The flow looked fine in review and tested badly with real people, which is the gap that costs the most to find late. I owned the usability work through several rounds — watched where people stalled, changed one thing at a time, and stayed on it until the numbers moved.

Reworked the flow around what usability testing surfaced; task success rose from 8% to 19%.

Every screen was designed from scratch and nothing felt like it belonged to the same product. I built the system — tokens, components, and the usage guidance — plus the rituals that keep it alive after the launch enthusiasm wears off.

Rebuilt the flow around the first real win; early drop-off fell 45% without adding a feature.

We were guessing what users wanted and calling it intuition, which meant the loudest opinion usually won. I ran the studies, brought back evidence instead of preference, and turned the findings into decisions the team could trust and repeat without me in the room.

Redesigned the loading, empty, and error states across the product; the app felt faster with no backend change.

The team was busy shipping screens and nobody could say whether any of it worked, because we measured output instead of outcomes. I put the metrics next to the design work — defined success up front and tracked it after launch — so we could tell a real improvement from one that just felt fresh.

Simplified the flow and clarified the copy; completion rose from 15% to 25% in the test.