Linked Open Data (LOD) is a term used to identify a set of principles for publishing ad interlinking structured data. In recent years, several EU initiatives have sought to encourage the development of LOD technology and the publication of data as LOD (e.g. the European Open Data Portal initiative). As the major provider of Official Statistics on Europe, Eurostat is also currently improving its agility in responding to new user needs by making it easier for Eurostat's statisticians and data analysts to search and integrate Eurostat data. One of the main obstacles to achieving these goals is the fragmented nature of Eurostat's current data architecture. Eurostat's data and metadata remain largely confined in separate data silos with a low degree of interoperability, which makes finding and combining data from different domains cumbersome. Through few use case applications, the benefit of LOD is demonstrated to enable: (i) statisticians and data analysts within Eurostat finding data that can be used to answer questions or analytical requests from users, (ii) users exploiting the links to other data sources to enrich their analysis of Eurostat’s data or discover new facts about these data. Besides modelling Eurostat’s data and metadata as linked data using well known ontologies, this paper also explores new models for knowledge representation, derived from statistical domain expertise, to allow a harmonised (and “linkeable”) description of statistical concepts and other ontological constructs.