Your return dropped and nothing changed. Here is what actually happened.

Nothing changed. That is the part that makes it maddening.

Same audiences, same budget, same landing page, same offer. Three weeks ago the account returned 4x and this week it returns 2.1x, and every input you can name is identical. So the instinct is to go looking at the inputs you cannot name: the algorithm changed, iOS broke something, a competitor outbid you.

Usually it is none of those. Usually the ad got old.

Frequency is the number to look at first

Open the ad set and add the frequency column. If it has crept past about 3 on a cold audience, the people you are reaching have seen the same image three times and the third view does nothing the first one did not already do.

Meta will keep spending. It has no opinion about whether your creative is tired, only about which of the creatives you gave it performs least badly. If you gave it four ads in August and it is still October, it is choosing between four tired ads.

What the decline actually looks like

It is not a cliff. That is why it gets misdiagnosed.

• Click-through rate falls first, usually a week before the return does
• Cost per click rises to compensate
• Return slides gently, which reads as noise rather than signal
• By the time it is obviously broken, you have spent three weeks finding out

The useful habit is to watch click-through rate rather than return. Return is downstream of your product, your price and your checkout, all of which move for their own reasons. Click-through rate measures one thing: whether the creative still stops the scroll.

The ad above and the one below ran in the same ad set, same week, same audience and same offer. One of them took almost the entire budget.

The fix is not a new audience

The reflex is to go hunting for a fresh audience, because that feels like the lever you control. It rarely works. The audience was never the constraint. Targeting has been automated for years, and the account that beats yours is not beating you on a cleverer interest stack.

What changes the number is new creative, tested against whatever is currently winning rather than against nothing.

What a testing cadence looks like

The point is not volume. The point is never arriving at the week where the control is the only thing running.

1. Keep whatever is winning. That is the control, and it is the thing to beat.
2. Put new creative in against it every week, from a written brief rather than from whatever got made.
3. Give it enough budget to say something. An ad on 4 euro a day has not failed, it has not been tested.
4. Cut the losers without ceremony and write down why, because the reason is worth more than the ad.
5. When a challenger beats the control, it becomes the control.

Done consistently, you never meet the three-week slide, because there is always something new in the account before the current winner tires.

The part worth being honest about

This does not make a bad offer work. Creative amplifies demand, it does not invent it, and no amount of testing fixes a product nobody wants at the price you are asking.

But where there is real demand, fatigue is the most common reason a good account quietly stops being one. It is also the cheapest thing on this list to fix.