The effectiveness of gamification in educational technology and digital learning environments has been the subject of research for over a decade. Gamified interventions encourage active learning by promoting role-playing, mentalisation, and experimentation. Learner-centred approaches have been shown to increase learning outcomes compared to traditional methods in some domains. Although the behavioural drivers associated with gamified learning are now well documented, practical advice for game designers who produce workplace training is much less common. This paper proposes the RAD-SIM framework for behaviourally effective cybersecurity games design, whereby psychological and behavioural principles are combined with learning theories to facilitate practical game-based learning. Using the framework will increase the engagement and retention of knowledge of the end-user. Although designed with cybersecurity training in mind, there are implications for educational game design in general. Future avenues for research, including potential case studies in practice, increasing the customisability of gamified interventions, and utilising natural data from organisations are considered. Social Machines conducted this work funded by the Research Institute for Sociotechnical Cyber Security (RISCS