In large cultural heritage data aggregation systems such as Europeana, automatic and manual metadata enrichments are used to overcome the issues raised by multilingual and heterogeneous data. Enrichments are based on linked open datasets, which can be very beneficial for enabling retrieval across languages, adding context to cultural heritage objects and for improving the overall quality of the metadata. However, if not done correctly, the enrichments may transform into errors, which propagate to several languages and impacting the retrieval performance and user experience. To identify the different processes that impact the quality of enrichments, Europeana and affiliated projects’ representatives have organised a series of experiments applying several enrichment techniques on a particular dataset constituted of random metadata samples from several data providers from several domains, but mainly from library held cultural heritage digital objects. Comparing and analysing the results shows that selecting appropriate target vocabularies, fine-tuning enrichment rules are as important as defining evaluation methods. The development of flexible workflows will contribute to better interoperability between enrichment services and data, but might make individual enrichment processes more ambivalent. Efforts where users evaluate and correct enrichments as well as the enrichments’ impact on retrieval and user experience also need to be considered. The presentation will show how a better understanding of enrichment methodologies will help cultural heritage institutions and specifically libraries to get the semantics right.