Span of control is decided in a spreadsheet during a reorg and discovered eighteen months later in an exit interview. The number usually comes from whoever was available to manage rather than from any view of how much managing the work needs.
The interesting part is that over-extension shows up in observable patterns well before it shows up in attrition, and almost nobody looks.
Why the standard numbers do not help
You can find authoritative-sounding recommendations for six, eight, ten, fifteen. They are not wrong so much as detached from the variable that matters.
What consumes a manager's week is not the headcount, it is the coordination load per report. Eight senior engineers working on independent, well-specified problems need a fraction of the attention that five juniors on a shared codebase need. A manager with reports across four time zones spends more of their week on the same number of people than one with everyone in a single office — the time zones playbook covers why overlap design dominates here.
So the honest formulation is not "how many reports" but "does this manager still have time to do the parts of the job that only they can do". Those parts are 1:1s that are not status updates, unblocking, hiring, and thinking about where the team is going. They are also the first things to be cancelled when the week is full, because they have no external deadline.
The signals, and the order they arrive in
Over-extension has a fairly consistent progression, and each stage is measurable.
Meeting load climbs and focus time collapses. The manager's calendar fills with the per-report overhead — 1:1s, their reports' planning sessions, escalations — and the remaining time fragments into gaps too short to use. This happens first and it is the cleanest early signal. If a manager's largest uninterrupted block has fallen below an hour, they have no time to think, whatever the headcount says.
Individual contribution stops entirely. For engineering managers this is often reported as a positive — they have "stepped up into management" — and sometimes it is. But a manager who has gone from some hands-on work to none in a quarter, without that being a deliberate decision, has usually been squeezed rather than promoted.
1:1 cadence degrades. Cancelled, shortened, or moved so often that they become status reporting. This is the symptom with the most direct line to attrition, because the 1:1 is where someone says they are unhappy before they say it in a resignation.
After-hours work rises. The manager's own work, the thinking, the writing, the hiring loop feedback, gets pushed into the evening because the day is fully booked. This is the latest signal and the one closest to burnout — the early burnout signals piece covers reading it.
Then the team's numbers move. Delivery slows, or quality drops, or people leave. By this point the cause is several months old.
What remote work did to this
Spans that worked in an office quietly got harder without anyone changing the number.
In a co-located team a manager picks up a great deal for free: who looks stuck, who has gone quiet, who two people are arguing with. None of that survives distribution. It has to be replaced with scheduled contact, which means the same eight reports now consume meaningfully more calendar.
Most organisations never adjusted. They moved to distributed work, kept the org chart, and the compression landed on managers as a slow increase in meeting load that nobody attributed to the cause. If you want one place to look for a structural problem created in the last few years, this is a good candidate.
Deciding what to do
The data tells you a manager is over-extended. It does not tell you which fix applies, and there are three.
Add a layer when the headcount is genuinely too high for one person and the work is divisible. The cost is a layer, which has its own communication overhead; do not do this reflexively.
Reduce scope when the problem is that the manager owns four concurrent projects rather than eight people. This is frequently the real cause and is much cheaper to fix. Worth checking before you restructure anything.
Reduce coordination load when the team's work is more entangled than it needs to be. If every change requires three people to agree, the manager is the integration point for that, and fixing the entanglement helps more than changing who reports to whom. Capacity planning covers working the distribution question properly.
What rarely works is leaving the structure alone and asking the manager to prioritise better. They are already prioritising; that is why the 1:1s are being cancelled.
Where ProdView fits
ProdView measures the observable symptoms: meeting load, focus time and fragmentation, working-hours and after-hours patterns, and how load distributes across a team. Aggregate-first, activity metadata only, never screen content, with everyone seeing the same dashboard their manager does. For this question it is the manager's own week you are reading, which is a workload diagnosis rather than a performance one.
When not to pick us: 1:1 cadence and whether those conversations are any good lives in your calendar and in people's heads, not in activity data. We can show you that the meetings shrank; we cannot tell you what stopped being said in them.
Try before you commit
Three seats are free forever — enough to look at one over-stretched manager's week before deciding whether the problem is headcount or scope. The meeting cost calculator prices the coordination overhead if the answer turns out to be scope.