This paper describes how to employ multi-terminal binary decision diagrams (MTBDDs) for the construction and analysis of a general class of models that exhibit stochastic, probabilistic and non-deterministic behaviour. It is shown how the notorious problem of state space explosion can be circumvented by compositionally constructing symbolic (i.e. MTBDD-based) representations of complex systems from small-scale components. We emphasise, however, that compactness of the representation can only be achieved if heuristics are applied with insight into the structure of the system under investigation. We report on our experiences concerning compact representation, performance analysis and verification of performability properties.