Targeting on Meta is now mostly automated. That leaves creative as the main lever you still control. The problem is that most creative tests are too small, too slow or measure the wrong thing.
Test concepts, not colours
Changing a button colour rarely moves results enough to detect. Changing the core idea does. A concept is a different reason to buy: a pain point, a desire, an objection handled, a demonstration or social proof. Test concepts first, then iterate on the winners.
- Problem: show the frustration your customer has today.
- Proof: a result, a review or a before-and-after.
- Demo: the product working in under ten seconds.
- Offer: a clear reason to act now.
The structure I use
I keep one always-on campaign for proven ads and one testing campaign with a fixed daily budget. Each test round gets three to five new concepts, each with two hooks. New ads run for at least a week before I judge them, unless they are clearly failing.
Your testing budget is tuition. Spend it on questions worth answering.
Reading results honestly
Look at cost per result first, then at the leading signals that explain it: hook rate (people who watch the first seconds), hold rate and click-through rate. A high hook rate with low clicks means the opening works but the message or offer doesn’t.
Always check results against your own backend data, not only the ads dashboard. If purchases in your store don’t line up, fix tracking before you trust any winner.
When to scale a winner
When a concept beats your current best on cost per result for a full week, move it to the always-on campaign and build variations: new hooks, new formats, new creators delivering the same idea. Scale budget in steady steps rather than doubling overnight.
