Bloom

The problem

Funnel data showed a 66% drop before the first wind-down session. Interviews explained why: people arrived at 11pm, tired, and were met with a questionnaire about their mental health from an app they had known for forty seconds.

The questions existed to personalize the program — a real need — but the order was backwards. Trust had to come first.

Process

I prototyped a “session-first” flow: the app starts a two-minute wind-down immediately, and asks one gentle question at the natural pause after it ends. Personalization builds across the first week instead of the first minute.

We tested four variants with first-time users at actual bedtime — remote, dim rooms, real fatigue. Watching people use a calm app while exhausted reshaped the micro-interactions: bigger targets, slower motion, no timers counting at you.

The solution

First run now opens straight into a session with a single choice — “wind down” or “fall asleep”. Questions arrive one per night, framed as tuning rather than intake. Every screen respects a hard rule we set: nothing on screen should raise your pulse.

Outcome

Onboarding completion rose from 34% to 71%, and day-7 activation improved 27%. The one-question-per-night pattern became the template for every personalization flow in the product.