Recent extensions to process algebras can be used to describe performance or error rate properties of systems. We examine an abstract approach to the representation of time costs within these algebras that permits the efficient calculation of performance bounds on systems. In particular we avoid the ‘state explosion” caused by the parallel composition of the representations of probabilistic time distributions. A major advantage of one of our approaches is its uniformity, which allows the eventual approximation level to be easily predicted from quality of the approximations to the underlying distributions.