Abstract
We propose a method to set a daily selling price of a single item from historical sales data; for example, POS data. First, we estimate a generalized linear model to predict the store-level daily sales volume of the item. To estimate this model, we assume that the sales volumes are Poisson-distributed and that the price elasticity is constant. From this model, we can obtain a demand curve and an estimate of price elasticity of the item as the model parameters. The assumption of Poisson distribution enables us to estimate the price elasticity from low-level sales amounts. In addition, we clarify the relationship between gross profit and three variables: the price elasticity, the selling price and the purchase price. Using this relationship, we propose a method to set a selling price that maximizes the gross profit with respect to the price, and we indicate that there are two price policies that give us a local maximum gross profit: the regular price policy and the discount price policy. Considering long-term benefits, we show the existence of a scenario that provides a higher gross profit at the expense of short-term profits, if there is some method used to control the price elasticity. Finaly, we describe a case study using POS data collected from drug stores. We estimate the price elasticity of stores, and verify the policy of the stores.
