A software engineer's confidence in the profitability of a novel design technique depends to a significant degree on previous demonstrations of its profitability in practice. Trials of proposed techniques are thus of considerable value in providing factual bases for evaluation. In this paper we present our experience with a previously presented design approach as a basis for evaluating its promise and problems. Specifically, we report on our use of the mediator method to reconcile tight behavioral integration with ease of development and evolution of Prism, a system for planning radiation treatments for cancer patients. Prism is now in routine clinical use in several major research hospitals. Our work supports two claims. In comparison to more common design techniques, the mediator approach eases the development and evolution of integrated systems; and the method can be learned and used profitably by practicing software engineers.