Creative fatigue is the slow death of a winning ad. CPA creeps up. CTR drifts down. Frequency climbs. You're still spending, but every pound buys a little less. Most teams notice three weeks late, by which point the ad has already cost them real money.

The good news: fatigue is one of the most predictable patterns on Meta. If you watch the right signals, you can catch it early, brief iterations in time, and extend a winner's life by months.

The metrics that flag fatigue first

Frequency on its own is overrated as a fatigue signal. Plenty of ads run at frequency 6+ without losing efficiency, and plenty of ads die at frequency 2. The earlier signals are downstream of behaviour, not impression count.

Watch these, in this order:

  1. CTR drift over a 7-day rolling window. A 20% decline from the ad's peak CTR is the first whisper.
  2. Hook rate (3-second views / impressions). If this drops while CTR is also dropping, the issue is the opening of the ad, not the offer.
  3. CPA trend over 14 days. A 15-25% creep is the "act now" signal.
  4. CVR on the landing page. If CVR is stable but CPA is rising, the issue is the ad, not the funnel.
  5. Frequency by audience. Useful as a corroborating signal. Useless on its own.

If three of these are moving in the wrong direction simultaneously, the ad is fatiguing. Don't wait for the fourth.

What fatigue actually is

Two things, mostly:

  • Audience saturation: the people most likely to convert from this exact ad have already seen it and either bought or chosen not to.
  • Pattern recognition: the audience starts recognising the ad as an ad. The opening frame, the cadence, the creator's face. The interrupt stops interrupting.

Both are inevitable. They're not a sign you did anything wrong. They're a sign you have a winner that's reached the end of its first life.

What to do once you've spotted it

Don't kill the ad. Iterate around it.

  • New hook, same body. First, cheapest iteration. Often buys 4-6 weeks more life.
  • Same script, new creator. Different face, different voice, same angle. Often buys another 4-6 weeks.
  • Same angle, new format. Static to video, video to carousel, long to short. Tests whether the angle still works in a different shape.
  • Same creator, new angle. Riskier; you're changing more than one variable. Use when the creator themselves drives the performance.

Run iterations as a fresh test, not as a swap inside the scale campaign. You need clean data on whether the iteration is actually working.

When to actually kill

Some ads can't be saved. Signs that an iteration won't help:

  • The angle itself has stopped working across multiple creators
  • The offer has changed (price, bundle, promo) since the ad ran
  • The seasonal context the ad relied on has passed
  • Competitors are running near-identical positioning

In those cases, kill clean and brief a genuinely new concept. Iterating around a dead angle just delays the inevitable.

The system that prevents firefighting

Set up a weekly review of the top 10 ads by spend. For each, check the five signals above. If any ad triggers the fatigue pattern, brief one iteration that same week.

Done consistently, this means you've always got fresh iterations entering test before any single winner collapses. The account never has a "what do we run now" panic moment.

That's the difference between accounts that scale smoothly and accounts that lurch between viral ads and recovery mode.