You’re welcome.

DL is used for its ability to extract deeper features in areas where humans find it difficult. So, they are applied a lot in image and video problems. But the volume of training data needed and heavy processing power are big constraints. With structured data, the results are mixed. There are cases where it’s used for better accuracy and to extract a few more percentage points over ML techniques, but it depends.

For a new problem, the guideline is to first try simpler, explainable Stats/ML techniques which are usually faster as well. Deep learning can be taken up next and when the constraints play to its advantages.