Abstract The aim to raise customer loyalty in telecom market requires an emphasis on one-to-one marketing and personalized services. To this end, it is essential to understand individual customer preferences for services. In this paper an improved fast hierarchical clustering algorithm (IFHCA) is proposed firstly. Then a method for identifying dial-up user preferences based on IFHCA is presented, in order to discover the pattern of users' preferences and recommend the most appropriate services. Finally, this paper analyses the relationship between users' preferences with on-line duration and traffic. Experiment result shows that IFHCA is of advantages such as analyzing on preference patterns of asymmetric digital subscriber line (ADSL) users accurately, lower time complexity than the classical hierarchical clustering algorithm (CHCA) to mining large dataset efficiently. Besides, the result is provided for selective management and commercial package customization.