This paper presents an innovative approach for grouping retail loans into homogeneous risk pools, which adheres to the provisions of the revised Basel II framework. The authors interpret Basel II using an efficient classification tree algorithm (recursive partitioning) and test it on a real data set of approximately 413,000 German auto loans. By classifying loans according to selective predictors of default, it is found that banks may achieve significant savings in terms of a lower regulatory capital requirement. Alternatively, this provides the opportunity to increase lending capacity.