An AI visual-search feature for fashion e-commerce — and, more than anything, an exercise in defining what success looks like before you test. We set behavioural criteria up front, then measured against them.
TrendWear, an online fashion retailer, had falling retention and repeat purchases. Despite wide variety and frequent sales, users couldn't find items matching their taste. The goal: improve discoverability and encourage repeat purchase.
You photograph an outfit you like; the AI suggests matching items in the store; dot indicators let you select the specific piece you mean (an outfit is many items — which one?); you add to cart and check out without leaving the app.
We set the success criteria up front — at least one add-to-cart per participant and 75% checkout completion — rather than asking "does it look nice?" We hit 80%. The 20% who dropped off were missing a back-navigation path to return to other products: a fixable problem we identified but didn't iterate on, since TrendWear was scoped as a methods exercise, not a full case study.
“We set behavioural success criteria before testing — add-to-cart and 75% checkout — not 'does it look nice.' We hit 80%. The 20% who dropped off were missing a back-navigation path. The real value was learning to define what success looks like before you test.”
© 2026 Marcos Lourenço
UX & Service Design · Lisbon, Portugal