The method
Every round runs the same five steps in the same order. Keeping the process fixed is what lets one round's result be compared with the next.
The five steps
- 01
Brief
We start from the number your team is judged on and the audience you are buying. One page per concept: product facts, allowed claims, hooks to film and the deciding metric.
Mock layout - 02
Hooks
We write a bank of openings per concept from different angles, and one body that every hook cuts into. If two hooks need different bodies, they become two concepts and go into a separate test.
Mock layout - 03
Production
Cast creators film the hooks as separate takes and the body once, in vertical, where they usually film. We edit every hook onto the shared body and cut lengths and ratios per placement, with text inside the safe zone.
Mock layout - 04
Launch
Variants launch with matched budgets and a spend threshold each, from the creator's handle, the brand's account or both. Disclosure is checked on every ad before it goes live.
Mock layout - 05
Readout
Once every variant has reached its threshold, we read the decision metric first and the creative metrics second. Losers are cut, winners get more budget and a fresh challenger, and the readout becomes the next brief.
Mock layout
One variable per test
A test that changes the hook and the creator at once can't say which one moved the number. We change one variable and hold the rest, and when two variables both matter we cross them in a matrix.
- Hook
- The opening: first frame, first line and on-screen text.
- Creator
- The same script performed by different people.
- Identity
- The creator's handle against the brand's account.
- Length
- The same concept cut to different durations.
- Offer
- The incentive or the price framing in the body.
- End card
- The closing frame and call to action.
Reading a test matrix
- Read across a row to judge a hook. One that wins with every creator is an angle worth scaling.
- Read down a column to judge a creator. One who lifts every hook is worth a second concept.
- Read single cells last, because one strong cell on its own can be a presenter carrying a weak hook.
| Hook / creator | Creator A | Creator B | Creator C |
|---|---|---|---|
| Problem | ScaleHook 31.2%CPA $18.40 | IterateHook 28.4%CPA $21.10 | ScaleHook 33.9%CPA $17.60 |
| Result | CutHook 24.7%CPA $24.90 | CutHook 22.1%CPA $27.30 | IterateHook 26.3%CPA $23.80 |
| Demo | CutHook 35.8%CPA $29.70 | CutHook 30.6%CPA $31.20 | CutHook 38.2%CPA $26.40 |
| Objection | ScaleHook 21.3%CPA $19.60 | IterateHook 19.8%CPA $22.80 | ScaleHook 23.5%CPA $18.90 |
Decision rules
Rules are written into the test plan before launch, so a result is read the same way whoever reads it.
- Spend threshold
- Each variant spends twice the target CPA before it is judged.
- Scale
- CPA 10% or more under target at threshold. Budget rises in steps, with a fresh challenger beside it.
- Iterate
- CPA between 10% under and 10% over target. The body stays and a new hook or a new creator goes on it.
- Cut
- CPA more than 10% over target at threshold. The budget moves to the leaders.
- Fatigue
- Frequency rising while hook rate and click-through fall. The next challenger rotates in.
Example rules. Thresholds are set per account and agreed before launch.
Terms we use
Platforms report the underlying metrics under their own names. These are the definitions behind our readouts.
- Hook
- The opening of an ad, before a viewer decides to stay: the first frame, the first spoken line and the on-screen text.
- Body
- Everything after the hook. In a hook test every variant shares one body.
- Concept
- One message for one audience, expressed as a body with several possible hooks.
- Variant
- One finished ad in a test, differing from its siblings in exactly one variable.
- Hook rate
- 3-second views divided by impressions. It shows whether the opening stops the scroll.
- Hold rate
- Views to 15 seconds, or to the end for shorter ads, divided by 3-second views. It shows whether the hook leads into the body.
- Decision metric
- The one business number, agreed before launch, that decides a test. Usually CPA, ROAS, cost per install or cost per lead.
- Spend threshold
- The minimum spend a variant must reach before it is judged, set in the test plan as a multiple of the target CPA.
- Spark Ad
- A TikTok ad that promotes an existing organic post, from a brand account or from a creator's account with an authorization code.
- Partnership ad
- A Meta ad on Instagram or Facebook that runs with a creator's or partner's identity alongside the advertiser's. Formerly called a branded content ad.
- Safe zone
- The part of a vertical frame that the platform interface does not cover, where text, faces and the product should sit.
- Creative fatigue
- A winning ad losing performance on the same audience over time, usually seen as frequency rising while hook rate and click-through fall.
Plan a first test
Send the product, your target metric and a monthly budget range. We reply with a proposed first test.
Plan a test