Research Data Management (RDM)

Research data management (RDM) is the set of practices that make research data findable, usable, and trustworthy across its lifecycle — from the plan written before data is collected through to its long-term preservation and reuse by others. For a research office, RDM is also a compliance surface: funders increasingly require a Data Management Plan (DMP) at proposal stage, journals require a data availability statement at submission, and repositories are audited against certification standards before an institution will point researchers at them. This hub collects CASRAI’s current RDM guidance and the vocabulary that underpins it, and links out to the real standards bodies — NIH, NSF, the Research Data Alliance (RDA), CoreTrustSeal, DataCite, FORCE11 — that define the actual requirements rather than restating them informally.

The research data lifecycle, briefly

Most institutional RDM guidance frames the work as a lifecycle rather than a single task: plan, collect, document and describe, store and preserve, then share and enable reuse. Each stage produces a different kind of commitment — a Data Management Plan is written before data exists; a dataset landing page and its metadata are produced once it does; a data availability statement is written at publication, after the data is already deposited. Treating these as one continuous chain, rather than isolated deliverables, is what keeps a DMP’s stated commitments actually true by the time a paper is submitted.

Data management plans: the funder-facing starting point

A DMP is the document a funder or institution uses to hold a project accountable for what it says it will do with its data — see CASRAI’s Data Management Plan (DMP) definition for the operational detail. The two most consequential U.S. federal requirements differ in ways that matter to how a plan gets written and reviewed:

• The NIH Data Management and Sharing Policy has been in effect since 25 January 2023 and requires a DMS Plan for all NIH-funded research generating scientific data.
• NSF’s plan was renamed from “Data Management Plan” to “Data Management and Sharing Plan” effective 22 January 2026 under PAPPG 24-1 Supplement 2, and is capped at two pages.

CASRAI’s NIH vs. NSF Data Management Plans guide covers exactly where these two requirements diverge in practice. Beyond the narrative plan a human reviewer reads, a growing number of DMP tools (DMPTool, DMPonline, DMPRoadmap) can also produce a machine-actionable DMP (maDMP) conforming to the RDA DMP Common Standard (version 1.2, published November 2025) — a JSON-based structure designed to let a DMP’s commitments be checked and cross-walked against ORCID, ROR, and CRediT identifiers automatically rather than only read by a person.

Choosing and trusting a repository

Where a dataset is deposited determines whether it stays discoverable, whether its identifiers resolve, and — for funder compliance — whether the deposit counts. CASRAI’s How to Choose an Open Data Repository guide walks through the practical selection criteria; the underlying vocabulary distinguishes a generalist repository from a domain repository, and either of those from an institutional repository run by a university itself.

Trust in a repository is not self-declared. CoreTrustSeal certifies a trusted digital repository against 16 requirements; it formed from the 2017 merger of the Data Seal of Approval and the World Data System certifications. When a funder or journal asks for a repository that meets recognized trustworthiness criteria, CoreTrustSeal certification (or an equivalent, such as ISO 16363) is the concrete thing being asked for, not a general reputation.

Making data FAIR — and, where relevant, CARE

The FAIR principles (Findable, Accessible, Interoperable, Reusable), set out by Wilkinson et al. in Scientific Data (2016), describe properties a dataset should have — not a specific format or platform. In practice, FAIR is implemented through the mechanisms covered elsewhere on this hub: a persistent identifier that resolves (see persistent data identifier and DOI), machine-readable metadata (see the next section), a clear reuse license, and deposit in a repository indexed by discovery tools such as FAIRsharing.

For Indigenous data specifically, FAIR is explicitly understood to be necessary but not sufficient. The CARE Principles for Indigenous Data Governance (Collective Benefit, Authority to Control, Responsibility, Ethics) were published by the Global Indigenous Data Alliance in September 2019 to center governance and self-determination alongside FAIR’s openness-and-reuse focus. CASRAI’s dictionary already covers the underlying concepts — Indigenous Data Governance, Indigenous Data Sovereignty, and GIDA itself.

Metadata, identifiers, and discoverability

A dataset is only as findable as its metadata is complete and machine-readable. Project metadata and a dataset’s own landing page are what a search engine, a repository index, or a CRIS actually reads — not the narrative description a researcher writes for a human. DataCite’s metadata schema, DOI registration, and persistent identifiers are the connective layer: a DataCite DOI issued for a dataset carries structured metadata (title, creator, publisher, resource type, and — since recent schema versions — richer relation and funding-reference fields) that a repository, an indexing service, and a CRIS can all read the same way. CASRAI’s DataCite metadata schema entry has the underlying detail.

Reporting data availability and citing data

Once data is deposited, a data availability statement is how a publication tells a reader where to find it — and it is checked, not decorative. PLOS, for example, explicitly rejects “available upon request” as a sufficient statement on its own. CASRAI’s How to Write a Data Availability Statement for Reproducibility guide and the underlying data availability statement definition cover what a compliant statement actually needs to say. Citing a dataset the same way a paper is cited — with a persistent identifier, not just a URL in a footnote — follows the Joint Declaration of Data Citation Principles; see also data citation principle.

Data sharing agreements and licensing

Not all research data can simply be deposited in an open repository. Where data involves identifiable participants, commercial partners, or cross-institutional collaboration, a Data Sharing Agreement or narrower Data Use Agreement sets the terms before data moves at all. Even for data that is openly shareable, the reuse license attached to a deposit determines what “open” actually permits — a dataset without an explicit license is not usable the way a CC0- or CC-BY-licensed one is, regardless of where it’s hosted. See CASRAI’s open data definition for how these pieces fit together.

Where this cluster connects to CRIS/RIM systems

An institution rarely tracks a dataset in isolation — it’s usually one record type inside a broader Current Research Information System (CRIS). CERIF, maintained by euroCRIS, and OAI-PMH harvesting under the OpenAIRE Guidelines are the mechanisms that let a repository deposit surface automatically in institutional research-information reporting rather than being entered twice. See CRIS interoperability and CRIS for the underlying vocabulary.

Reproducibility: where RDM meets computational research

Good research data management is also what makes work reproducible in the first place — a dataset without documented methods, code, and environment is not reusable in any practical sense, however well it’s deposited. CASRAI’s Dictionary distinguishes several related concepts that get conflated in casual use: reproducibility (an independent team reaching the same result from the original data and code), replicability (an independent team reaching the same result from newly collected data), and computational reproducibility specifically, which depends on the computational environment — often captured today in a container image (Docker, Singularity, or Apptainer) — being preserved alongside the code itself.

The practical RDM commitments that support this: an open code deposit alongside open data, a code availability statement published the same way a data availability statement is, and — for methods and materials that aren’t data or code — open materials deposit and a RRID for uniquely identifying a reagent, tool, or resource used. Pre-registration of a study’s hypotheses and analysis plan, and participation in the broader open science movement, are the upstream practices that make all of this traceable rather than reconstructed after the fact. CASRAI covers why this matters at the field level in reproducibility crisis.

Frequently asked questions

Is a Data Management Plan always required?

It depends on the funder and, increasingly, the institution. Most major U.S. federal funders (NIH, NSF) require one at proposal stage; many non-U.S. funders and institutional research offices now require an equivalent regardless of whether a specific grant mandates it. Check the specific funder’s current policy — these requirements have changed materially even within 2026 (see the NIH vs. NSF guide above for what currently differs).

What makes a repository “trustworthy” rather than just convenient?

Independent certification — CoreTrustSeal is the most widely referenced standard — against defined requirements covering governance, technical infrastructure, and long-term preservation commitments, not simply that a platform is well known or free to use.

Does FAIR mean the data has to be fully open?

No. FAIR is about a dataset’s discoverability and machine-readability, not its access level — a dataset can be FAIR and still sit behind a controlled-access gate (common for sensitive or human-subjects data), provided the metadata describing it, and the conditions for requesting access, are themselves findable and clear.

How does a DMP relate to a data availability statement?

A DMP is a forward-looking commitment written before data exists, reviewed by a funder at proposal stage. A data availability statement is a retrospective, publication-stage record of where the resulting data actually ended up. A DMP that promised deposit in a named repository should be traceable to a data availability statement that confirms it happened.

This hub reflects CASRAI’s current published RDM guidance and will expand as new guides and comparisons are added to the cluster — the sections above link out to real, external standards bodies wherever CASRAI does not yet have dedicated coverage of a sub-topic, rather than implying coverage that doesn’t exist yet.