Why your ads fatigue —and your CPMs keep rising.

Your best ad died and you didn't touch anything. Same audience, same budget, same landing page — and the cost per result that held for six weeks quietly doubled. If you run paid social for a DTC brand, this has already happened to you, and it will happen to every winner you ever find. This note explains the mechanics: what fatigue actually is, why it shows up on your bill as rising CPMs and CPAs, and the arithmetic that tells you how much fresh creative your spend level actually demands.

One promise before we start: every figure here is either Meta's own documented definition (cited at the bottom), our own measured data labelled as ours, or arithmetic marked illustrative that you can redo in a spreadsheet. The internet is full of "creative dies after N days" claims that dissolve when you chase the citation. We won't add to them.

Strip the folklore away and fatigue is one sentence: the people your money can reach have seen this idea before, and they respond less each time they see it again. An ad is a message shown into a finite pool of attention. The first exposure catches the people most ready to act. Every exposure after that is fishing the same water with the same lure — the responsive fish are already caught, and the rest have learned to swim past it.

Meta measures this against your own history, not a calendar. Ads Manager flags an ad as creative limited when its cost per result is worse than what your past ads achieved, and as fatigued when cost per result reaches roughly twice that baseline — and it counts exposures of the same image or video across your other campaigns too, so "duplicating the winner into a new campaign" does not reset the clock.^[1] That last clause is the one most media buyers miss, and it's why the duplicate-and-pray move so rarely works.

The companion metric is frequency — the average number of times each person has seen your ad, defined by Meta as impressions divided by reach.^[2] Frequency isn't fatigue by itself; a strong concept can hold at a frequency that kills a weak one. But a rising frequency on a flat audience is the mechanical precondition for fatigue: it means the same eyes, again.

Here's the part that surprises founders: fatigue doesn't just lower your click-through rate. It raises the price of every impression you buy. Meta's auction doesn't sell impressions at a flat rate — it ranks advertisers by a blend of bid, how likely the person is to act on the ad, and the ad's quality signals. When your creative tires, fewer people click, save, share or buy per thousand impressions. The auction reads that as a weaker ad, and a weaker ad has to pay more to win the same slot.

So one tired creative produces a double penalty: response falls and the toll rises. Your CPA is the product of both. This is why accounts that "didn't change anything" watch their blended CPA drift up month over month — the account didn't change, and that's exactly the problem. Rising CPMs across the whole market get blamed (and market CPMs do move seasonally), but on the accounts we open, the fatigue component is usually the part the operator actually controls.

You don't need a tool for this. You need three columns and an honest hour:

1. Set your baseline. Take your best sustained cost per result from the last 90 days — not your best day, your best stable week. That's the number Meta compares you against, and it should be the number you compare yourself against.^[1]
2. Read each running ad against it. Anything running at roughly twice baseline is fatigued by Meta's own definition. Anything drifting past ~1.5× is on the road. Note the spend still flowing to those ads — that's your fatigue tax, in currency, per week.
3. Check frequency by ad, not by account. Impressions ÷ reach, per ad, over its lifetime.^[2] A climbing frequency with a flat audience size tells you the pool is exhausted for that idea — no amount of budget nursing fixes that.

Most founders who run this exercise for the first time find that a meaningful slice of their budget is buying impressions from ads that crossed the fatigue line weeks ago. Not because they're careless — because nothing in the default Ads Manager view makes the line visible. If you want the full nine-check version of this exercise, we published it: audit your own ad account.

Fatigue converts into a production requirement, and this is the arithmetic almost nobody does. Illustrative, so label it as such: suppose a winning concept holds for six weeks at your spend level before crossing the fatigue line, and suppose one concept in four that you test becomes a winner. To keep three winners live at all times, you need a new winner every two weeks — which at a one-in-four hit rate means shipping and honestly scoring eight or more new concepts a month. Change the assumptions to your own measured numbers; the structure of the conclusion survives: the higher your spend, the faster the water gets fished out, and the more concepts your account eats.

Now compare that against what most brands actually produce: a handful of variations a month, mostly re-crops and colour swaps of the incumbent winner. A variation is not a new concept — Meta counts exposures of the same creative idea across campaigns,^[1] and your audience is even less fooled than the system. New hook, new angle, new format: that resets attention. New background colour: that doesn't. The full working of this arithmetic, including how to measure your own hold time and hit rate, is here: the production maths nobody does.

Not "refreshing" the ad. Not raising the budget on the tired winner — that buys deeper into an exhausted pool at a rising toll. The only durable answer is a production system that ships genuinely distinct concepts at the rate your spend consumes them, scores them honestly, and kills losers without sentiment.

We say that with our own money behind it. We built the Crank engine on our own brand first: shipped concepts, scored every one, kept the losers on the record next to the winners. That run produced 247 winning creatives for one brand, moved click-through from 3% to 11% across the winning set, and grew site traffic +400% — our own brand (TITAN), our own accounts, stated as ours. On the client side, the same replacement discipline took one account from €22.14 to €4.56 a lead — 4.9× cheaper (−79%) — real client, name under NDA, figures from Meta Ads reporting in the account's own currency, and past performance, not a promise. And behind it all sits €10M+ of the founders' own Meta spend — self-reported, own-brand accounts — which is where these mechanics stopped being theory for us.

If you want the wider system this sits inside — what AI genuinely changes about creative production in 2026 and the claims to stop repeating — the free, ungated guide covers it: AI Marketing Hacks 2026.

Everything above, you can do yourself. If you'd rather have senior operators do it — baseline, fatigue map, frequency read, creative-gap analysis, plus free ads and a 30-day plan — that's exactly what the free Crank Audit is. Two minutes to request, findings in 24 hours, no call required.