Ph.D. Thesis - Forward, Back and Home Again
Analyzing User Behavior on the Web

Abstract

This thesis provides an integrative overview of theoretical insights and empirical findings on Web navigation, and extends this body of knowledge with results from a number of user studies. A framework was developed for capturing, cleaning and analyzing Web usage data.

In our laboratory studies we identified two navigation styles that were closely associated with perceived disorientation – an unstructured, flimsy navigation style, and a laborious navigation style that helps to prevent disorientation. Task-related link suggestions were shown to be positively received, to improve user navigation styles, and to decrease task execution times.

In a long-term study, we analyzed user interaction with Web browsers and various aspects of page revisits. From the results it becomes apparent that users often have difficulties in returning to pages that are visited on an infrequent basis, and that browsers lack backtracking support for multiple windows or tabs. We observed a shift of the Web toward a hybrid between hypermedia and interactive applications, which requires rethinking of the metaphor of Web navigation and the introduction of document management support.

Several design implications for navigation support are discussed. A branching history view, query modification histories, link relevancy indicators, and more context-adaptive result ranking will improve the orienteering process of information finding. Explicit history search and annotated paths from waypoints will provide better means for relocating the large amount of infrequently visited pages. The introduction of trails will help in reducing disorientation.

It is unlikely that the various aspects of Web navigation can be covered by one integrated model. Therefore, it seems to be more desirable to strive toward a comprehensive body of empirical data and theoretical insights, with explicit indications in which situations they are applicable. In our opinion, the current challenge for Web personalization is to design and evaluate straightforward mechanisms that address the various issues dealt with in this thesis.