Datasets are defined as items shared on repositories which are categorized as datasets. The data is sourced from DataCite and Figshare, including their hosted repositories, and includes items classified as datasets, software, physical objects, data papers, data collections and some other records. Preprints, posters and images are excluded from the index.
Datasets are an important research output and may be used in different ways, for example by 

• researchers who would like to find datasets for reuse;
• academic institutions who are interested in making their data more accessible and who want to get a more complete picture of the outputs of their researchers;
• funders who would like to analyze the impact of their grants in terms of datasets as a research output;
• publishers who want to analyze journals in terms of publications with associated datasets - and where datasets get deposited.

For more information, see also the article "Which data repositories are indexed in Dimensions?"