The observation of designers' behaviour in collaborative design activities and the analysis of protocols improved the understanding of how novel ideas emerge, what occurs among designers and, indirectly, what methods have a good impact on the outcomes. Yet, protocol analysis requires recording the design sessions, often in a simulated environment, thus introducing a bias in the observation. Moreover, the analysis takes up to 1000 times the duration of the observed design session. These limitations definitely hinder the scalability of this practice to large experiments in real operational environments.This paper investigates the possibility to use the data collected in log files, automatically recorded during collaborative design sessions assisted by an ICT design support tool, as a means to extract relevant information about the design process and ultimately to infer insights about co-designers' cognition during the session. In this perspective, the paper proposes a set of metrics tailored to an Augmented Reality-based collaborative design tool. The study has been carried about by processing the data collected in 5 real case studies conducted in three different design companies.