Should I open here? Predictive models for restaurant site selection

Abstract

Restaurant failure can be considered from two different perspectives, the pre- and post-opening decisions. While the post-opening factors include the day-to-day operating decisions, the pre-opening decisions are the ones that a restaurateur makes in the planning and design of a restaurant; with an initial decision of the site location (Chen & Tsai, 2016; Yang Yang, Roehl, & Huang, 2017). Site locations are often made by intuition, as there is a lack of comprehensive research about which elements are the most important for a restaurant's future success. The decision of where to open a restaurant is one of the most critical pre-opening decisions for restaurateurs (Egerton-Thomas, 2005). The current site selection method often includes engaging an expert with a 'gut feel' about a potential site's success (Clarke, Horita, & Mackaness, 2000). However, if data can be used to gain a more thorough understanding of success factors related to the site, the success rates may improve (Wood & Reynolds, 2012). Considering the knowledge gap between the pre-opening area factors and potential success of the site, the research questions are as follows: 1. Which socio-demographic attributes and site characteristics have the highest impact on restaurant success or failure? a. What is the most critical socio-demographic area attribute a restaurateur needs to pay attention to when selecting an area? b. What is the most critical restaurant site characteristic that a restaurateur needs to consider when selecting a site? 2. Considering the potential influence of the overall site characteristics, can a model be created to aid in restaurant site selection? Prior studies on site selection for independent restaurants have looked at a limited number of area attributes or site characteristic elements. This is the first study to investigate multiple pre-opening characteristics of the area attributes and site characteristics. The model created is the first scientific method assisting restaurateurs in deciding on the site. The study, conducted in Hong Kong, focused on the survival rates of 6,710 newly opened restaurants in 2016, 2017, and 2018. Nineteen individual variables, including eight socio-demographic elements, eight site characteristics, and three restaurant characteristics, were combined to investigate the research questions. A logistic regression model and the corresponding marginal effects were used in investigating the first research question. Locating within a mall provides the single most significant increase in potential success at 11.229 percent. The second greatest impact was the average number of individuals living in a household, as for every additional person, the potential for success increases by 10.098 percent. Two predictive models, logistic regression and artificial neural network were used to investigate the second research question. Both models can be used to predict restaurant success with the logistic regressions accuracy rate of 71.27 percent and the artificial neural network accuracy rate of 72.55 percent. The artificial neural network is selected as the preferred model due to the marginally higher accuracy and greater area under the ROC curve. The study expands theoretical contributions to the principle of minimum differential and central place theory. The principle of minimum differentiation showed that the clustering of restaurants is essential if the competitor is of a similar price point. The central place theory found that individuals still prefer not to travel far distances within a dense urban environment. While transportation is ample, restaurant patrons still prefer to dine near home. The research outlines the importance of site selection and that the mantra of 'location, location, location' is critically important to restaurants. Restaurateurs cannot rely solely on differentiated concepts, low fixed costs, or quality food and service. The location is a crucial ingredient in the overall success and failure of a restaurant, and the artificial neural network model created will aid all restaurant industry stakeholders in selecting the ideal site for their endeavor.