1
Load two snapshots of the same dataset
A model is only valid on data that looks like what it was trained on. Give me the baseline (what you trained on) and the current data (what you want to score), and I'll tell you whether that assumption still holds.
Baseline training data
Drop baseline CSV
the data your model learned from
Current new data
Drop current CSV
the data you want to score
Or load a worked example:
2
Verdict
3
Column by column
Sorted by Population Stability Index — the measure risk teams actually monitor. Below 0.10 is stable, 0.10–0.25 warrants watching, above 0.25 is a significant shift.