The Semantic Web has matured from a mere theoretical vision to a variety of ready-to-use linked open data sources currently available on the Web. Still, with respect to application development the Web community is just starting to develop new paradigms in which data as the main driver of applications is promoted to first class status. Relying on properties of resources as an indicator for the type, property-based typing is such a paradigm. In this paper, we inspect the feasibility of property-based typing for accessing data from the linked open data cloud. Problems in terms of transparency and quality of the selected data were noticeable. To alleviate these problems, we developed an iterative approach that builds on human feedback. Our experiments show the benefit of the approach and its applicability for entities from a variety of domains.