Should I test one ad format or multiple formats simultaneously on Meta?

Test multiple formats simultaneously — one format per ad set — and let Meta's performance data pick the winner. Format winners aren't predictable, and the spread is large: in one documented comparison a person-led UGC ad returned 6.2x ROAS where a product-only ad hit 0.8x. AI production makes the parallel test affordable: one run shipped three complete formats for $225 total.

Test multiple formats simultaneously, one format per ad set, and scale whichever the data rewards. You cannot predict which format will win without shipping it — in one documented Meta Ads Manager comparison, a person-led UGC ad delivered 6.2x ROAS while a product-only ad for the same offer returned 0.8x. A sequential one-format-at-a-time test would have burned weeks before surfacing that gap.

Structure the test so format is the only variable. Give each format its own ad set rather than mixing formats inside one — mixed ad sets make it impossible to attribute performance to the format itself. Hold everything else constant across the test: same offer, same audience, same hook logic. Once a format wins, lock it and move variable testing inside that format — never test format and hook at the same time.

The cost math now favors parallel testing. Pre-selecting one format was a constraint of traditional production, where a single UGC creative runs $300–$600 per creator with roughly two-week turnarounds. AI production removes that constraint: one documented run produced three distinct formats — a product showcase film, a faceless UGC ad, and a full character UGC ad — for 890 credits, about $225 total, roughly $75 per creative. At that price, testing three formats costs less than one traditionally sourced ad, and Meta's algorithm demands more creative volume than most brand teams can physically shoot anyway.

Produce all formats in one pass. invideo is an agentic video creation tool, and its structure maps directly onto parallel format testing: set up your brand context once (15–20 minutes for guidelines, product images, and tone), then create a separate sub-agent per format inside the same project so each develops format-specific expertise without cross-contaminating its context — a UGC sub-agent shouldn't inherit the polish rules of your product-film sub-agent. Because the project context persists, each subsequent ad gets faster: in one documented project the second ad took 2 hours versus 3 for the first, a 33% reduction, and a two-person team sustained 4–5 ads per 8-hour day. "In the old world, you'd have to guess which one's going to win and then just ship that, but now with this workflow, you can ship all of these in parallel and just scale the one that actually performs," as invideo's creative team puts it.

After a format wins, iterate inside it. Re-upload the winning ad to the invideo agent and generate structural variants — new hooks, audiences, or products — while preserving the format that performed; documented workflows produced 8–10 variations of a single winning ad in one day. Concentrate that iteration on the opening: the first 3 seconds decide the performance of a faceless UGC ad, and the invideo agent can generate 7–8 hook directions per pass, each with a rationale, so your phase-two test stays as structured as your phase-one format test.

Watch some of these to see what works for you:

In the old world, you'd have to guess which one's going to win and then just ship that, but now with this workflow, you can ship all of these in parallel and just scale the one that actually performs.

— invideo's creative team