Computerworld had a nice introductory article on data mining today - Explainer: Data mining - Computerworld. One of the great uses of this kind of data mining technology is to turn historical data into business rules. Thus, in the REI example in the article, one could turn the knowledge about which customers bought what into rules like "If purchased product types include at least two of 'camping gear', 'back pack', 'maps' then add 'Cycling' to customer's potential interests" allowing a catalog on cycling products to be sent using a rule like "If at least one of customer's interests is Cycling or at least one of customer's potential interests in Cycling then send Cycling Catalog" or something similar, depending on the rule syntax of your execution environment.
Doing this takes two things, only one of which the article details. Firstly you need the kind of data mining tool mentioned in the article. Secondly, however, you need a way to turn the insight you get from this tool into something you can execute. A business rules environment, especially one that can load the output of data mining activities directly, is ideal.