Photo: Shutterstock (Shutterstock)
There are so many good restaurants out there, you don’t want to waste your time and money eating at a bad one. But how to ensure you’re going to like where you go? Use this great restaurant recommendation tip from Reddit user Sauwa, sharing on the lifeprotips board, and your odds of a satisfactory meal will go way up—it’s so simple and yet so genius that I’m embarrassed I haven’t been doing it for years.
Here’s how it works: When you’re trying to find a decent place to eat using Yelp or another review site, don’t rely on the collective rating or number of stars. Instead, find a single positive review of a place you already like, then read what that user enjoyed and didn’t. If it matches your taste, dig into their post history and eat at other restaurants the user recommends.
You’re essentially appointing a stranger to a role that used to be filled by newspaper restaurant critics, but without having to rely on the tastes of the editorial board of The Sheboygan Press. Your online stalkee’s recommendations are more likely to match your specific taste—and lead you toward a decent meal—than an algorithmic aggregation of all users’ opinions.
Why crowds are not necessarily wise when it comes to restaurants
Collecting many users ratings and averaging them is a variation of the “wisdom of crowds” theory first detailed by Marquis de Condorcet in 1785. Here’s a simplified explanation: Imagine an obscure, specific question with a true or false answer. You’d have no way of knowing if one person’s answer was correct, but if more people answered, you could rely on the crowd’s wisdom, even if only a small percentage of respondents actually knew the correct response. Theoretically, everyone who didn’t know would split evenly between “true” and “false,” canceling out each other’s votes and leaving the correct response obvious.
Relying on the wisdom of crowds works great for some kinds of reviews, particularly products that have a specific function. If 90% of people who buy a hammer report that it drives nails reasonably well, it’s probably a good hammer. But how much you might enjoy a restaurant, a movie, or a novel is a different thing altogether, because that’s about personal taste. While there are some things most of us agree on when we eat out—restaurants shouldn’t serve raw chicken, for example—the finer points vary. My idea of a great burger and yours might differ wildly, and a fantastic hole-in-the-wall rib joint would still get a terrible review from people who like frou-frou cuisine.
Napoleon Dynamite and the “love it or hate it” effect
Back in 2006, Netflix started offering a million bucks anyone who could embetter its movie recommendation system. Improvements were made—most people’s movie tastes are scarily predictable—but algorithm after algorithm got hung up on Napoleon Dynamite. There was seemingly no way to predict peoples’ opinions of the quirky 2004 indie comedy (and a handful of other movies) based on other films they liked. But people have strong opinions about Napoleon Dynamite: they either love it or hate it with little middle ground. The result, in terms of review aggregation is something like 2 1/2 stars out of 5. Average. Which is the least likely response you’d have to seeing the movie.
It can work the same way with restaurant recommendations, particularly for “non-traditional” foods or anything experimental. If you love spicy food, that place that makes authentic Korean Galbi jjim is getting 5 stars. If you’re not into it, though, the stuff is inedible—one-star. Average it up, and a we’re right in the middle. That helps no one.
The potential dishonesty of review aggregation
I don’t know for sure whether the reviews on popular restaurant ratings sites accurately reflect the opinions of users, but I’d put a lot of money on “no.” Leaving aside whether the sites themselves are honest, individual businesses often live or die on positive ratings, and it’s not difficult for a business to either goose its own reputation with fake positive reviews or sandbag the competition with negative ones. It’s estimated that 20% of online reviews are fake—enough to affect the overall rating, especially for newer places with few reviews.
It’s difficult for large platforms to weed out fake reviews (although they do try), but it’s easy as an individual to find another genuine individual. To weed out fakes, be suspicious of reviews written with generic language, especially the same language applied to more than place. If you want to get all internet-detective, do a reverse-image search on profile and food pics to see if they were lifted from elsewhere. Once you’ve done this, you’ve found your own, personal food influencer and your city’s most delicious burritos will become clear to you.