The Effects of Customisation on the Usability of Visual Analytics Dashboards: the Good, the Bad, and the Ugly

DOI

Visual analytics dashboards have become key tools for decision-making. Yet, information overload and high expectations of the actual graph literacy of users hinder their effective use. Customisation features such as global filters (filters that update all views), visualisation hiding, and graph switching have been proposed to improve dashboard usability and accommodate user diversity. However, time constraints and a lack of awareness of the available customisation functionalities are known to be barriers to engaging with customisations. In this paper, we examine how visual analytics dashboard customisation affects usability and identify existing risks and barriers. We found that short-term surface customisations improve the usability of dashboards, both objectively (users are quicker and more correct) and subjectively (users report a lower cognitive load). For this to happen, the engagement with customisations must not be exploratory but meaningful (i.e., longer than five seconds). We also identified a set of customisations that, if not implemented carefully, are detrimental to cognitive load and completion times. Informed by our findings, we provide insights for building user models that predict the usability of dashboards-in-use and associated adaptive interaction techniques to assist users.