Which creative actually works? +73% CVR, answered.
After leaning into the winners
Repeatable, not a one-off
Video, text, and image mining
The problem.
The Lip Bar was running a wide set of paid-social creatives without a rigorous read on which ones actually drove conversions. Spend was spread across ads on gut feel and surface metrics, not on a real understanding of what was working, so budget followed whatever felt good that week.
The approach.
I analyzed the creative library to isolate the true performers, and there's a real toolkit for this beyond eyeballing a dashboard. Video analysis reads scene pacing, hook timing, and on-screen text density. Text mining of captions and hooks surfaces the language that lands. Image analytics on thumbnails picks up dominant color, composition, and framing. Pattern-matching the winners against those features turns "this ad felt good" into "these specific attributes convert."
The outcome.
Leaning spend into the attributes that actually converted lifted paid-social ad CVR by 73%. More valuable than the number: the analysis became a repeatable method, a way to score the next batch of creatives against what wins, instead of relitigating the last batch.
From the work.
If you're spreading spend on gut feel, I'll tell you which creative attributes actually convert, and give you a way to keep scoring them. One slot open for Q3 2026.