Social innovation (SI) is an ambiguous concept with multitude of definitions debated vigorously. While the primary impact of the definitional debate is on academic research, it also significantly shapes the scope and design of public policy. The motivation of this study is to analyse the correspondence of the different definitions of SI in social innovation practice and draw policy lessons. Through an extensive review of literature of the multitude of definitions and based on previous research, the researchers identify four overall components (objectives; actors and actor interactions; output; innovativeness) that the definitions use in different combinations. For the analysis it has been used the European Social Innovation Database (ESID), a comprehensive database of social innovation projects utilising machine learning and text-mining to collect data. ESID does not adopt a specific definition of SI but scores projects based on the four components mentioned above, which enables its users to construct a query based on a preferred definition. The researchers consider these components as different types of SI, each indicating to a set of specific features. Based on over 6,000 SI projects included in ESID, University of Strathclyde analyse the prevalence of these types of SI as well as their relationship with each other, geography and topics. The policy brief is supported by a web app (https://bit.ly/ESIDapp) which enables the reader to view the underlying data and interactively engage with the analysis.