A creator sent me a Figma file last month with nineteen thumbnail versions on it. Nineteen. Same face, same crop, same video. The differences between version 12 and version 19 were a stroke weight and two shades of orange. He asked me which one was best. The honest answer was that none of them were, because none of them were different enough to matter.
That file is the reason I wanted to write this. "How many thumbnail revisions are normal" is one of those questions where the number people want is not the number that helps. So I will give you both: the counts I actually see across the work we do, and the more useful framing, which is what each round is supposed to accomplish before you are allowed to ask for the next one.
How many thumbnail revisions are normal?
Two to three rounds per video is normal, and about one video in six needs a fourth. That is the honest distribution from our own queue. Round one lands the concept, round two fixes legibility and hierarchy, round three is polish. Anything past four rounds is almost never a design problem. It is a decision problem, and no amount of redrawing solves it.
When a video goes past four rounds, one of three things has gone wrong. Either the video itself has no clear promise, so there is nothing to draw. Or two people are giving contradictory feedback and the designer is being pulled between them. Or the creator is chasing a feeling instead of naming a specific problem. All three are fixable, and none of them are fixed by another version.
Worth saying out loud: we offer unlimited revisions at ClickTheory. We still average two to three. Unlimited is there so nobody ships something they hate on a technicality, not because we expect to use it. If a package needs eight rounds, that is a signal I failed at the brief stage, and I would rather fix the brief.
What should each revision round actually fix?
Each round should have one job, and the jobs should get smaller. Round one decides what the image is about. Round two decides whether a stranger can read it at 168 pixels wide on a phone. Round three decides the small stuff that separates fine from good. If your rounds do not narrow, you are not revising, you are wandering.
Here is the ladder I use with clients:
- Round 1, the idea. Is this the right subject, the right emotion, the right visual promise? Are we showing the outcome, the conflict, or the object? This is the only round where "start over" is a reasonable note.
- Round 2, the read. Does the focal point survive shrinking? Is the text three or four words at most? Does it fight the title or complete it? Is there enough contrast between subject and background at thumbnail size, not at full size?
- Round 3, the finish. Edge cleanup, colour balance, glow, the exact crop, how it sits next to your last six uploads in a browse feed.
- Round 4, rare. Usually because new information arrived. The edit changed. The title changed. A competitor posted something adjacent and the whole shelf shifted.
If a note does not belong to the round you are in, park it. "Can we try a different background" during round three is how a two round job becomes a nine round job.
A revision should remove a specific problem. If you cannot name the problem in one sentence, you are not requesting a revision, you are requesting reassurance.
Why do some channels need ten rounds and others need one?
Because the number of rounds tracks brief quality, not designer skill. Channels that give me the video, the title options, the audience, and one sentence on what the viewer gets tend to approve in one or two rounds. Channels that send a link and say "make it pop" tend to need five, and the fifth one is usually close to the first.
The other big variable is how many people are in the approval chain. One decision maker is fast. Two is fine if one of them is clearly the tiebreaker. Three with no tiebreaker is a machine for generating rounds. Before a project starts, I ask who signs off. It sounds bureaucratic for a $250 job. It saves days.
Format matters too. A talking head essay with a clean promise usually resolves fast. A multi-topic video, a compilation, or anything where the creator is not sure what the video is about will churn, because the packaging is being asked to make a decision the edit refused to make. That is also why we bake title consultation into the work rather than treating the title as someone else's problem. Half of the thumbnail arguments I have seen were actually title arguments in disguise.
If you want to cut rounds before the design even starts, the fastest lever is a tighter brief. I put the exact questions we ask into our pre-publish packaging checklist, and most of them take under a minute to answer.
Is unlimited revisions a real benefit or a marketing line?
It is real, but it is not the feature people think it is. Unlimited revisions does not mean you should use them. It means the price is not a hostage situation. You are never staring at a thumbnail you dislike doing arithmetic about whether the fix is worth another invoice line.
The places it genuinely earns its keep:
- The video changes after the thumbnail is drafted, which happens more than anyone admits.
- You upload, watch the first two hours, and want a different swing. That is a legitimate second attempt, not a revision in the design sense, but nobody should be charged extra for reacting to data.
- A style is still settling. New clients often spend the first two or three videos calibrating, then drop to one round for the rest of the batch.
What unlimited revisions does not fix is an absent decision maker. If nobody will say yes, unlimited just means unlimited waiting. Our turnaround is 48 to 72 hours per thumbnail, and that clock is on our side of the work. The approval clock is on yours, and it is usually the longer one.
How do you give feedback that ends the round?
Name the problem, name where you see it, and leave the solution to the designer. "The text is competing with the face" is a note I can act on in twenty minutes. "Make it more punchy" is a note that costs a full round and often lands somewhere you did not want.
Three habits that collapse revision counts fast:
- Look at it small first. Shrink the image to about 200 pixels wide before you write a single note. Half of all feedback disappears at that size, because it was about details nobody will ever see.
- Compare it to your own shelf, not to a Mr Beast frame. Put it next to your last eight uploads. Your job is to look like the best version of you, not like a channel with a different audience and a different promise.
- Write notes as a numbered list, once. Drip fed feedback across a day is the single most expensive habit in this whole process. One list, one round, done.
The same principle applies downstream. When a designer hands files to your editor, the ambiguity tax shows up again, which is why I wrote up how to hand a thumbnail off to your editor with the file naming and safe area rules we use.
When should you stop revising and start testing?
Stop when two rounds have produced changes you cannot confidently rank. That is the point where opinion has run out and only the audience can tell you anything new. Past that line, another round is procrastination with a design tool open.
This is where an A/B variant earns its $15. Instead of arguing between two directions for a day, we build both properly and you run them. YouTube's own test and compare feature is documented in their creator resources, and it settles arguments in a way that no group chat ever has. Two genuinely different swings beat six near identical ones, every time.
Also worth remembering: for niches where the visual language is dense and specific, like hardware or software coverage, the first round often needs to be more literal than creators expect. We go into that in more depth on thumbnail design for tech reviewers, where the product itself is usually the strongest asset on the frame.
Do revisions on old videos work differently?
Yes, and they are usually faster. Repackaging a back catalogue takes about one round per video, because the video already exists, the performance data already exists, and you are not guessing about the promise. You know what the video delivers. You just have to say it better than you did eighteen months ago.
This is the cheapest growth most channels are sitting on. A video with 40,000 views and a weak frame is a paid asset you already own. New packaging on it costs about $31 in a batch of eight and takes 48 to 72 hours. A new video costs a week. I made the fuller argument in why old videos deserve new thumbnails, and it is the first thing I look at when a channel tells me growth has flattened.
Seasonality shifts the maths too. Browse behaviour in the back half of the year is different enough that the same frame can underperform against the same competitors, which is the reasoning behind our Q4 packaging approach.
What does a healthy revision process look like end to end?
A brief that answers who the video is for and what they get. A first draft inside 48 to 72 hours. One consolidated round of notes. A second draft. Usually an approval, occasionally a small third pass. Then, if you are genuinely torn, a variant to test rather than a fourth round to argue about.
That is the shape of nearly every batch we run. Eight custom thumbnails for $250, one per video, drawn for that video rather than dropped into a template, with a dedicated designer who learns your face, your lighting, and your audience over the batch instead of starting cold each time. Styles get tested on our own channels before they touch client work, which is a big part of why round one usually lands close. You can see the full breakdown on our pricing and services page.
If you are counting rounds and feeling bad about it, the number to worry about is not three or four. It is when the rounds stop getting smaller. That is the tell. Fix the brief, name the decision maker, and let the audience settle the ties.
And if your channel is a podcast, the packaging conversation usually starts one step earlier, at the edit, which is the work my sister studio does over at Pixel Samy.
If you want a second pair of eyes on a batch you are stuck revising, send me the links and I will tell you what I would change.