Guide
Creative is the biggest lever left in Facebook ads — targeting is mostly algorithmic now. The problem is that testing creative the standard way is expensive. Here's a framework that finds winners without funding a graveyard of losers.
A live test only means something once each variant has enough conversions to separate signal from luck — a common rule of thumb is around 50 conversions per variant. At a $20 CPA, that's roughly $1,000 per variant; test five creatives and you've committed $5,000 to find out which one wins. And most won't: practitioners consistently report that only a small minority of creatives — often one to three in ten — become winners. The standard playbook asks you to pay full price to identify the seven duds. The full cost comparison is in pre-testing vs A/B testing.
Three to five per test is the practical sweet spot. Below three, you can't spot patterns — you learn that an ad lost, not why. Above five, the budget per variant gets so thin that nothing reaches significance and you end up extending the test or making gut calls. The more important number is cadence: because winners fatigue, you need a refresh pipeline, not a one-off test. Test variants that differ on one meaningful axis — hook, format, angle — rather than five near-identical cuts, or you'll get five near-identical results.
The cheapest variant to test is the one you never launch. Before any budget is committed, pre-test each creative: PreTestAds predicts viewer engagement from the video or image itself and scores it against a benchmark of 76 curated top-performing ads — Hook Strength in the first 3 seconds, where attention drops, the peak moment, and whether engagement is still alive in the CTA window. Variants that score weak get cut or re-edited before they ever touch your ad account. Your paid test then compares three contenders instead of five lottery tickets, and the same per-variant budget buys cleaner data.
Every winner dies. The pattern is consistent: frequency creeps up, CTR slides, cost per result rises — the audience has seen the ad and already voted. At higher spend this happens faster, sometimes within weeks. Teams that scale smoothly treat creative as inventory: while the current winner runs, the next batch is already scored, edited, and queued. Pre-testing makes that pipeline cheap to maintain, because refreshing creative doesn't require a fresh round of paid validation for every candidate. (If you're also weighing a post-spend creative-reporting layer for reading the account side, see Motion and its alternatives.)
Predicted engagement tells you whether people will watch — it doesn't know your offer, your price, or your landing page, so it can't promise conversions. Use it to screen creative, then let the live test on real spend make the final call. If an ad holds attention but still doesn't convert, the creative usually isn't the culprit — see why ads don't convert for where to look next. Testing a multi-card format? The same screen applies to the lead frame — see carousel ad best practices.
Upload your creatives — video or static — and see which ones are worth a paid test. Up to 8 free credits, no credit card.
Score Your CreativesThree to five creatives per test is the practical sweet spot — fewer than three limits pattern-spotting, more than five spreads budget too thin to reach significance. Most teams find only a small minority of creatives become winners, so a steady testing pipeline matters more than any single test.
A defensible live test needs enough conversions per variant to trust the result — a common rule of thumb is roughly 50 conversions per variant. At a $20 CPA that is about $1,000 per variant, so a five-variant test can run $5,000. Pre-testing creative before launch reduces how many variants need paid validation.
Creative fatigue is when an ad's performance decays because the audience has seen it too often: frequency climbs, CTR falls, and cost per result rises. The fix is a pipeline of fresh tested creative ready to rotate in — not squeezing more budget into a tired ad.