Guide
How to test ad creative on Meta
To test ad creative on Meta, run different ads against the same audience with the same budget, and judge them on cost per result, not on clicks. Test big differences first: a new idea beats a new button colour. Give each test about a week, keep the winner, and build the next test from it.
Last updated
Two ways to run a test
| A/B test in Ads Manager | Several ads in one ad set | |
|---|---|---|
| How it works | Meta splits the audience so each person sees only one version. | All ads share one audience and Meta decides which to show. |
| Fairness | Each version gets an equal share of budget. | Spend flows to the early leader. Some ads barely run. |
| Speed and cost | Slower, and needs more budget. | Fast and cheap. |
| Result | A declared winner with a confidence level. | A ranking you interpret yourself. |
| Use it when | The decision matters and you want a clean answer. | You are screening many ideas to find the few worth a proper test. |
Most teams use both: several ads in one ad set to screen ideas quickly, then an A/B test between the two or three that survived.
What to test, in order
The bigger the difference between two ads, the sooner the result is clear and the more it teaches you. Work from the top of this list down.
- 01
The idea
Who the ad speaks to and what it promises. A testimonial against a product demo against a problem statement. This is where the large differences are. - 02
The opening
The first line of text, the headline, the image, or the first two seconds of a video. Most people decide here whether to keep looking. - 03
The format
The same idea as a single image, a carousel and a short video. - 04
The details
Button label, colour, a line of copy. Test these last. The differences are small and need a lot of budget to detect.
Setting up a fair test
- Same audience, same placements, same budget for every version. Only the creative differs.
- Three to five ads per ad set when screening. More than that and the budget is spread too thin for any of them to be judged.
- About a week. Meta recommends seven days for an A/B test, so that every day of the week is included. Results from two days are mostly noise.
- Enough results to mean something. Meta's guidance is that an ad set settles after roughly 50 of its optimisation events in a week. If your budget cannot reach that for purchases, optimise the test for an earlier step such as add to cart.
- Do not edit mid-test. Changing budget or creative restarts learning and spoils the comparison.
Reading the results
| Metric | What it is | What it tells you |
|---|---|---|
| Cost per result | Spend divided by purchases, leads or whatever you optimise for. | The verdict. Decide on this. |
| Link click-through rate | Link clicks divided by impressions. | Whether the ad earns the click. |
| Hook rate | Three-second video plays divided by impressions. | Whether the opening stops the scroll. Video only. |
| Hold rate | ThruPlays divided by three-second plays. | Whether the video keeps people once they stop. |
| Conversion rate | Results divided by link clicks. | Whether the landing page delivers what the ad promised. |
| Frequency | Average times each person saw the ad. | Rising frequency with falling results means the ad is wearing out. |
Hook rate and hold rate are not columns Meta gives you by name. You build them as custom metrics in Ads Manager from the video-play columns.
Use the supporting metrics to explain the verdict
- High click-through rate but poor cost per result: the ad attracts the wrong people, or the landing page lets it down.
- Low click-through rate but good cost per result: the ad filters well. Keep it.
- Low hook rate: the opening is the problem. Change the first frame or first line and leave the rest.
- Good hook rate, low hold rate: the opening works and the middle loses people. Shorten it.
After the test
- Keep the winner running and move the budget to it.
- Write down what won and your best guess at why. In three months this list is worth more than any single ad.
- Make the next test a variation of the winner against one completely new idea.
- Expect every winner to fade. When frequency climbs and cost per result rises for a week, replace it.
Questions people ask
How long should I run a creative test on Meta?
About seven days. Meta recommends a week for an A/B test so that every day of the week is included, since results on a Tuesday and a Saturday often differ. Stopping after two or three days usually means deciding on noise.
How many ads should I test at once?
Three to five ads in an ad set is a practical range when screening ideas. With more than that, the budget is spread so thinly that most ads never get enough impressions to be judged.
Should I use Meta's A/B test or put several ads in one ad set?
Use several ads in one ad set to screen many ideas quickly and cheaply, accepting that Meta will favour an early leader. Use the A/B test tool when you need a clean answer between two or three finalists, because it splits the audience evenly.
Which metric decides the winning ad?
Cost per result, meaning what you pay for each purchase, lead or other outcome you optimise for. Click-through rate, hook rate and conversion rate explain why an ad won or lost, but an ad with a high click-through rate and a poor cost per result is not a winner.
Try it on your own competitors.
$9 for 7 days. Paste a competitor's Page and see their best ad rebuilt for your brand.
Keep reading
Mockingbird is an independent product from Onwards Analytics and is not affiliated with, endorsed by or sponsored by Meta. Meta, Facebook, Instagram, Meta Ads Manager and the Meta Ad Library are trademarks of Meta Platforms, Inc. Meta changes its tools often; check Meta's own help pages for the current detail before you rely on it.