Suggestion: Churn prediction based on historical usage pattern changes

Current behavior: At-risk scoring is configurable by pillar weights but tuning is manual with no historical data to validate which signals actually precede churn. Desired behavior: Data-driven churn prediction identifying which feature usage changes historically precede churn — enabling health scoring tuned to real behavioral signals, not arbitrary weights.

Notes: Flagged as a potential differentiator and marketable GoCSM capability. PushPress focus is usage-based signals, not revenue intelligence.