Abstract. This paper presents an approach for monitoring several important aspects related to user behaviour during the execution of Web tasks1 . The approach includes the tracking of user interactions with the Web site and exploits visual data mining to highlight important information regarding Web application usage. In particular, our approach intends to be a natural heir of the approaches based on clickstream visualization, by integrating them with the visualization of page-level data and by improving them with the definition of ad-hoc zoom and filter operations. Furthermore, we present a theoretical framework to formally define our proposal. Lastly, in order to test the approach, a simple casestudy for a particular practical usability evaluation has been carried out. To this aim, we built a prototypal system composed of a tracking tool, responsible for tracking user interactions and a visualization tool for task analysis.