Creative fatigue or a bad week: telling them apart

01What does creative fatigue actually look like in the numbers?

Three properties together, and all three have to be there. The decline is gradual, spread over a week or more rather than landing on one day. It is confined to one creative and its close relatives while other ads in the same account hold their numbers. And it starts at the top: the hook rate goes first, then hold, then clicks, then purchases, in that order, over days.

The mechanism is simple and it is about the audience rather than the ad. The people most likely to respond respond early. What is left over is a harder crowd seeing a familiar frame, and the ad has not changed at all - the population it is being shown to has. That is why fatigue never arrives overnight and why pausing an ad for two weeks sometimes revives it.

Fatigue is a slope. Almost everything else is a step.

The first thing we check on any account

02What looks like fatigue and is not

Your creative sits in the middle of a stack, and every layer above and below it can move without anybody touching a frame. A seasonal shift changes who is browsing. An auction shift changes what an impression costs. A landing page deploy changes what happens after the click. A tracking break changes what you can see at all. And a competitor launching into your audience changes the price of the same attention.

The one people underestimate is the delivery layer. Meta documents a learning phase that an ad set re-enters after a significant edit, which means the fix somebody applied on Monday is a plausible cause of the dip they are reading on Thursday. If a person edited the ad set to respond to a drop, the response is now inside the data they are using to diagnose the drop.

The second underestimated one is arithmetic rather than advertising. An ad that had an unusually good week is likely to have a worse next week for no reason other than regression toward the mean. If the ad you are worried about was your best performer last month, some of what you are seeing now is the same selection effect described in why the winner of a twenty-creative test is usually the luckiest one.

03Rule them out in cost order

The first four steps here are free and take about twenty minutes together. The last one costs a production cycle. Doing them in this order is the whole method, because the expensive step is the one everybody reaches for first and it is the one that most often changes nothing.

04Is frequency a reliable signal of ad fatigue?

On its own, no. Frequency counts impressions divided by reach, and it rises whenever your audience is small, your budget is large, or your delivery narrows - none of which is fatigue. A frequency of four against a tight retargeting pool is normal. The same number against a broad prospecting audience means delivery has collapsed onto a fraction of it.

Read it alongside two other things and it becomes useful. Frequency rising while reach is flat means you are re-serving the same people, which is the actual mechanism of fatigue. Frequency rising while hook rate falls at the same time is the pair worth acting on. Frequency alone, with reach still growing and the hook rate steady, is a number going up for a boring reason.

05The two responses point in opposite directions

This is what makes twenty minutes of diagnosis worth it. If it is fatigue, the right move is to spend money making something new and to leave the tired ad running quietly while you do. If it is a bad week, the right move is to change nothing about the creative and to fix the layer that moved - and specifically to avoid editing the ad set, because an edit resets delivery and buys you a second bad week to explain.

06How often should you refresh creative?

There is no interval that is right for every account, and anyone quoting you one is quoting a blog. The honest version is a rate rather than a schedule: refresh fast enough that a fatigued ad is never the thing carrying your budget. Motion's own dividing line is instructive - they treat a creative as a loser if it is turned off, or never reaches real spend, before day 28. Most creatives do not survive to have a fatigue problem.

Their volume figures set the rate. Motion reports an advertiser testing four ads a week surfacing roughly 0.2 winners a week, which is about one a month, and eighteen a week surfacing roughly 0.9, which is about four. The hit rate itself barely moves: about 5% across the whole dataset, running from roughly 3.8% at accounts under $10,000 a month to roughly 8.2% at accounts over $1m. More winners come from more swings, not from better luck per swing.

07The constraint is the cost of a usable asset, not the ideas

Here is the part that is ours rather than the platform's. Nobody has ever run out of ideas for ads. What runs out is the money and the calendar between an idea and a file good enough to put spend behind. If four shippable creatives a month is the cadence you need, the real question is what a shippable one costs you, including the ones that do not make it.

Our own numbers say plan for the waste rather than pretend it away. Fewer than half our films survive a human eye after clearing every automated check, so we build a surplus by default and the last creative act is selection rather than repair. A bad take is re-rolled, never patched, which is re-roll, never repair, and the accounting is in what failed renders actually cost.

The reason a cadence is affordable at all now is that the cost of a variant collapsed while the cost of judgment did not. That asymmetry is the whole business model of a modern creative pipeline, and the same asymmetry is what makes ads that survive contact with an audience longer - covered in why some ads resist fatigue and sized in how many creatives your budget can actually teach you.

The expensive mistake is rebuilding creative in response to a bad week, watching the rebuild fail against a number that was never about creative, and concluding that new creative does not work for you. Misdiagnosing one week is cheap. The conclusion you draw from it costs a year.