Current adaptive systems acquire information about users mainly by simple tracking of resources, a user has requested and by asking users to supply the needed information. In this paper, we discuss user modeling based on observing a user’s interaction with the system. We propose to collect usage data on the server side as well as on the client side. Collected data are then processed into knowledge about user’s intentions and preferences. This processing relies on a set of heuristics, which help to interpret the usage patterns found in the collected data.