Utilisation rate

The instinct is that unbillable time is waste and the target should be as close to full as possible. That instinct produces agencies that cannot pitch, cannot train anybody, and cannot absorb a single problem without something else failing.

Consider what lives in the gap. New business: proposals, pitches, the conversations that fill next year. Internal work: the case studies, the process improvements, the hiring. Training: the reason a junior becomes useful. And slack, which is what absorbs the account that goes wrong, and every agency has one going wrong at any given moment.

None of those is optional, and all of them are unbillable. An agency at ninety-five per cent utilisation has not eliminated them; it has moved them into evenings, which is a different thing entirely and one that shows up later as turnover.

The usual targets reflect this. Delivery staff somewhere in the seventies, with more senior people lower because management, mentoring and client relationships are real work that nobody bills for directly. An agency whose account directors are at eighty-five per cent billable is not managing anything.

Sustained high utilisation is the most reliable leading indicator an agency has for people leaving, and it arrives months before the conversation that confirms it.

The mechanism is simple enough. A person running at ninety per cent billable for a quarter has no room for anything unexpected, and the unexpected is constant. The only available absorber is their own time, so the overrun goes into evenings and weekends, and it goes there quietly because everybody is busy.

The economics are worth stating plainly, because they are frequently inverted. Replacing a mid-level person costs months of recruitment, several months of reduced output while they learn the accounts, and a quantity of client relationship risk that is hard to price. Protecting a few percentage points of utilisation is not worth any of that.

Which makes utilisation a number to watch from both ends. Too low and the agency is paying for capacity it is not using. Too high and it is borrowing against people it will have to replace, at a rate it never recorded.

Utilisation measures whether time was charged to a client. It says nothing about whether that time was charged at a rate that covers what it cost.

An agency can be fully utilised and losing money on every account. Underpriced retainers produce exactly this: the team is busy, the timesheets are full, the utilisation report is green, and the gross margin per client is thin or negative. Every hour is billable and none of them helps.

The number that completes the picture is per-client gross margin, and it is the one most agencies never calculate. Utilisation answers whether capacity is being used; margin answers whether using it was worth doing.

Read together they diagnose. High utilisation and poor margin is a pricing problem. Low utilisation and good margin is a sales problem. Both low is a business with two problems, and reading either number alone will send the work at the wrong one.

The first is the denominator. Measuring against hours worked rather than hours available makes overtime invisible: somebody who worked fifty hours and billed forty shows eighty per cent, and the ten hours of unpaid overrun have vanished from the report. Against contracted hours less holiday, the same week shows over a hundred per cent, which is the signal you wanted.

The second is the numerator. Time spent fixing the agency own mistakes is not billable work, however much a client is technically paying for it. Counting it flatters utilisation and margin together, and it conceals the rework that was the actual problem.

Both errors run in the same direction, which is the dangerous part: they make a struggling team look efficient. An agency using utilisation to decide whether to hire will conclude it has room when it does not.

The correction is a definition written down and applied consistently, and a rework category in the time tracking that is explicitly not billable. Neither is difficult; both are usually missing.