Your initial structure needs enough detail to show a clear plan, but not so much that you spend several minutes designing the perfect tree before doing any analysis. The structure is a starting hypothesis, not the finished answer.
A useful standard is two or three meaningful top-level branches with enough sub-points to show what you would actually investigate. If a bucket is just one vague word like ‘customers,’ it may be too shallow. If it has six layers before you have seen any data, it is probably overbuilt.
Good structures also prioritize. You can briefly tell the interviewer where you would start and why, which makes the structure feel more practical and less like a checklist.
With practice, you should be able to produce something clear in roughly a minute or two. The exact timing varies, but if structure-building consistently consumes a large part of the interview, simplify. A useful reference is the case interview structure flow.