Creative fatigue or a market shift: how to tell the difference

A number alone does not prove fatigue

An ad's click-through rate falls for a week and the obvious read is that people have seen it too many times. It might also be that CPMs rose across the whole account, that a competitor started bidding on the same audience, or that the platform's delivery simply shifted where the impressions landed. Frequency climbing is not proof either: a healthy ad can run at a higher frequency than a tired one if the account around it is doing the same thing.

The question worth asking is never "did this number move." It always moved. The question is whether it moved more than the account it sits inside, over more than one comparison window, in more than one place at once. Answering that needs a baseline, not a single reading.

Subtracting the account from the ad

Start with the account's own numbers, but with this one ad's own spend and delivery removed first, so the ad is never quietly explaining itself by being compared against a total that includes its own contribution. That gives an account baseline that moves for reasons outside this ad entirely: seasonality, competition, a platform-wide CPM shift, an audience getting saturated across every ad running against it.

Then compare the ad's own efficiency change against that baseline's change, over the last 7 days against the prior 7, again against the prior 14, and again against the prior 30. A decline that only shows up in one of those three windows is weaker evidence than one that shows up in two or three. What is left over after the account's own move is subtracted out is the creative-specific change, the part of the decline this ad cannot blame on the account it runs in.

One long-running, founder-story ad in the demo workspace behind this site makes the point concretely. Over the last 7 days against the prior 7, its own ROAS fell 12 percent, from 4.19x to 3.69x. Over that exact same stretch, the account's ROAS, with this ad's own delivery taken out, rose 3 percent, from 3.95x to 4.06x. The account did not get worse. This ad did, by a wider margin than the plain 12 percent suggests once the account's own improvement is netted out: a creative-specific change of negative 16 percent.

The account did not get worse. This ad did.

Fig. 1 · one founder-story ad's ROAS against the account's ROAS with that ad's own delivery removed, prior 7 days versus the last 7, Northwind Sleep demo workspace, fictional data

If the account's own CPM had simply risen for every ad that week, the honest conclusion would run the other way: a market or platform story, not a creative one. The subtraction above is exactly what stops that mix-up. It nets the account's own move out before anything gets called creative-specific, in either direction.

Seven checks, by name

No single comparison decides the outcome. Seven separate checks run against every active ad, each one a plain question with a yes or no answer behind it: a sustained decline across more than one baseline window, an account CPM shift, an account CTR shift, the account's own seasonality, a conversion rate change relative to the account, a frequency increase alongside other signals, and the creative-specific change itself, the last one asked as "Does the ad's efficiency decline exceed what the account explains?" (fatigue check)

Four statuses, one of them a second chance

The seven checks resolve to one of four plain statuses.

Status	What it means
Healthy	No sustained or creative-specific decline found. Nothing to act on.
Watch	An early or unconfirmed signal. Keep running, and check again in a few days.
Likely fatiguing	A sustained, creative-specific decline, usually backed by a falling CTR, CVR or a climbing frequency. Time to test new creative.
Revive	The same decline, on an ad that was a genuine winner before it faded.

The founder-story ad above lands on Watch, not Likely fatiguing: the creative-specific change is real and worth tracking, but on its own it has not yet cleared the bar that a sustained decline, corroborated by more than one baseline window, requires. The recommendation follows the status rather than the raw percentage: keep it running, recheck in a few days, and name the softest signal so there is something concrete to watch for.

The relaunch that keeps the original as the control

Revive is the status that matters most, because it changes what happens next. An ad that reaches it was not a lucky guess that stopped working, it was a proven winner whose decline is creative-specific rather than explained by the account around it. That is worth a second act, not a burial.

Keeping the original in the mix is the whole point. A relaunch that only shows new variants has nothing to measure them against except each other, so a mediocre new idea can look like progress simply by comparing well to a worse one. Measuring every new direction against the ad that already worked is a harder bar, and a more honest one.