Every ad has a shelf life. The same creative shown to the same audience for long enough will eventually stop working, not because the offer changed, but because the audience has simply seen it too many times. Here's how to tell when that's happening, and what to do about it.
Watch your frequency metric
Frequency tells you the average number of times a person in your audience has seen a specific ad. As a rough guide, once frequency climbs past 3 to 4 within a short reporting window for a cold audience, performance usually starts to soften. For warm retargeting audiences, fatigue can set in even faster since the pool of people is smaller.
Rising CPM, falling CTR
When cost-per-thousand-impressions climbs while click-through rate declines, that's the platform telling you engagement is dropping, which pushes up the cost of reaching the same audience. This combination, more than any single metric on its own, is the clearest sign that creative needs to change.
A comment section that's gone quiet, or turned negative
Early on, a good ad generates some engagement: likes, shares, the occasional comment. When that dries up, or when comments shift toward "seen this already" or "again?", it's a signal worth acting on even before the numbers move.
A simple testing cycle
Rather than waiting for an ad to visibly fail, the more reliable approach is a standing testing cycle: two to three new creative concepts introduced every few weeks, tested against the current best performer, with the losing variations paused and budget shifted to whatever's actually converting. This keeps the account ahead of fatigue instead of reacting to it after cost-per-result has already climbed.
What counts as a genuinely new concept
Changing the background colour or swapping one product photo for another of the same product usually isn't enough to reset fatigue with the same audience. A genuinely new concept changes the angle: a different hook, a different format (video instead of static, UGC-style instead of studio), or a different value proposition entirely.