Question

You deployed a demand-forecasting model to a Vertex AI endpoint and enabled model monitoring for feature skew and drift. The business expects gradual, legitimate shifts in customer behavior over the coming months as a new product line ramps up. Your goal is to detect training-serving skew that indicates a data-pipeline problem, while avoiding a flood of alerts caused by the expected slow behavioral changes. How should you configure the monitoring baselines?

Accepted Answer

Use the original training dataset as the baseline for skew detection, and use a rolling window of recent production traffic as the baseline for drift detection.

Answer

Use a rolling window of recent production traffic as the baseline for both skew and drift detection.

Answer

Use the original training dataset as the baseline for both skew and drift detection.

Answer

Disable drift detection entirely and only compare live traffic against the training dataset for skew.