The revision policy that stops scope creep

Scope creep is almost never a client behaving badly. It is two people using the word "revision" to mean different things and not discovering it until round four. You mean "change something in the thing we made." They mean "keep going until I like it." Both readings are reasonable, neither is written down, and the gap between them is absorbed entirely by you.

The fix is four sentences in a proposal and one uncomfortable conversation you only have to have once. This is what those four sentences should say and how the conversation goes.

Revision versus new deliverable

Start here, because every other clause depends on this line being drawn somewhere specific. A revision changes an existing element within the approved concept. A new deliverable changes what the thing is.

Request	Classification	Why
"Make the text bigger"	Revision	Same asset, same concept
"Try the other music track"	Revision	Swap within the approved brief
"Trim it to 20 seconds"	Revision	Re-cut of existing material
"Can we see it in 1:1 as well?"	New deliverable	A second asset, separately usable
"Use a different presenter"	New deliverable	Re-shoots or re-generates every shot
"Actually let's make it funny instead"	New deliverable	New concept, new script, new everything
"Add our new product at the end"	New deliverable	New material not in the approved brief
"Same video, but for the German market"	New deliverable	New audio, new captions, new usage

The pattern underneath: if the request can be satisfied by re-cutting or adjusting what already exists, it is a revision. If it requires generating material that was not in the approved brief, it is a new deliverable.

Print this table into the proposal. Not as an appendix. In the body, above the price.

The four clauses

One: cap the rounds at three. Three is not arbitrary. Round one catches the misunderstanding, round two catches the taste disagreement, round three catches the detail. A fourth round is almost always a briefing failure that a fourth round will not fix.

Two: set a feedback window. Five business days from delivery. Notes that arrive after the window are welcome and are priced as a new round. This clause is not about being difficult; it is about the fact that a project reopened six weeks later costs you the entire context reload, and the context reload is most of the work.

Three: require consolidated feedback. One document, one voice, per round. Three stakeholders sending contradictory notes across two channels is not one round, it is three rounds arriving simultaneously and cancelling each other out. The clause names a single approver by role.

Four: price the overrun before it happens. A stated per-round fee, in the proposal, from day one. A reasonable rule of thumb is 25–35% of the original deliverable fee per additional round, which is high enough to be a real signal and low enough that a client who genuinely needs a fourth round can say yes without a renegotiation. The point is not the revenue. The point is that a priced fourth round converts an awkward conversation into an ordinary one.

Reported market rates give you a sanity check on what that percentage turns into. Two 2026 aggregations put the average single UGC ad deliverable at $198 and $212, and a 2026 rate survey puts most working professionals at $500–$1,200 a video. Run 25–35% across that spread and an additional round comes out at roughly $50–$75 on an average deliverable and $125–$420 on professional-tier work. Do the multiplication against your own fee rather than copying a band from a post, because a fixed number that is 35% of one project is 8% of the next and stops being a signal. Usage rights are priced separately in the same reporting, roughly €200–€500 for 30-day paid ad usage and €1,000 or more for an unlimited buyout, which is worth remembering when a "small revision" is actually a request to run the asset somewhere new. Pricing AI video services as an agency covers the full ladder.

Policy text for a proposal

Paste this, adjust the numbers, and put it above the price rather than below it.

Revisions. This quote includes three rounds of revisions per deliverable. A revision is a change to an existing element within the approved brief: timing, text, music, colour, pacing, ordering, or a re-cut of material already produced.

A request that requires new material not in the approved brief — a different concept, a different presenter, a new product, a new language, or an additional aspect ratio — is a new deliverable and is quoted separately.

Feedback. Each round of feedback is consolidated into a single written document from [role], delivered within five business days of receiving the preview. Notes received after that window are welcome and are treated as a new round.

Additional rounds. Rounds beyond the third are billed at [amount] per round per deliverable, agreed in writing before work begins.

Approval. Written approval of a preview, or the absence of consolidated notes within the feedback window, constitutes acceptance of that round.

The last clause is the one most people leave out and the one that actually closes projects. Without it, a deliverable is never finished, it is only un-objected-to, and un-objected-to has no invoice date attached.

Why previews make the cap enforceable

A round cap is easy to write and hard to hold when every round costs you money, because at round three you are choosing between an unhappy client and an unprofitable job. The economics change if iteration is cheap and only the confirmed output is charged.

That is how the editor is built. The cut is one timeline, preview: true renders a free 480p pass with a short per-user cooldown, and the export charge lands once on the version you confirm, regardless of clip count. The full billing behaviour is in previews and the final export. Practically, it means "warmer, then back, then warmer but keep the crop" is a sequence of free preview renders rather than three billable exports.

So the round cap is not protecting your credit spend. It is protecting your calendar and your attention, which are the genuinely scarce inputs. That reframing matters when you explain the policy, because "each round costs me money" invites a client to offer to pay, and "each round costs both of us two weeks" does not.

Build the loop so the client experiences rounds as fast and finite: send a share link rather than a file, so review takes a tap and no login, collect one consolidated response, and run a self-check before anything leaves. A client review loop built on previews and share links is the operational version, and sharing a generation with a link is the primitive it runs on. Running an AI check on the video against the brief before you send is what stops a round being spent on something you would have caught yourself.

The round-four conversation

You will have this conversation. Have it early, in writing, in a tone that assumes good faith, because the client almost always has good faith and no idea they are at round four.

"Quick note before I start on these — this is round four, and the quote covered three. That's completely normal on a first project together, so I want to flag it rather than surprise you on the invoice.

Two options. I can do this round at the additional-round rate we agreed, which is [amount], and we carry on.

Or, if it's useful, we take fifteen minutes first. Looking at the notes across all four rounds, most of them are about [the specific pattern: tone, pacing, how much product is on screen]. If we lock that down together now, I think round four is the last one instead of the fourth of six. That call costs nothing and I'd rather do it that way.

Either works. Which do you prefer?"

Three things that script is doing. It states the fact before doing the work, so the invoice is never the first mention. It offers the paid path first, which makes the free path read as generous rather than as a retreat. And it names the pattern in the notes, which is what actually ends the loop, because a fourth round of the same disagreement is a briefing problem and the only fix is talking about the brief.

If the pattern appears across clients rather than within one, the problem is upstream of the policy, and the fix is usually in the brief template rather than the contract. Reroll rates and the shots you throw away is the production half of that diagnosis, and client reporting for content freelancers is where the round count belongs in the client-facing document.

FAQ

Is three rounds too few for a new client?

For a first project, quote three and expect to give a fourth without charging for it. The value of the clause on project one is that it exists and gets mentioned, not that it gets enforced. Enforce it from project two, when you have a shared reference for what a round actually produced.

What if the client refuses to consolidate feedback?

Consolidate it yourself and send it back for confirmation before starting. "Here's what I've captured from the three threads, confirm this is the complete list for round two" takes ten minutes and converts an unbounded conversation into a bounded one. It also creates the written record that makes the round-four conversation straightforward.

How do I handle a revision request that arrives after approval?

Treat it as a new round, priced, with no argument about whether the approval was real. The approval clause exists precisely so this is a rules question rather than a relationship question. If the client is good and the request is small, do it anyway and say you are doing it as a one-off — but say it, so the exception stays an exception.

Does an aspect-ratio variant really count as a new deliverable?

Yes, and holding this line is worth more than it looks. A second aspect ratio is an asset the client can run in a separate placement, with separate media spend behind it, which makes it a separate deliverable no matter how fast it is to produce. Price it below the original if you like, since the incremental production genuinely is cheaper once the concept is approved, but price it.