How Agency Overhead Actually Lands on a Project

Of the three components of true project cost, overhead is the one with the strangest status. Labour at real cost gets computed badly but at least gets computed. Unbilled scope gets forgiven but at least gets noticed. Overhead — routinely a quarter of an agency's entire cost base — most commonly gets handled by not being handled: subtracted from the year's revenue in one line, allocated to nothing, owned by no project, visible in no project review.

The result is an agency where every project looks healthy and the company doesn't. Forty projects report 45–55% margins; the year lands at 8%; and the gap between those two facts has no owner, because the cost that explains it was never put anywhere anyone looks. This is the second of the three standard mistakes in margin accounting, and unlike the other two it has a genuinely clean fix. Here is the method we ran at a 120-person agency for two years, and the two places it needs care.

What counts as overhead (and what doesn't)

First, the boundary. Overhead is what the agency pays because the agency exists: rent and facilities, the salaries of people who don't bill (operations, finance, marketing, most of leadership), company-wide software, insurance, recruitment marketing, the long tail of unavoidable small costs.

It is specifically not the per-person costs that follow an individual — their laptop, their licences, their employer's NI. Those belong inside each person's fully-loaded hourly cost. The rule that keeps the two apart: if the cost would stop when a specific person left, it lives in their rate; if it would stop only when the agency closed, it is overhead. Keeping the boundary clean matters because the two pools are managed differently — rates through hiring and pay decisions, overhead through the harder conversation about what the company carries.

For scale: a 50-person agency typically carries £150–220k of monthly overhead; a 100-person agency £350–500k. If your figure is proportionally far above these ranges, the allocation method below will faithfully report the consequence — which is rather the point.

The method: pro-rata to revenue, monthly

The allocation itself is deliberately unclever:

1. Take the month's actual overhead — one number, from the management accounts.
2. Take each project's share of the month's revenue.
3. Each project absorbs that share of the overhead, for that month.
4. A project's lifetime overhead allocation is the sum of its monthly absorptions.

If the agency turned over £600k in April and a project produced £60k of it, the project absorbs 10% of April's overhead — £20k, if overhead ran at £200k. Nothing more sophisticated is needed, and sophistication here is usually a warning sign: allocation by headcount, by hours, or by negotiated "fairness" all invite gaming and argument. Revenue share is blunt, proportional to benefit, and impossible to lobby.

Why monthly rather than annually: an annual allocation against full-year revenue smooths the slow months out of existence — and the slow months are exactly where the information is. A project that ran through February needs to carry February's reality.

The slow-month problem, and the floor

Run the naïve method for a while and one month will produce an absurdity. Turnover halves in a quiet August; the only projects running are two small retainers; and the formula dutifully assigns each of them a monstrous share of the full overhead. The retainers report catastrophic margins, their owners object — correctly — that they did nothing different, and confidence in the whole system dies in one review meeting.

The fix is a minimum-revenue floor. Establish the agency's break-even monthly turnover — the revenue level at which the agency covers its costs — and use it as the denominator whenever actual turnover falls below it. Projects then absorb their share of the floor, not of the shrunken actual month. The arithmetic consequence is exactly right: in a slow month, the projects still running absorb a fair share of overhead, and the unabsorbed remainder — the true cost of the slow month itself — stays visible at company level as the price of an underfilled pipeline, instead of being smeared across whichever clients were loyal enough to still be there.

One project, two months

Take a £60k/month retainer at an agency with £200k monthly overhead and a £500k break-even floor.

Strong month — agency turns over £600k. The retainer is 10% of turnover and absorbs £20k of overhead. With labour at real cost running, say, £26k, its real margin for the month is (£60k − £26k − £20k) / £60k ≈ 23% — a healthy month, honestly measured.

Weak month — agency turns over £400k. Naïve pro-rata would charge the retainer 15% of overhead (£30k) and crush its margin to ~7%, for sins entirely not its own. With the floor: the denominator is £500k, the retainer absorbs 12% — £24k — and reports ~17%. The remaining £40k of that month's overhead (the share nobody's revenue covered) stays at company level, labelled what it actually is: the cost of a £100k pipeline shortfall.

The floor is the difference between an allocation system the delivery leads trust and one they spend review meetings litigating. It is a one-line refinement, and it is not optional.

The two classic mistakes

Allocating nothing. The default, and the expensive one. Without allocation, project margins run 25–30 points flattered, and every downstream decision — pricing, renewals, which work to seek more of — inherits the flattery. If your projects report in the fifties and your year lands in single digits, this is almost certainly why.

Baking overhead into hourly rates. The over-correction: load the whole overhead pool into everyone's cost rate and let allocation happen implicitly through hours. It feels tidy and it fails twice. It distorts — overhead lands in proportion to hours logged rather than revenue, so labour-heavy low-rate work absorbs more company cost than a high-value engagement that used fewer hours. And it hides — once overhead lives inside forty individual rates, it stops existing as a number anyone reviews, and an overhead problem quietly becomes a "our people seem expensive" problem, which sends leadership hunting in exactly the wrong place.

Overhead should be visible, allocated, and boring: one pool, one method, one floor.

Keeping it running without a spreadsheet ceremony

Everything above fits in a spreadsheet — for one month. The reason agencies abandon it is not the arithmetic but the upkeep: revenue shares shift, projects start and finish mid-month, the overhead figure lands weeks later with the close, and by the time the sheet is reconciled it is reporting on a month everyone has stopped managing.

This is the part Saldo automates: overhead accrues onto live projects pro-rata as the month runs, the floor applies itself in the slow months, and each project's margin on screen already contains its honest share of the company it belongs to — continuously, next to labour at real cost and unbilled scope. The 15-minute demo shows the full allocation on your own Jira data, and pricing is a flat monthly fee, not a percentage of anything it finds.

The shorter version: overhead is a quarter of your cost base, and a margin report that ignores it is not conservative — it is fiction with good posture. Allocate it pro-rata, floor it in the slow months, keep it out of the rates, and the number on every project finally means what everyone assumes it means.