Security managers face the challenge of formulating and implementing policies that deliver their desired system security postures — for example, their preferred balance of confidentiality, integrity, and availability — within budget (monetary and otherwise). In this paper, we describe a security modelling methodology, grounded in rigorous mathematical systems modelling and economics, that captures the managers’ policies and the behavioural choices of agents operating within the system. Models are executable, so allowing systematic experimental exploration of the system-policy co-design space, and compositional, so managing the complexity of large-scale systems.