Usability is a software attribute usually associated with the “ease of use and to learn” of a given interactive system. Nowadays usability evaluation is becoming an important part of software development, providing results based on quantitative and qualitative estimations. In this context, qualitative results are usually obtained through a Qualitative Usability Testing process which includes a number of different methods focused on analyzing the interface of a particular interactive system. These methods become complex when a large number of interactive systems belonging to the same context of use have to be jointly considered to provide a general diagnosis, as a considerable amount of information must be visualized and treated simultaneously. However, diagnosing the most general usability problems of a context of use as a whole from a qualitative viewpoint is a challenge for UE nowadays. Identifying such problems can help to evaluate a new interface belonging to this context, and to prevent usability errors when a novel interactive system is being developed. From a quantitative viewpoint, condensing results in singles scores, metrics or statistical functions is an acceptable solution for processing huge amounts of usability related information. Nevertheless, QUT processes need to keep their richness by prioritizing the “what” over the “how much/how many” questions related to the detection of usability problems.

To cope with the above situation, this paper presents a new approach in which two datamining techniques (association rules and decision trees) are used to extend the existing Qualitative Usability Testing process in order to provide a general usability diagnosis of a given context of use from a qualitative viewpoint. In order to validate our proposal, usability problems patterns belonging to academic webpages in Spanish-speaking countries are assessed by processing 3450 records which store qualitative information collected by means of a Heuristic Evaluation.