you configure a processing job to generate statistical rules and constraints (referred to as your baseline) against which the model quality drift can be detected. Model Monitor suggests a set of default baseline statistics and constraints.