You will choose a different data structure for different use scenarios—for IoT data it will be very granular:
{Vector1}/{Vector2}/{Vector3}/{YYYY}/{MM}/{DD}/{HH}/{mm}
On the other hand, for storing user data, the structure may be completely different:
{AppName}/{UserId}/{YYYY}/{MM}/{DD}
It all depends on your current requirements. The data structure is extremely important when you plan to perform an analysis on the files stored—it directly affects the size of files and their number, which can further affect the possible toolset for your activities.
Another important thing here is the legal requirements—if you use any kind of sensitive data as a folder or a filename, you will have to be able to perform a clean up efficiently ...
