The worst project I ever ran was not a hard build. It was a homepage.
We had scoped it properly, or so I thought. Fixed fee, clear deliverables, "revisions included" in the proposal because that felt generous and closing-friendly. Round one came back with structural feedback, which was fair. Round two was refinement. Round three was the client's co-founder, who had not been in any meeting, deciding the whole thing should feel "more premium". Round seven was a colour.
Nobody was being unreasonable. There was simply nothing in the document that told any of us when it was supposed to stop.
Before starting Scopeyard I spent years running a product studio, and most of the last three delivering AI automation into healthcare, recruitment and operations teams. In that time I have watched more margin die inside revision rounds than inside bad estimates. Bad estimates you notice. Revisions leak quietly, one polite email at a time, and they never show up on an invoice.
Revision Profitability = Rounds Included × Definition of a Round × Willingness to Invoice Round Four
Most agencies get the first number roughly right and lose money anyway, because the other two are missing.
1. The answer is two, and the answer is not the point
Industry practice is well settled. Most agencies include two to three revision rounds in the project fee, with anything beyond that billed hourly or per round. For design work two is standard. For development one is typical, because in development you are correcting against a spec rather than against taste.
Two works because of what each round is actually for. Round one is structural — the layout is wrong, the flow is wrong, this section shouldn't exist. Round two is refinement — the copy in that block, the spacing, the tone. By round three you are in polish. If a fourth round exists, someone made a mistake earlier, and it is usually in the brief, not in the work.
So the number is two. Write it down and move on, because the number is the least important part of this. I have seen agencies with "two rounds included" lose forty hours on a project and agencies with "unlimited revisions within the sprint" finish on budget. The difference was never the number. It was whether they had defined what a round is.
2. Define a round or the client will define it for you
A revision round is not "a change". A revision round is one consolidated set of written feedback, from one named approver, delivered within five working days of presentation.
Every clause in that sentence is there because I lost money without it.
One consolidated set stops the trickle — five emails over two weeks, each one arriving after you've already actioned the last. That trickle is the single most expensive pattern in agency delivery and it is almost never malicious. Clients send feedback as they think of it. Unless you tell them not to.
From one named approver stops the co-founder problem. If three people can send feedback, you are not doing two rounds, you are doing six. Name the approver in the statement of work. Other stakeholders route through them.
Within five working days stops the project going cold and coming back at a point where your team has lost all context and has to relearn the codebase to change a heading. Late feedback is a new round, and it should cost like one.
Write those three constraints into the statement of work, and — this matters more than the contract — into the kickoff deck. A revision limit buried in Appendix C guarantees an argument later. A revision limit on a slide everyone looked at on day one guarantees a shrug.
3. What this is costing you right now
I would rather show you the leak than argue about it.
PMI has long put scope creep at around 52% of projects, with 85% of the projects that creep exceeding budget by an average of 27%. More recent industry data puts the rate higher — figures of around 70% appear across software development and creative services, where the deliverable is subjective and therefore infinitely arguable.
For agencies specifically, a 2025 survey of agency managers and executives found 57% lose between $1,000 and $5,000 every month to unbilled work, and a further 30% lose more than $5,000 a month. Only about 1% successfully bill for all of their out-of-scope work.
Take the middle of that. $3,000 a month of unbilled work is $36,000 a year. On a five-person agency running at a typical 15–20% net margin, that is not a rounding error. That is your margin.
Here is what it looks like on a single $20,000 project:
| Scenario | Rounds delivered | Extra hours | Effective margin impact |
|---|---|---|---|
| Two rounds, consolidated feedback, one approver | 2 | 0 | Baseline |
| Two rounds "included", feedback trickles in | 4–5 | 25–40 | Roughly 8–15% of project value gone |
| "Revisions included", no definition | 6+ | 50–80 | Project breaks even or worse |
The hour ranges are from our own delivery data, not a survey, so treat them as illustrative. The direction is not in dispute.
4. Price round three before you need it
The mistake is not that agencies run out of rounds. It's that they run out of rounds with no price sitting next to the next one, so the conversation becomes a negotiation held in bad conditions — mid-project, client already frustrated, you already over budget.
Put the price in the proposal, at the point where the client is calmest and least invested in that specific colour.
Two models work. Per-round is cleaner for design and content: a flat figure, commonly around $500 a round at the smaller end, scaled to the size of the deliverable. Hourly is cleaner for development and anything where the change size varies wildly: professional web and product rates broadly sit between $75 and $200 an hour, freelance design between $60 and $150.
Whichever you choose, make it a fair number rather than a punitive one. A penalty rate reads as a threat and invites the client to argue about whether their request was really out of scope. A fair rate reads as a menu, and clients buy from menus. Some of my best-margin projects had three paid extra rounds on them, and the client was happy, because at no point did they feel ambushed.
One phrasing that has never once caused friction: "Two rounds are included. If you want more we're glad to do them — additional rounds run at X, and we'll always tell you before the clock starts."
The last clause is the one that matters. Never invoice a surprise.
5. Cut round two by fixing round zero
The cheapest revision is the one that doesn't happen, and most avoidable revisions trace back to the same cause: 45% of teams identify unclear objectives as the most frequent trigger behind scope creep. Not difficult clients. Unclear briefs.
Three things reduced our revision volume more than any contract clause ever did.
Show work earlier and uglier. A wireframe review at 30% costs an hour. The same conversation held against a finished, polished screen costs a week, because now you're throwing away work and the client can feel it. Agencies that hide work until it's beautiful are buying themselves round three.
Ask what "done" looks like at kickoff, in writing. Not "does this look good" — what specifically has to be true for you to sign this off? Clients rarely have this answer ready. Making them produce it at kickoff is the whole trick.
Present with a decision, not with options. "Here is the direction and here is why" produces feedback. "Here are three concepts, what do you think?" produces taste. Taste has no round limit.
6. Make approval a state, not a vibe
Rounds only work if everyone can see, without asking, which round they are in.
This sounds trivial and it is where most agencies actually fail. The contract says two rounds. Six weeks later nobody on either side can tell you whether the feedback that arrived on Thursday was round two or round three, because the record is scattered across email, Slack and a call nobody wrote up. When it is genuinely ambiguous, you will eat it. You will eat it every time, because arguing costs more than the work.
So the round has to live somewhere the client can see it. Each deliverable has a state — in review, changes requested, approved — with a timestamp and a named person attached. Round two closes when the approver approves it, visibly, and round three opens with an explicit note that it is round three and what it costs.
This is exactly the problem Scopeyard was built for: deliverables that carry their own approval state, so the round count is a fact both sides can see rather than a memory both sides dispute. But the principle stands whatever you use — if your revision policy lives only in a PDF and not in the place the work is reviewed, it isn't a policy.
Related reading: why slow client approvals hurt agency cash flow and how to run a better client review process.
7. When to give one away
I am not arguing for rigidity. Rigidity on revisions is how agencies win $400 and lose a retainer.
Give a round away when the miss was yours — you misread the brief, you shipped something rushed, you were late. Give it away when the client has been fast, decisive and pleasant all project and something genuinely changed on their side. Then say clearly that you are giving it away. An unpriced favour is worth nothing; a favour that names its own value buys goodwill you can spend later.
Hold the line when the request is a new idea wearing a revision's clothing. "Can we just also add a pricing page" is not a revision. It is a deliverable, and it gets an estimate, a date and an invoice. The moment you let that through as a revision, you have taught the client that the word means whatever they need it to mean.
Final thoughts
Two rounds. One consolidated set of feedback per round. One named approver. Five working days. A price for round three, printed before anyone needs it. That's the entire policy, and it fits on a slide.
The agencies that lose money on revisions are almost never the ones being exploited. They are the ones who never said when it stops, and then felt awkward about it for eleven weeks.
Generosity you didn't decide on isn't generosity. It's just an unpriced invoice you're too polite to send.