Abstract
Various advance planning takes place before a new retail store is opened. Since the ultimate goal is a high volume of sales, an accurate sales prediction is desired in the planning stage. In this study, an arbitrary population was prepared for a chain of retail stores deployed in an particular area, and an attempt was made to predict sales at new stores by using patterns of 100 variables, including geographical information, demographics, and consumption trends. Training signals were created by using all 100 variables for areas in which stores had been opened in the past, and the Mahalanobis-Taguchi system of quality engineering was used to create a polynomial expression with 100 terms to estimate the sales of the new stores. The error in the sales prediction was calculated from the S/N ratio obtained from the polynomial expression, and proved to be about ±10%.
