Competitive facility location strategies for retail firms

Retail firms in competitive markets need location analysis strategies that correctly classify outlets according to how well the outlets attract consumer demand. These location analysis strategies describe a sequence of decision making tasks. At each step in the sequence, these strategies specify which variable or method the decision makers should use to evaluate the performance of the outlets. At the end of the sequence, the strategies inform the decision maker of the classification of each outlet. Neither location theory nor location analysis provide insight about the construction of "optimal" location analysis strategies. A computer program is developed to provide a computational environment to generate location analysis strategies from data collected from simulated retail markets. The purpose of this computational environment is to provide a mechanism for combining the methods of location analysis with the more traditional research associated with location theory. Realism is further enhanced by incorporating consumer choice strategies based on ideas from choice theory. Data are obtained from these simulated retail markets concerning the locations of retail outlets, consumer populations, retail activity, and the choices made by consumers regarding their shopping destinations. In addition, output from two models from location analysis are included in the data. These data are used to generated location analysis strategies. The location analysis strategies are produced using a technique from artificial intelligence called decision tree induction. These location analysis strategies are shown to classify unseen outlets nearly as well as they classify outlets used to generate the strategies. Furthermore, the form and predictive ability of the location analysis strategies appears quite stable over a range of parameters. Implications of this work for theoretical and applied studies are discussed.