Consumer App
Creatives that scale past the hero concept.
Why hero creatives stop working at week four, and the concept system that replaced them across three English-speaking markets.
CASE / 02
Confidential
Every ad was a slightly worse version of the last one that worked. The team shipped on time, volume was fine, but CAC drifted back up within a month of every new winner. That was the problem we were hired to solve.
GEO
US · CA · UK
Setup
Concept system
Duration
9 weeks
Shipped
Q1 2026
The diagnosis took four weeks
The diagnosis took four weeks
The paid team was not the problem. Production was on schedule. The issue was structural. The concept that made the hero ad work had never been written down anywhere. Every new variant copied the surface (shot list, script beats, colour grade) and missed the three ingredients that actually moved CAC.
We spent a month sitting in weekly creative reviews and reading the last 180 days of asset data. Three things came out of that.
Concept was tacit. Winners clustered into four patterns. Five of the losing variants were copying pattern surface, not pattern structure.
Iteration was reactive. New creative was only commissioned once CAC drifted past threshold. By that point the team was two weeks behind the decay curve.
Regional blind spots. US and Canada ran the same creatives with swapped accents. Platform algorithms flagged the duplicates. CVR held for a week, then dropped.
Pattern extraction
Pattern extraction
Three highest-performing creatives from the last six months got reverse-engineered. Not the script, the structure: hook type, visual anchor, proof shape. Winners fell into four clear patterns. One of them had never been documented.
Concept matrix
Concept matrix
Twenty-four concepts tested. Each one was a pattern crossed with a regional cue. Batches of four shipped per week. Every concept had a written hypothesis and a kill threshold before the ad went live. Three got killed in the first two weeks. Six moved to scale.
Regional rhythm
Regional rhythm
Instead of shooting in the US and translating, UK and Canada ran in parallel from week one. London-shot UK creatives with UK proof points. CA used Canadian financial references. Same concept, different region. Platform algorithms stopped flagging duplicates. CPM came down as a side effect.
Launch With Us