Scenario

You shipped a model that looked strong in validation, but it is underperforming after launch. The team wants you to figure out why the results are worse and whether the issue is in the model, the data, or the decision threshold.

Question

What would you do if your model is underperforming?