Onboarding rebuilt around activation, not a feature tour
People signed up for a mental-health app and never reached their first real session. The product explained itself instead of getting them to do one thing.
An early-stage app in a crowded category. Sign-ups were fine; what happened after wasn't. Nothing in the funnel told us where people gave up, because the first five minutes were a tour rather than a task.
Mapped the CBT, journaling and mental-health landscape and pulled out the patterns that get people to value fastest. Then rebuilt onboarding around one completed action — seven screens instead of six, but every one of them doing work — prototyped it in Figma and took it through the founder and the engineers to implementation. I also reworked the App Store listing, where the funnel actually starts.
As co-founder I also own how the product reaches people: positioning, the landing page, and the funnel that feeds it. Which means I design onboarding knowing exactly what promise the user arrived with.
Step by step, old above and new below
Onboarding · before / afterRoughly the same number of taps, a different job. Every step now either sets an expectation or changes what the user gets — and the two screens that only made them wait are gone.
The old first screen was a stock landscape and a slogan. The new one names what the app is for, in the visual language of calm the product actually delivers.
Sign-in now comes with a reason — keep your streak, journal and check-ins — plus a way to continue without an account, so the wall isn't the first thing you meet.
The 25%-to-100% "personalising your journey" loader was two screens of waiting that taught the user nothing. It's gone; the same time now goes into the first check-in.
The funnel starts before the app does
App Store product pageThe old screenshots sold a different product to the one we had built: bold black type, a hard purple-and-yellow illustration, and captions competing with each other for the loudest claim. Nothing in them looked like the calm the app is actually for.
I rebuilt the set to match the new visual language and the real feature order — one claim per screen, in the sequence a user meets them: check in, get today's plan, talk it through, see progress. Conversion from product page to install went from 23% to 29%.
