Ad fatigue and when to refresh

The same people seeing the same thing until it stops registering. It is a function of three things: how large the addressable audience is, how fast you are serving it, and how distinctive the asset is.

Which means fatigue is a supply problem before it is a creative problem. A small audience served hard will exhaust a brilliant asset quickly; a large one served gently will run an ordinary asset for months. The asset is the variable people reach for because it is the one they can change, not because it is usually the binding one.

Frequency rising. The same people are seeing it more often. On its own this is not fatigue — it is a description of supply.

Engagement falling. Hook rate first, because the opening is what stops registering first, then hold rate. This is the signal that turns rising frequency into fatigue, and it arrives before the money signal does.

Cost per result rising. The last to arrive and the one most people notice first, by which point the diagnosis is weeks late.

The shape that says fatigue is the first two together: frequency rising while engagement falls, on the same audience, with the same creative. Frequency rising with engagement stable is a small audience being served efficiently, and the correct response is to leave it alone or widen the audience — the opposite of a refresh.

Read hook rate and hold rate is how to read the middle signal properly, including why an empty ratio is undefined rather than zero and why reach cannot be summed across days.

A shrinking audience. Exclusions accumulating, a retargeting pool that is not refilling because prospecting was cut, a seasonal contraction in who is in market. Frequency rises because the denominator fell. A new asset served to the same exhausted pool fatigues on the same schedule.

An auction getting more expensive. Competitors arriving, a seasonal peak, a category-wide bidding shift. Your cost per result rose because the price of attention rose. Nothing about your creative changed and nothing about a new one will change the price.

Something after the click. A landing page edit, a price change, a stock problem, a form that broke. Engagement holds and results fall — which is precisely the pattern that should send you past the ad rather than back to it.

A seasonal floor. Everything is down at once. A new creative launched into it will underperform and be blamed for the season, and you will have spent a good asset on a bad fortnight.

The discriminator for all four is whether the problem is confined to one asset. Fatigue is asset-specific: a fresh asset on the same audience recovers, and that is the test. If everything moved together, the cause is not the creative.

Three reads, and they take about ten minutes.

Is the week real. A range ending today includes a day that is not over. Platforms revise the recent past, and a connection whose sync rewrote a lot of stored history is marked restated. A partial day against whole ones has manufactured more fatigue diagnoses than any real mechanism.

Is it one asset or all of them. Compare the asset against its siblings in the same period, on the same audience. This is the single decisive check.

Did we change something. Every applied change is a row on the record with what it moved and from what, so the question of whether a budget or a bid moved just before the decline has an answer rather than a discussion. A large change to either is exactly the sort of condition change a delivery system reacts to, and the stretch of unsettled delivery that follows looks a great deal like fatigue.

The arrangement that record sits inside, stated once: your connections read your accounts, a change is written as a draft on the record with the state it would change from and the checks it passed, applying it is gated on approval — by you, or by a rule you wrote — and no platform write reaches a customer account until that platform has granted write access. Applying a cleared change to a platform is in private beta.

If the diagnosis holds, three decisions.

Change the part that decayed. The opening is usually what stops registering first, and a new opening on a body that still works is a cheaper and more informative change than a whole new concept. It is also a variable you can write down, which makes the refresh a test rather than a replacement.

Do it as a new entity rather than an edit where you can. Creative writes are complete replacements rather than field patches, which is why the undo for one is the previous specification put back exactly, and editing a live asset merges two periods into one set of numbers. A new entity keeps the history readable.

Do not stack it with a budget change. Two changes at once produce a result attributable to neither, and both cost a stretch of unsettled delivery.

And keep the old asset's numbers. The next time this audience fatigues, how long the last one lasted is the only sourced estimate you will have of how long the next one will.

The accounts that do not have a fatigue problem are the ones with something ready before it is needed, and the reason is arithmetic rather than diligence: producing an asset takes longer than noticing you need one.

A modest standing pipeline — one new concept in production at all times, with the variable it is testing written down — turns fatigue from an emergency into a scheduled swap. Size a test before you run it is what stops that pipeline producing unreadable comparisons, and the creative brief template is what stops it producing assets nobody can test.