Service mash up is the act of integrating the resulting data of two complementary software services into a common picture. Such an approach is promising with respect to the discovery of new types of knowledge. However, before service mash up routines can be executed, it is necessary to predict which services (of an open repository) are viable candidates. Similar to Knowledge Discovery in Databases (KDD), the authors introduce the Knowledge Discovery in Services (KDS) process that identifies mash up candidates. In this paper, the KDS process is specialized to address a repository of open services that do not contain semantic annotations. In these situations, specialized techniques are required to determine equivalences among open services with reasonable precision.