This paper describes an experiment about the relevance of using Cognitive Work Analysis (CWA) to improve the system definition process. CWA is a method for analyzing complex sociotechnical systems from a cognitive point of view. We submitted a set of exercises involving some specification tasks to a group of professional system engineers, some were asked to build on CWA data, while others to rely on some classical method results (Hierarchical Task Analysis). Beyond the results obtained from this specific experiment, this paper aims at demonstrating that controlled experiments are not only possible but also relevant in system engineering field.