Every account has the ad that worked. It carried the quarter, everyone quotes it, and then over about six weeks it quietly stops performing. The instinct is to conclude that the creative was wrong, or that the team lost their touch, or that the platform changed something.

Usually none of that happened. What happened is that the ad ran out of people who had not already seen it.

The maths nobody writes down

An audience has a finite size. Spend divided by cost per thousand impressions gives you impressions; impressions divided by reach gives you frequency. Hold spend steady inside a fixed audience and frequency climbs every single week, mechanically, whatever the creative says.

Response falls as frequency rises because the people most likely to react react early. What is left in the audience after four exposures is, by definition, the part of it that did not respond to three. Efficiency does not fall because the ad got worse. It falls because the remaining audience is a harder audience.

You are not competing with your competitors for attention. You are competing with the version of your own ad that this person already ignored twice.

Why this matters more than it used to

Broad targeting and automated placements made this faster. When the platform is free to find your buyer anywhere, it finds the easiest ones first and moves outward — so the decay curve is steeper than it was in the era of tightly defined audiences. Smaller markets feel it hardest, because the pool refills more slowly.

It also means creative volume is now a media variable, not a design preference. An account with four live concepts and one new one a month has a structurally lower ceiling than an account with the same budget and four new ones a month. Nothing about the copywriting explains the difference.

How to run creative as a system

Produce on a cadence, not on a crisis

The worst time to commission new creative is the week performance drops, because you then wait two weeks for assets while the account bleeds. A fixed monthly cadence — a known number of new statics and videos, scheduled before anyone needs them — turns a recurring emergency into a routine.

Vary the thing that matters, not the thing that is easy

Changing a background colour produces a new asset, not a new test. Meaningful variation happens at the level of the angle: a different problem, a different objection answered, a different person speaking, a different format entirely. Three genuinely different angles will teach you more than thirty colour variants.

Give every test a threshold before it starts

Decide in advance what result would make you scale it and what result would make you kill it, and at what spend you will make that call. Without that, tests do not end — they linger, absorbing budget, defended by whoever suggested them.

Retire deliberately

Watch frequency and click-through rate together on your top spenders. When frequency is climbing and click-through is sliding on the same asset, that asset is finished in that audience. It may well work again in three months, or in a different market, or against a colder audience — fatigue is specific to an audience, not permanent to an ad.

The staffing consequence

If creative volume is a media variable, then creative capacity belongs in the media budget conversation, not beside it. The most common failure we see is an account with enough money to scale and not enough assets to scale into — spend goes up, frequency goes up faster, and efficiency falls in a way that looks like the media buyer’s fault.

It is not a media buying failure. It is an arithmetic one, and it is solvable a month in advance by anyone willing to plan production instead of reacting to a dashboard. The media side and the creative side are one system; budgeting them separately is what makes them look like they are failing each other.