Skip to content
GKgkml.dev
← All scenarios
MLOpsbrutal~30 min

AUC fell 14 points and every pipeline is green

A fraud model retrains weekly and is monitored on data drift (PSI per feature), prediction distribution, and service metrics. Ground-truth labels arrive with a 30-day lag, so offline AUC is computed monthly.

This month's AUC came in at 0.79, down from 0.93. Every data-validation check passed. Every PSI value is below 0.1. Prediction-score distribution is within its usual band. Service latency and error rate are normal. No alert fired at any point.

The last four weekly retrains all completed successfully and each new model beat the incumbent on its own validation split before promotion.

Every guardrail passed and the model got substantially worse. Tell me what class of failure is invisible to all four of those monitors, and what you would compute first to distinguish between the candidates.

Choose how you want to be interrogated

The same case under two modes is two different exercises. If you are unsure, RCA drill is the one that trains root-cause analysis most directly.

Before you start: you begin locked. The Mentor will give you nothing — not a nudge, not a category — until you state a specific position and the mechanism you think produces it. Four genuine attempts unlock the resolution. There is no shortcut.