Cutting Customer Case Volume 20x with Petabyte-Scale Health Signals

Salesforce's Customer Success Group built a Technical Health Score system that processes petabytes of platform telemetry to give customers an objective, measurable view of their Salesforce implementation health. The system uses a five-pillar taxonomy (Security, Efficiency, Operational Excellence, Customization, Observability), normalizes diverse metrics onto a 1–100 scale using distribution-based peer comparison, and runs analytics on an off-core platform to avoid impacting live workloads. A multi-stage ML pipeline using Partial Least Squares regression prioritizes explainability over black-box accuracy, ensuring customers can trace any score change to a root cause. Validation shows customers who improve from Fair to Excellent scores see support case volumes drop ~20x and support costs fall ~35x, while internal data-gathering cycles shrink from weeks to hours.