Organizational design frameworks: Galbraith Star, McKinsey 7S, Mintzberg, spans & layers, holacracy
The five org-design lenses every HR pro should know — Galbraith's Star, McKinsey 7S, Mintzberg's five configurations, spans & layers math, and holacracy vs…
On this page▾
- Org design = deciding who decides what, who reports to whom, and how work flows. Five frameworks dominate the field.
- Galbraith's Star Model: strategy → structure → processes → rewards → people. Use it as a diagnostic — if any point is misaligned, the org will drift.
- McKinsey 7S: hard elements (strategy, structure, systems) + soft (shared values, style, staff, skills). Use it to check whether a restructure will actually stick.
- Mintzberg's 5 configurations (simple, machine, professional, divisional, adhocracy) explain *why* startups feel like adhocracies and Fortune 500s feel like machines.
- Spans & Layers: the math behind 'too many managers'. Healthy span 6–10 ICs per manager; healthy layers ≤ 7 between CEO and frontline. Most orgs fail layers, not spans.
- Holacracy works for ~50–200 people doing knowledge work with high trust. Below 50 it's overhead; above 200 it collapses without strong governance discipline.
Organizational design is the most under-taught skill in HR. It is also the one that separates an HRBP who 'runs processes' from one who shapes how a company actually works. This guide covers the five frameworks every people leader should be fluent in — what they are, when to use them, and where each one breaks.
Org design is the deliberate choice of who decides what, who works with whom, and how work flows. Every other people decision — hiring, leveling, comp, performance — is downstream of org design.
Why org design is HR's job
When a company isn't shipping, the CEO usually says 'we need to hire harder' or 'we need to reorg'. Both are symptoms. The underlying question is almost always: have we drawn the boxes and lines wrong? Org design answers that. HR owns it because no one else has both the people data and the cross-functional view.
The five frameworks below are not competitors — they're different lenses. Galbraith tells you what to align. 7S tells you whether change will stick. Mintzberg tells you what kind of organism you are. Spans & layers tells you whether the plumbing works. Holacracy tells you whether to throw out the hierarchy entirely.
1. Galbraith's Star Model
Developed by Jay Galbraith at MIT in the 1970s, the Star Model is the most-used diagnostic in serious org-design practice. It says any organization is the alignment of five points. Misalign one and the system drifts.
- 1StrategyWhat we're trying to win at. Differentiation? Cost leadership? Speed? Strategy dictates everything below.
- 2StructureReporting lines, business units, geographies. Functional? Divisional? Matrix? Should follow strategy, not precede it.
- 3ProcessesHow decisions flow, how information moves, how work hands off. Includes RACI, governance, planning cycles.
- 4RewardsComp, recognition, promotion criteria. Whatever you reward is what you'll get more of — regardless of what strategy says.
- 5PeopleCapabilities, levels, hiring criteria, development. The org is only as capable as the humans inside it.
Pick any goal your company keeps missing. Walk the five points. One of them is misaligned. (Most often: rewards. You say 'collaboration matters' but you only promote individual heroics.)
Galbraith's most useful insight: structure follows strategy, not the other way around. Most reorgs fail because someone redesigned the org chart before redesigning the strategy — they're now optimized for last year's bet.
2. McKinsey 7S
Created at McKinsey in 1980 by Tom Peters, Robert Waterman, and Julien Phillips. The 7S framework extends Galbraith by separating hard from soft elements. Use it when a change is planned and you want to know whether it'll actually stick.
- Strategy — the plan
- Structure — the org chart
- Systems — processes, tools, workflows
- Shared values — the centre; the core beliefs that bind the rest
- Style — how leaders lead, what they tolerate
- Staff — who's actually here, their capability and culture
- Skills — collective competencies, what the org is known for
The 7S insight: a restructure that changes only the hard elements will not deliver the result. If you flip the structure from functional to product-led but the shared values still reward individual function-leader heroics, the matrix will collapse within 18 months. Every senior HR person has lived this story.
Use 7S as a pre-mortem. Before a reorg, score each S today and in the target state. The gaps tell you what change-management work you have to do *before* the restructure, not after.
3. Mintzberg's five configurations
Henry Mintzberg (McGill, 1979) catalogued every type of organization into five archetypes. This is the most useful framework for explaining why your org *feels* the way it does.
| Configuration | Coordination mechanism | Best example | Typical pathology |
|---|---|---|---|
| Simple structure | Direct supervision by founder/owner | Startup of 5–30 people, family business | Doesn't scale past the founder's attention bandwidth |
| Machine bureaucracy | Standardization of work processes | Manufacturing, fast-food chain, government agency | Slow, rigid, hostile to innovation |
| Professional bureaucracy | Standardization of skills (long training) | Hospitals, universities, law firms | Individual professionals over-power admin; reform is slow |
| Divisional form | Standardization of outputs (P&L targets) | GE, Unilever, multi-BU conglomerates | Duplicated overhead; corporate becomes a holding company |
| Adhocracy | Mutual adjustment (everyone talks to everyone) | Ad agencies, R&D labs, early-stage startups, IDEO | Doesn't scale; burnout from constant coordination |
Why this matters for HR: each configuration needs a different HR operating model. A machine bureaucracy needs robust grading, job descriptions, and union relations. A professional bureaucracy needs partner-track frameworks and dual ladders. An adhocracy needs fluid roles and tolerance for ambiguity — kill the job description.
Most startups are adhocracies that try to scale by becoming machine bureaucracies overnight (usually after a Series C). The right answer is usually professional bureaucracy with adhocracy pockets in R&D — but very few CHROs know how to design for that.
4. Spans & layers (the math)
Span of control = how many direct reports one manager has. Layers = how many levels between the CEO and the frontline IC. These two numbers determine more about how a company feels than any vision statement.
The two diagnostics
- Span check: count how many of your managers have fewer than 4 direct reports. If more than 20% of managers do, you have too many managers. They will invent work to justify themselves — usually meetings.
- Layers check: take any frontline IC. Count the layers up to the CEO. If it's more than 7 in a company under 5,000 people, decisions are taking 3x longer than they should.
Companies almost always restructure by adding a layer ('we need a VP of X to coordinate'). The right answer is usually the opposite: widen spans and remove a layer. Painful, political, correct.
5. Holacracy vs hierarchy
Holacracy, created by Brian Robertson and popularised by Zappos in 2013, replaces job titles and reporting lines with 'circles' of accountabilities. The promise: faster decisions, less politics, more autonomy. The reality is more nuanced.
- Clear reporting lines; one boss per person
- Decision rights flow with the org chart
- Slower change; easier to onboard outsiders
- Politics is about access to your boss's boss
- Works at any scale, from 5 to 500,000
- Roles (not jobs) are filled by people; one person holds many roles
- Decisions made within a circle by 'integrative decision-making'
- Faster pivots once people internalise the rules
- Politics is about controlling the governance meetings
- Works best 50–250 people, knowledge work, high trust. Often breaks above 300.
The honest take from companies that tried it (Zappos, Medium, Github, Buurtzorg): holacracy front-loads enormous training cost and demands extreme governance discipline. Medium reverted in 2016. Zappos retained it but admits 18% of employees left during the transition. Buurtzorg (Dutch home-care, 14,000 nurses) is the canonical success — because the work is intrinsically autonomous and the model fits.
Knowledge work, 50–250 people, high trust, work that genuinely benefits from autonomous decisions. Otherwise, use a flatter hierarchy with strong RACI — you'll get 80% of the benefit without the governance overhead.
How to pick the right framework
| If your question is… | Reach for… |
|---|---|
| Why isn't our strategy working? | Galbraith Star — find the misaligned point |
| Will this reorg actually stick? | McKinsey 7S — score current vs target on soft elements |
| Why does our org feel like X? | Mintzberg — identify your archetype, accept it, design for it |
| Do we have too many managers/layers? | Spans & layers diagnostic — count the numbers |
| Should we go flat / self-managing? | Holacracy lens — but only if you fit the 50–250 / knowledge / trust criteria |
Most senior HR practitioners run Galbraith continuously, deploy 7S before a planned change, use Mintzberg in their head to set expectations, count spans & layers once a quarter, and quietly avoid the holacracy debate until the right moment.
FAQ
Frequently asked questions
Is 'matrix' a Mintzberg configuration?
No. Matrix is a structural choice that sits inside Mintzberg's divisional or adhocracy types. It's not a configuration on its own — it's a coordination overlay, usually added when functional excellence and product focus both matter.
What's the difference between Galbraith and 7S?
Galbraith is older and tighter (5 points; strategy is causal). 7S is broader (7 elements; shared values are central). In practice, use Galbraith for diagnosis and 7S for change planning.
How often should HR redesign the org?
Lightly every year; deeply every 2–3 years or after a major strategic shift. Continuous tinkering destroys trust; never reorganizing accumulates structural debt.
Read next
All playbooksJohn Kotter found that most change efforts fail because steps are skipped. Here's the sequence, where most leaders break it, and the diagnostic to find which…
Kotter and Lewin work at the org level. ADKAR is the model for the human level — what each person needs to actually change, and which letter is your current…
Five major change models compared on a single page — Kotter's 8 steps, ADKAR, Lewin's unfreeze-change-refreeze, Bridges' transitions, and the Satir change…