Sohum
A breathwork companion that teaches you to slow down
Meditation apps kept asking beginners for ten quiet minutes they did not have. Sohum starts at ninety seconds and earns the rest.
- Role
- Research, UX, UI, prototyping
- Timeline
- 14 weeks
- Platform
- iOS and Android
- Year
- 2025

(The brief)
My final-year capstone, built with a yoga therapist who ran beginner classes at a community centre in Jayanagar and kept losing students after week two.
(The problem)
Every app we audited opened with a ten-minute guided session and a streak counter. For a first-timer that is two forms of pressure at once. Nine of the twelve people I interviewed had downloaded a meditation app, used it twice and never reopened it. The barrier was never interest, it was the size of the ask.

What people actually told me
I ran twelve interviews and three diary studies over two weeks. The pattern was blunt: beginners did not want to meditate, they wanted to stop feeling wound up before a meeting. They also hated being watched. Streaks, badges and weekly reports read as judgement, and the moment someone missed a day they deleted the app rather than face the broken chain.
The decision that shaped everything
We cut the streak. In its place the home screen asks one question, how much time have you got, with ninety seconds as the default and the largest tap target on the screen. Longer sessions exist but you have to go looking for them. It felt risky, retention mechanics are the industry default, but it tested as the single most-liked change in round two.
Designing a pacer you can follow with your eyes closed
A visual breathing guide is useless the moment someone shuts their eyes, which is exactly what you want them to do. I paired a slow expanding form with a haptic pulse at the turn of each inhale and exhale, so the phone can sit face down in a lap and still keep time. Rive let me prototype the easing properly, and the therapist tuned the ratio until it matched how she paces a room.
Testing and what broke
Eight moderated tests on the clickable prototype. The first build hid the session length behind a settings sheet and four of eight participants never found it. The second put it on the home screen as a row of three tiles and everyone completed the task in under six seconds. The audio permission prompt also had to move: asking on launch got refused, asking after the first completed session got accepted every time.

(Key decisions)
- Ninety seconds as the default session, not ten minutes
- No streaks, no badges, no weekly performance summary
- Haptic pacing so the screen can be ignored entirely
- Permission requests deferred until after the first success
- A palette tested at WCAG AA in both light and dark rooms
(Outcomes)
- Testers completed a first session unaided
- 8/8Testers completed a first session unaided
- Median time from open to breathing
- 6 secMedian time from open to breathing
- Pacer iterations before the final easing
- 11Pacer iterations before the final easing
- Contrast compliance across both themes
- AAContrast compliance across both themes
What I took from it
The best feature I designed was the one I removed. Cutting the streak counter made every other decision simpler, and it is the change the therapist still mentions when she recommends the prototype to new students.
(Next project)
Kirana
Bringing the corner shop online without losing what makes it work
