The World Wide Web has become a major source of information dissemination for academia, business and government organizations. Hence, the usability and effectiveness of these websites is increasingly important. User behavior modeling is an important element of such evaluations. We have developed a tool, WebAnalyzer, that lets website administrators select the best parameters (number of clusters, distance measures) for clustering user sessions, representations of user behavior while interacting with a web site. Clustering of labeled session data is performed, and both running times and cluster quality measures such as sensitivity and specificity are reported. Website administrators can then select the parameters that achieve the most desirable combination of clustering quality and running time for the labeled data, and apply these parameters to similar but unlabeled datasets to form high-quality user models that permit improved evaluation of website effectiveness.