As Linked Open Data originates from various sources, lever- aging well-defined ontologies aids integration. However, oftentimes the utilization of RDF vocabularies by data publishers differs from the in- tended application envisioned by ontology engineers. Especially in large- scale datasets as presented in the Billion Triple Challenge a significant divergence between vocabulary specification and usage patterns can be observed. This may impede the goals of the Web of Data in terms of dis- covering domain-specific information in the Semantic Web. In this work, we identify common misusage patterns by employing frequency analysis and rule mining and propose re-engineering suggestions.