This is an extremely tricky subject, there are just too many variables for anyone to get too dogmatic. Some general guidelines follow. However, one approach is to sacrifice the first few frames of the roll as an experiment:
1. Take three or four test shots at the best guess ISO.
2. Shoot the rest of the roll.
3. When processing: pull off the test shots (about 8" including the leader) and process it separately. (See suggestions below.)
This will give you an idea if your processing scheme is at least in the ballgame. You can adjust your time/dilution for the remaining frames.
For Tri-X, FP4 and C41 film, we'd recommend exposing at 1/2 the rated ISO. With E6, we got the best results shooting at rated speed.