In my opinion, data drift detection methods are very useful when we want to understand what went wrong with a model, but they are not the right tools to know how my model's performance is doing.