Test ad creative by changing one idea at a time in a separate campaign, giving each ad enough delivery, and picking the winner on cost per result. Move the winner into your main campaigns once the result is clear.
TL;DR
- Change one thing per test: the hook, the angle or the format, not all three.
- Run the test apart from your best campaigns, so a bad ad does not drag them down.
- Keep each test to a few ads. More ads split delivery too thin.
- Decide the winning metric and the minimum results before you launch.
- Do not edit ads during the test. Edits can restart learning.
What do you need first?
- Two to four versions of one idea, built from a creative brief.
- A budget you can afford to lose, set per ad set.
- A single result to judge on, such as cost per purchase or cost per lead.
- A date when the test ends, written down.
Let Madgicx's AI do the work
Its AI marketer finds what's working and prepares the next move. You approve.
Try for Free ($0 Trial)The steps
- Pick the question. For example: "Does a problem hook beat a result hook for our serum?" How to check: you can say the question in one line.
- Build the versions. Change only the hook, keep the rest the same. How to check: the versions differ in one clear way.
- Set up a separate test. Use a new campaign or ad set for the test ads. How to check: your main ad sets are untouched.
- Pick the metric and the stop point. For example: cost per purchase after at least 50 results, or seven days, whichever comes later. How to check: you wrote both numbers before launch.
- Launch and wait. Meta says ad sets usually exit the learning phase after about 50 results in the week after the last significant edit, so do not change budgets or ads in that window. How to check: delivery is steady and no edits were made.
- Read the result. Compare cost per result, not clicks or likes alone. How to check: each ad reached your minimum.
- Roll out the winner. Add it to your main ad sets in one planned change. How to check: you recorded what won and why.
A worked example
A made-up pet store tests two hooks for a dog treat. Both ads use the same photo, text and offer.
- Version A opens with the question "Does your dog turn down treats?"
- Version B opens with "Gone in seconds. Even picky dogs ask for more."
Each ad gets the same daily budget in a new ad set. After seven days, A has 61 purchases at $9.40 each and B has 54 at $8.10 each. Both passed the 50-result mark, so the team picks B and builds the next test on the result hook, with new images. The numbers are made up, but the habit is real: one change, one metric, one decision.
What are the common mistakes?
- Testing too many things. If the image, text and offer all change, you cannot say what worked.
- Stopping early. An ad that leads on day two may not lead on day seven.
- Editing during the test. Changes can send an ad set back into learning.
- Tiny budgets with many ads. Delivery goes to one or two ads, and the others are never judged.
- Calling a win on clicks. A cheap click that never buys costs more than it looks.
How do you test cheaply?
Test the opening. A new first second or first line is cheap to make and often moves results more than a new offer. Make several hook versions of one video by changing only the first few seconds. For static ads, an AI image generator can produce several backgrounds quickly. Then review them for accuracy before launch.
What to do next
Write down the result, then run the next test on the winner. Our post on creative testing tools compares software for running and reading tests, and the answer on how to track ad creative performance lists the metrics to compare.
Where Madgicx fits
Madgicx helps before and after the test. The Madgicx AI Ad Generator makes new image ads from a prompt, an image or any ad in the Meta Ad Library, so versions are ready fast. Creative Insights breaks down every creative, including video, to show what makes winning ads win and flags creative fatigue. You approve what launches.
FAQ
How long should an ad creative test run?
Run it until each ad reaches your minimum number of results, which is often about a week. Meta says ad sets usually need about 50 results to leave learning.
How much budget does a creative test need?
Enough to reach your minimum results for each ad within the test window. If your cost per purchase is $10, 50 results per ad costs about $500 per ad.
Can I test creative inside my main campaign?
You can, but new ads can pull delivery from your best ones. A separate test campaign protects them.
What should I test first?
Test the hook, because the first second decides whether people keep watching.
Related
- Answers: Tracking ad creative performance
- Answers: Ad creative testing tools
- Answers: Ad creative hooks
Go deeper
Our post on ad testing goes through test structures in full.
Sources
- About the learning phase, Meta Business Help Center
