Having done a fair amount of research on this - there isn't a perfect way, as a user I just have to find something that I can apply consistently and that makes sense.
Two things are helping. Hazel (a great program for automating file moves - I hope the any sandboxing introduced in mountain lion doesn't kill it) and aliases.
For live stuff I am working a two level filing system using the documents folder as the root.
The first level is the company, the second is the project for that company.
For new files that I create I am using the format:
yyyy-mm-dd - company - project - good description of the file
I then save them in the documents folder. Hazel monitors that folder, renames everything to lower case and then sweeps the file into the appropriate project folder
That deals with most of my stuff. For some particular clients I need a third level so I have a slightly different format for that:
xxyy - yyyy-mm-dd - company - good description of the file
where xx is the first two letters of the individuals first name and yy the first two letters of the surname. Hazel sweeps the file into the appropriate individuals folder and re-arranges the filename to put it back into the 'standard' format.
I am using textexpander snippets (another great program) to speed up the filename generation.
Any other tricks or ideas I should be considering?