At the same budget, every dollar a new ad spends comes out of one that sells. And Meta counts adding a new ad to an ad set as a significant edit that sends the ad set back into learning. So new ads never go into the ad set your winners live in.
- Tests get their own budget. New ads go in their own ad set or campaign, with their own budget.
- The cap is a fifth. All test spend together stays at or under 20% of your daily budget. The other four fifths stay on what already sells.
- Each idea gets about one sale's worth of spend a day, for 7 days. Less than that and you're judging noise.
- More ideas than slots: the most different go first. A new buyer, a new reason or a new format goes before a small twist on something already running. The rest wait for a slot.
- A test that passes both [winner tests](/resources/winner-tests) graduates. It moves into the main budget, and its slot goes to the next idea in line.
Test budget
Test budget per day = daily budget x 0.2 Ideas testing at once = test budget per day / your cost per sale Each idea: about 1 sale's worth of spend a day, for 7 days
This matters more with AI ads, because you can make more than you can test. Above a fifth, I've watched tests drag a whole account below its target. Under it, a losing idea costs you one small slot for one week.
Testing only pays when the winners can keep funding it.
