Guides / Creative / 7 min read / Updated 1 August 2026

How many ad creatives do you need?

Most accounts do not fail because their ads were bad. They fail because there were four of them, three did nothing, and the fourth was asked to carry a year.

The hit rate sets the volume

Nobody sensible claims a reliable strike rate for ad creative, because it varies by account, category and how good the last batch was. What every operator agrees on is the shape: most new ads do not beat the control, and the ones that do are not predictable in advance by the people who made them.

That has one operational consequence. If you want a new winner this quarter, the question is not "which ad should we make" but "how many concepts can we get in front of the auction at a cost we can absorb". Volume is not a substitute for judgment; it is what makes judgment affordable.

What counts as a new creative

Three levels, and confusing them is the most common reason a testing programme goes flat:

• A new concept. A different reason the product matters — a different problem, buyer, or moment. This is the only level that can produce a step change.
• A new execution of a concept. Same argument, different format: a demo instead of a talking head, a comparison instead of a testimonial.
• A new variation of an execution. New first three seconds, new headline, new thumbnail, different cut length.

Most accounts do the third level exclusively, call it testing, and conclude after two months that creative has stopped working. What stopped working is one idea, tested eleven ways.

A cadence that survives contact with reality

• Every cycle, ship a batch, not an ad. A batch is several concepts, each with a couple of executions. One ad launched alone gives you nothing to compare it to.
• Weight it towards net-new. Iterating winners is cheap and necessary and will not find you the next thing. Most of the batch should be arguments you have not made yet.
• Kill on a rule, not on a feeling. Decide before launch what spend without a conversion means the ad is dead. Then honour it, including on the one you liked.
• Never judge on the first day. Early numbers are delivery noise. This is the discipline that separates a testing programme from a slot machine.

What to do when something wins

Two things at once, and the order matters less than doing both.

First, protect it. Do not edit a winning ad, do not rename it, do not restructure the ad set around it. If you want a variant, make a new ad; the running one keeps its history.

Second, mine it. A winner tells you which argument the market accepted, and that is a much better brief than anything you started with. New first-three-seconds on the same body, the same argument in a different format, the same argument to an adjacent buyer. Winners are a source of concepts, not just a source of revenue.

When you have no data of your own

New accounts have no winners to mine, and the usual response is to guess in a meeting. There is a better free option: the public ad libraries.

Meta's Ad Library and Google's Ads Transparency Center let you see what your competitors are running right now, and roughly how long each ad has been live. Longevity is the signal — nobody keeps paying to run an ad that loses. An ad that has been up for months is a message the market has already voted on.

Take the structure, never the asset. Copying a competitor's footage or their exact copy is both a legal problem and a bad idea, since you would be shipping a worse version of something the audience has already seen.

Fatigue, and what it actually is

When frequency climbs and results decay, the instinct is to blame the ad. Often the ad is fine and the audience is exhausted — the same people have seen it enough times.

The distinction matters because the fixes differ. A tired audience is fixed by widening targeting or by new creative aimed at people the current ad never spoke to. A genuinely worn-out ad is fixed by a new one. Look at frequency and reach before deciding which you have.