There are a few different models. RateBeer uses this system (the link takes you to ratings of Supris):
Aroma 1-10Scores work out to .5-5.0. The problem when you have a large sample set, though, is that ratings tend to flatten out. Westvleteren 12 gets the highest user rating 4.51, whereas 2,500 beers in, and the ratings are still around 3.4--which seems not very useful. But I digress.
The Beer Advocate has a similar system for its members (again, link is to Supris). A ten point scale for appearance, smell, taste, mouthfeel, drinkability. The site operators then convert users scores into a score similar to wine ratings, ranging from the 70s to 100, which seems more useful.
Homebrewers, for what it's worth, have a similar system.
Somehow I'm not so high on the aggregate based on component elements of the beer. You get averaging that way: a beer that smells awful but is actually kind of unexpectedly delightful might get a 3.5, along with a completely forgettable beer you wouldn't pay a plug nickel for. Finally, big beers always get better ratings because there's so much more to push the envelope higher (every beer in the RateBeer top ten is a big beer).
I'm at a loss. Any suggestions?