health-model-review

To audit CS health model distribution anomalies, this Skill analyzes portfolio health data to detect tier distribution shifts and quantifies how much ARR sits in each tier. It flags skewed Green, Yellow, or Red classifications to identify misclassification issues.

Customer health models show too many Green accounts due to stale signals, low component data coverage, or incorrect component weights. This Skill separates missing-data issues from true scoring problems and evaluates calibration against actual renewal outcomes.

To recalibrate health model component weights for a CS Ops review, run a full audit to evaluate component weight validity and signal freshness. The Skill provides evidence-based guidance on governance for recommended scoring changes.

You can check health model predictive accuracy by correlating tiers with churn outcomes when historical renewal data is available. The Skill evaluates alignment and surfaces false positives versus false negatives in Green, Yellow, and Red classifications.

Running a predictive accuracy check on a portfolio health model requires health model configuration data and portfolio health data. To assess calibration against churn outcomes, historical renewal data must also be available for the analysis.

This health model audit does not modify live scoring configurations because it enforces strictly read-only analysis. It provides governance guidance for any recommended calibration changes, ensuring you control implementation.