By Kristians, Founder, AutoAdy. Six years buying Meta ads — first for agency clients, then for his own ecommerce and lead-gen brands.
Last updated:
What changed: Made the H2s question-form, linked the citations, and added the variation engine.
A/B testing Facebook ads means running two variants with one variable changed. Requires 100 conversions per variant, one variable at a time, and 95% significance before declaring a winner.
Key Takeaways
Use Meta's Experiments for clean splits. One variable only. Budget for 100+ conversions per variant in 7-14 days. Split 50/50, run simultaneously.
At $30 CPA: need $6,000+ total (100 x $30 x 2). Run 7-14 days for weekly patterns. Under 7 days misses weekday/weekend skew of 20-40%.
Wait for 95% significance in Experiments dashboard. After 14 days without 95%, difference is negligible — pick lower CPA variant. Run 2-3 tests monthly.
Figures below come from Meta ad accounts AutoAdy has audited or monitors. They are directional, not a controlled study.
FAQ
7-14 days minimum. Under 7 misses weekly patterns. No 100 conversions by day 14? Increase budget or treat as directional.
Not in standard A/B — you can't isolate causation. Multivariate needs 4-8x more budget.
Creative (image/video) — biggest impact, wins 62% of tests. Then headlines, audiences, landing pages.
Related
Lower your Facebook CPA with 3 high-impact levers: creative refresh, audience targeting, and landing page optimization.
Creative fatigue kills ad performance. Learn the 5 detection signals: frequency >3.0, CTR declining 2+ periods, CTR below 50% of peak. Median creative lifespan is 14-18 days.
Spin one winning ad into fresh angles worth testing.
Start free. Connect your Meta account. See your wasted spend in 60 seconds.