Agency Utilization Rate: How to Measure and Fix It

Utilization rate is the number agency owners quote most confidently and define least carefully. It is one of the five metrics I use to run a delivery week, and I covered it in passing in my piece on the KPIs that actually run agency delivery. This is the deep version: how to calculate it honestly, what a rising number is really telling you, and the levers that move it without damaging the work or the people doing it.

My basis for that is twelve years running delivery operations for marketing agencies and 92 completed engagements. One thing to flag before you read on: there is no benchmark percentage anywhere in this article, and the section on targets explains why.

What is agency utilization rate, and how do I calculate it?

Utilization rate is the share of your team’s available working time that goes to billable client work. Billable hours over available hours, per person, then rolled up. The formula is trivial. Everything difficult about this metric lives in the denominator.

You have three plausible denominators, and they produce three different numbers from the same timesheets:

• Contracted hours. What people are paid for on paper. Stable, simple, and flattering, because it ignores holiday and pretends nobody is ever away.
• Available hours. Contracted hours minus holiday, sick leave and public holidays. Harder to compute, and the only version that answers the capacity question you are actually asking.
• Hours actually worked, overtime included. This is the one nobody publishes, because when a team is quietly working late it is the only denominator that shows it. A figure that looks healthy against contracted hours can be built entirely out of unpaid evenings.

Pick one, write it down, and use the same one every quarter. My preference is available hours as the headline figure, with hours actually worked computed quietly alongside it. The gap between those two is a burnout gauge, and it costs nothing once you are already collecting the data.

Which hours count as billable?

This is where the metric usually goes soft. In the KPI piece I called it definition drift. The deeper problem is that every agency has a list of grey categories, and if leadership never rules on them, each person rules for themselves. Your rolled-up figure is then an average of a dozen private definitions.

The categories that need an explicit ruling, in my experience:

• Rework and revisions beyond the agreed round. Hours that look billable and earn nothing.
• Pitch and proposal work, which is client-facing, unpaid, and easy to log against whichever account is open.
• Account management, status calls and client email.
• Internal review and QA on client deliverables.
• Onboarding, training and internal tooling work.
• Time spent on your own agency’s marketing.

There is no universally correct ruling on any of those. There is only a written one. The test I apply: given the same week, would two people on your team produce the same timesheet? If not, you do not have a metric, you have a mood.

Is a high utilization rate a good thing?

Not on its own, and this is the misreading that costs the most money. Utilization tells you that hours were pointed at client work. It tells you nothing about whether those hours produced anything sellable.

Three ways a rising number hides a falling business:

• Rework inflates it. Producing the same deliverable twice books twice the hours against the same fee. Utilization climbs while margin falls.
• Over-servicing inflates it. Hours poured into a fixed-fee account past the point of scope are billable by category and unpaid in fact.
• Under-priced work inflates it. A service line sold too cheaply keeps everybody fully occupied on work that cannot fund the agency. Full and unprofitable is a real state, and it is the one utilization is worst at detecting.

So read utilization next to what you actually collected. Track billable hours logged against hours invoiced and paid for, which professional services firms call realization, and the difference between the two is your write-off. An agency with a strong utilization figure and a weak realization figure is not busy earning. It is busy giving work away, and every hour of that still counts as billable in the top-line number.

Why does quality fall as utilization climbs?

Because slack is what absorbs variability, and a schedule booked to the edge has none. Agency work is full of variability: a sick day, a client approval that lands four days late, an urgent request from your biggest account. When there is room in the week, those events get absorbed. When there is not, every one of them turns into a missed date or a rushed deliverable, and waiting time rises far faster than the extra load would suggest.

The second effect is worse because it is invisible on the dashboard. The activities that get cut first when everyone is fully booked are the unbooked ones: QA, documentation, training, and the quiet thinking time that catches a bad brief before production starts. Those are precisely the activities that prevent rework. Cut them and rework rises, rework books more billable hours, and utilization goes up again. The metric rewards the behaviour that is degrading the work.

That loop is the reason I treat utilization as a dial with a ceiling rather than a score to chase.

What utilization target should I set?

Your own trend, and I will be direct about why that is the only answer I will give. I have no benchmark data on record, and a percentage I invented for a healthy utilization rate would be quoted in somebody’s leadership meeting a week later as though it had been measured. So here is the construction instead of a number.

1. Baseline one full quarter with the definitions frozen in writing. Nothing you measure before that point is comparable to anything.
2. Look at the weekly spread, not the quarterly average. The average conceals the fortnight that broke people.
3. Set a band with a floor and a ceiling. The floor protects the business. The ceiling protects the work, and almost nobody sets one.
4. Judge direction against companion metrics. Utilization rising while rework and cycle time hold flat is a genuine gain. Utilization rising while either of those climbs is borrowed, and you repay it next quarter.

One caveat: compare like periods. A seasonal client base produces a swing that says nothing about how well the agency is run, and reading it as a performance signal leads to hiring at the wrong moment.

What actually moves utilization without burning the team?

The levers, in the order I would pull them:

1. Fix intake first. Most non-billable time is created upstream of the people logging it. A brief that enters production without a signed-off exit condition generates rework, chasing and rescoping, and every hour of that lands on somebody’s timesheet weeks later.
2. Attack waiting time, not working time. Split cycle time into working and waiting, and go after the waiting half. Asking people to work faster raises utilization on paper and raises rework in practice.
3. Limit work in progress. A person holding six live projects loses hours to switching that appear on no line item. Fewer concurrent projects per person usually raises both utilization and throughput, which sounds contradictory until you watch it happen.
4. Automate the assembly, not the craft. The recoverable non-billable hours are in status reporting, timesheet chasing, report building and copying data between systems. I work inside ClickUp, Asana, Monday.com, Airtable and Notion, and automate across them with n8n, and that is where the hours come back. Nothing there speeds up the actual work, which is the point.
5. Change the mix or the price. The denominator lever nobody uses. If one service line consumes disproportionate capacity for its fee, repricing it or declining it moves your utilization economics more than any scheduling change will.
6. Buy flex capacity for peaks. Subcontracting the spike protects the ceiling without adding permanent salary to the denominator, which is what makes the next quiet month survivable.
7. Protect deliberate non-billable time instead of deleting it. Training, QA and process work are the cheapest rework prevention available. Cutting them raises the number this quarter and costs you the next one.

Notice that only one of those seven is a scheduling change. Utilization is an output of how the agency is designed, which is why I treat it as a systems question rather than a resourcing one when I work with an agency as an operations consultant.

Should I hold individuals to a utilization target?

No, and this is the fastest way to ruin the metric. A person’s utilization mostly measures what got assigned to them, and assigning work is a management responsibility. Publishing a per-person leaderboard converts a capacity signal into a personal performance grade, and the rational response to a personal grade is to log hours that make the grade look right. You will have destroyed the data inside a quarter.

Look at it per person by all means, in private, as a workload distribution check. One person consistently at the ceiling while the team sits below it is a scheduling failure, not a character difference.

So what is my verdict on utilization rate?

Measure it, define it in writing, and never manage to it on its own. Utilization is a capacity dial, and the only decision it should trigger by itself is whether to hire, subcontract or decline the next project. It cannot see quality, it cannot see profit, and it rises for bad reasons as readily as good ones. Read it beside rework, realization and cycle time, hold it against your own frozen baseline rather than an imported figure, give it a ceiling as well as a floor, and keep it off individual scorecards. That is the whole of my position, and it sits inside the wider system I set out in my agency operations guide.

Where to start with your own utilization number

If you want an outside read on what your figure is actually made of, that is part of what the Agency Ops Audit covers. It is $1,500 fixed, runs two weeks, and ends in a written 90-day roadmap and a 60-minute readout call. If your capacity is fine and the real problem is pricing or intake, I would rather say so in the readout than sell you a scheduling project.