The past few years have seen important advances in the domain of Semantic Web Services (SWS), especially in data mediation. Most work in the area has focused on the semantic alignement of input/output concepts at design time, and on schema-level integration [1]. Correct communication is not guaranteed even when two services are connected to each other with compatible input/output concepts. Indeed, conceptually compatible data may not be usable when data representation and scaling conflicts occur. To address this problem, which we refered to as the contextual heterogeneity problem, we rely on the Minimal Service Model [2] and operate with Linked Services as SWS that offer explicit semantics. In this paper, we present our Mediation as a Service (MaaS) architecture and demonstrate its applicability with a running scenario and a prototype.