Testing online navigation recommendations in a web site Juan D. Vel´ asquez 1 and Vasile Palade 2 1 Department of Industrial Engineering, University of Chile, Chile, jvelasqu@dii.uchile.cl 2 Computing Laboratory, University of Oxford,UK Vasile.Palade@comlab.ox.ac.uk Abstract. An online navigation recommendation system provides the prospective web site visitor with a set of pages that could be of his/her interest. Because the recommendations are given during the user session in the web site, it could be very damaging for the overall business of the company owning the web site, if the recommendations are erroneous. In this paper, we introduce an a priori method to estimate the success of an online navigation recommendation. The methodology was tested in a recommendation system that works with the data generated in a real web site, which proved the effectiveness of our approach. 1 Introduction An efficient way to improve the relation between the web site and its users is by the personalization of the site structure and content, i.e., “actions that tailor the web experience to a particular user, or a set of users ” [6, 10]. The web site personalization can be implemented by recommendations done directly to the web site users, which can be web masters, web designers, anony- mous web visitors; in short, any users of the web site [5,13]. Depending on the web user, the recommendations can be: – Online - Provide the web site visitor, during his/her session in the site, with recommendations about interesting topics, for instance a web page, text contents, etc. – Offline - Done mainly to the web master, web designers and, in general, to any person in charge of maintaining the structure and the content of the web site. This paper is concerned with how to do online recommendations, especially those that are relative to the user navigation in the web site. When a user visits a web site, the browsing behavior is related to what he/she is looking for. Identifying the user’s behavior usually becomes a true challenge, because the site doesn’t consider the particular needs of each user, e.g. if the user is an amateur in the web and needs assistance to find the desired information.