Research Agency Utilisation Rates: Why You Are Fully Booked and Still Missing Budget

Here is a conversation that happens in insight agencies more often than anyone admits.

It is the end of the quarter. Everyone has been flat out. Late debriefs, a topline deck built over a weekend, a fieldwork window that slipped and swallowed two people's diaries. Nobody has had a quiet week since February.

Then the numbers come in, and the margin is down.

The instinct is to look for a villain. A bad project. A difficult client. A pricing mistake on the tracker. Usually there is not one. What there is instead is a utilisation problem nobody could see, because nobody was measuring the right thing, against the right denominator, early enough to do anything about it.

Utilisation is the least glamorous number in an agency and the one that quietly decides whether you make money. This piece covers how to calculate it properly, what "good" actually looks like for a research agency (which is not what it looks like for a creative agency), and the three numbers that matter more than utilisation once you have got it right.

What utilisation actually measures, and where agencies fool themselves

The formula is not the hard part:

Utilisation = billable hours ÷ available hours

The hard part is that most agencies get both halves of that fraction wrong, in opposite directions, and the two errors cancel each other out just enough to look plausible.

Error one: the denominator

Plenty of agencies divide by contracted hours. A full-time researcher on 37.5 hours a week gives you 1,950 hours a year, and that becomes the denominator. It is a tidy number and it is fiction. Nobody is available for 1,950 hours.

Take the same researcher and work it through honestly:

• 1,950 gross contracted hours
• less 25 days annual leave (187.5 hours)
• less 8 bank holidays (60 hours)
• less roughly 5 days of sickness, training and all-agency time (37.5 hours)
• = 1,665 genuinely available hours

If that researcher bills 1,100 hours, they are at 56% against contracted hours and 66% against available hours. Same person, same year, ten points of difference and two completely different conversations. One of them triggers a performance review that should not be happening. The other tells you the truth.

Error two: the numerator

The second trap is subtler and more expensive. Billable is not the same as billed.

Hours logged against a project code are not hours a client paid for. If you scoped a project at 200 hours and delivered it in 260, every one of those 260 hours looks billable in your time data. Utilisation reads beautifully. The margin on that project has gone.

This is why utilisation on its own is a dangerous number to manage against, and why the section further down on realisation rate matters more than this one.

What good looks like, by role

There is a reasonable amount of published data on utilisation in professional services generally. SPI Research's 2025 Professional Services Maturity Benchmark, covering just over 400 firms, put billable utilisation at 68.9% for 2024, its lowest reading in five years and below the 75% the sector generally treats as healthy. Broader guidance tends to place research and consultancy services around the 70% mark, and treats sustained utilisation above 85% as a warning sign rather than a win.

Those are whole-firm numbers across professional services. They are a useful anchor, but a blended agency figure hides the thing you actually need to manage, which is that a research executive and a research director should not be anywhere near the same number.

A blended agency figure of 60 to 65% is realistic for a boutique carrying a heavy pitch load. Larger firms with dedicated operations support can reasonably plan for closer to 70%.

A necessary caveat: these are planning ranges we use when we sit down with agencies, not a published industry benchmark. The MR sector has no equivalent of SPI's professional services dataset, and anyone presenting one should be asked where it came from. Treat the table as a starting hypothesis to argue with, not a scorecard to be marked against.

One thing matters more than the absolute number anyway: direction of travel. A researcher who has sat at 85% for two consecutive quarters is not your best performer. They are a resignation risk with a good attitude.

Why creative agency benchmarks do not transfer

Almost every article and tool in this space is built for creative, marketing and software agencies. The underlying assumption is that work burns down at a roughly even rate across a project's life. Research does not behave that way, and five specific differences break the model.

1. The fieldwork gap

A project can be live for ten weeks and consume sixty hours of researcher time, most of it clustered in weeks one and two and then again in weeks eight to ten. In between, that project is "active" in every status report and consuming almost nothing. Plan capacity against project duration and you will systematically believe you are busier than you are, then wonder why nobody could take the extra piece of work that would have paid for the quarter.

2. Qual and quant load in opposite directions

A quant project front-loads hard: questionnaire design, scripting, soft launch checks. Then it goes quiet through fieldwork before a heavy analysis and reporting tail. Qual is more evenly loaded but carries immovable spikes, because a group recruited for Tuesday happens on Tuesday.

Two projects with identical fee values can have entirely different capacity profiles. It also means an agency with a balanced qual and quant book genuinely has more usable capacity at the same headcount than a pure-qual shop, because the peaks interleave. Very few agencies model this, and it is one of the more valuable things to know when deciding what to pitch for.

3. Recruitment lag is capacity you have already spent

When a recruiter slips a week, the researcher's calendar does not refill itself. That week is usually lost. It never appears in a utilisation report because those hours were never logged against anything. It shows up months later as a number nobody can explain.

4. The proposal tax

Research agencies write a lot of proposals, and the serious ones are substantial: methodology, sample design, often a costed field quote. A senior researcher can lose eight to ten hours to a single competitive pitch. At a one-in-four win rate, three quarters of that time is non-billable by definition. That is not overhead leakage, it is a predictable and plannable cost of being in the market, and it belongs in the capacity model rather than in the "where did the week go" conversation.

5. The debrief tail

The project is finished when the report ships. Then the client comes back six weeks later wanting a cut by region for their board, and a couple of slides reworked for a different audience. Nobody logs it against the project. It happens on essentially every study, and across a year it is a meaningful share of a senior person's time.

Three numbers that matter more than utilisation

Once utilisation is calculated honestly, it becomes a diagnostic rather than a target. These three do more work.

Realisation rate

Utilisation tells you how busy people were. Realisation tells you how much of that the client actually paid for.

Realisation = fee earned ÷ (hours worked × standard rate)

An agency running 75% utilisation and 80% realisation has an effective utilisation of 60%. That gap is where over-servicing lives, and most agencies have never put a number on it. The quickest way in: take your last ten completed projects and compare scoped hours to delivered hours. If you are consistently 20% or more over, you do not have a capacity problem. You have a scoping problem, and adding people will make it more expensive rather than better.

Ninety-day headroom

Not "are we busy right now" but "how many sellable hours exist in the next quarter, and where are they sitting".

This is the question an MD asks constantly, in the form of "can we take this on?", and it is almost always answered from memory by whoever has the best sense of the diary. It should be answerable in about ten seconds, broken down by month and by person, with provisional pipeline included at a weighted probability. Getting this right is what lets you say yes to the right work in May instead of scrambling in June.

Concentration risk

How much of next quarter's committed revenue depends on one or two individuals?

Every agency has someone who is the only person who can run the healthcare tracker, or moderate in the client's second language, or handle the account that pays for a third of the overhead. If that person is at 85% and climbing, that is not a strong utilisation figure. It is a single point of failure with a fortnight booked off in October.

Why the spreadsheet cannot answer any of this

None of these calculations is difficult. The problem is not the maths, it is keeping the inputs true and connected. A spreadsheet can hold every one of these numbers. What it cannot do is stay honest without someone paying constant attention.

• It records intent, not reality. Someone allocates Sarah sixty hours in October. October happens differently. Nobody goes back and reconciles it, so the historical data you would need to plan next year is quietly wrong.
• It does not know what changed. When fieldwork slips a week, the downstream analysis and reporting time should shift with it. The sheet does not move on its own, so the shape of the next two months is wrong from the moment the recruiter emails.
• It cannot answer a "what if". "If we win the tracker, are we covered?" requires a model you can run twice. A snapshot cannot do it, so the answer comes down to instinct.
• The person maintaining it is the constraint. It is usually one operations manager, which means the quality of your capacity planning is capped by how much of their week they can spend on maintenance. That is exactly the time you also need them spending on delivery.

This is not really a failure of spreadsheets. Spreadsheets are excellent at recording things. They are poor at reconciling things, and capacity planning is a reconciliation problem wearing a recording problem's clothes.

The version you can do this week

You do not need to buy anything to find out where you stand. Set aside an hour with your last two quarters of data.

1. Fix the denominator, once, properly. For each person: contracted hours, less leave actually taken, less bank holidays, less training and all-agency time. That is their available hours. Most agencies have never written this number down.
2. Divide. Pull billable hours per person from wherever they currently live and calculate real utilisation. Compare against the table above. Flag anyone above 85% and anyone below 50%. Both are problems, and only one of them looks like one.
3. Check your realisation. Last ten completed projects, scoped hours against delivered hours. The percentage over is your over-servicing rate. Above 15% and scoping is the thing to fix first.
4. Then the hard one. For the next ninety days, write down committed hours by person by month, and available hours by person by month. Subtract one from the other.

Step four is where most agencies stop, and it is the only step that changes a decision. The first three tell you about a quarter that has already been billed. The fourth tells you whether you can say yes to the thing that lands in your inbox on Thursday.

If you cannot complete step four in under an hour, that is not a failure of the exercise. That is the finding.

What you are actually aiming for

Utilisation is not a number to maximise. An agency running at 90% is not winning. It is one resignation, one slipped fieldwork window or one difficult client away from a bad quarter, with no absorption left anywhere in the system.

What you are aiming for is a number you trust, that updates without anybody being a hero about it, far enough ahead that it changes what you do rather than explaining what already happened. Most agencies have the first part and neither of the other two.

If you have read this and recognised your own operation, we would be glad to walk you through how we build capacity planning for insight agencies. We would look at your team structure, your qual and quant mix, and how you currently answer the "can we take this on" question, and show you what it would look like built around that. No generic product tour.