I like to consider it a tool for ordering theories, before testing them (in hard science, of course).
It is not a tool that finds truth. It does not concern itself with truth values, it is a heuristic.
Let us suppose that, while fixing a snack for my good friends hamster bong and hamstergirl, I hear a rustling noise that comes from the living room. My question is: What made that noise ?.
Possible answers are:
- My pet hamster has been freed by hamster bong and hamstergirl, and is currently frolicking in the wastepaper basket.
- A burglar has entered the house.
- Martians have landed in my backyard and they are paying a visit.
- The ghost of Microsoft is rattling a paperclip chain.
Notice the trick: I have ordered some of the (infinite) possible theories in inverse order of assumptions number and strength necessary. In detail:
- Hamsters are frolicsome.
- Someone broke and entered.
- Living beings exist on other planets.
- Corporate entities return as ghosts and suffer for their sins.
The scientist in me (and in you, and in a lot of other rational people) will start with the first theory, and shout "Put the little bastard back in the cage, will ya ?". That tests the theory. If the test fails, then I will perform more tests. Only ultimately I will have to pull out the Martian-Zap-o-Mat, or the Blessed Linux CD in order to banish the noisy uninvited guests.
Really, Occam's Razor is not anything strange: what is strange is how people occasionally stop applying it.