Reading Your Own Structure

Some organizations come close to one of the explorer's forms and others mix several; the explorer is built for both.

Why so many organizations are hybrids

Here a hybrid is any organization that mixes forms; the explorer's hybrid structure is one such mix, units on different bases side by side, each under a single boss. Mintzberg (1980), a synthesis of earlier research offering propositions without a new test, set out five configurations, each resting on one prime coordinating mechanism. He claimed that effective organizations favor a logically consistent configuration, while contradictory pressures or transitions drive some to hybrids. He later qualified the pure types: some forms, he expects (Mintzberg 1991, a practitioner article), are most common in hybrids, and a 1992 book version of a similar text (Mintzberg 1992) finds no truly effective organization in pure form. Most recently he wrote that many organizations come close to one of four forms, while others are hybrids of them or in transition among them (Mintzberg 2024, a conceptual article).

The measurement tradition reached a similar place: the Aston group, in a multi-organization comparative survey of 52 work organizations in England, built scales for five dimensions of structure and found that they resolved into four underlying factors, not one; the authors concluded that the idea of a single bureaucratic type was no longer useful (Pugh and others 1968). Hinings and Lee (1971), a small multi-firm replication in nine manufacturing organizations, found the same groupings: specialization, formalization, and standardization went together and with size, while centralization went with dependence on other organizations. Wagner (2021), a meta-analysis of 346 correlations from 155 published articles, found that many links among these dimensions have weakened since an earlier meta-analysis found them significantly intercorrelated, which he reads as failing to give consistent support to the bureaucratic tenet that they go together. How structuring relates to centralization is disputed: Donaldson, Child, and Aldrich (1975), a discussion ungraded here, report moderately high negative correlations between the two in Child's national study of 82 organizations, lower ones at Aston, and on that relation call the Aston results somewhat discrepant with both that study and Hinings and Lee's. The machine bureaucracy's profile is contested on that point.

Whether a hybrid is a lapse or a choice is unsettled. Doty, Glick, and Huber (1993), a quantitative test across a sample of organizations, found, against their own expectations, that organizations resembling Mintzberg's ideal types were not more effective, while fit to Miles and Snow's types predicted 24 percent of the variance in effectiveness. Ketchen and colleagues (1997), a meta-analysis of 40 tests, found that configuration does partly explain performance, though it pooled many strategy types with structures, and tests of pure structural types stay mixed. Krabberød (2015), reviewing the 218 articles that cited the Doty test, argues that the test has not been shown to be exhaustive. Gresov and Drazin (1997) make theoretical room for a mixed answer: different designs can be functionally equivalent, so more than one can fit.

Against mixing, Williamson (1975) doubted that a mixed enterprise, with divisions run variously as a holding company, as an M-form, and under close supervision, could reliably be expected to last, though he called the judgment that it lacks rationality properties probably too strong. Only our reading ties two other works to mixing. Dessein, Garicano, and Gertner (2010), a formal model with no new data, take business units paired with company-wide functions as given and show the pairing builds in a conflict: functional managers want too much standardization, unit managers too much local adaptation. Whittington and colleagues (1999), a panel survey of European firms, found that piecemeal changes in structure, processes, and boundaries, information technology excepted, delivered little performance benefit, while the few firms that made the full set enjoyed high-performance premia; it tested sets of changes, not mixed forms.

Setting the profile

Every form is profiled on eleven dimensions, each built on a named source; where the source gives no bands, the bands are this site's reading. Set your organization on the same scales:

• Basis of grouping. What are the units just below the top formed around: a function or process, a product or purpose, a client or market segment, a place (Gulick 1937; Mintzberg 1979), or two or more at once, as in the combined forms Galbraith (1971) describes? Thompson's rule, as Van de Ven, Ganco, and Hinings (2013) cite it, is to group the most interdependent positions together first.
• Specialization. How many functions are held by full-time specialists: low, medium, or high (our bands on the scale of Pugh and others 1968)?
• Standardization and formalization. How much of the work is covered by written procedures, job descriptions, and checked rules: low, medium, or high (Pugh and others 1968; Hage and Aiken 1967; our bands)?
• Layers and span of control. Flat, moderate, or tall: our bands, drawn on Graicunas's count of a superior's relationships (1933) and Jaques's strata of time span (1990).
• Decision rights. Where formal and informal power sits: at the apex; at the apex, with informal power flowing to the analysts who write the standards; delegated to the heads of market-based units; spread selectively to groups at many levels; or spread down the line and out to the operators (Mintzberg 1980).
• Prime coordinating mechanism. Mutual adjustment, direct supervision, or standardization of work processes, outputs, skills, or norms: the one the organization leans on most (Mintzberg 1979, 1989).
• Lateral mechanisms. The highest rung of Galbraith's ladder in regular use: hierarchy, rules, and goals only; direct contact and liaison roles; task forces and teams; integrating and managerial linking roles, the latter with budget power; or a matrix (Galbraith 1973, 1974; the first level is his base, and pairing his seven lateral processes into the other four is ours).
• Differentiation against integration. How sharply units differ in goals, time horizons, interpersonal style, and formality, and what holds them together (Lawrence and Lorsch 1967).
• Mode of exchange. Market, bureaucracy, or clan (Ouchi 1980, a conceptual article).
• Boundary permeability. How freely work, information, and people cross the vertical, horizontal, external, and geographic boundaries (Ashkenas, Ulrich, Jick, and Kerr 1995): closed, selective, or permeable, in our bands.
• Centralization. Most decisions at the top, mixed, or low, in our bands on the Aston scale. It is profiled but not counted, since decision rights carry the vertical aspect.

First, decide which of two things you are describing: the chart or the practice. Lincoln, Hanada, and McBride (1986), a survey of 55 American and 51 Japanese manufacturing plants, found the Japanese plants had greater formal but less de facto centralization. Aghion and Tirole (1997), a formal model, show that real authority, unlike the formal right to decide, follows information. Wulf (2012), summarizing for managers her panel studies with interviews in seven firms, argues that flattened firms show more control at the top; the panel study itself (Rajan and Wulf 2006, more than 300 large firms) read the flattening partly as delegation, not simply centralization or decentralization. Meyer and Rowan (1977), a work of theory, hold that formal structures are often adopted for legitimacy and decoupled from the work. Fligstein (1985), a multi-firm historical statistical study, and Palmer, Jennings, and Zhou (1993), a multi-firm statistical study of adoption, found imitation among the causes of the multidivisional form's spread. Donaldson (1995), a book-length critique, argues the evidence for fit-driven adjustment is stronger, and Kraatz and Zajac (1996), a longitudinal study of 631 private liberal arts colleges, found little support for neoinstitutional predictions, though the change they studied was curricular. The form profiles describe formal designs, mostly by our reading of their sources, so they sit closer to the chart; if your practice differs, profile both.

Second, if parts of the organization answer differently, one level for the whole hides them. Our reading: profile each major unit, read the forms each is nearest, and treat the whole as their arrangement.

Reading the nearest forms

The explorer's printed rule: for each form, count the ten counted dimensions on which your level equals the form's level; where the form spans two levels, a match on either counts; ties are shown as ties; nothing is weighted.

The levels are not equally firm: of the 451 profile cells, 301 rest on judgment, 137 on theory, and 13 on evidence, and 27 are contested. A match against a judgment cell is a match against our reading, and the counts below take each contested cell at its first view.

Where the sources set no level, a profile carries a judgment middle or span: the co-determined, distributed, and boundaryless forms overlay any operating structure, and the hybrid structure mixes mechanisms unit by unit. Process organization, project-based organization, and the distributed form each span two levels on five counted dimensions, most of those spans by our judgment. A high count against a broad profile says less than the same count against a narrow one.

Nine pairs overlap on all ten counted dimensions, so some profiles match both equally, among them the professional bureaucracy and the professional partnership. No pair is identical: other profiles separate each, and so does each entry's distinguishing feature.

Illustrative. A manufacturer with three product divisions, profiled as product grouping, medium specialization and formalization, moderate layers, power delegated to division heads, standardized outputs, direct contact and liaison roles, moderate differentiation, bureaucratic exchange, and closed boundaries, matches the hybrid structure on all ten counted dimensions and the multidivisional form on nine, though the hybrid's profile is mostly judgment middles. Under the second view of the multidivisional's contested lateral-mechanisms cell, from one historical case study and a survey of many firms, the liaison roles match too, a tie at ten. A high-volume plant inside one division (functional grouping, high specialization and formalization, tall layers, informal power with the analysts, standardized work, no designed lateral links, low differentiation, bureaucratic exchange, closed boundaries) matches the machine bureaucracy on all ten, with the functional structure and two others at seven. The functional count falls to six under the second view of its differentiation cell, from the same Lawrence and Lorsch study, and the algorithmically managed count under the second view of its decision-rights cell, a judgment from two studies of platform work. The research lab (functional grouping, high specialization, low formalization, moderate layers, power spread selectively, mutual adjustment, task forces and teams, moderate differentiation, bureaucratic exchange, selective boundaries) ties the boundaryless and co-determined forms at eight. Under the second view of its contested boundary-permeability cell, from two case studies, boundaryless leads alone at nine.

A count says which forms your structure looks like, not which fit your situation. Fit is the explorer's situation view, which uses only the selection form of fit: which structures go with which situations, not what a pairing adds to performance. The method itself is disputed. We read the Aston factors and Wagner's meta-analysis, which show some dimensions varying apart from others, as a reason to count them one at a time; Wagner himself concludes that dimensional research on structure now lacks a viable theoretical foundation. Configuration theory holds that the elements of a structure reinforce one another, and Meyer, Tsui, and Hinings (1993), an editorial overview, set the configurational approach apart from the contingency approach; our reading is that a count of separate matches can miss how those elements work together. Population ecology doubts adjustment itself: Hannan and Freeman (1984), a work of theory, hold that inertia keeps most organizations from adapting, so selection produces the correlation of structure with context. Donaldson (1995) replies that misfit lowers performance and prompts incremental adjustment toward fit, and that ecology understates managers' capacity to adapt; Donaldson (2001), that fits form a continuum, with most organizations at intermediate points, not in a few configurations.

Structure, process, people, or culture

The explorer covers macro structure; process and workflow design is in Workflow Topologies, job design and pay in People & Talent, and culture as such in Culture & Organizational Development. No study read here gives a sorting test; the records offer signs.

Some symptoms name no form. A strategy that keeps missing, whatever the structure, is in theory a misfit between contingencies that no structure removes (Burton and Obel 2004, the most developed scholarly fit model), and Child's (1972) strategic choice reads its remedy as a change of strategy, domain, or standard. A unit whose every redesign fixes one problem and creates another, because part of its work calls for fixed procedures and part for working things out as people go, is our description of the conflicting demands behind what Gresov and Drazin call tradeoff equifinality: several designs answer them, each at its own cost. A design blocked by laws, agreements with employees, or deeply embedded systems fails the feasibility test of Goold and Campbell (2002), a practitioner checklist, not a tested model.

Process or grouping. Work that stalls at the handoffs between departments has structural remedies, and the symptom list is contested on which ones. One view names process organization, the matrix, and team-based organization, its case for process organization resting on Hammer and Champy (1993), a practitioner book with no test. The other drops process organization: in Byrne and colleagues (2004), a survey of every medical center in one national health system, joined to administrative data, service lines, which group clinical work across departments, had no positive and several negative effects; taking service lines as a process organization is our reading. Our reading, from Thompson's rule: if the most interdependent work already sits inside one unit and still stalls, look first at the flow; if it straddles units, the grouping is a candidate.

People or design. Goold and Campbell count the people test among their four fit tests: "Does the design adequately reflect the motivations, strengths, and weaknesses of the available people?" The reverse also happens: a formal change that people have not yet enacted. Lee (2024), one ethnographic case study, concludes that decentralization is an ongoing collective process, not a one-time structural change.

Culture or differentiation. Departments that differ in goals, time horizons, and style can look like a culture clash. Lawrence and Lorsch (1967), a multi-firm comparative field study of ten organizations in three industries, found that the high performers matched that differentiation to the diversity of their environments and integrated with devices proportioned to it; the sample is small and the measures are managers' perceptions. Tests of the contingency line it belongs to are split. Drazin and Van de Ven (1985), a quantitative study of 629 work units in two states' employment security agencies, supported the selection and systems forms of fit but not the interaction form. Pennings (1975), a quantitative field test of the structural-contingency model, found that fit did not explain effectiveness, a failure Donaldson (2001) puts down to methodological limitations. Schoonhoven (1981), a quantitative study of hospital operating room suites, found traditional contingency hypotheses unsupported but more precise interactive, nonmonotonic ones better supported. Our reading: such a difference is a design fact to integrate.

When the signs point to structure, the entries give each form's costs and failures; Change & Transformation covers the change.