Taxipilot

Taxipilot prepares candidates for taxi and private-hire licensing assessments — topography, regulation, and route knowledge.

The problem. The platform had been designed as a desktop course. Its actual users were drivers studying in ten-minute fragments on a phone, in a parked car, often on a patchy connection. Session analytics showed a large number of starts and very few completions, and the obvious reading — low motivation — was wrong.

What we did. Talking to candidates made the real constraint clear: a study session had to survive being interrupted at any moment. Anything that lost progress when a call came in was effectively unusable. We restructured the material into units that fit inside a realistic gap, made progress durable across interruptions and connection loss, and designed the review system around spaced repetition of the specific items each candidate got wrong.

What shipped. A phone-first learning flow where sessions resume exactly where they stopped, content works through connection drops, and revision targets individual weak areas rather than replaying whole modules.

Related: UX research · product design