Meta ad fatigue: refresh on frequency or on CTR and CPA trends?

Take a dental clinic running a prospecting ad set at a frequency of 2.5 over seven days, with CPA still flat. Some people say leave it alone, others say the audience is tired. I don't think frequency is a good trigger on its own. A 3 on a big audience with steady CPA is usually fine, while a 1.8 on a small retargeting list can already be worn out. Frequency only tells you how many times people saw the ad, not whether they still care. I'd watch CTR falling and CPM creeping up together, plus CPA on a rolling weekly view, and refresh when two of those move the wrong way. What triggers do others use, especially on small local audiences where frequency climbs fast?

Two signals make sense, but what decides which one to trust is audience size relative to spend. On a small local audience, frequency climbs fast and CTR often drops before CPA moves, so CTR is your earlier warning. On a big prospecting set, CPA is the number that actually pays the bills, so I'd let frequency ride longer.

CTR is the earlier warning, but I'd measure it against the ad's own launch week rather than a fixed benchmark. Say a dental clinic's ad starts at 1.4% CTR and settles around 1.1% after a month, which is normal for a local audience. If it drops to 0.7% while frequency is still under 2.5, that tells you more than the frequency number does. Rising CPM on its own often just means more competition in the auction, especially around holidays, so I'd only call it fatigue when CTR is falling at the same time.

Launch week almost always flatters you, because the first people to see an ad are the warmest. I'd set the baseline at the first stable week after the learning phase instead. Then the rule of thumb is simple: refresh when CTR sits about a third below that baseline for a full week and CPA is climbing with it.

Agreed on the baseline, but the CPA half of your rule depends on how many conversions you get a week. Say a dental clinic books about 4 consultations a week from a prospecting set. One bad week can swing CPA by 40% from noise alone, so "CPA is climbing" will fire on random days and you'll end up refreshing creative that's still working. At that volume I'd lean on CTR against baseline and only read CPA over a two or three week window.

Flip it to 30 leads a week and CPA moves mean something within a few days, so your rule works as written.

So it comes down to weekly conversion volume. Under roughly 10 a week, CTR should carry the decision and CPA is mostly noise. Above that, the CPA trigger earns its place. That cutoff is a rough number, not a hard line, so test it against your own weekly data before you lock in a rule.

I'm not sure CTR should carry the decision alone at low volume. On a small local set, CTR can slide because delivery drifts toward colder, less relevant people, not because the creative is worn out, so you end up refreshing an ad that still works. Look at the CTR drop next to a three-week CPA trend instead.

The cleanest way to separate the two is to look at reach and frequency side by side. If frequency is flat but weekly reach keeps shrinking, you're burning through a small pool. If reach is growing and CTR still drops, that's usually new people who just don't care, which is a targeting problem, not creative fatigue.

It depends on how much of the audience you've already reached. A growing reach number means little if the pool is huge. But on a local radius of 5,000 people, if you've reached 3,000 of them in two weeks, the people left are the colder ones, and CTR drops for that reason even with a good ad.

So divide weekly reach by the estimated audience size from the ad set. If you're still under about half and reach keeps growing, a CTR drop is more likely down to targeting or colder leftovers. Past that point the ad is mostly showing to people who've already seen it, and the drop is closer to real fatigue.

A quick test sorts out either case. Put a fresh creative into the same ad set. If CTR comes back with the same people, the creative was the problem. If it stays flat, look at the audience.

A creative swap only tells you something if the new version is different enough to stop people in a new way. Change the headline on the same image and you'll mostly get the same reaction, so a flat result doesn't prove much. How far the new hook, format or opening frame sits from the old one is what decides whether the test means anything.