A/B Testing Ad Hooks: What to Log When Meta and TikTok See Different Audiences
A lightweight spreadsheet: holdout cells, cost per 15-second view, and when to call a "winner" a mirage.
Hook tests are not "which word is cuter" — they are about which problem framing pulls the right people from cold traffic. That means isolating the first 1–2 seconds, keeping the body identical, and spending enough to exit noise without burning months of budget on noise.
Columns that matter
- 15-second/75% view rate or equivalent attention proxy
- Thumb-stop rate if available
- CPC to landing — not the same as creative quality, but tells you if curiosity matched intent
Rapid ideation
Script ten first lines from a single offer; film once if you are on camera, or generate variants for off-camera VO tests.