Data Stewardship

Data stewardship is the responsible management and governance of research data throughout its entire lifecycle, from planning and collection through documentation, storage, analysis, sharing, and long term preservation. Good data stewardship follows the FAIR principles, ensuring that data are Findable, Accessible, Interoperable, and Reusable.^[1] It encompasses not only the technical aspects of data management but also the ethical, legal, and organizational responsibilities that come with handling research data.

As research produces ever larger and more complex datasets, the need for systematic data management has become urgent. Many funding agencies now require data management plans, either with grant applications or early in the funded project, and journals increasingly expect evidence that data have been managed according to recognized standards. Poor data stewardship leads to data loss (studies estimate that the availability of research data declines sharply within years of publication)^[2], wasted resources when data must be collected again because originals are unusable, and missed opportunities for secondary analysis and data reuse. Good stewardship extends the value of research investments by ensuring that expensive datasets remain accessible and usable long after the original project ends.

Data stewardship begins before data collection, with the creation of a data management plan (DMP). The DMP describes what data will be collected, how they will be organized and documented, what standards and formats will be used, how quality will be assured, where data will be stored during and after the project, who will have access, and how the data will eventually be preserved or shared.^[3] Tools such as DMPTool and DMPonline provide structured templates aligned with funder requirements.^[3]

During the research, stewardship involves implementing the DMP through consistent file naming conventions, version control, regular backups, documentation of all data transformations, and maintenance of data dictionaries and metadata records. Data security measures, including encryption, access controls, and secure storage environments, protect sensitive data from unauthorized access.

After the research, stewardship focuses on preparing data for deposit in an appropriate repository, selecting licenses that govern reuse, and ensuring that documentation is comprehensive enough for other researchers to understand and use the data without consulting the original team.^[1] For data that cannot be shared due to privacy constraints, stewardship includes maintaining secure access mechanisms and documenting the conditions under which access might be granted.

Organizations increasingly designate data stewards, individuals or roles responsible for overseeing data management within a research group, department, or institution. Data stewards may be librarians, data managers, or researchers with specialized training. Their responsibilities include advising on data management planning, ensuring compliance with institutional and funder policies, providing training on data management tools and practices, and facilitating data deposit and sharing.

Data stewardship applies to every research project that generates or uses data. It is not limited to large scale projects or data intensive fields; even small qualitative studies benefit from organized data management, clear documentation, and thoughtful planning for data preservation. Stewardship planning should begin at the grant application stage, when the DMP is first drafted, and continue throughout the project lifecycle.^[3] It is especially critical for longitudinal studies, multisite collaborations, and any project where data will be shared or preserved for future use.

A research consortium studying environmental exposures and child development across five countries designates a central data stewardship team comprising a data manager, a metadata specialist, and a privacy officer. The team develops a comprehensive DMP specifying ISO 8601 date formats, a harmonized variable naming convention, DDI metadata standards, and a tiered access model. During the five year data collection period, the team conducts quarterly quality audits, maintains a centralized data dictionary, and provides training to site coordinators. After the project, the harmonized dataset is deposited in a domain repository with a 12 month embargo and comprehensive documentation: the open tier carries a Creative Commons CC BY 4.0 license, while the more sensitive child level variables stay under controlled access, following the tiered access model.