A Catalog Item has weekly scheduled full profiling that also runs anomaly detection. However, it is a dynamic CI, meaning the logic can be edited depending on requirements. In short, it means that the ‘truth’ can vary depending on the requirement, meaning A, B, C was not an anomaly a year ago, but it is now. Currently, every profile ever run is stored, which makes anomaly detection less effective. Trends that were previously valid may no longer apply. Furthermore, retaining profiles from 7+ years ago is unnecessary and not allowed by law I suppose.
A useful improvement would be an automatic limit on the number of stored profilings for a CI. When a new profiling run occurs, the oldest profiling would be removed once the limit is reached. A similar feature already exists in MP:
Advanced settings → Results storage → Remove results after X runs.
Applying this concept to Catalog Items would allow setting a limit (e.g., 25 runs), which should be sufficient for anomaly detection to set a trend and find anomalies while still providing enough historical context.
Is such a feature available for Catalog Items?
Note: Running anomaly detection in Monitoring Projects is not desirable, as it would execute on every run.