Span of Control

What does a recorded span leave out?

Everything that does not sit beneath the manager in the reporting field. Dotted-line and matrixed reports are excluded, so a manager who shares direction of a function across regions carries a load the record never shows. Contractors, agency staff and outsourced teams are usually excluded too, even where they are managed daily and their work is the manager's to answer for. So are the people a manager effectively leads without any line at all: a founder-era colleague who still routes decisions through them, a team whose nominal manager defers on everything that matters. The recorded span and the felt span therefore diverge, often widely, and the recorded one is what gets analyzed, presented and used to justify a change. Anyone reading spans should know which of these their data includes before drawing a conclusion from it.

What actually determines an appropriate span?

Four things, none of which is company-wide. How variable and non-routine the work is: a manager overseeing standardized, repeatable output can hold far more people in view than one whose reports each face a different unfamiliar problem every week. How experienced and autonomous the reports are: a team that needs direction consumes management time that a team of long-tenured specialists does not. How much individual delivery work the manager still carries, since a working manager has only a fraction of a week to spend managing. And how much the reports depend on the manager to coordinate between them, because work that has to be sequenced across the team routes through one person by default. These differ enough between two teams in the same company that any single number applied across both will be too wide in one place and too narrow in the other.

What is the trade-off between spans and layers?

They move against each other. For a fixed population, widening spans removes layers, and removing a layer shortens the distance between a decision and the work it affects, which is usually the reason the change is made. What the change does not do is make the management work disappear. Coaching, prioritizing, unblocking, reviewing, hiring and handling the difficult conversations still have to happen; they are simply divided across fewer people. That is a transfer of load onto managers' attention, not a saving, and the transfer is invisible on a chart, which shows only that a box has gone. The cost surfaces later and somewhere else, as slower feedback, thinner development, decisions waiting on one calendar, and eventually the departure of people who never had a manager with time for them. Naming the transfer when the decision is taken is what keeps it honest.

How should span be read across a whole organization?

As a distribution, not an average. The average is the least informative figure available here, because it is produced by two opposite problems canceling each other out. A company can report a perfectly reasonable middle number while containing both managers with a single report and managers with more people than anyone could plausibly support. Sorting every manager by span and looking at the ends of that sorted list takes minutes and tells you far more than any summary. The two ends are different problems with different remedies and should not be discussed together. Reading the shape also survives disagreement about what a correct span is: whatever range a function believes in, the managers well outside it are still the ones worth a conversation.

The single-report manager is worth understanding before it is corrected, because the reasons vary and only some are errors. A title was given in place of a raise the budget could not carry. A structure was left behind after a reorganization and nobody closed it out. A senior individual contributor was given one report so their level would make sense on paper. A team shrank through attrition and its manager was never re-pointed. Each needs a different response, and removing the layer is right for only some. The wide end deserves the same care. Pulling the reporting field straight from the [employee database](/employee-database-software) rather than from a maintained slide is what makes either reading trustworthy, since a stale structure produces confident wrong answers at both ends.

Why does a target span produce the wrong changes?

Because it is applied where it is easiest to apply, not where it is warranted. A target arrives as a company-wide instruction, and the units that meet it first are the ones whose work was already routine enough to support wide teams. The units that cannot meet it honestly meet it anyway, by re-pointing reporting lines so the record satisfies the rule while the actual supervision stays where it was. That leaves a structure that looks compliant and describes nobody's real working relationships, which is worse than the original problem because every later analysis reads the fiction. The other predictable outcome is that the target is met by deleting the narrow spans, since those are the visible, countable savings, and whether that layer was doing anything is a question a numeric target never asks.

The workable alternative is to set the expectation locally and test it against evidence. Each function agrees what its own managers are expected to spend time on, which fixes what a supportable span means there, and states the range as a guide rather than a threshold anybody is measured against. Outliers are then examined individually, and the examination asks what the manager actually does rather than how many boxes hang beneath them. Modeling a proposed shape before committing helps, since the effect of collapsing a layer is easier to argue about when the resulting spans can be seen. That is straightforward where structure lives in [HCM software](/hcm-software) alongside position and reporting data, and painful where it lives in a deck. Either way the baseline is the organization's own distribution over time, not a borrowed range.

What happens to a manager's attention as their span widens?

It divides, and not evenly. A manager with more reports than they can hold does not give each a proportionally smaller share; they give almost everything to whoever is loudest, newest or in trouble, and the quietly competent get nothing. That is a rational response to scarcity, and it is invisible in any structural view, because the reports appear beneath the same box. The people it costs most are those performing steadily in the middle, who receive no development, no feedback they can act on, and no advocacy when opportunities are allocated. They rarely complain, since nothing has gone wrong, and they leave for a role where someone will pay attention. The signal to look for is not workload but coverage: which reports the manager can describe in specific terms, and which only by role.

Widening a span is survivable when something else absorbs what the manager can no longer do, and unsurvivable when nothing does. Reports experienced enough to run without direction absorb a great deal, which is why the same span works in a senior team and fails in a graduate one. Senior individuals who take on coordination without a title absorb more, until they ask why they do a manager's work without the standing. Written practice absorbs some, because a team with clear expectations needs less individual steering. Where none of those exist, the load stays with the manager and comes out of whatever they did before, usually their own delivery work first and their people work second. Deciding in advance which absorber is being relied on turns a widened span from a hope into a plan.

Why does span get raised in cost exercises rather than design ones?

Because it converts easily into a headcount argument. Managers are countable, they cost more than the people beneath them, and a span number turns a structural question into an arithmetic one a spreadsheet can settle. So the topic tends to surface when a company is looking for reductions rather than when it is designing how work should run, and the two conversations reach different answers from the same data. A cost exercise asks how few managers the structure can be made to contain. A design exercise asks what each manager is expected to do and how many people that allows them to do it for. Only the second produces a number anyone can defend afterwards, and only the second leaves a record of the reasoning the next reorganization will inherit.

There is also a class of organization where the measure explains very little. Where work runs through projects, products or accounts, a person's day is shaped by whoever is directing the work this quarter, and the formal manager may do nothing more than hold the pay and performance relationship. Their recorded span can be large without corresponding to any supervision load, while the project leads carrying the real coordination appear to manage nobody. Reading that structure by span alone produces conclusions that are precisely backwards. The check is cheap: compare the reporting data held in [org chart software](/org-chart-software) against how work is actually assigned, and see whether the two describe the same organization. Where they do not, span measures the administrative tree rather than management effort.