Photo by Kelly Sikkema on Unsplash
The default creative refresh on most paid social accounts is monthly: three new ads land, two die in week one, the rest run on borrowed time. Meta's own ad system trips fatigue alerts when frequency hits 3 to 4 plus, which happens for most ecommerce ads inside 7 to 10 days (Growth Jockey, 2025; Madgicx, 2025). The refresh cadence has to match the fatigue cadence.
A weekly cadence is the practical floor. AdAmigo's 2025 benchmarks recommend testing 3 to 5 creatives per phase with at least 50 conversion-tracked events before calling a result (AdAmigo Meta Ad Creative Testing Benchmarks, 2025). Motion's 2025 guide adds a 1,000-impression-per-variant or 100-conversion floor for statistical significance in competitive ecommerce auctions (Motion, 2025). Below those volumes, you are not testing; you are guessing.
The harder discipline is concept diversity, not volume. Motion's 2026 Creative Benchmarks report finds that only about 5% of ads spend 10x their account median, meaning real winners are rare and the cost of testing the same concept five different ways is paying for the same losing bet five times (Motion Creative Benchmarks 2026). Pixis' 2024 framework for scaling teams allocates 30 to 50 weekly creative tests against roughly 20 always-on winners absorbing the bulk of spend (Pixis, 2024).
A 20% lift over baseline is the conservative win-call threshold (Motion, 2025); below that, creative differences usually wash out in auction noise. Combine it with Meta's automatic fatigue alerts and you have a decision rule: launch weekly, hold each variant to 1,000+ impressions or 100+ conversions, scale anything 20%+ above baseline, kill the rest, and refresh against fatigue rather than the calendar.