Recommender Systems in E-Commerce

Electronic commerce (EC) is, at first sight, an electronic means to exchange large amounts of product information between users and sites. This information must be clearly written since any users who accesses the site must understand it. Given the large amounts of information available at the site, interaction with an e-market site becomes an effort. It is also time-consuming, and the user feels disoriented as products and clients are always on the increase. One solution to make online shopping easier is to endow the EC site with a recommender system. Recommender systems are implanted in EC sites to suggest services and provide consumers with the information they need in order to decide about possible purchases. These tools act as a specialized salesperson for the customer, and they are usually enhanced with customization capabilities; thus they adapt themselves to the users, basing themselves on the analysis of their preferences and interests. Recommenders rely mainly on user interfaces, marketing techniques, and large amounts of information about other customers and products; all this is done, of course, in an effort to propose the right item to the right customer. Besides, recommenders are fundamental elements in sustaining usability and site confidence (Egger, 2001); that’s the reason why e-market sites give them an important role in their design (Spiekermann & Paraschiv, 2002). If a recommender system is to be perceived as useful by its users, it must address several problems, such as the lack of user knowledge in a specific domain, information overload, and a minimization of the cost of interaction. EC recommenders are gradually becoming powerful tools for EC business (Gil & García, 2003) making use of complex mechanisms mainly in order to support the user’s decision process by allowing the analogical reasoning by the human being, and avoiding the disorientation process that occurs when one has large amounts of information to analyse and compare. This article describes some fundamental aspects in building real recommenders for EC. We will first set up the scenario by exposing the importance of recommender systems in EC, as well as the stages involved in a recommender-assisted purchase. Next, we will describe the main issues along three main axes: first, how recommender systems require a careful elicitation of user requirements; after that, the development and tuning of the recommendation algorithms; and, finally, the design and usability testing of the user interfaces. Lastly, we will show some future trends in recommenders and a conclusion.