Expressing preferences when querying databases is a natural way to avoid empty results and information flooding, and in general to rank results so that the user may first see the data that better match his tastes. In this paper we outline the main research issues to be faced in order to develop a system for handling user preferences on OLAP cubes.