Most approaches to examining the effects of marketing variables on market shares treat the market as a single aggregate. It would be managerially more useful to examine the marketing effects at the segment level. A market segment is defined in this paper to be a group of consumers who are homogeneous in terms of the probabilities of choosing the different brands in a product class. The market thereby is broken up into brand loyal segments (one corresponding to each brand) and switching segments (each of which corresponds to a subset of consumers choosing from its own subset of brands). From merely the aggregate matrices of cross-classification of brands purchased at different time periods, the proposed approach estimates (i) segment sizes and (ii) market shares period by period for each segment. The changes over time in within-segment market shares are related to marketing variables such as price and promotion through a Logit Model. Application of the model to data from a scanner panel shows that the proposed approach produces a modest improvement in predictive validity compared to the traditional approach of treating the market as a single aggregate.