A/B testing creatives: how to find ads that work

No one knows in advance which creative will "pop" — not even an experienced team. So strong ads are found through tests, not intuition. Let's break down how to run A/B tests correctly.

What to test

• The hook — the first seconds of a video or the headline.
• The format — video vs static, length, style.
• The offer — different value wordings.
• The audience — different segments on one creative.

The main rule: change one element at a time, or you won't know what made the difference.

How to test correctly

• A sufficient sample. A conclusion from 50 impressions is noise. You need volume for statistical significance.
• Equal conditions. One period, one audience, one budget per variant.
• One success metric set in advance: CTR, CPA, completions — not "liked it."
• Don't stop too early. Give the test time to gather data.

Common mistakes

• Changing everything at once — you can't isolate the cause.
• Drawing conclusions from small numbers.
• Comparing variants at different times / on different audiences.
• Confusing engagement with sales.

What to do with the result

The winning variant — scale it; the losers — turn them off, and launch a new test against the winner. It's an endless improvement cycle, not a one-off action.

Takeaway

A/B tests turn ads from a lottery into a manageable process: one element, a sufficient sample, a clear metric. We test creatives systematically and scale what actually pays off.