The Power of Digital Word of Mouth

Almost 90% of consumers trust online reviews as much as they trust personal recommendations. And before deciding to visit a local business, half of all 18-54-year-olds seek out and read online reviews. For better and worse, customer ratings have become a fundamental part of doing business.

“Empirical research shows that user-generated reviews can significantly impact revenues,” write Ishita Chakraborty, Joyee Deb and Aniko Öry, all from SOM. The question, then, which they ask in a recent working paper, is when do consumers decide to write a positive or negative review? What inspires consumers to talk about certain experiences but not others? What motivates word-of-mouth communication?

The researchers have developed a pared down model to explore the basic contours of this question. The model rests on the straightforward assumption that people write online reviews “because they want to think that somehow their review is helpful to others,” Öry says. “They would like it to change the behavior and decisions of others.”

With this premise, she and her colleagues establish a striking result. The theoretical prediction is at that at well-known national chains, people have no incentive to leave a positive review; they write exclusively negative. At non-chains, reviews can be positive or negative depending on which part of an experience felt most representative of the establishment, and thus, most informative for other potential customers.

Öry, Deb, and Chakraborty use Burger King to clarify this logic. The strength and ubiquity of the restaurant’s brand means that loyal customers, whether in Seattle or Charlotte or Shanghai, know what to expect when they drop in for a meal. Given that, a positive review that praises the food or service at Burger King is neither informative nor helpful, as regular customers get a consistent product and already know that they like it. But if a new Burger King outlet opens somewhere and provides a subpar experience, then this is noteworthy for customers who have come to trust the brand. This could inspire a customer to write a negative review, informing others who might want to eat there about the experience. At a comparable burger restaurant that isn’t a chain, the same is not true. Reviewers will write a positive review if they had an exceptionally good experience, suggesting the restaurant is above average; after a mediocre experience, a review is arguably less informative.

The authors then take this model to data. They collate a dataset with Yelp reviews from several cities between 2004 and 2017 and show that the empirical evidence is consistent with their model’s results: Among major chains, nearly half of all reviews were one-star—an overwhelming negative bias. Smaller and newer chains, where the brand was not as well established, had fewer one-star reviews. And independent restaurants received five-star reviews 41% of the time. In fact, being a chain restaurant results in approximately a 1-star reduction in rating relative to a comparable independent restaurant.

This work, Chakraborty noted, raises interesting questions for both consumers and reviewing platforms. For consumers, these findings suggest that reviews alone tell only a partial—in both senses of the word—story. Bad averages don’t necessarily make for a bad restaurant, as they may say more about reviewer incentives than restaurant quality. For platforms, Deb wonders if there are ways to present results that would be more informative than simply highlighting average star ratings. Among chain restaurants with a strong brand, for instance, would it make sense to lessen the weight of negative reviews given consumer predilections to write them? Is there, ultimately, a way for companies like Yelp to provide more valuable information?

“In the end, the selection of who writes a review matters, and the decision to write a review depends on interactions between firms, consumers, and reviewers,” Öry says. “I see this as a first step to understanding this strategic motive, and how strategic thinking and interaction can lead to selection of different types of reviews.”