The problem
CSMs discover churn risk in the renewal call, when it is too late to intervene. The signals were there weeks ago: usage dropped, support tickets spiked, the champion left.
The system
- Define churn patterns. Extract signals that preceded past churns from usage data, support logs, and call recordings.
- Build monitors. Automated detection on the same patterns for current accounts.
- Score and route. Risk score with evidence delivered to the CSM as a Slack DM.
- Track saves. Measure intervention success rate to refine monitors.
Field notes
Churn signals use the same infrastructure as buying signals. The scraper fleet and routing layer are identical; only the patterns differ.