We develop a class of demand models for differentiated products. The new models facilitate the BLP method (Berry et al. 1995) by avoiding numerical inversion of the demand system. They accommodate rich patterns of substitution and complementarity while being easily estimated with standard regression techniques and allowing very large choice sets. We use the new models to describe markets for differentiated products that exhibit segmentation along several dimensions and illustrate their application by estimating the demand for cereals in Chicago.