Key takeaways

Start from a few hooks and rewrite each many ways, instead of 120 unrelated ideas.
Budget follows purchases per spend, not clicks.
Too many variants with too little spend each makes results unclear. Test in smaller batches.
AI can write the rewrites faster now. I did this by hand, so that part is a plan I haven't run.

For Color Street’s holiday campaign at The Cirqle in 2021 and 2022, my team tested 120 variations of ad copy. The ones that sold got the budget.

The campaign beat its target four times over (the case study has the numbers). But the testing also wasted money in a way I’d avoid next time.

I ran part of the Meta ads, under 5% of the spend. So this is what I saw and learned from the team.

What happened

I worked on the ad copy tests. The team started from a few hooks: the angles we thought would land with mothers and beauty fans in the US and Canada. We rewrote each one many ways until there were 120 variations in the A/B tests.

Budget followed purchases per spend. An ad that got attention but didn’t sell lost its budget.

The waste was in the spread. With 120 variants running, each got too little budget to tell us much, and some results were noise. Next time I’d run a small batch, cut the losers and write the next batch from the winners.

Try it

Pick three or four hooks. One sentence each, with the reason a buyer would care.
Rewrite each hook five or six ways. Change the opening line and the offer wording, but keep the hook.
Run a small batch, with enough spend on each variant to read.
Judge on purchases per spend, never clicks, and move budget to the winners.
Write the next batch from the winners. Repeat.

I did this by hand, so the AI part is a plan I haven’t run. I’d use AI for step 2, the rewrites, and nothing else. A person still picks the hooks and checks every line before it runs, especially any claim about the product.

Where does it go wrong?

Too many variants for the budget. You can’t tell a winner from noise.
Judging on clicks. An ad can win attention and lose sales.
Letting AI write the hooks. The hook is the thinking, so it stays with you.

This week, write three hooks for your best-selling product and four rewrites of each. Run them with equal spend and compare purchases per spend once the budget is used.

The full story is in the Color Street case study. Part of AI in performance marketing. Next: Google and Meta run your ads now.

Sources

Color Street case study: the campaign figures.

Frequently asked questions

How many ad variations should I test at once?

Fewer than we did. The team ran 120 over the campaign. Spread across that many, each one had too little budget to read. Start with a small batch, kill the losers, and write the next batch from the winners.

What should decide which ad wins?

Purchases per spend. On the Color Street campaign, budget moved to what sold, not to what got the most clicks.

Three-minute lesson on this Ads: more variants, better tests Marketing course, medium. Free, no sign-up.

More from the blog

All articles →