Online reviews can make or break a business. This holds true for the restaurant industry, where diners can quickly become online food critics, talking about flies in the soup on restaurant review sites like Yelp, TripAdvisor, Google, Facebook, Urbanspoon, and Foursquare, among others.
In fact, according to a Harvard Business School study, a one-star decrease in a restaurant’s overall Yelp rating could lead to a 5 percent loss in revenue. Indeed, Yelp seems particularly influential – to a degree that its reviews and ratings of restaurants could determine if and when a badly performing restaurant could shut down.
At least that’s what the results of a new University of Maryland study have shown. After analyzing approximately 130,000 Yelp reviews, researchers from the University of Maryland’s Robert H. Smith School of Business have found a way to predict if a restaurant is doomed and when it might close.
And it’s not just about having low numerical ratings. The researchers Shawn Mankad and Anandasivam Gopal have developed a text-analysis tool which identifies linguistic patterns and predicts, with about 70 percent accuracy, which restaurants are bound to close within the next quarter based on what their Yelp reviews are saying.
“The star ratings are definitely important, but to look at the text as well brings a more nuanced ranking,” Mankad said. “The text is bringing new information to the table.”
The study, entitled “More than Just Words: Using Latent Semantic Analysis in Online Reviews to Explain Restaurant Closures,” shows how the researchers leveraged unstructured data to identify words that could serve as potent signifiers of quality. These words include “food,” “good,” “place,” “like,” “friend,” “order,” “time,” “service,” “great,” and “nice” – words that are in Yelp reviews associated with restaurants that are likely to survive.
Mankad explained: “Constructing the variables, putting it into a predictive model – this is something that has never been done before.”