When designing a publicly available Web service, a service designer has to take care of costs and revenue caused by this services. In the very beginning possible partners might only be vaguely known, or the service behavior contains arbitrary repetitions. Then the estimation of costs for running this service is difficult and decisions based on them can hardly be made. We propose a static analysis of the service's behavior. We over-approximate possible runs and therefore costs of the service. Our approach provides a basis for reasoning about nonfunctional properties as shown for costs. Key words: Nonfunctional properties, state equation, open nets, service behavior 1 Motivation and approach A service provider in a service-oriented architecture (SOA) [1] has only limited knowledge about the future interaction partners of its published service. Then determining nonfunctional properties such as estimating the costs of a service run (or any other performance measure) in the interaction...