League Pass churn-to-campaign pipeline
A retention workflow that flags at-risk League Pass subscribers, drafts a personalized save campaign with generative AI, and holds every campaign for human review before it goes out.
- Context
- Independent concept prototype. Not affiliated with or endorsed by the NBA.
- When
- Feb 2026
- Built with
- Lovable, React
The problem
Streaming subscriptions churn quietly. By the time a League Pass subscriber cancels, the signals have usually been there for weeks: watch time falling off, games skipped, a downgrade to a cheaper plan, no logins. A retention team needs to act on those signals early, at scale, without sending something generic or wrong to a paying fan.
What it does
The tool is a four-stage board. A churn model flags high-risk subscribers, generative AI drafts a campaign for each one, a person audits the draft, and only then is it deployed. Each subscriber card shows the churn probability, the plan value at stake, their team, and the signal that triggered the flag. A running total at the top tracks revenue at risk, interventions pending, and revenue addressed.
Before drafting, you can interview a “digital twin” of the subscriber: a simulated version of that fan you can question about why they’re drifting away, so the campaign answers a real reason instead of guessing.
How it works
- Scoring: the churn risk score comes from a random forest. Each tree estimates the probability of churn from behavioral features such as watch time decay and how the subscriber’s local team is trending, and the score is the average across trees.
- Drafting: the language model doesn’t freelance. It receives the model’s outputs as structured context in a constrained system prompt, so the campaign is anchored to the signals that flagged the subscriber.
- Human audit: nothing deploys straight from the model. Every draft passes through a review column first.
- Compliance mode: a toggle masks personal information on screen, for demos and for anyone who doesn’t need to see it.