The brief
A fitness app with a modest performance marketing budget needed to find which messaging angle actually worked, weight loss, strength, habit building, or convenience, before spending on a full production shoot built around the wrong hook.
How we would approach it
Separate the hook from the production value
Early testing does not need expensive footage, it needs to isolate which opening line and framing stops the scroll. AI generated video lets a wide set of hook variations go into testing in days rather than the weeks a full shoot for each angle would require.
Build a structured test matrix, not random variations
Variations generated against a deliberate matrix, different opening hook, different core promise, different pacing, so the results actually isolate which variable is driving performance rather than producing a pile of ads with no clear read on why one won.
Run the test on a small, controlled budget
A limited spend spread evenly across the variations, run only long enough to reach a reliable read on hook performance, not a full campaign budget, since the goal at this stage is information, not scale.
Reserve the real production budget for the proven winner
Once a hook and angle is clearly outperforming the others, that is what the full production shoot gets built around, so the more expensive, polished asset is backed by evidence of what actually works for this specific audience.
What this approach is built to deliver
A validated creative direction before the real production budget is spent, cutting the risk of commissioning an expensive shoot around an angle that was never going to perform.