Optimal Pricing Metrics for Digital Goods

The recent advances of information technologies make price discrimination more prevalent and more complicated. Flooded by a large number of variables found by data mining algorithms, pricing managers are perplexed by the task of selecting variables for price discrimination to maximize profits. However, relevant literature remains scarce. This paper attempts to investigate how a monopoly seller should determine the optimal combination of pricing variables. The criterion found is similar to the selection of independent variables for linear regression; it is revenue-maximizing to select the variable that best reduces the residual variance of buyer’s willingness-to-pay. When incorporating costs associated with pricing metrics into this model, I propose a linear-running-time algorithm to solve this problem with any general cost structure. As to the implications for public policy, the welfare analysis shows that the monopoly seller will use an excessive number of metrics compared with the socially optimal level. By linking this theoretical model to a probit model, I demonstrate how to use this model to solve the price discrimination problem of an electronic greeting cards website.