This paper examines how information on the purchasing patterns of different customer segments can be used to more accurately evaluate the economic impact of mergers. Using a detailed dataset for the leading manufacturers in the US during the late nineties, I evaluate the welfare effects of the biggest ($25 billion) merger in the history of the PC industry between Hewlett-Packard and Compaq. I follow a two-step empirical strategy. In the first step, I estimate a demand system employing a random coefficients discrete choice model. In the second step, I simulate the postmerger oligopolistic equilibrium and compute the welfare effects. I extend previous research by analysing the merger effects not only for the whole market but also for three customer segments (home, small business and large business). Results from the demand estimation and merger analysis reveal that: (i) the random coefficients model provides a more realistic market picture than simpler models, (ii) despite being the world's second and third largest PC manufacturers, the merged HP-Compaq entity would not raise postmerger prices significantly, (iii) there is considerable heterogeneity in preferences across segments that persists over time, and (iv) the merger effects differ considerably across segments.