Creative fatigue is a supply problem

creative fatigue is what happens when the people in your target audience have already seen the ad. click-through rate falls, cost per acquisition climbs, and the creative that carried the account for three weeks stops working inside a few days.

most accounts respond by changing the bid, the budget, or the audience. those are usually the three things that are not broken.

the part everyone gets right

the diagnosis is not hard, and every article ranking for this term covers it. fatigue looks like this inside an ad account:

• click-through rate drops while impressions hold or rise
• frequency climbs against the same audience
• cost per acquisition rises without a matching change in conversion rate on site
• hold rate in the first few seconds falls before anything else does

that last one moves first. people stop watching before they stop clicking, so retention in the opening seconds is the earliest signal you have.

none of that is controversial. you can read the same list on ten other pages. the useful question is what you do on the morning it shows up.

three things that look identical to fatigue

before you throw out a working ad, rule out the things that produce the same chart.

audience saturation. fatigue is the audience being tired of the ad. saturation is you having run out of audience. the tell is whether a genuinely new creative recovers performance against the same targeting. if a fresh concept lands and the numbers come back, it was fatigue. if a fresh concept dies the same way, your pool is spent and the creative was never the constraint.

auction pressure. if cost per mille rose across every ad set in the account on the same day, including ones you did not touch, that is the auction, not your video. check a campaign you have not edited in a month before blaming the creative.

a measurement change. a tracking break, a consent change, or an attribution window edit will move reported cost per acquisition without anything real happening. reconcile against your own backend numbers before you act.

the order matters. rule out measurement, then the auction, then saturation. what is left is fatigue.

work out your own refresh rate

there is no universal frequency threshold, and any article that gives you one is guessing at your account. the number you need is your own, and you already have the data to compute it.

take the last quarter and pull two things:

1. how long each winning creative held before its cost per acquisition rose past your target
2. how many winners you were running at once

that gives you a burn rate. if your winners historically last three weeks and you need four running concurrently, you are consuming roughly four new winners a month. then apply your own hit rate. if one concept in five becomes a winner, you need to be testing twenty concepts a month to stand still.

run that arithmetic with your figures rather than mine. the output is usually uncomfortable, and it is the actual reason the account keeps stalling. not the bid.

the constraint is production, not diagnosis

this is where the standard advice stops being useful. it tells you to refresh creative. it does not tell you how to produce twenty concepts a month, which is the only part that was ever hard.

most teams hit the same wall. a shoot is booked weeks out. a creator costs a few hundred dollars per video. so the refresh rate the arithmetic demands is four or five times the refresh rate the budget allows, and the account runs tired creative because the alternative is running nothing.

dan has been public about the sourcing side of this since 2025.

I get real ugc models for $3/vid in bulk and distribute 100+ videos a days

that is from his post on 27 september 2025. the point is not the price on its own. it is that per-unit cost is what sets your maximum test count, and a test count is what beats fatigue.

here is what a single clip at that cost did once it was distributed:

that screenshot is from his post on 1 november 2025, reporting a $4 reaction clip that reached 500k views at a cost per mille of $0.008. one clip, one result, published by the person who ran it. treat it as one data point rather than a forecast for your account.

on whether the footage has to be filmed by a real person at all, we have published a separate read on ai versus human reaction clips with a customer's hook rate comparison. that page owns the number.

fresh does not mean a new edit

this is the trap that makes a refresh fail. teams take the winning clip, change the caption, recut the first second, and call it a new creative. the audience has seen that face in that room saying that thing. a new text overlay does not reset anything.

what actually reads as new to someone scrolling is a different person having a different reaction somewhere different.

three clips, three people, three rooms, three reactions. confusion, curiosity, laughter. that is the axis to vary, and it is the axis a caption swap leaves untouched. three worked examples of that axis, each with its emotion, runtime and score, are in ugc video examples.

when you plan a batch, vary these deliberately:

• the person, so the face is not recognisable from the last ad
• the emotion, because a reaction is the hook, not a face
• the setting, so the background does not read as the same shoot
• the length, so the pacing changes

our b-roll library is built for exactly this shape of problem: reaction clips bought per unit so the test count is a budget decision rather than a scheduling one. pricing is per clip, and the packages are sized around running a batch instead of a single video.

what to do this week

1. reconcile your tracking, then check an untouched campaign for auction movement. rule out the impostors first.
2. pull your winner lifespan and concurrent winner count for the last quarter. compute the burn rate.
3. divide by your hit rate to get the number of concepts you need to test monthly.
4. compare that to what you actually shipped last month. the gap is the problem.
5. close the gap on unit cost, not on effort. the refresh rate you need is a function of what one test costs you.

creative fatigue is not a signal that your creative was bad. it is a signal that it worked long enough for the audience to memorise it. the accounts that keep scaling are not the ones with better ads. they are the ones that can afford to keep making them.