Test hooks without changing anything else
This template exists to make paid-social experiments interpretable. The body beats and closing instruction are written once and reused across every variant, while the opening scene is the only thing that is regenerated.
This template exists to keep experiments honest. The body beats and the closing action are written once and reused, while the first scene is regenerated per variant. Because only one scene changes, performance differences can reasonably be attributed to the hook rather than to pacing, voice or edit.
Run five to ten angles — pain, outcome, objection, curiosity, social proof — and archive the losers instead of deleting them; an angle that fails for one audience often works for another. The template is not meant to be published as-is without at least one real test.
Keep a short log of which angle each variant used and what it produced. The archive is what turns a month of spending into a reusable message, rather than a folder of files nobody can tell apart.
Scene outline
4 scenes · ≈15s of 15s
- 01~3s
Hook (variable)
The only scene that changes between variants; one angle, one line.
- 02~5s
Mechanism (fixed)
How the product produces the result, in a single beat.
- 03~4s
Believability (fixed)
The detail that makes the mechanism credible.
- 04~3s
Action (fixed)
The same closing instruction in every variant.
Example prompt
Create a 15-second vertical ad for a meal-planning app, structured for hook testing. Keep this body fixed: the app builds a week of meals from what is already in the cupboard, generates a single shopping list, and adapts when plans change. Keep this ending fixed: start a free week today. Then write the opening hook as a pain-led line about throwing away food bought for a recipe never cooked. Conversational tone, first person, no exclamation marks.
Best for
- Paid social testing
- Creator ad variants
- Landing-page message research
Frequently asked questions
- How many scenes are regenerated per variant?
- One. That is the whole point: the mechanism, believability and action scenes are reused unchanged, so the credit cost per variant is the cost of a single opening scene.
- Can I vary the call to action instead?
- Yes — the structure works for any single-variable test. Just change one thing per round, or the result stops attributing cleanly.
- What makes a good set of first angles?
- Pain, outcome, objection, curiosity and permissioned social proof. They fail in different ways, which tells you more than five wordings of the same idea.
Run a clean hook test
Fix the body, generate the openings you want to compare and see the estimate for the whole batch first.