Architecture models are used in enterprise management for decision support. These decisions range from designing processes to planning for the appropriate supporting technology. It is unreasonable for an existing enterprise to completely reinvent itself. Incremental changes are in most cases a more resource efficient tactic. Thus, for planning organizational changes, models of the current practices and systems need to be created. For mid-sized to large organizations this can be an enormous task when executed manually. Fortunately, there's a lot of data available from different sources within an enterprise that can be used for populating such models. The data are however almost always heterogeneous and usually only representing fragmented views of certain aspects. In order to merge such data and obtaining a unified view of the enterprise a suitable methodology is needed. In this paper we address this problem of creating enterprise architecture models from heterogeneous data. The paper proposes a novel approach that combines methods from the fields of data fusion and data warehousing. The approach is tested using a modeling language focusing on cyber security analysis in a study of a lab setup mirroring a small power utility's IT environment.