1985.

Michael Tushman and Charles O'Reilly made it a design, arguing (1996) that firms facing incremental change punctuated by discontinuous change must pursue both kinds of innovation at once, and describing (2004) separate exploratory units tightly linked at the senior executive level. The problem was sustaining today's business while creating tomorrow's, when the two need opposite structures.

Where it fits

One rating, size, rests on evidence, the rest on theory or our judgment; three are contested, none is a Strain, and Partial means neither for nor against.

The ambidextrous organization against the twelve contingency factors

Each factor at its three levels, Fits, Partial, or Strain; select a row for its reason, sources, and grade.

Select a factor to read its reason, sources, basis, and grade.Size. A meta-analysis of 114 studies found that structural separation helps firms of all sizes balance exploration and exploitation, so size gives no reason against separate explore and exploit units. Basis: Evidence. Grade: several independent quantitative studies or a meta-analysis. Sources: O'Reilly and Tushman 2004; Gibson and Birkinshaw 2004; Jansen and colleagues 2009; Fourné and colleagues 2019.Age. The form protects new ventures from an established business's routines, so it is aimed at established and old firms, where the pull toward exploitation is strongest. Basis: Theory. Grade: conceptual only, no empirical test. Sources: O'Reilly and Tushman 2004; Tushman and O'Reilly 1996; March 1991; Hannan and Freeman 1984; Amburgey and colleagues 1993. View 1: Young partial, Established fits, Old fits. The form protects new ventures from an established business's routines, so it is aimed at established and old firms, where the pull toward exploitation is strongest. Sources: O'Reilly and Tushman 2004; Tushman and O'Reilly 1996; March 1991. View 2: Young partial, Established fits, Old partial. Inertia rises with age, and attempts at reorganization raise death rates, so reorganizing an old organization into this form is itself a hazard at the old level. Sources: Hannan and Freeman 1984.Environmental uncertainty (unpredictability, dynamism). Separate exploring units let a firm prepare for discontinuous change while exploiting today's business; in a stable environment they add cost with little to explore. Basis: Theory. Grade: conceptual only, no empirical test. Sources: Tushman and O'Reilly 1996; O'Reilly and Tushman 2004; Luger and colleagues 2018; Ossenbrink and colleagues 2019; March 1991; Tushman and colleagues 2010; Pennings 1975. View 1: Stable partial, Shifting fits, Unpredictable fits. Ambidextrous designs let firms facing punctuated change explore and exploit at once, and they executed innovation streams better than other designs. Sources: Tushman and O'Reilly 1996; O'Reilly and Tushman 2004; Tushman and colleagues 2010. View 2: Stable partial, Shifting fits, Unpredictable partial. Firms chose structural ambidexterity when opportunities called for culture and capabilities fundamentally different from their own, contextual ambidexterity when opportunities were numerous and uncertain, and a hybrid of the two when both held. Sources: Ossenbrink and colleagues 2019.Environmental complexity. Separating exploring from exploiting answers a change in technology or markets, not the number of factors in the environment, so complexity does not settle it. Basis: Judgment. Sources: Tushman and O'Reilly 1996; Ossenbrink and colleagues 2019.Munificence and hostility. Separating exploring from exploiting is not shown to pay more or less with resources: in the study nearest the point, structural differentiation did not condition the ambidexterity to performance link. Basis: Judgment. Sources: Jansen and colleagues 2012.Diversity of products and markets. Explore units do best when they can lean on the parent's existing strengths, which a single or related business supplies and an unrelated portfolio rarely does. Basis: Judgment. Sources: Hill and Birkinshaw 2012; O'Reilly and Tushman 2004.Strategy type. Separate exploit and explore units under one senior team suit the analyzer's mix of a stable core and new domains, but other studies reach the same mix inside one unit, across alliances and acquisitions, or in sequence. Basis: Judgment. Sources: Tushman and colleagues 2010; He and Wong 2004; Junni and colleagues 2013; Tushman and O'Reilly 1996; Miles and Snow 1978; Burton and colleagues 2020a; Gibson and Birkinshaw 2004; Stettner and Lavie 2013; Siggelkow and Levinthal 2003; Jansen and colleagues 2012; Gresov and Drazin 1997. View 1: Defender (exploitation) partial, Analyzer (exploitation and exploration) fits, Prospector (exploration) partial. Separate exploit and explore units under one senior team executed innovation streams, incremental and discontinuous innovation together, better than functional, cross-functional, and spinout designs, which matches the analyzer's mix of a stable core and new domains. Sources: Tushman and colleagues 2010; Tushman and O'Reilly 1996; He and Wong 2004; Miles and Snow 1978; Burton and colleagues 2020a. View 2: Defender (exploitation) partial, Analyzer (exploitation and exploration) partial, Prospector (exploration) partial. The analyzer's mix can be reached without separate internal units: through contextual ambidexterity inside a business unit, by exploring through alliances or acquisitions while exploiting internally, which did better than balancing within one mode, or by decentralizing for a time and then reintegrating; and structural differentiation did not condition the payoff of unit ambidexterity. Sources: Gibson and Birkinshaw 2004; Stettner and Lavie 2013; Siggelkow and Levinthal 2003; Jansen and colleagues 2012; Gresov and Drazin 1997.Technology and routineness. It houses routine work in its exploit units and novel work in its explore units, so it is built for an organization that has both, which a single level of this scale does not describe. Basis: Judgment. Sources: O'Reilly and Tushman 2004; Perrow 1967.Interdependence. Explore and exploit units kept apart and integrated mainly at the top suit units that only share resources; handoffs or back-and-forth work between them load the senior team, though cross-functional interfaces can carry part of the integration. Basis: Judgment. Sources: Jansen and colleagues 2012; O'Reilly and Tushman 2004; Thompson 1967; Jansen and colleagues 2009.Knowledge intensity and professionalization. Separating explore and exploit units answers a tension between today's and tomorrow's business, not a level of training, so this factor gives no reason for or against. Basis: Judgment. Sources: O'Reilly and Tushman 2004.Global integration pressure. No work read ties the ambidextrous form to integration pressure across countries, so this factor gives no reason for or against. Basis: Judgment. Sources: O'Reilly and Tushman 2004.Local responsiveness pressure. No work read ties the ambidextrous form to local responsiveness across countries, so this factor gives no reason for or against. Basis: Judgment. Sources: O'Reilly and Tushman 2004.

With the three contested cells counted by their first view, the 36 settings show Fits at 11 and Partial at 25, with none at Strain. Judgment is the basis of nine of the twelve cells.

Source: O'Reilly and Tushman 2004, a practitioner's report of a multi-unit comparison; Gibson and Birkinshaw 2004, a survey of 4,195 people in 41 business units; Jansen and colleagues 2009, a two-wave survey of 230 firms; Fourné and colleagues 2019, a meta-analysis of 114 studies; Tushman and O'Reilly 1996; March 1991, formal models, with no field data; Hannan and Freeman 1984; Amburgey and colleagues 1993, a quantitative study of organizational change and failure rates; Luger and colleagues 2018, a longitudinal study of many firms in one industry; Ossenbrink and colleagues 2019, four case studies by scholars; Tushman and colleagues 2010, a longitudinal field study of 13 business units; Pennings 1975, a quantitative field test of the structural-contingency model; Jansen and colleagues 2012, a quantitative study of 285 units; Hill and Birkinshaw 2012, a panel study of 95 venture units; He and Wong 2004, a statistical study of 206 firms; Junni and colleagues 2013, a meta-analysis of prior studies; Miles and Snow 1978; Burton and colleagues 2020a; Stettner and Lavie 2013, a quantitative study of 190 firms; Siggelkow and Levinthal 2003, a computer simulation, with no field data; Gresov and Drazin 1997; Perrow 1967; Thompson 1967.

Your own situation against the ambidextrous organization

Set any factor to outline its cell in the strip above; the lines under the switches give the counts, the form's place among the 41, and the misfit rules that fire.

Size
Age
Environmental uncertainty (unpredictability, dynamism)
Environmental complexity
Munificence and hostility
Diversity of products and markets
Strategy type
Technology and routineness
Interdependence
Knowledge intensity and professionalization
Global integration pressure
Local responsiveness pressure

Set: 0 of 12 factors. Fits on 0, Partial on 0, Strain on 0, Contested on 0.

Its place in the explorer's rankingEvery factor is on Any: all 41 forms tie, and the explorer lists them alphabetically.1st of 411st of 41, tied with one other1st of 41, tied with 0 othersSunk by a misfit rule: 1st of 41Sunk by a misfit rule: 1st of 41, tied with one otherSunk by a misfit rule: 1st of 41, tied with 0 others

  • Flagged (its sources disagree): A prospector bound by heavy rules. Disputed: Burton and Obel 2004; against it, Sine and colleagues 2006; Adler and Borys 1996.
  • Flagged (its sources disagree): A defender without written standards. Disputed: Burton and Obel 2004; against it, Mintzberg 1980.
  • Flagged (its sources disagree): An analyzer whose stable core has no standards. Disputed: Burton and Obel 2004; against it, Burton and Obel 2004; Mintzberg 1980.
  • Flagged (its sources disagree): A large organization with few written rules. Disputed: Burton and Obel 2004; against it, Mintzberg 1980.
  • Flagged (its sources disagree): A small organization with a large one's rules. Disputed: Burton and Obel 2004; against it, Sine and colleagues 2006.
  • Flagged (its sources disagree): Routine work without written standards. Disputed: Burton and Obel 2004; against it, Trist and colleagues 1963; Pasmore and colleagues 1982; Beekun 1989; against it, Hickson and colleagues 1969; against it, Child and Mansfield 1972.
  • Flagged (its sources disagree): Nonroutine work bound by procedures. Disputed: Burton and Obel 2004; against it, Adler and colleagues 1999; Adler and Borys 1996; against it, Hickson and colleagues 1969; Child and Mansfield 1972.
  • Flagged (its sources disagree): Few written rules in a stable environment. Disputed: Burton and Obel 2004; against it, Burton and Obel 2004; against it, Mintzberg 1980.
  • Flagged (its sources disagree): Heavy rules in an unpredictable environment. Disputed: Burton and Obel 2004; against it, Sine and colleagues 2006; Bigley and Roberts 2001; Adler and Borys 1996.
  • Flagged (its sources disagree): Reactor: a defender's structure kept after the environment changed. Disputed: Miles and Snow 1978; Burton and Obel 2004; against it, Sine and colleagues 2006; Bigley and Roberts 2001; Adler and Borys 1996.

Ranking rule: by count of Fits on the factors the reader has set, then fewest Strains, then alphabetical; a fired misfit rule that no source disputes sinks the form to the bottom when every level of the form's profile meets it, a disputed one, or one that meets the form only through one end of a spanned level, one view of a contested profile cell, or one part of a two-part form, is flagged on the row (with both views where it is disputed), and a misfit between two of the reader's own settings is shown once above the list; a contested cell counts only where its views agree at the reader's level, and otherwise shows as Contested; factors on "Any" count for nothing; with every factor on "Any" the list is alphabetical and says so.

Open these settings in the explorer

The misfit rules that can fire for this form

  • A prospector bound by heavy rules fires when Strategy type at Prospector (exploration): flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Sine and colleagues 2006; Adler and Borys 1996.
  • A defender without written standards fires when Strategy type at Defender (exploitation): flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Mintzberg 1980.
  • An analyzer whose stable core has no standards fires when Strategy type at Analyzer (exploitation and exploration): flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Burton and Obel 2004; Mintzberg 1980.
  • A large organization with few written rules fires when Size at Large: flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Mintzberg 1980.
  • A small organization with a large one's rules fires when Size at Small: flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Sine and colleagues 2006.
  • Routine work without written standards fires when Technology and routineness at Routine: flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Trist and colleagues 1963; Pasmore and colleagues 1982; Beekun 1989; against it, Hickson and colleagues 1969; against it, Child and Mansfield 1972.
  • Nonroutine work bound by procedures fires when Technology and routineness at Nonroutine: flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Adler and colleagues 1999; Adler and Borys 1996; against it, Hickson and colleagues 1969; Child and Mansfield 1972.
  • Few written rules in a stable environment fires when Environmental uncertainty (unpredictability, dynamism) at Stable: flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Burton and Obel 2004; against it, Mintzberg 1980.
  • Heavy rules in an unpredictable environment fires when Environmental uncertainty (unpredictability, dynamism) at Unpredictable: flagged (its sources disagree). Disputed: Burton and Obel 2004; against it, Sine and colleagues 2006; Bigley and Roberts 2001; Adler and Borys 1996.
  • Reactor: a defender's structure kept after the environment changed fires when Strategy type at Defender (exploitation) and Environmental uncertainty (unpredictability, dynamism) at Unpredictable: flagged (its sources disagree). Disputed: Miles and Snow 1978; Burton and Obel 2004; against it, Sine and colleagues 2006; Bigley and Roberts 2001; Adler and Borys 1996.

Of the 39 contingency misfit rules, ten can fire for this form, each through one end of its span on formalization. All ten are disputed, so each would flag the form and none would sink it.

Source: Burton and Obel 2004; Burton and colleagues 2020a.

  • Size. On evidence: fits at every level, since structural separation helped firms of all sizes balance exploration and exploitation, an outcome of balance, not performance (Fourné and colleagues 2019, a meta-analysis of 114 studies).
  • Age. Contested, theory. Partial young, fits established and old, where exploitation pulls hardest (O'Reilly and Tushman 2004; Tushman and O'Reilly 1996; March 1991). Or partial old too, since inertia rises with age and attempts at reorganization raise death rates (Hannan and Freeman 1984, theory), though Amburgey, Kelly, and Barnett (1993), a quantitative study of organizational change and failure rates, report that the effects of change on failure vary by type.
  • Environmental uncertainty. Contested, theory. Partial stable, our judgment from March (1991); fits shifting and unpredictable (Tushman and O'Reilly 1996; Tushman and colleagues 2010). Or partial unpredictable: in four case studies by scholars, firms chose contextual ambidexterity when opportunities were numerous and uncertain, a record of choices, not performance (Ossenbrink, Hoppmann, and Hoffmann 2019). Pennings (1975), a quantitative field test of the structural-contingency model, found environment and structure fit failed to explain effectiveness, a failure Donaldson (2001) attributes to method; it bears on the three environment misfit rules.
  • Diversity. Our judgment: fits single and related, partial unrelated; in a panel study of 95 venture units, units lasted longer building new capabilities while using the parent's strengths (Hill and Birkinshaw 2012).
  • Strategy type. Contested, our judgment. Fits the analyzer, partial elsewhere (Tushman and colleagues 2010; He and Wong 2004; Miles and Snow 1978; Burton, Obel, and Håkonsson 2020, the most developed scholarly fit model). Or partial throughout, since the mix can be reached inside one unit (Gibson and Birkinshaw 2004, the survey above), across modes (Stettner and Lavie 2013, a quantitative study of 190 firms), or in sequence (Siggelkow and Levinthal 2003, a computer simulation with no field data); differentiation did not condition the payoff (Jansen, Simsek, and Cao 2012, the study of 285 units); and designs can be functionally equivalent, a conceptual argument (Gresov and Drazin 1997).
  • Interdependence. Our judgment: fits pooled, partial sequential and reciprocal, since handoffs between units integrated mainly at the top load the senior team, though cross-functional interfaces can carry part (O'Reilly and Tushman 2004; Thompson 1967; Jansen and colleagues 2009).
  • Complexity, munificence, and technology. Our judgment: partial throughout (Ossenbrink and colleagues 2019; Jansen, Simsek, and Cao 2012, whose munificence is internal resources; Perrow 1967, since the form houses routine and novel work).
  • Knowledge intensity, global integration, and local responsiveness. Our judgment: partial throughout; no work read ties separation to training levels or international pressure (O'Reilly and Tushman 2004).

The dissent panel applies; the computational-design school's simulations show the best design depends on interactions among choices, which bivariate fit cannot capture (Rivkin and Siggelkow 2003).

What it does well

It protects breakthrough work from the business that pays for it. In O'Reilly and Tushman (2004), a practitioner's report of a multi-unit comparison of 35 attempts, more than 90 percent of ambidextrous designs reached their goals, against none of the cross-functional or unsupported teams and a quarter of functional designs, which matched them only for breakthroughs substituting for existing products. In Tushman and colleagues (2010), a longitudinal field study of 13 business units, ambidextrous designs executed innovation streams better than functional, cross-functional, and spinout designs. The two reports share authors, the form's originators among them, so neither independently confirms the other.

What it costs

You pay for two organizations, some functions duplicated, and a senior team able to run both. Integration must be built: in a two-wave survey of 230 firms, separation paid off through senior team integration and cross-functional interfaces (Jansen and colleagues 2009). In a quantitative study of 190 firms, balancing exploration and exploitation inside one mode, internal organization included, undermined performance (Stettner and Lavie 2013).

How it fails

Our judgment ties each mode to Coordination Failures.

How the ambidextrous organization fails, and the coordination failures behind it

The entry's failure modes at the left, each tied to the Coordination Failures patterns it names at the right; a pattern links its entry there.

Regulator overloadGoal displacement and metric gamingSubsystem self-perpetuationEnvelope-edge collapseSiloed knowledgeJurisdictional gap or overlap
Regulator overloadGoal displacement and metric gamingSubsystem self-perpetuationEnvelope-edge collapseSiloed knowledgeJurisdictional gap or overlap

How many ties the failure has. Highest: The senior team cannot hold the contradictionThe new unit is judged by the core business's yardsticksThe core business absorbs every senior hourA balance tuned for gradual change breaks at a discontinuityThe separated unit is cut off from what it needsInitiatives fit no unitRegulator overloadGoal displacement and metric gamingSubsystem self-perpetuationEnvelope-edge collapseSiloed knowledgeJurisdictional gap or overlap

Select a failure to read its place in the network.The senior team cannot hold the contradiction. Degree 1The new unit is judged by the core business's yardsticks. Degree 1The core business absorbs every senior hour. Degree 1A balance tuned for gradual change breaks at a discontinuity. Degree 1The separated unit is cut off from what it needs. Degree 1Initiatives fit no unit. Degree 1Regulator overload. Degree 1Goal displacement and metric gaming. Degree 1Subsystem self-perpetuation. Degree 1Envelope-edge collapse. Degree 1Siloed knowledge. Degree 1Jurisdictional gap or overlap. Degree 1
FailureDegree
The senior team cannot hold the contradiction1
The new unit is judged by the core business's yardsticks1
The core business absorbs every senior hour1
A balance tuned for gradual change breaks at a discontinuity1
The separated unit is cut off from what it needs1
Initiatives fit no unit1
Regulator overload1
Goal displacement and metric gaming1
Subsystem self-perpetuation1
Envelope-edge collapse1
Siloed knowledge1
Jurisdictional gap or overlap1

Six failure modes name six patterns, and no pattern is named by more than one.

Source: Tushman and colleagues 2010, a longitudinal field study of 13 business units; Thompson 1967; Jansen and colleagues 2009, a two-wave survey of 230 firms; Gibert and BenMahmoud-Jouini 2025, a longitudinal study of four units in established firms; March 1991, formal models, with no field data; Luger and colleagues 2018, a longitudinal study of many firms in one industry; Hill and Birkinshaw 2012, a panel study of 95 venture units; Fang and colleagues 2010; Burgelman 1983, one inductive case study.

The senior team cannot hold the contradiction. In a longitudinal field study of 13 business units, the senior team's handling of contradictory architectures was crucial (Tushman and colleagues 2010). Handoffs between units load it, our judgment from the interdependence cell: regulator overload.

The new unit is judged by the core's yardsticks. Gibert and BenMahmoud-Jouini (2025), a longitudinal study of four units in established firms, found exploration unit managers in tension over their charter and how the parent's performance management applied to them: goal displacement and metric gaming.

The core absorbs every senior hour. Adaptive processes refine exploitation faster than exploration and are likely to become self-destructive in the long run (March 1991, formal models with no field data): subsystem self-perpetuation.

A balance tuned for gradual change breaks at a discontinuity. In Luger, Raisch, and Schimmer (2018), a longitudinal study of many firms in one industry that measured an orientation, not separate units, ambidextrous firms kept reinforcing their balance, which helped under incremental change and hurt under discontinuous change: envelope-edge collapse.

The separated unit is cut off. In a panel study of 95 venture units, units lasted longer by becoming ambidextrous themselves, through strong ties to parent executives, business unit managers, and venture capitalists (Hill and Birkinshaw 2012): siloed knowledge.

Initiatives fit no unit. Management found operating-level initiatives hard to place (Burgelman 1983, one inductive case study): a jurisdictional gap.

Misfit rules. Spanning formalization, the form fires ten misfit rules, built on Burton and Obel's (2004) propositions; Burton, Lauridsen, and Obel (2002), a quantitative study of 224 Danish small and medium-sized firms, tested the earlier model's rules only as a set, finding lower return on assets with any misfit and no further loss from added ones. The reactor rule joins two factors; the other nine pair one factor with one structure, the interaction form of fit, unsupported where the selection and systems forms were supported (Drazin and Van de Ven 1985, a quantitative study of 629 work units in two states' employment security agencies, public-sector units, as Donaldson notes). Firms in quasi fit gained by improving either fit; those in perfect institutional fit slightly lost performance striving for contingency fit (Volberda and colleagues 2012, a survey of 3,259 respondents in 1,904 companies), so a firing rule does not by itself show that fit would help a firm whose structure its field treats as legitimate. All ten are disputed, so each flags with its views and none sinks:

  • A prospector bound by heavy rules, a small organization with a large one's rules, heavy rules in an unpredictable environment, and, our judgment from Miles and Snow (1978), a reactor keeping a stale defender's rules there: Sine, Mitsuhashi, and Kirsch (2006), a quantitative study of new ventures in one sector, where more formal founding teams outperformed organic ones; Adler and Borys (1996), a conceptual argument, on enabling formalization, except for the small organization; and Bigley and Roberts (2001), one inductive case study, for the two environment rules.
  • A defender without written standards, few written rules in a stable environment, and a large organization with few written rules: Mintzberg's (1980) professional bureaucracy, little formalized in large units and stable environments; Burton and Obel's own 2004 rule pointing the other way on the stable one; and the 2002 test, which held no large firms.
  • An analyzer whose stable core has no standards, a rule for the analyzer that does not innovate, carried by the table's merged level: the same 2004 book allows an innovating analyzer a simple configuration when small.
  • Routine work without written standards, and nonroutine work bound by procedures: the Aston reappraisal, where technology bore only on structure nearest the workflow (Hickson, Pugh, and Pheysey 1969, a multi-organization comparative survey), and size related to structure far more than technology (Child and Mansfield 1972, a multi-firm comparative study). For routine work, also self-regulating groups (Trist and others 1963, several field studies in one industry; Pasmore and colleagues 1982, a review of 134 reported experiments; Beekun 1989, a meta-analysis of 17 studies); for nonroutine work, NUMMI's metaroutines (Adler, Goldoftas, and Levine 1999, one case study of a single plant) and Adler and Borys. Limits: Mohr (1971), a quantitative field study of technology and structure, found no support for effectiveness resting on their consonance, his structure being participation in decisions, though Donaldson attributes this to method and reports later studies of the job unpredictability Mohr studied finding support; Glisson (1978), a path analysis of 30 human service organizations, found structure shaping routinization; and Schoonhoven (1981), a quantitative study of hospital operating room suites, found the traditional hypotheses unsupported, so the nonroutine threshold simplifies.

What the research found

Most evidence concerns ambidexterity in general, not separated units: Birkinshaw and Gupta (2013), a review, warn of unclear meaning and measurement, and Raisch and colleagues (2009), an editorial framing a special issue, list four open tensions.

For separation: the two reports above; Jansen and colleagues (2009), a two-wave survey of 230 firms, where separation worked through integration; and Fourné and colleagues (2019), a meta-analysis of 114 studies, the only meta-analysis here of structural separation: it helped balance, more so in high technology, though service firms did better without it.

Against or qualifying it: Jansen, Simsek, and Cao (2012), a quantitative study of 285 units, found differentiation did not condition the ambidexterity to performance link. Junni and colleagues (2013), a meta-analysis of prior studies, found the link moderated by industry, level, and method. Gibson and Birkinshaw (2004), a survey of 4,195 people in 41 business units, showed ambidexterity without separation. Stettner and Lavie (2013), a quantitative study of 190 firms, favored balance across modes; Luger and colleagues (2018), a longitudinal study of many firms in one industry, found the benefit contingent. Ossenbrink and colleagues (2019), four case studies by scholars, set structural and contextual approaches on one continuum. Boumgarden, Nickerson, and Zenger (2012), a reanalysis of several published cases that also finds the two approaches complements, and Siggelkow and Levinthal (2003), a computer simulation with no field data, are two dissents from simultaneous separation. In the simulation, with pervasive interactions among activities, temporary decentralization then reintegration beat permanent decentralization; it models decentralized divisions, not exploring and exploiting units.

Gibert and BenMahmoud-Jouini (2025), a longitudinal study of four units in established firms, is the most recent. It studies the separated unit's managers, not firm results, and shows separation negotiated over time.

The evidence ledger holds 21 studies filed under this form.

In practice: one published case

O'Reilly, Harreld, and Tushman (2009), in California Management Review, is one case study. It concerns IBM, as the source names it, and one of its three authors, J. Bruce Harreld, was an IBM executive.

IBM used its Emerging Business Opportunity process to explore new businesses alongside its existing ones. The authors present the process as a deliberate and repeatable version of multi-level selection, a frame taken from evolutionary theory.

On results, the authors report that the process enabled IBM to generate more than $15 billion in growth between 2000 and 2005.

The source concludes that multi-level selection processes, together with organizational ambidexterity and dynamic capabilities, may help organizations adapt to technological and market change and survive over long periods.

The record of the case rests on the abstract of the Stanford research paper version, and that abstract does not say how the opportunities were organized or how the growth was counted.

As evidence about the form, it covers one firm, in a frame wider than structural separation, and it does not show that separate units, rather than the process around them, produced the growth. The search ran through Crossref and the web for published cases of ambidextrous organizations and corporate venture units.

Source: O'Reilly and colleagues 2009. Organizational ambidexterity: IBM and emerging business opportunities, California Management Review 51(4), 75 to 99; one case study.

Variants and near relatives

Dual operating system (Kotter 2012, an article, and 2014, a practitioner book). A volunteer network for strategic initiatives, linked to the hierarchy and staffed from all its levels. Its volunteers serve in addition to their regular jobs, so they belong to the hierarchy and the network at once, where a Hypertext organization member belongs to one layer at a time; its antecedent, the collateral organization, is taught there.

Contextual ambidexterity (Gibson and Birkinshaw 2004). Individuals in one unit divide their own time between alignment and adaptability; the structure does not separate them.

Sequential ambidexterity (vacillation). The whole structure shifts over time, decentralizing to explore and then reintegrating (Siggelkow and Levinthal 2003, a computer simulation). Boumgarden, Nickerson, and Zenger (2012), a reanalysis of several published cases, find that vacillation may give higher long-run performance than ambidexterity, which helps at the margin within it; the two are complements.

New venture division (internal corporate venturing). A standing unit that hosts new ventures apart from the operating divisions; Burgelman (1983) studied one in the field and later (1984) set out alternative designs for them; Miles and Scaringella's (2012) futures group is a later, untested design of the kind.

Semi-isolated subgroups (Fang, Lee, and Schilling 2010). Moderate linking performed best in simulation.

Venture teams, skunkworks, and idea incubators (Daft 2016, 2021). Team-scale separation of exploring work, presented in a textbook.

The Hybrid structure, the nearest neighbor, sets different grouping bases side by side for efficiency, while this form separates units by learning mode and integrates them at the top. Unlike the Multidivisional (M-form), whose general office allocates capital and judges results, here the senior team holds two learning modes together.

Near relatives: the Adhocracy, a whole pool of innovating experts, though Parikh (2016), a literature review, argues that ambidexterity now covers it; the entrepreneurial university (Professional bureaucracy); amoeba management (Internal market organization), wrongly conflated with ambidexterity, argue Adler, Hiromoto, and Suzuki (2019), a conceptual review. Pointed to Change & Transformation: the promotor roles (Kieser and Kubicek 1992) and Daft's (1978) dual-core model.

The ambidextrous organization beside its nearest neighbor, the hybrid

Select a number to fill what it marks in both drawings; the list under them gives the dimensions on which the two profiles differ.

It keeps exploring and exploiting units structurally apart, each with its own processes, structures, and cultures, and integrates them at the top through a senior team that holds the contradiction, a shared intent, and targeted links; it does not mix grouping bases side by side for efficiency.

Ambidextrous organization

Targeted linksSenior teamSeparate unitsSenior teamExploitingunitExploitingunitExploringunitBy purpose: exploitBy purpose: exploreDN11223
Targeted linksSenior teamSeparate unitsSeniorExploitExploitExploreBy purpose: exploitBy purpose: exploreDN11223

Reports toAdjust to each other directlyDDirect supervisionNStandardization of normsDifferentiation contested: one unit can explore and exploit at once, with no separate unit (Gibson and Birkinshaw 2004).

Hybrid structure

Strategic apexFunctions anddivisionsChiefexecutiveFunctionaldepartmentFunctionaldepartmentProductdivisionProductdivisionBy functionBy productWO11223
Strategic apexFunctions and divisionsChiefFunctionFunctionDivisionDivisionBy functionBy productWO11223

Reports toAdjust to each other directlyWStandardization of work processesOStandardization of outputs

  1. Each ambidextrous unit is grouped by purpose, exploiting or exploring, and holds the functions its mode of learning needs (our judgment); the hybrid sets central functions beside product divisions.
  2. The senior team joins the two sides by direct supervision and a shared intent, our judgment; the hybrid's departments use standardized work and its divisions standardized outputs, also our judgment.
  3. The ambidextrous units meet through formal cross-functional interfaces: task forces and teams (Jansen and colleagues 2009); the hybrid's functions and divisions settle their dealings directly, our judgment.

Differs on 7 of 11 dimensions

  • Basis of grouping Product or purpose; Hybrid structure: Function or process and product or purpose
  • Standardization and formalization Low and high; Hybrid structure: Medium
  • Decision rights and the locus of authority Limited vertical decentralization; Hybrid structure: Limited horizontal decentralization to limited vertical decentralization
  • Prime coordinating mechanism Direct supervision and standardization of norms; Hybrid structure: Standardization of work processes and standardization of outputs
  • Lateral mechanisms Task forces and teams; Hybrid structure: Direct contact and liaison roles
  • Differentiation against integration High differentiation, heavy integration; Hybrid structure: Moderate differentiation, some integrating roles
  • Boundary permeability Selective; Hybrid structure: Closed

Each sets more than one kind of unit under one top. The ambidextrous form keeps exploring units apart from exploiting units, each with its own processes, structures, and cultures, and joins them at the senior team; the hybrid sets functional departments beside product divisions for efficiency.

Source: O'Reilly and Tushman 2004, a practitioner's report of a multi-unit comparison; Jansen and colleagues 2009, a two-wave survey of 230 firms; Daft 1995; Mintzberg 1980, a synthesis of earlier research, offering propositions without a new test; Dessein and colleagues 2010, a formal model with no new data.

Moving into or out of it

Our reading: moving in means giving the exploring work its own processes, culture, and head under the senior team, and building integration at the top (O'Reilly and Tushman 2004; Jansen and colleagues 2009). In O'Reilly and Tushman's (2004) practitioner's report, seven of eight initiatives moved into the design improved substantially; in Tushman and colleagues' (2010) longitudinal field study of 13 business units, transitions in went with more innovation, shifts away with less. In four case studies by scholars of alliances, not separate units, frontline managers defined a charter from below, complementing the mandated one (Zimmermann, Raisch, and Birkinshaw 2015). Hannan and Freeman (1984) argue, as theory, that attempts at reorganization increase death rates, though Amburgey and colleagues (1993), a quantitative study of organizational change and failure rates, report effects varying by type. Change & Transformation covers running the reorganization.

The thinking behind it

Robert Duncan (1976): dual structures and the term, at second hand. James G. March (1991): exploration and exploitation. Michael Tushman (1996, 2010): the design, a practitioner-facing argument whose systematic evidence is mixed, and its test; buffering (Benner and Tushman 2003). Charles O'Reilly (2009, 2013): ambidexterity at IBM, and a review. Cristina Gibson and Julian Birkinshaw (2004): contextual ambidexterity. Justin Jansen (2009): integrating separated units, later qualified. Nicolaj Siggelkow (2003, with Levinthal): temporary decentralization; with Rivkin (2003), balancing search and stability. Peter Boumgarden (2012): vacillation. Robert Burgelman (1983, 1984): internal corporate venturing and designs for corporate entrepreneurship. Sebastian Raisch (2009, 2018): the open tensions and dynamic balancing. Richard Burton (2004, with Obel): the misfit rules.

Also called

Structural ambidexterity (Ossenbrink and colleagues 2019). Dual structure, after Duncan (1976). Organizational ambidexterity, the wider term the reviews use (O'Reilly and Tushman 2013).

Works cited

On this page, “read” means read by this series: where a line says a work, or part of one, was not read, the series did not read it.

Back to the explorer