In short
- Test one variable at a time.
- Run fewer tests, with bigger differences between them.
- Judge every ad on its cost per order, not its clicks.
- Stop a test at [50] orders or [14] days, whichever comes first.
Most creative testing advice assumes a budget you probably don’t have. Run twenty variants, wait for a clear winner, keep it. At [$10k] a month, twenty variants means none of them ever gets enough spend to tell you anything.
Here is how we test when the budget is small. It is slower, and it works.
Why small budgets need different tests
An ad needs enough orders before its cost per order means anything. With a small budget the spend spreads thin across variants, and every result looks like noise. The answer isn’t more patience. It’s fewer, bolder tests.
Step 1: One variable at a time
Change the hook and keep everything else. Or change the offer and keep the hook. When two things change at once, a winner can’t tell you why it won.
Same video, two hooks · [14] daysSample data
A · Product firstCPA [$38]
B · Problem firstCPA [$27]
Step 2: Fewer tests, bigger ideas
A small budget can’t afford small differences. Test ideas that could move cost per order by [20]% or more: a new hook, a new offer, a new format. Save button colours for later.
| Monthly spend | Tests | Ads | Test first |
|---|---|---|---|
| Under [$5k] | [1] test | [2] ads | The biggest idea you have |
| [$5k] to [$10k] | [2] tests | [2–3] ads each | Hooks, then offers |
| [$10k] to [$25k] | [3–4] tests | [3] ads each | Hooks, offers and formats |
| Over [$25k] | Weekly tests | [4–6] ads a week | The full backlog |
A small budget can’t afford small differences.
Step 3: Judge on orders, not clicks
Click-through rate tells you which ad people noticed. Cost per order tells you which ad paid for itself. On a small budget only the second one matters, so it’s the only number on our test sheet.
Step 4: Know when to stop
A test ends when one of these is true. Write the rule down before the test starts, so nobody argues about it on Monday.
- It reached [50] orders.
- It ran for [14] days.
- After [7] days it costs [30]% more per order than the current best.
What to do next
Pick your biggest idea, write down the one variable it changes, and give it [14] days. If you’d rather we ran it, book a strategy call and bring your ad account.
Sources
- [About the learning phase] · [Meta Business Help Center]
- [Creative best practices] · [TikTok Business Center]
- [Our experiment log: E02, E04, E13] · Hypential
Questions
Two or three per test, plus the current best. More than that and none of them gets enough spend to tell you anything.
Once you know what works. Meta’s automated creative is good at recombining proven pieces, but it hides which piece made the difference. Test hooks and offers by hand first, then let it mix the winners.
Yes. The rules are the same: one variable, bigger differences, orders over clicks. Hooks matter even more there, so start by testing the first second of the video.
