Most small and mid-size ecommerce brands should launch 3 to 6 new ads a week. Give each new ad roughly 2x your target CPA in spend before judging it, and keep testing to about 20 to 25% of total budget. If that maths doesn't fit your budget, test fewer, stronger concepts.
The budget maths
If your target CPA is $40 and you give each new ad $80 to prove itself, five new ads a week costs $400. On a $5,000 monthly budget, that's about 35% in testing, which is too much. On $8,000 a month it's about 22%, which is right.
Test angles before variations
A new angle (a different reason to buy) teaches you more than a new colour or caption. Prioritise:
- Angle: problem, proof, founder story, comparison, gifting.
- Hook: the first three seconds.
- Format: video, static, carousel, UGC.
- Offer: bundle, gift, guarantee.
When to kill, keep or scale
| Result after 2x target CPA spend | Action |
|---|---|
| No purchase at 1.5x target CPA | Kill |
| 1 to 2 purchases under 1.5x target | Keep testing |
| 3+ purchases at or under target | Move to scale campaign |
Keep winners running
The most expensive mistake in creative testing is switching off a winner because it's "old". If an ad is still at target CPA, keep it. Refresh only when frequency climbs above 3 or CPA rises 30% week on week.
Creative testing calculator
Frequently asked questions
How long should I test a Meta ad?
Judge on spend, not days. Around 2x your target CPA in spend is usually enough to see whether an ad can convert, often 3 to 7 days.
Should I test ads in a separate campaign?
Yes. A separate testing campaign stops new ads from stealing budget from proven ones and makes results easier to read.
How many ads per ad set?
3 to 6 is a good range. Too many ads in one ad set and Meta spends on only one or two of them.
Written by Shimeez Bok, founder of Growth Alchemy, an ecommerce growth studio in Auckland, New Zealand.
