Wolf: an activation loop for a live social app

How it works

You cannot rank by behaviour for a user who has no history. That was the insight everything else followed from. Onboarding had to bootstrap the interest graph immediately, and the discovery feed had to use those interests from the very first open. Onboarding and discovery stopped being two features and became one loop.

The feed ranks live rooms on six signals: the interest graph captured at signup, behavioural history, country-level cultural relevance, trending, social proof (which friends are in this room), and room quality (how many people, gifts sent, activity level). The same targeting surfaces in-app games, which were an engagement surface of their own.

The third surface is behavioural notifications. They fire on the user's own history: a relevant event starts, a friend joins a room, activity picks up somewhere they care about. That is the re-engagement half of the loop.

I ran product discovery, wrote the specs, prioritised the backlog and worked hands-on with two native squads, one per platform, as the single product owner across both.

Key features

• Onboarding designed to capture interests, not just create an account
• Six-signal discovery ranking with social proof and room quality as first-class inputs
• Behavioural notifications as the re-engagement mechanism
• In-app games targeted with the same interest and behaviour signals
• Retention tracked two ways in parallel: app opened, and core action (room joined) within N days

Step by step

1. Sign upInterests are captured during onboarding.
2. Interest graphBootstrapped from onboarding so ranking works from the first open.
3. Feed rankingSix signals combined to surface live rooms and games.
4. ActivationThe user joins a room or a game in session one.
5. NotificationsTriggered by the user's own history and social graph.
6. ReturnMeasured as core-action retention at day 1, 7 and 30.