Every colony has a morning walking group. Nobody in it is an athlete, yet the group survives for years. The reason is simple: if Sharma ji does not turn up, someone rings his doorbell. A fitness app has no doorbell. That was the real problem behind the VitaFit brief, which asked for two features to lift engagement and repeat usage on a platform offering on-demand yoga, zumba and similar classes.
The brief named two retention problems: users lose motivation, and they feel no personal connection with instructors. I treated these as two separate leaks that need two separate fixes, and wrote the PRD around that split.
Start with who is leaving
I used three personas. The Drifter signs up with energy and fades within weeks. The Duo works out only when a friend does. The Loyalist is regular but treats the app like a video library and would leave for any cheaper library. Six feature options went through RICE scoring. Two came out on top because each one attacks one named problem directly.
Feature 1: Squads and Streak Challenges
Users join a squad of six. The squad's goal is a weekly streak: each member completes three sessions in seven days. One freeze per month covers travel or illness. XP is awarded at +10, +25 and +50 for rising levels of effort. Six is deliberate. In a group of fifty, nobody notices your absence. In a group of six, everybody does. That is the doorbell.
The weekly 3-in-7 rule matters as much as the squad. A daily streak punishes one bad day and the user quits in shame. A weekly streak forgives Monday and still asks for a habit.
Feature 2: Instructor Circles and Coach Notes
After three classes with the same instructor, a user unlocks that instructor's Circle, which includes a live session capped at 25 seats. Users can also send a Coach Note and receive a personal reply within 48 hours. Each instructor has a cap of 40 notes a week, and if the instructor cannot reply in time, the fallback reply is clearly attributed to whoever wrote it. No pretending.
A promise in a PRD is only real when it has a number, an owner and a fallback.
The cap looks like a restriction on users. It is actually protection for the feature. Without it, popular instructors drown in notes, replies become one-liners, and the personal connection we promised turns into a helpdesk.
How we would know it worked
| Element | Decision in the PRD |
|---|---|
| North Star | Weekly active engagement (WAE), with a metric tree under it |
| Target | +15% relative improvement in week-4 retention |
| Proof method | 50/50 holdout for 8 weeks, so the lift is measured against users who never saw the features |
| Rollout | Alpha, then beta, then 50%, then general availability, with kill criteria at each gate |
The second half of the assignment was visual: six flowcharts and eighteen wireframes covering squad creation, streak states, Coach Note states and the edge cases, such as a member leaving mid-week or a freeze being used on the last day. Drawing the states exposed gaps the prose had hidden.
What I took away
A PRD is closer to a rent agreement than a brochure. The brochure says "community and coaching". The agreement says six people, three sessions, one freeze, forty-eight hours, forty notes. Engineers, designers and instructors can only hold you to the second kind of document, and that is exactly why it is worth writing.
Evidence used: public writing on Strava's group-activity data, Peloton Teams and Duolingo streaks informed the two levers. Baselines in my workbook are placeholders awaiting real data. The squad simulation is synthetic and declared as such.