Most Customer Health Scores Are Never Validated Against Actual Churn

A majority of customer success teams build a health score once at rollout and never back-test it against actual churn or expansion outcomes, resulting in scores that quietly lose predictive power as the product and customer base evolve.

• Health scores should be back-tested against actual 90-day-forward churn/expansion outcomes at least twice a year
• Usage-based signals typically carry more predictive weight than sentiment signals, but the ratio should be validated per-business rather than assumed
• CSMs trust and act on health scores more when the scoring logic is transparent rather than a black-box weighted average
• A health score that hasn't been revalidated in over a year should be flagged as a governance risk, not treated as reliable

• • Auditing an existing customer health score model for predictive validity
• • Building a governance checklist for periodic health-score revalidation