Every founder I speak to who wants better margins reaches for the same lever first: make the team bill more hours.
It is the most tempting move and the most expensive one. You push utilisation from 68% to 85%, margins improve for two quarters, and then your best senior leaves. Now you are paying somewhere between 50% and 200% of their salary to replace them, according to Gallup's long-running turnover research, and the person who knew how three of your accounts actually worked is gone. The margin you bought was a loan.
Before starting Scopeyard I spent seven years running a product development studio and delivering AI automation work across healthcare, recruitment and operations. The projects that made real money were almost never the ones where people worked hardest. They were the ones that were priced correctly, scoped tightly, and moved through client approvals without stalling. The hours were ordinary. The margin was not.
So here is the framing I use:
Agency Margin = Price Discipline × Delivery Efficiency × Sustainable Utilisation
Three multipliers. Most founders only ever pull on the third one, which is also the only one with a human ceiling. The first two are effectively uncapped.
1. Know what "good" actually looks like before you optimise anything
You cannot improve a number you have not benchmarked. The current picture across agencies in 2025–2026 looks roughly like this:
| Metric | Weak | Healthy | Strong |
|---|---|---|---|
| Delivery (gross) margin | Under 40% | 50–60% | 60%+ |
| Net profit margin | Under 10% | 10–20% | 25%+ |
| Billable utilisation | Under 60% | 65–75% | 70–75% held steady |
| Non-billable time | 40%+ | 25–35% | Under 25% |
Promethean Research put the average digital agency at around 13% after-tax net margin in 2025, down from 14% the year before. The split by size is the interesting part: studios under ten full-time staff averaged 19%, while agencies of 50 or more averaged 8%. Bigger did not mean better. It meant more overhead absorbing the same delivery output.
Specialisation shows up just as sharply. Generalist shops tend to land at 15–20% net; specialists at 25–40%. That gap is not effort. It is pricing power and repetition.
Parakeeto's guidance is a useful target to hold yourself to: 50–60%+ delivery margin at the P&L level, 70%+ on individual projects. If you do not know your delivery margin per project, that is the first thing to fix — everything below depends on it.
2. Fix realisation before you touch utilisation
Utilisation is how much of your team's available time is billable. Realisation is how much of that billable time you actually get paid for. Founders obsess over the first and ignore the second, which is backwards, because realisation costs the team nothing to improve.
Ignition's 2025 survey of 273 agency managers and executives found that 57% lose between $1,000 and $5,000 every month to unbilled work, and another 30% lose more than $5,000 a month. Only 1% bill for all out-of-scope work.
Run the arithmetic on your own shop. A team of eight, average blended cost of $45 an hour, losing four unbilled hours per person per week:
8 × 4 × $45 × 48 weeks = $69,120 a year, gone.
That is roughly a mid-level hire's fully loaded cost, being given away in quiet increments by people who did not want to have an awkward conversation. Nobody worked less hard. The money simply never got invoiced.
The fix is not a policy document. It is making out-of-scope work visible at the moment it happens, in the same place the client can see it, so the conversation is about a logged item rather than someone's memory of a Slack thread three weeks ago.
3. Treat scope leak as a margin line, not a personality problem
Industry surveys put scope leak somewhere between 15% and 27% of project budgets. On a $60,000 project at a 55% delivery margin, a 20% leak does not cut your profit by 20% — it roughly halves it, because the leak comes entirely out of the margin, not out of costs you have already committed.
Three things that actually move this, in order of impact:
Cap revisions explicitly. Two rounds included, a stated rate for the third. I have written about how many revision rounds agencies should include — the number matters less than the fact that it is written down and priced.
Define "done" per deliverable, not per project. A project-level definition is too coarse to argue with. A deliverable-level one is specific enough that both sides know when it is finished.
Log the small stuff. The killers are never the big change requests. Those get quoted. It is the twenty-minute favours, forty times over.
4. Cut the non-billable tax instead of raising the billable ceiling
The average agency spends 25–35% of total available hours on non-billable activity. That is the number to attack — not the billable side of the equation.
Move a team of ten from 32% non-billable to 25% and you free up roughly 2.8 hours per person per week. At a $120 blended rate, that is around $168,000 of recoverable annual capacity, and nobody worked a single extra hour.
The recurring offenders in every agency I have looked at:
- Status meetings that exist because nobody trusts the board
- Rewriting the same brief, SOW or QA checklist from scratch each time
- Chasing clients for feedback and approvals
- Time entry reconstructed on Friday afternoon from memory
The third one is worth its own paragraph. Approval delay is not just a cash-flow problem — it is a margin problem, because a stalled deliverable gets re-opened, re-familiarised and re-explained. Every restart costs delivery hours you cannot bill for.
5. Point AI at delivery hours, not at headcount
87% of marketers now use generative AI in at least one workflow, up from 51% in 2024, and practitioners report recovering around six hours a week on average — more for seniors, less for juniors. A 2025 Productive.io survey of 180+ agencies found timelines compressing by 3–4x at some shops.
The margin question is what you do with the recovered hours. There are two strategies and only one of them holds.
The first is to cut junior roles. 23% of agencies cut junior copy roles in 2025 and 31% planned cuts for 2026. It improves margin for about four quarters and then you have no bench, no succession, and seniors doing work they resent.
The second is to keep the headcount and change what the hours are spent on: pre-filling first drafts, generating test cases, drafting SOWs, summarising client calls into structured updates, doing the first-pass QA. The seniors get their judgement time back. Delivery margin per project rises. Nobody is doing more hours.
If you are running an automation practice, the same logic applies to your own internal delivery — I covered the scoping side in how to scope AI automation projects.
6. Hold utilisation in the 70–75% band and stop there
Here is the part most margin advice skips. Utilisation and margin are not linearly related. They are a curve with a peak, and past the peak the relationship inverts.
| Utilisation | What happens |
|---|---|
| Under 60% | Overstaffed or underpriced; margin bleeds through idle cost |
| 65–75% | The productive band; margin peaks here |
| 75–85% | Overtime creeps in, quality slips, rework starts eating the gain |
| Over 85% | Burnout, errors, turnover; margin goes negative on a 12-month view |
Median agency utilisation sits around 68% against a common target of 75%. Closing that seven-point gap is worth real money. Pushing past it is not.
The cost of getting this wrong is specific and measurable. 71% of agency employees report burnout. Replacing a marketing agency employee costs roughly 1.5 to 2 times their annual salary. And 89% of burnout-related cost shows up as presenteeism rather than absence — people at their desks, billing hours, producing work that needs redoing. That last figure is the one that should worry you, because it means burnout damages margin long before anyone resigns. You do not see it in the headcount report. You see it in rework.
7. Raise your price before you raise your effort
Every lever above is operational. This one is a decision.
The 25-point margin gap between specialists and generalists is not an efficiency gap. It is a pricing gap, funded by the fact that a firm which has done the same category of work forty times can quote confidently, scope tightly, and reuse most of its thinking. If your positioning is "we do digital", you will be compared on price, and you will win on price, which means you will lose on margin.
Pick the two things you are genuinely best at. Price those at what a specialist charges. Let the rest go. If you want the diagnostic that tells you whether your leverage is improving or you are just adding cost, revenue per employee is the cleanest single metric for it.
Final thoughts
Every one of these moves — realisation, scope leak, non-billable drag, AI-assisted delivery, sustainable utilisation, specialist pricing — has the same shape. Margin improves because the work gets clearer, not because people get squeezed.
That clarity has to live somewhere both you and the client can see. Scoped deliverables, a definition of done, out-of-scope items logged when they happen, and approvals that do not sit in an inbox for nine days. That is exactly what we built Scopeyard for, and why the marketing agencies using it tend to catch scope leak in the week it happens rather than at the end of the quarter.
Only about a third of agencies hit every key operating benchmark. The rest leak 15–30% of possible revenue through loose time tracking and unbilled scope. That leak is not a people problem. It is a systems problem you have been solving with overtime.
Your team is not your margin lever. Your systems are — stop borrowing against people to pay for the difference.
Sources: Promethean Research — How Profitable are Digital Agencies?, Parakeeto — Definitive Guide to Agency Profitability, Ignition 2025 Agency Pricing & Cashflow Report, TMetric — Marketing Agency Benchmarks, Scoro — Billable Utilization Benchmarks, Gallup via Qooper — Cost of Employee Turnover, Digital Applied — AI Marketing Statistics 2026