How many ad creatives should you test per month?

A straight answer, the reasoning behind it, and why volume only helps when every concept is grounded in something real.

Top advertisers on Meta test somewhere between 20 and 40+ genuinely different creatives a month. If you're shipping two or three polished pieces, that's the ceiling on your account, not the algorithm's.

Why the number is that high

Meta's algorithm improved faster than most creative teams did. About 80% of performance in 2026 comes from creative, 20% from media buying. Broad targeting is the default, so the creative does the filtering. To find winners in a broad audience the machine needs a lot of distinct swings, not minor recolors.

"Genuinely different" is doing the work

Twenty near-identical variations isn't twenty tests. Five real approaches (a UGC clip, a product demo, a testimonial, a text-heavy explainer, a lifestyle shot) give the algorithm room to actually learn. Volume without variety just spends money faster.

The bottleneck is rarely the algorithm's appetite for creative. It's your team's ability to produce concepts you can defend, fast enough to feed it.

Volume only pays off when it's grounded

Thirty ungrounded concepts a month is noise with a budget. Thirty grounded ones, each anchored to a winning pattern, a competitor gap, or a buyer insight, is a testing program. That second version is AgentMark's whole job: generate the batch, grounded, in your format, so the volume is real.