Below the shoot date folder you might segregate raw, tif, web output, etc. into other folders.
I have around 5 different major topics, where the subject matter is very unique, plus a "misc" major topic for general work. I file date-subject rather than subject-date because in most cases the image I am looking at has an EXIF date, or I know an approximate date or "era". If I know the shot date I can easily find the original, even if the file name has been changed.
There is no right way here; this is just the way I like to do it and why.
Another great feature of PM is the ability to open an entire folder structure on one contact sheet. Or to randomly aggregate folders into one contact sheet, one at a time.
This feature, however, does not scale up endlessly. I would say that much over 5-10,000 images in one contact sheet becomes problematic. But that number rises over time as hardware improves.
If you have a "small" folder structure of perhaps 10,000 images or less then you can do certain "cataloging" operations such as keyword searches. If you want to search across 50,000 or 100,000 images, or more, then you really need a catalog app such as perhaps lightroom or iMatch.
I'm not sure how well LR scales up into sizx figure photo counts. I know that iMatch does a very good job with that and it's a difficult point of research I would have to do if I ever contemplated a switch to LR.
Another filing variant would look like this, and I have done this for my earlier years but haven't gotten around to doing the recent years:
The select tree contains raw images of high interest. In my case I retain the original image in the original structure simply for backup purposes and for cohesion when I rummage through the originals (I don;t want my best shots missing when I do that).
If you are like me, you might want to store many originals, but the images you really care about are a tiny percentage of the originals. The Select structure then contains a more manageable number of images to aggregate into PM for a keyword search.
The downside of this approach is that the number of file versions proliferates, creating difficulties or more work to keep the various versions consistently keyworded. I find that there is nothing easy when it comes to image management
The bible of digital image management is "The DAM Book" by Peter Krogh. Available from your favorite bookseller. I have to admit I do not own the book but everyone else seems too . I'll get around to it some day. But if you are unsure how you want to do things then that book may help.
The theory of Digital Asset Management is that the OS file structure should be unimportant. The filing structure is in the keywording. In theory the OS filing structure should not matter at all and is unimportant. In my opinion, though, the folder structure is a backup to the DAM system.
Stuff happens and if I ever lost my image catalog I would not want to be left with an undecipherable pile of images. I think there should be a coherent basic structure to the image folders, just in case.
Personally I do not do much keywording in PM, except for very important images that I would have difficulty locating in a date based folder structure. As a wildlife photographer I do try to directly keyword rare or unusual birds, for example.
I do everything else in iMatch because it is easier, and it avoids a problem I have with a complex hierarchical keyword structure. That is something that I don't think PM does well, although some have tried to convince me otherwise. But as I said, there are some things I really like about iMatch, and one of them is the keywording functionality.