Théo Garnier
View profile
Renewal-risk scoring
Monthly scoring of accounts based on usage signals + GPT-4 narrative summary. CSMs get a prioritized list.
Renewal rate up 6 pp quarter-over-quarter.
0
Problem
CSM team was reactive — churn signals only surfaced when an account was already disengaging. By then, the save-rate was low.
Flow
Monthly cron → pull usage + support + login data → compute composite score → GPT-4 narrative → rank top-15 per CSM → Slack delivery.
Cadence
Monthly, first business day at 09:00
What I built
- Monthly pull of usage + support-ticket + login frequency per account
- Composite score with weighted signal contributions
- GPT-4 generates narrative per account: why-risky + suggested next step
- Slack-delivered prioritized list of top-15 at-risk accounts per CSM
Limits
- Lags real-time signals — refresh is monthly, not live
- Only B2B contracts — consumer plans are not covered by this scoring
AI tools used
- OpenAI API
- n8n
- Slack
@theo.garnier