Usability can be defined as the degree of a product’s fitting to the characteristics of a person or of a group of people. The concept of usability includes aspects of using a product that are closely linked to the user’s degree of satisfaction and preferences. As a multidisciplinary concept, definitions may vary depending upon the specific area on which one focuses. However, common terms can be found throughout literature. Parameters such as the difficulty and steepness of the learning curve for the end-user, or flexibility and adaptability are commonly evaluated. In the context of Assistive Robots, factors taken into account are related to user acceptance, security, precision of task execution, and overall system efficiency. Hence, it is also closely linked to the concept of dependability. Boundary conditions related to the environment and to the user must be taken into account. In this paper, the importance of the role of benchmarking the usability of Assistive Robots is discussed, and a methodology for obtaining usability data from experiments is proposed. The proposed methodology is part of a continuous improvement framework that is based on the System Knowledge Space, which will be described within the text. Then, a general view at results extracted from experiments performed with an Assistive Robot and real potential system end-users in realistic scenarios is given. This exemplary usability benchmarking assessment follows the guidelines of the methodology that is proposed. The experiments that are described were developed as part of the ASIBOT program at the Universidad Carlos III de Madrid in collaboration with the National Paraplegic Hospital in Toledo (Hospital Nacional de Parapléjicos de Toledo). The last part of our paper deals with results of how these experiences have influenced actual and future research efforts and discusses how this should positively affect the scientific research and developer community.