01
Five parts of the app, five different looks.
Onboarding, account creation, home and sessions each used their own colours, fonts and buttons.
For someone deciding whether to open up to an app, that inconsistency reads as unreliable.
02
Onboarding became the first conversation.
The old flow listed credentials, star ratings and permission requests.
Now the coach introduces herself, asks your name and explains privacy in her own voice. New users send their first message before onboarding ends, and that first message is what we measured.
03
One design system, light and dark.
I designed each component once and built light mode from the same colour rules, so the two themes never drift apart.
Text has a minimum contrast on photo backgrounds, enforced in code, so it stays readable on every screen.
Core colours
#115430
#66BA24
#F5F0ED
#1E2220
#F2994A
Motion
Every animation shares one timing curve. Old content leaves quickly and new content arrives a beat later, so the whole app moves the same way.
Inside the app
Screens from the working prototype, as a returning user sees them.
The A/B test
New users were split 50/50 between the old and new onboarding, and measured over their first 7 days. About 600 iOS and 120 Android users.
Still open
Too early to call retention.
People who got the redesign are coming back more often so far, but they have only had the app a short time. I'm treating it as a good sign, not a result.
We also found a tracking bug that made the old version look better than it was, and fixed it before reporting any numbers.