Customer Valuation: A Hierarchial Bayes Approach for Estimating Recency, Frequency and Monetary Value in Direct Marketing

摘要: Measures of recency, frequency and monetary value are extensively used by direct marketers to predict customer behavior and to identify their best customers. Problems emerge in estimating these measures because of the limited information that typically exists about customers. This paper presents hierarchical Bayes models which overcome this information deficit. These models use continuous distributions of unobserved heterogeneity and result in more accurate individual-level measures of recency, frequency and monetary value than those obtained from other approaches. In an analysis of purchase records of a business-to-business direct marketing firm, we demonstrate that the likelihood function of our model provides a natural way of combining recency and frequency measures, leading to a conceptually simple procedure for customer valuation.