- What “Machine-Actionable” Actually Means
- The RDA DMP Common Standard: What It Changes
- Which Tools Implement the Standard
- Why Funders Are Moving in This Direction
Data management plans (DMPs) have traditionally been static, prose documents written once at proposal stage and rarely opened again. That is changing. Funders, repositories and institutional systems are converging on machine-actionable data management plans (maDMPs) — DMPs structured so that software, not just people, can read and act on them. The shift is being driven by the RDA DMP Common Standard, a specification from the Research Data Alliance that turns free-text plans into structured, exchangeable data. This article explains what “machine-actionable” means in practice, what the standard actually changes, which tools implement it, and why funders are pushing the sector in this direction.
What “Machine-Actionable” Actually Means
A conventional DMP is a Word document or PDF: a human writes prose describing what data will be collected, how it will be stored, and where it will end up. A reviewer reads it once, files it, and rarely revisits it. Nothing in that document can be queried, validated automatically, or passed to another system without someone re-typing it.
A machine-actionable DMP replaces (or accompanies) that prose with structured fields — dataset descriptions, distribution details, metadata standards, licences, repository identifiers — encoded so that a repository, funder portal, or research information system (CRIS) can parse them directly. The foundational framing paper, Ten Principles for Machine-Actionable Data Management Plans (Miksa, Simms, Mietchen & Jones, PLOS Computational Biology, 2019, cited over 130 times), describes the goal as embedding DMPs in existing research workflows so parts of the plan can be generated, validated and updated automatically rather than retyped at every stage.
- Structured, not free-text — fields for dataset type, format, volume, access conditions and repository are discrete and machine-parseable.
- A living document — updated through the project lifecycle rather than filed once and forgotten.
- Interoperable — exportable between DMP tools, repositories, CRIS platforms and funder systems without manual re-entry.
- Partially automatable — some fields (e.g. ORCID iDs, grant metadata, repository policies) can be pre-filled from connected systems.
Definitions of related research-data terms are catalogued in the CASRAI Dictionary.
The RDA DMP Common Standard: What It Changes
The RDA DMP Common Standard for Machine-actionable Data Management Plans, developed by an RDA working group, defines a shared JSON schema for representing a DMP’s core elements: project and funder metadata, one or more datasets, each dataset’s distribution (repository, licence, access level), and the metadata standards applied to it. The schema is published and version-controlled openly on GitHub, so any tool builder can implement it without licensing constraints.
Before a common schema existed, each DMP tool stored plans in its own proprietary structure. A plan created in one system could not be meaningfully exported to another, and funders could not aggregate structured data across grant portfolios without manual extraction. The Common Standard changes that by giving every participating tool the same underlying data model, so a DMP authored in one platform can, in principle, be exported as valid maDMP JSON and ingested by another.
This matters most at the points where a DMP currently has to be re-keyed: submitting to a funder portal, registering a dataset with a repository, and reporting compliance at project close. A structured, standard-conformant DMP removes several of those manual hand-offs.
Which Tools Implement the Standard
Three tools dominate current maDMP practice, each maintained by a different non-profit research-infrastructure organisation:
| Tool | Maintaining organisation | Primary user base | maDMP support |
|---|---|---|---|
| DMPonline | Digital Curation Centre (DCC), University of Edinburgh | UK and international institutions | API and structured export aligned to the RDA Common Standard |
| DMPTool | California Digital Library (CDL/UC3) | US universities and federal-grant researchers | Templates mapped to funder requirements; RDA-aligned export in progress |
| ARGOS | OpenAIRE, originally built under the EU FAIRsFAIR project | Horizon Europe and EOSC-affiliated researchers | Native maDMP JSON, direct repository and metadata-standard linking |
DMPonline and DMPTool both originated as template-driven questionnaires aligned to specific funder wording, then layered structured export on top as the Common Standard matured. ARGOS was built later, directly on the RDA schema, as part of the EU-funded FAIRsFAIR (“Fostering FAIR Data Practices in Europe”) project, which is why it links more natively to repositories and metadata standards rather than treating them as free-text fields. Institutions choosing between them should check which one their funder or repository already exchanges data with, rather than assuming full interoperability across all three.
Why Funders Are Moving in This Direction
Funders adopted DMP requirements originally to make researchers think about data stewardship before, not after, the fact. Horizon Europe requires a DMP as a formal deliverable for data-generating projects, due within six months of the project start and updated at least at the mid-term and final reporting points — a recurring obligation that is far easier to track programmatically than by re-reading prose each time. The US National Institutes of Health introduced its Data Management and Sharing Policy in 2023, requiring a DMS plan for every funded project involving scientific data, which has pushed US institutions toward tools that can validate plans at scale rather than review them manually.
For funders managing thousands of active grants, machine-actionable plans mean compliance can be checked computationally — flagging, for instance, a dataset with no named repository or an access licence inconsistent with funder policy — instead of requiring programme officers to re-read each document individually. For research administrators, the practical benefit is fewer duplicate data-entry tasks across grant systems, repositories and institutional CRIS platforms, and DMPs that can be audited at renewal or close-out without starting from scratch.
Common Questions About Machine-Actionable DMPs
What is a machine-actionable data management plan?
A machine-actionable data management plan (maDMP) is a DMP whose content is structured — typically as JSON conforming to the RDA DMP Common Standard — so that repositories, funder systems and research information platforms can read, validate and act on it automatically, rather than relying on a human re-reading free-form prose.
What should a data management plan include?
A DMP typically describes the types and volume of data to be generated, metadata standards applied, storage and security arrangements, ethical and legal considerations, roles and responsibilities, and the data-sharing and long-term preservation plan, including the intended repository and access licence.
Why is research data management important?
Sound research data management improves the integrity, reproducibility and reuse value of research outputs. It ensures data remain findable and accessible after a project ends, satisfies funder and publisher mandates, and reduces the risk that valuable data become unusable or unrecoverable once the original team disperses.
The direction of travel is clear: DMPs are moving from a one-off compliance document to structured metadata that persists and updates across a project’s life, feeding repositories, funder reporting and institutional systems without re-transcription. Institutions that adopt an RDA-aligned tool now — DMPonline, DMPTool or ARGOS — are better positioned as more funders begin to require, rather than merely accept, structured plans.
Leave a Reply