In 2016, the restaurant industry generates more than $766 billion in sales and jobs for one in 10 workers in America. Despite its high impact on U.S. economy, the restaurant industry is also well known for its high failure rate. Nevertheless, research on restaurant survival has been sparse. In this paper, we examine whether user generated content including reviews and images could be used to predict restaurant survival. In particular, we use deep learning methods to analyze 1.3 million Yelp reviews and 0.8 million images from 24,415 restaurants. Tracking the survival of these restaurants over the last decade (from 2004 to 2015), we find that both the volume and the valence of images are strong predictors of restaurant survival. Nevertheless, when it comes to consumer reviews, only the valence (not the volume) matter. Interestingly, even after controlling for content analysis of review text and other review-related variables such as star-rating and review length, consumer sentiment extracted from review text is still strongly associated with the survival of restaurants. We also find that chain restaurants and restaurants from larger categories have better chances to survive. Restaurants from all price levels also appear to have equal chance of survival in the marketplace. To our knowledge, this is among the first large-scale empirical research on restaurant survival. We are also among the first to introduce both text- and image- based deep learning into the marketing literature.
Joint work with Mengxia Zhang, PhD student of Marketing, Marshall School of Business, University of Southern California.
Bienvenue à tous!