
Row Count Funnel
coverageRaw inputs, training windows, and optional scored rows at a glance.
Curated EDA, feature-engineering diagnostics, model interpretation, slice analysis, and monitoring drift views. Click any thumbnail to enlarge; interactive Plotly versions open in a new tab when available.

Raw inputs, training windows, and optional scored rows at a glance.



Sampled numeric feature histograms across feature-engineering groups.



Top category counts for categorical engineered features.

High-variance numeric features correlation structure.


Hour-of-day and weekday coverage across training windows.




ROC, PR, calibration, and score-by-class view.


LightGBM gain importance from trained artifacts.

LightGBM split-count importance across engineered features.

Permutation importance on held-out rows.

Aggregated feature gain by registry group.

Global SHAP distribution across sampled held-out rows.

Mean absolute SHAP contribution ranked by feature.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Top local contributors for a representative held-out case.

Top local contributors for a representative held-out case.

Top local contributors for a representative held-out case.
Full model analysis bundle for this target.
single_class
Full model analysis bundle for this target.
single_class
Full model analysis bundle for this target.
single_class

Predicted-vs-actual, residual histogram, and residual scatter.

LightGBM gain importance from trained artifacts.

LightGBM split-count importance across engineered features.

Permutation importance on held-out rows.

Aggregated feature gain by registry group.

Global SHAP distribution across sampled held-out rows.

Mean absolute SHAP contribution ranked by feature.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Feature value vs SHAP contribution.

Top local contributors for a representative held-out case.

Top local contributors for a representative held-out case.
Full model analysis bundle for this target.
not_enough_rows

Prediction vs target behavior across common product slices.

Prediction vs target behavior across common product slices.
Prediction and feature PSI from the latest monitoring report.
no_drift_metrics

Current scored distributions against training hold-out predictions.