Companies want to meet the requirements of as many heterogeneous customers as possible on the one hand and offer a small variety as possible to save costs on the other hand. To resolve the resulting dilemma, this paper presents a procedure that provides a quantitative and objective decision-making basis for an efficient portfolio using data mining. Valuable information can be generated from the data that is overlooked by previous techniques by using various data mining tools. The application of the procedure in a case study shows that a similar customer satisfaction can be reached with just 40 % of the formerly available variants.