The aim of this article is to present the work and the findings of an effort to improve the response rate of direct mail. The purpose was achieved applying data mining techniques to the available data. The resulting predictive model can, given a set of characteristics of a person, determine probabilistically, whether the person is likely to respond positively to a direct marketing campaign with a purchase of a product or service. The work is based on the data of a Dutch insurance company which was published for the purpose of the CoIL Challenge 2000. The contest had the same aspirations as this article. Its results are discussed briefly in the last chapter.