The production maths nobody does before hiring a creative team

Every conversation about creative fatigue ends up in the wrong place. Someone says the ads have gone stale, someone else proposes a new angle, and three weeks later there are four new creatives in the account. Then it happens again.

The reason it keeps happening is that fatigue is being treated as a creative problem when it is arithmetic. An ad account consumes creative at a rate. If your production is slower than your consumption, you will run out — not once, permanently. No amount of talent fixes a throughput deficit, and no amount of new angles changes the division.

You need two numbers. C is how many distinct creatives you want live and earning at any one time. L is the effective life of a creative, in weeks — how long it goes before its cost per result drifts back past where it started.

The creatives you must ship per week, just to stand still, is C ÷ L. That's it. That's the whole model.

Take the four-week column. A brand holding 20 live creatives with a four-week effective life needs five new creatives every week, forever, and that is the steady state — not the launch push, not the Q4 sprint. Twenty a month, replacing themselves, indefinitely.

Now put that against how creative is usually bought. A traditional agency retainer that produces ten to twenty assets a quarter is, against that account, running at roughly a quarter of replacement rate. It isn't that the work is bad. The work can be excellent. It is arithmetically incapable of keeping the account fed, and everyone involved experiences that as "the creative keeps fatiguing" rather than as a division that doesn't work.

You'll have noticed the table offers three values of L rather than one answer. That's deliberate. There is a widely-circulated set of figures about how many days ad creative survives — a ten-day decay curve, a peak at days four to seven, a CTR collapse by day ten. We went looking for the source of those numbers and could not find one. They trace to vendor blogs citing other vendor blogs, with no account count, no spend level and no stated definition of what was measured. Several of them contradict each other about when the peak even is.

The same is true of the frequency thresholds you'll see attributed to Meta — that 2.5 is a warning and 3.5 means refresh. Meta publishes no numeric frequency threshold for fatigue anywhere. Its own documentation says frequency may average one to two per ad set or run much higher depending on budget, audience size and schedule, and treats a rising frequency alongside falling performance as a signal to investigate, not a number to obey.^[2]

So the honest answer to "how long does creative last" is that the platform itself defines it against your account's own baseline — which means your L is measurable, specific to you, and beats any category average by a distance.

1. Open your ads manager and pick your last 8–12 creatives that ran with meaningful spend. Ignore anything that never got out of learning — a creative that was starved isn't a creative that fatigued.
2. For each one, set the date range to its first 7 days and note its cost per result. That's its baseline: the number it proved it could deliver. This is the same quantity Meta's own fatigue status is computed against, so you're measuring the thing the platform measures.
3. Now chart the same creative week by week and find the first week where cost per result crosses back above that baseline and stays there. Not a single bad week — a crossing that persists.
4. The gap between launch and that crossing is that creative's effective life. Write the number down.
5. Do it for all 8–12 and take the median, not the mean. One freak winner that ran for five months will otherwise drag your planning into fantasy.

Two things usually fall out of this exercise. The first is that the number is shorter than people expect, and shorter at high frequency than low. The second is more useful: the spread is enormous. Most accounts have a couple of creatives that ran for months and a long tail that died in a fortnight, and the median sits uncomfortably close to the tail.

That spread is the actual argument for volume. You are not trying to produce twenty mediocre ads instead of four good ones. You're trying to buy enough attempts to find the two that run for months, because those two are carrying the account and nobody — not us, not you, not the agency pitching you — can reliably pick them in advance.

Once you have C ÷ L as a monthly number, the question of who makes your creative answers itself, and it answers differently for different brands.

• Under ~4 a month. Do it yourself or use a freelancer. The volume genuinely doesn't justify a system, and a founder with taste and modern tools will outrun a slow agency at this level. We say the same thing on our comparison page, including the part where you shouldn't hire us.
• Roughly 4–15 a month. This is the awkward middle where most £5k–50k/mo brands actually live. Too much for evenings and weekends, not enough to justify a senior in-house hire, and above what a quarterly retainer produces. This is the gap productised creative exists to fill.
• Above ~20 a month, sustained. Now an in-house team starts to make sense on unit cost, if you can hire senior, wait out the hiring, and keep them. That's three conditions, and the third one is where most in-house creative teams quietly fail.

Notice that none of those brackets is decided by how good anyone's portfolio is. Portfolio decides whether the creative is worth running. Arithmetic decides whether you'll have any.

If your replacement rate is ten a month and you're producing four, you don't have a creative strategy — you have a slow decline with periodic bursts of optimism. Every new angle lands into an account that is structurally short of material, performs for a few weeks, and then joins the pile.

That's why we lead with volume and scoring rather than with taste. Volume is what makes the search possible; scoring is what stops volume becoming slop; a senior human signing off is what stops the scorer becoming the boss. When we scored our own fifteen launch concepts, the spread ran from 45 to 70 and we published the losers next to the winners — because the point of producing at rate is that most of what you produce is supposed to lose.