What a Good Project Margin Looks Like for a Digital Agency
On a real-margin basis — labour at true cost, overhead allocated, unbilled scope counted — a healthy agency project lands around 20–30%. Most agencies quoting 50–60% are quoting labour margin, a different number. The honest benchmark ranges, and why the popular ones mislead.
The direct answer first. On a real-margin basis — labour costed at each person's fully-loaded rate, a fair share of agency overhead allocated to the project, and unbilled scope counted — a healthy digital agency project lands somewhere around 20–30%. Projects in the mid-teens are acceptable if they buy something else (a new capability, a strategic logo). Below 10% is thin enough that one generous change request tips it underwater. Negative is more common than anyone admits.
If those numbers look shockingly low against the "50–60% margin" you have heard quoted — including, possibly, inside your own agency — the explanation is almost never that your agency is twice as good. It is that the 50–60% figure is a labour margin: invoice minus timesheet at role rates, with no overhead and no unbilled work in it. Labour margin runs 25–40 percentage points above real margin at a typical agency. Both numbers are real. Only one of them is the one your bank account experiences.
This piece gives the honest ranges for both, explains why the industry benchmarks you find in surveys mislead, and lists what actually moves the number — so you can tell whether yours is a pricing problem, an overhead problem, or a measurement problem.
"What should our project margin be?" is the most reasonable question an agency owner can ask, and it is surprisingly hard to get a straight answer to. Industry surveys quote figures that turn out, on inspection, to be measuring different things. Peers quote numbers in the pub that are labour margins wearing a project-margin costume. And the honest answer — it depends which margin you mean — sounds like an evasion, when it is actually the whole point.
So let me try to give the straight answer, with the definitions attached, the way I wish someone had given it to me when I was running delivery for a 120-person agency and confidently quoting a number that was wrong by a third.
First: which margin are you quoting?
There are two numbers that get called "project margin" in agencies, and they differ by 25–40 percentage points.
Labour margin is invoiced revenue minus logged hours at role rates (or at some blended cost figure). It falls out of the timesheet system with no extra work, which is why it is the number most agencies actually have. It contains no overhead — no rent, no operations salaries, no software stack, no marketing — and no unbilled work.
Real project margin is revenue minus true cost: each named person's hours at their fully-loaded cost, plus the project's pro-rata share of monthly agency overhead, plus the cost of everything delivered that never made it onto an invoice. The full definition, with a worked example, is in what project margin actually is.
A project that reports 55% on the first definition typically lands between 15% and 25% on the second. We learned this the uncomfortable way: the first project we recalculated properly went from a celebrated 55% to a sobering 18%.
Any benchmark discussion that doesn't state which definition it is using is noise. Here are both.
The honest ranges
These are drawn from our own portfolio over two years — roughly forty active projects at any time — and from the agencies we have since run the same calculation for. Ranges, not decimals, because false precision is its own kind of lie.
| Reading | Labour margin | Real project margin |
|---|---|---|
| Strong | 60%+ | 30%+ |
| Healthy | 50–60% | 20–30% |
| Acceptable with a reason | 40–50% | 12–20% |
| Thin — one favour from underwater | 30–40% | 5–12% |
| Underwater (and usually undetected) | below 30% | below 5% |
Three notes on reading the table.
"Acceptable with a reason" is a real category. A mid-teens real margin can be a fine outcome if the project bought something: a first project in a new sector, a case study you needed, a team learning a stack you intend to sell. It is only a problem when it has no reason — when it was priced as if it would land at 30% and quietly didn't.
The agency-level number is different again. Project margin rolls up, then the costs that no project should absorb (genuine one-offs, financing) come out, and you arrive at agency EBITDA — typically low-to-mid teens at a decent agency, 20%+ at a genuinely well-run one. Which number your board actually wants is its own discussion.
Dispersion matters more than the average. An agency averaging 22% real margin might be forty projects all near 22%, or half at 35% and half near zero. The second agency has a much better business hiding inside a much worse one — and only sees it if margin is computed per project, not as a roll-up.
Why the benchmarks you've seen mislead
Most published agency benchmarks come from surveys, and surveys inherit whatever definition each respondent used. An owner who tracks labour margin answers "55%". An owner who tracks real margin answers "20%". Both land in the same averaged column of the same industry report, and the result is a benchmark that measures nothing except how respondents do their accounting.
The practical consequence: comparing your real margin to a survey benchmark that is mostly labour margins will make a healthy agency feel like a failing one. I have watched an owner with a genuinely strong 26% real margin conclude the agency was in trouble because "the industry runs at 50". The industry does not run at 50. The industry's timesheet reports run at 50.
The only benchmark that is fully trustworthy is your own portfolio, computed one way, over time. Direction beats comparison: an agency moving from 17% to 23% real margin over a year is doing well regardless of what any survey says.
What actually moves the number
Four levers explain almost all of the variance between projects, in roughly this order of weight.
- Realised rate. Not the rate card — the revenue actually collected per hour actually delivered, after discounts, overruns and goodwill. This is the heaviest lever and the most commonly untracked one.
- Role mix. Every hour a senior spends on work scoped and priced for a mid is margin leaving the building at a rate of the cost difference. It compounds quietly across a portfolio.
- Overhead ratio. An agency whose overhead has grown faster than turnover can price perfectly at the project level and still land thin — the projects are carrying a heavier share than the pricing assumed.
- Unbilled scope. Unpapered change requests, extended "warranty" support, founder time at zero cost. Typically the smallest of the four in any single project and the most corrosive across a year, because it is invisible by construction.
If your number is below the healthy range, these four are the diagnostic sequence: check the realised rate first, the role mix second, the overhead ratio third, and only then start the harder conversation about what gets given away.
What to do with your number
If you have read this far and realised you do not actually know which range your projects sit in — that is the normal case, not the embarrassing one. Most agencies have never computed a real project margin, because no tool in the standard stack produces it: the hours are in Jira, the true employee costs are in payroll, the overhead is in the management accounts, and nothing joins the three.
Joining them is what Saldo does — read-only on top of the Jira you already run, each person's hours at their real cost, overhead allocated pro-rata, reported continuously rather than at year-end. The 15-minute demo runs on your own data, which means the question this article answers in ranges gets answered, for your agency, in specifics.
The shorter version: a good real project margin for a digital agency is 20–30%, a good labour margin is 50–60%, and the most expensive mistake available is believing you have the first because a report shows you the second.
Going deeper: How Saldo calculates margin — estimate by role rate, actual by employee cost
Continue inside Saldo