PRODUCTION SCENARIO
A SageMaker Model Monitor data quality schedule reports that three input features have moved away from the training baseline. Endpoint latency and error rate are unchanged, and confirmed outcome labels for recent predictions have already been collected in Amazon S3.
Which measurement should the team make FIRST?
Answering here is anonymous. Nothing is saved unless you sign in.
Show answer and explanation
Answer: Model quality metrics against the baseline, using the labels already collected
The labels are already in hand, so model quality metrics measure the thing in question directly: whether predictions became less accurate. Feature attribution drift says which inputs carry more weight without saying whether accuracy followed. Comparing against last month swaps a fixed baseline for a moving one, and schema violations catch malformed inputs, not a shift within valid ranges.