How we test ads
The loop behind every sprint and every month on retainer: research, angles, a matrix, production on 72-hour cycles, a fair test, and a decision for every ad. Then round again, briefed from what the last round showed.
The steps
- 1
Research
Your reviews, your best sellers, the ads competitors keep paying to run, and your own past results.
Day 1 - 2
Angles
Four or five reasons to buy, each one specific enough to be wrong.
Days 1–2 - 3
Matrix
Angles crossed with executions, one ad per cell, so every result says something.
Day 2 - 4
Produce
Scripts approved, ads made and checked, on 72-hour cycles.
Days 2–6 - 5
Test
Every ad gets a fair budget in your account, against your current best.
Week 2 - 6
Read
The numbers in order: hook, hold, click, purchase.
Weeks 2–3 - 7
Decide
Cut, iterate or scale, with the reason written down.
Week 3
How the twenty ads are planned
Rows are angles, columns are executions. A winning row tells you what to say; a winning column tells you how to make it.
| Presenter, to camera | Presenter, demo | Hands only | Product hero | |
|---|---|---|---|---|
| Problem first | V01cut | V02cut | V03iterate | V04cut |
| The routine | V05cut | V06scale | V07iterate | V08cut |
| The ingredient | V09cut | V10cut | V11cut | V12cut |
| The unboxing | V13iterate | V14cut | V15cut | V16cut |
| The myth | V17cut | V18cut | V19cut | V20cut |
- Cut: stopped after the test
- Iterate: new versions of the angle
- Scale: gets the budget
The numbers we read
- Hook rate
- Did they stop? If it’s low, the opener is the problem. New first two seconds.
- Hold rate
- Did they stay? If it’s low, the middle drags. Tighter edit, product on screen sooner.
- Click-through rate
- Did they want more? If it’s low, the reason to click isn’t there. Sharper offer or ending.
- Cost per customer
- Did they buy, sign up or subscribe? If clicks are high and customers few, look past the ad: the page, the price, the offer.
How we decide what to cut
- When to judge
- Each ad is judged once it has spent a set multiple of your target cost per customer, agreed up front.
- Cut
- Weak on early signals with no sign of recovering: stopped, and the reason written down.
- Iterate
- A strong hook with weak conversion: keep the opener, change one thing.
- Scale
- At or under your target cost per customer: more budget, raised in steps.
Want to see what we’d test first?
Tell us what you sell and what you’re running now. We’ll come back with the angles we’d test first and what it costs.