Once you get an alert or a user has complained about model performance, time to resolve is critical. You must identify your data's problematic areas and switch to your Python environment to find the root cause.
Deepcheck helps you quickly understand the root cause of the issue throughout the model lifecycle.
Why Root Cause Analysis?
Where Is the Issue
You were just informed by either the customer success team or the Deepchecks monitoring system that your mission-critical deployed model is acting weirdly. You would like to know if a data pipeline, a new model version, or a data change caused the issue.
Understand the Urgency
The issue shows a model’s abnormal behavior. You need to know the impact: is there a significant impact on all of the data, a specific segment, or does only a minor portion of the data need to be monitored? Understanding the scope of impact is crucial when understanding your issue’s urgency.
Get to the Granular Reason
Once you know the issue, you need to dive in and figure out the source of the problem — a necessary step for mitigation. In many cases, this would require you to switch from the monitoring system UI to your code.