card marketplace · 1m+ sessions · rolled out
Both variants beat control by about 10% on session yield. I recommended the one that engaged people less — because its clicks were qualifying and the other variant's weren't.
Read the case study →product page system · four products · mixed result
Business shoppers convert when they scroll into the page body. Premium shoppers convert when they don't scroll at all. A single template quietly assumes those are the same person.
Read the case study →Design shipped. Results went to analytics and never came back as design decisions. So I built the verdict taxonomy, the intake rule, and the audit that found half the portfolio had never been validated live. Nobody asked me to.
Read the case study →case study 03
Card listing redesign. Won on three of four templates — and I wrote the fourth up as an open question instead of letting the average carry it.
case study 04
A starter card for people with no credit history. The only project here where preference research and live behaviour agreed.
Powered by fountain pens and spite for bad UX.