Counting what is easy to count produces a defensible structure that nobody tested against the plan.
Why it matters when the plan changes
Structural decisions are among the most consequential a leadership team makes and are usually made from precedent, benchmark and preference. Bringing evidence to them is a real improvement. The risk is that the available data measures the organisation's shape, its span of control, layer count and cost per head, rather than its behaviour, so the design optimises the countable and leaves the causal untouched. Formal design is only the first of several layers; real ownership and dependencies sit beneath it and rarely appear in a benchmark.
The tension is between measurability and relevance. Layers, spans and cost per head are precise, comparable and only loosely related to whether a plan gets delivered. Decision routes, dependency structure and where work crosses boundaries are causal and harder to assemble, so most exercises use the first set and describe themselves with the second's language. Harvard Business Review reported in 2015 that only 9% of managers can rely on colleagues in other functions all the time, evidence of where delivery fails.
In practice
A design exercise benchmarks spans and layers against peers, finds the organisation has one more layer than the median, and removes it. The plan's actual constraint is three decisions that cross a boundary the design did not change. The structure is now benchmark-compliant and the constraint is untouched.
Evidence
Span of control is the direct determinant of layer count and is the most readily measured structural property.
Span of control, Wikipedia (2026)Cross-unit commitments are the least reliable part of execution, which is what structural benchmarks do not measure.
Donald Sull, Rebecca Homkes and Charles Sull, Why Strategy Execution Unravels and What to Do About It, Harvard Business Review (2015)
What it cannot tell you
Data-driven organisational design as commonly practised measures shape, spans, layers, cost per head, not behaviour. It cannot tell you whether decisions sit with the right people, where dependencies cross boundaries, or whether the plan will be delivered. A structure can be fully benchmark-compliant while the causal constraint on delivery remains completely untouched.
Questions
Which decisions the plan depends on and who currently holds them, which interfaces the work crosses, where decisions queue or reopen, and where capacity is already committed. All are assemblable from records that exist, and none of them is a staff number.
Because they are available, comparable and easy to defend in a board paper. Harvard Business Review reported in 2015 that only 9% of managers can rely on colleagues in other functions all the time, a gap benchmarks of spans and layers never surface. Their comparability is what makes them attractive; their distance from execution is what makes them weak.
No, as a prompt. A structure well outside peer norms is worth explaining, and the explanation is sometimes a deliberate choice and sometimes drift. The error is treating the benchmark as the answer rather than as a question about why this organisation differs.
Whether the proposed structure puts the decisions the plan depends on with people who can make them, and whether it reduces interfaces on the critical path. Span of control, as described in the Wikipedia entry updated in 2026, determines layer count but says nothing about whether those decisions sit with the right people.
The target model is the output; this is the question of what evidence shaped it. A model produced from benchmark data describes a defensible shape, and one produced from what the plan demands describes an arrangement that might deliver it.