Check a predictive model for input drift
CDP
Check a predictive model for input drift
Compare the live scoring population’s feature + prediction distribution against the model’s training distribution using the Population Stability Index (PSI). Query params: window_days (default 30, how far back the recent window reaches), bins (default 10, equal-frequency buckets), and threshold (default 0.25, the PSI value that counts as drift). Returns 200 with a per-feature PSI report plus an overall drift verdict, or 422 when the recent window has too few active contacts to judge. Owner / admin / developer only.
GET
Check a predictive model for input drift
Path Parameters
Model key from GET /api/v1/cdp/predictive-models (also uplift for the persuadability endpoints).
Available options:
churn_propensity, conversion_intent, lifetime_value, engagement_fatigue Query Parameters
Required range:
1 <= x <= 365Required range:
2 <= x <= 20Required range:
0.05 <= x <= 1