In the last years both industry and academia have shown a great interest in ensuring consistent cooperation for business-critical services, with contractually agreed levels of Quality of Service. Service Level Agreement specifications as well as techniques for their evaluation are nowadays irremissible assets. This paper presents Puppet (Pick UP Performance Evaluation Test-bed), an approach and a tool for the automatic generation of test-beds to empirically evaluate the QoS features of a Web Service under development. Specifically, the generation exploits the information about the coordinating scenario (be it choreography or orchestration), the service description (WSDL) and the specification of the agreements (WS-Agreement).