OBJECTIVES: Patients with mild stroke or transient ischaemic attack (TIA) are at high risk of recurrent stroke and other cardiovascular events. The Assessment of Real-world Evidence in Stroke/TIA (ARES) programme aims to characterize the management and outcomes of patients with mild stroke/TIA in real-world clinical practice using the most suitable data sources. METHODS: In an initial Systematic Understanding of Real-world Evidence (SURE) assessment, suitable data sources (cohorts, registries and databases) were identified and characterized by systematic literature and web searches supplemented with e-mail and telephone contact. Data sources were recommended if they were active, representative, accessible, recorded National Institutes of Health Stroke Scale (NIHSS) scores or ABCD2 scores, and reported health resource utilization, ischaemic events and death during follow-up of at least 90 days (either direct or via linkage). The programme of included studies was finalized with input from principal investigators. RESULTS: More than 2900 publications and 300 websites were screened, and 16 registries, 17 cohort studies and 43 databases were reviewed. Nine data sources from seven countries were recommended, of which six complementary sources were included: Get With The Guidelines-Stroke in the USA (an in-hospital database including about 1600 hospitals); National Stroke Registry in China (132 hospitals); Fukuoka Stroke Registry in Japan (seven stroke centres); Clinical Research Centre for Stroke – 5^thDivision Registry in South Korea (12 stroke centres); Riks-Stroke in Sweden (all Swedish hospitals admitting patients with acute stroke); and Erlangen Stroke Registry in Germany (Erlanger community). Based on a globally agreed study design concept, protocols for each data source have been developed locally and are now being implemented. CONCLUSIONS: The ARES programme will provide global, observational data from contemporary populations with mild stroke/TIA in real-world clinical practice. Studies will be presented individually owing to differences in the nature of the data sources.