Question

Your team deployed a demand-forecasting model on a Vertex AI endpoint with model monitoring enabled for prediction drift. Since launch, the monitoring job triggers drift alerts almost every day, but on investigation most alerts correspond to normal seasonal fluctuations in traffic patterns rather than genuine model problems. Engineers have started ignoring the alerts. Which action best restores the usefulness of your monitoring while still catching real drift?

Accepted Answer

Increase the drift detection threshold (alerting distance) so that only larger, sustained distribution shifts trigger alerts, and validate the new threshold against historical seasonal data.

Answer

Disable prediction drift monitoring entirely and rely on periodic manual model re-evaluation instead.

Answer

Reduce the monitoring sampling rate so fewer prediction requests are analyzed, which will produce fewer alerts.

Answer

Switch the monitoring baseline from the training dataset to the most recent hour of serving traffic to make the comparison stricter.