Churn Signal Feature Calculator

Scenario

Use after normalization to turn recency, usage, support, sentiment, NPS, and renewal timing into explainable features without requiring a trained model.

Input and output

Input is a normalized customer health contract. Output contains feature values, risk contributions, missing-signal warnings, and a preliminary score from 0 to 100.

Execution and recovery

Run python scripts/features.py examples/input.json. Tune thresholds in the input policy and rerun. This skill only computes evidence; it does not send alerts or start retention campaigns.

Privacy boundary and acceptance

Use redacted account references and coarse bands. Acceptance requires contributions that can be recomputed from the input and a warning whenever a key feature is unavailable.