Alternative data for restaurant bankruptcy prediction

Abstract The restaurant industry has a high failure rate and predicting restaurant bankruptcy is an important task to mitigate economic losses. Restaurant consumer review data contains rich information and influences future restaurant demand. We investigate the value of publicly available data for predicting bankruptcies of individual restaurants. Our unique data set includes alternative consumer data from the two most frequently used online restaurant platforms in Switzerland and official business data from Swiss government websites. In evaluating the accuracy of predictive bankruptcy probabilities, we quantify the potential of the alternative data sources and we analyze the relevance of novel predictor variables. We find that alternative consumer data is relevant for restaurant bankruptcy prediction and the highest prediction accuracy is achieved when alternative data and traditional business data are combined.