Data are at the heart of policy decisions and represent the most valuable asset, after people, at the European Central Bank (ECB).2 This recognition has driven the institution towards the adoption of a data management strategy which is oriented towards more common and integrated data processes and to a data centric architecture.
Data quality management is core to the ECB statistical function. In the last two decades tremendous efforts have been undertaken in order to fill the gaps in aggregated and standardised data. Today the focus shifted on managing high volumes of data, in producing high-quality and granular data3 and in sharing data products. This has driven the ECB towards the adoption of large-scale IT systems that allow to combine more efficiently data sources and data models and to perform data analysis with different analytical tools. However, combining data in an IT platform, although necessary, is not sufficient to achieve high quality integrated data.4
Indeed, the quality of integrated data depends not only on the quality of the individual data sources, but also on the quality of all interrelated components of data management. These components encompass the univocal identification of business entities (master data) for which economic transactions and positions need to be analysed. Further components are the conceptual definitions and the methodology of compilation of data collected and provided from different sources; as a consequence, contextual integration requires congruence in the definition of concepts and in the codification of data. Additionally, clear and structured information (metadata) that clarify the meaning and the structure of the data are relevant to integrate data efficiently. Finally, a state-of-the-art technical infrastructure is essential to enable the integration of all these components.
Integrating data involves different but interdependent data management activities. Therefore, to reach the desired level of quality, not only data but also models, processes and tools must respond to measurable quality indicators, according to a defined and compatible maturity model. In this sense, the fitness for purpose of integrated data is a richer concept than the fitness of a single dataset individually considered.