I am unaware of any process in nature more volatile than that of rating a restaurant. The wildly-diverse experiences that wildly-diverse people can have even at the same establishment mimics the sporadic nature of earthquake predictions, snowflake patterns, or the formation of galaxies. To truly gauge a restaurant, all potential reviewers would need to dine on the same night with the same crowd, be cooked for by the same staff, served by the same server, at the same table, have the same tastes, having had the same prior dining experiences; already a staggering number of variables.
Another factor, as any professional restaurant reviewer can tell you and most people discover on their own, is that one's impression of a restaurant can change radically even between visits. Chefs make mistakes, servers have “off nights”, moods of diners vary wildly, and what someone is craving should be aligned (ore or less) with the cuisine of the restaurant chosen. If you're in the mood for a steak, dinner at "Salad Salad Salad!" probably isn't going to do much for you.
And herein lies the volatility and inaccuracy of Yelp...
There is a great article about how nearly every restaurant scores 4 out of 5 stars, whether flagship or "dive", whether the food is far better than the atmosphere, or the other way around. A review must be tempered with expectations; if you’re in a tiny taqueria, then expect amazing tacos rather than stellar service or Le Cirque-like ambiance. If you’re at Le Cirque, don’t complain that they made you wear a jacket and wouldn’t serve you ketchup with your duck fat fries.
The average of 4 makes sense when you think about it. Flaws can be found in nearly every dining experience. Maybe there was a draft at your table, or lighting fixtures were aimed at the floor rather than on your table. A dirty fork, a chipped wine glass, or a poorly-timed cooking duration can leave just enough of a mark to cause someone to withhold that extra star.
On the other hand, most people instinctively avoid a restaurant to which they would likely award only a single star. Factors of such restaurants are visible even before the menus are.
And what about those little stars? What do they actually mean? Half the reviewers go into lengthy detail about “what the stars mean to them.” I'd rather listen to people discuss horoscopes.
The daily creation of a high-end restaurant menu depends on a thousand variables; availability and timely-delivery of ingredients, capable and thorough preparation of those ingredients, timely and accurate cooking and combining of them on a plate, timely and orchestrated delivery of that plate to the correct table. It mimics the complexity of a mobile phone call in and is miraculous in its very existence; a complexity which only becomes visible when it fails.
Unfortunately, Yelp's greatest flaw is that inaccurate or unfounded reviews cannot be "voted off the island". The only thing you can do with an errant review is not say anything positive about it. That's right. If a vegetarian gives an otherwise 5-star steakhouse a single star, the other Yelpites can't then vote that attention-seeking jackass off the review list. If you're not into meat, why were you at a steakhouse and, more pressingly, why would you choose to write about it? I don't negatively review Tom Waits concerts nor the new Dodge Charger, and I'm pretty sure that people interested in either don't care about my opinion of them either.
Let me put this in another scenario. Let's say you invite 100 people to a party, and one of them takes a dump on your piano. In Yelp land, the only thing you can do is NOT say anything NICE about that person. You can't kick them out, you can't blacklist them from other parties, and you can't really tell them off. Despite the fact that everyone can smell the obvious deed in the room, they are similarly powerless against cleaning it up or preventing it from happening again.
All those budding restaurant reviewers (the verbal equivalents of "producers" of YouTube videos) treat reviews the way eighth-graders treat a book report. “Chez Blah is located on the corner of…” Stop. We know where the restaurant is, that information is given even before the reviews start. Then they go into long, tedious dissertations about how they were meeting someone for their birthday after having just won the world record in backward roller skating, etc. On and on we read, looking for useful content to make or break the decision. Through stories of newborn puppies, Frisbee championship wins, parking lot fender-benders, and “too many coktails(sic) before dinner”, are tiny nuggets of information. In short, a star rating really doesn't mean a whole lot. To truly consider a restaurant, you need to read through the reviews, which puts us back to square one.
The presumed goal of Yelp was to have a large cross-section of people share their opinion of a given restaurant (which makes sense), but reading through the reviews is a bit like reading a newspaper review where every paragraph was written by a different person and they repeat the obvious over and over again.
The actual goal of Yelp is to be a successful and profitable website and, to do that, you need "traffic". And what gets traffic? Vanity. Seeing one's own words, having them read by others, commented upon, and easily referenced. (In my case, the vanity more than the traffic.) Quantity of Yelpites is better than quality of them.
I need to explore Yelp further to know how (or if) it can actually be useful. The fact remains that determining whether a restaurant is worthy of your money is still reliant on a number of pieces of information; word of mouth (which carries much farther and more reliably than electrons), web research, “pedigree”, and perhaps most tellingly, how busy are they? I instituted a policy some time ago called “I don’t dine in empty restaurants”. It has been the most reliable indicator of quality to date - with only one gaping flaw; the Cheesecake Factory is always full. So is The Olive Garden (I think). Likewise for Chili’s, Chevy’s (which I like as a guilty pleasure), McDonald’s, etc. Being full - by itself - is not an indication of quality, it is simply an indicator of popularity. Again, trust (at least partly) your instincts. If you see three seemingly desirable restaurants in a row, two are full, and one is empty at 8:30 on a Saturday, you at least know which one to avoid. And you don't even need to Yelp.