This paper investigates the use of a two-stage scoring process with the aim of improving the accuracy of the scoring system. The two stages are: (1) pre-segmenting the population and (2) building separate scorecards for each segment. The segmentation is done using a combination of a novel subspace selection technique proposed by Dey and Roberts (2002) and a BIRCH (proposed by Zhang et al. (1996)) like two-step cluster detection technique. Empirical results obtained from real data demonstrate the strength of the two-stage approach. The segmentation stage is able to separate the population into parts with widely different risk profiles (measured by the population odds). In the second stage, different scorecards are built for each of these population segments thereby significantly improving the overall accuracy of the scoring system. Further, the proposed approach is easy to implement in credit card account management systems currently used in the credit card industry.