Skip to content

How to Write a Data Management Plan for Grants

Updated 23 official sources

Quick Answer: NIH, NSF, DOE and NASA all require a data management and sharing plan with research proposals, and they treat it differently.

Full answer and how we checked it

NIH recommends 2 pages or less and keeps the plan out of scored peer review; NSF builds it in a Research.gov tool (required since April 27, 2026) and peer reviewers evaluate it; DOE's Office of Science asks reviewers to comment on it; NASA counts its 2-page plan toward intrinsic merit. This guide is for researchers and research administrators writing a plan for any of these funders.

Data note: Grantsights checked these requirements on October 18, 2026 against NIH notices NOT-OD-21-013, NOT-OD-21-014, NOT-OD-24-175, NOT-OD-24-176 and NOT-OD-25-101; NSF Policy Notice 26-202 and NSF's Engineering directorate guidance; the DOE Public Access Plan pages and Office of Science guidance; and NASA's ROSES plan FAQ. Repository fees change, so confirm them before you budget.

In this guide

Data management plans used to be an afterthought. A paragraph about storing files on a lab server was enough for many reviewers. That changed in 2023.

NIH's Data Management and Sharing (DMS) Policy took effect January 25, 2023, and since then the approved plan has been a term and condition of every NIH award that generates scientific data, per notice NOT-OD-21-013 (opens in a new tab). NSF has required a plan since 2011. In 2026 it moved the plan into a Research.gov tool and began requiring that data supporting NSF-funded publications be shared at the time of publication. DOE put new plan rules into solicitations starting October 1, 2025, and NASA's Science Mission Directorate made its plan part of the merit evaluation.

Many plans fail for the same reason: they're generic. They describe intentions without naming the data types, standards, repositories and dates that reviewers and program officers need. This guide covers the cross-agency picture. If you're only applying to NSF, the NSF data management plan guide goes deeper on that one tool.

What a Data Management Plan Does in Your Application

A plan does 3 jobs, and the weight of each depends on the agency.

Compliance. At NSF, a missing plan gets the proposal returned without review, according to NSF's Engineering directorate guidance (opens in a new tab). At DOE's Office of Science, a proposal without a plan may be rejected without merit review, according to the Office of Science digital data management page (opens in a new tab). At NIH, program staff must find the plan acceptable before award.

Review. This is where agencies split. According to NIH's DMS policy overview (opens in a new tab), peer review won't see or review the plan, though reviewers will consider related budget items. NSF's Engineering page says plans are subject to peer review. NASA's ROSES guidance says the plan is evaluated as part of intrinsic merit. A weak plan costs you score points at NSF and NASA and costs you time at NIH.

Post-award accountability. The approved plan becomes a promise you report against. NIH's progress reports now ask what data you generated, where you shared it, and under which identifiers. If you promised a deposit within 12 months of collection, program staff can check.


Looking for your next grant?

About once a week, we email a short list of open grant calls with real deadlines from different funders. It lists open calls only.

No spam. Unsubscribe anytime.

Agency Requirements at a Glance

Here's what 5 major research funders expect, with links to the policy text.

NSF: Data Management and Sharing Plan

NSF has required a supplementary plan since 2011. According to NSF Policy Notice 26-202 (opens in a new tab), which took effect January 22, 2026:

  • Proposals must include a plan created with the Research.gov DMSP tool, released April 27, 2026. Proposals submitted earlier used a PDF upload.
  • The notice struck the two-page PDF limit, but NSF's Engineering guidance (opens in a new tab) still refers to a two-page limit that solicitations can change. The plan isn't part of the 15-page project description, and it can't be used to get around that limit.
  • Recipients must share all data supporting NSF-funded publications at the time of publication. Exceptions must be described and justified in the plan.
  • Collaborative and subaward proposals submit one combined plan.
  • The plan may cover 5 kinds of content: types of data and materials produced, data and metadata standards, access and sharing policies, reuse and redistribution policies, and archiving.
  • The Research.gov form is directorate specific. NSF's Engineering page says it asks about each data or research product across 8 items, from product category and provenance to public archiving, timeline and accountability.

Important: Each NSF directorate publishes its own expectations, and the form fields change with the primary directorate you pick when you create the proposal. A plan written for Social, Behavioral, and Economic Sciences won't match the Geosciences fields. The NSF-specific guide walks through the tool.

NIH: Data Management and Sharing Plan (DMS Policy)

The DMS Policy applies to NIH-funded research that generates scientific data, for competing applications with receipt dates on or after January 25, 2023. According to NIH notice NOT-OD-21-014 (opens in a new tab), the plan should address 6 elements in 2 pages or less:

  • Data type
  • Related tools, software and/or code
  • Standards
  • Data preservation, access, and associated timelines
  • Access, distribution, or reuse considerations
  • Oversight of data management and sharing

Other rules from NOT-OD-21-013 and NIH's policy overview:

  • Data should be shared as soon as possible and no later than the time of an associated publication or the end of the performance period, whichever comes first
  • Program staff review the plan; it isn't part of scored peer review unless the funding opportunity says so
  • The approved plan becomes a term and condition of the award
  • Data sharing is reported in Research Performance Progress Reports (RPPRs)
  • Data management and sharing costs are allowable and belong in the budget request

DOE: Data Management and Sharing Plan

According to DOE's policy for digital research data management (opens in a new tab), all DOE-funded research and development awards and contracts are subject to a DOE-approved plan, and the 2023 Public Access Plan requirements entered solicitations starting October 1, 2025. Solicitations issued before then still follow the 2014 rules. The plan must cover 5 requirements:

  • How data will be shared and preserved, or how results could be validated and replicated if they won't be
  • Timely and fair access, including making data behind peer-reviewed publications open at publication
  • Repository selection that matches federal desirable characteristics (DOE doesn't usually name a repository, but a sponsoring office can)
  • Data management and sharing resources, including written approval when a plan commits a DOE user facility beyond its normal support
  • Limits on sharing for confidentiality, privacy, security, and intellectual property

Sponsoring offices can add requirements in the solicitation. Publicly shared data must be reported to DOE's Office of Scientific and Technical Information (OSTI). DOE's page sets no page limit.

NASA: Open Science and Data Management Plan

According to NASA's ROSES plan FAQ (opens in a new tab), Science Mission Directorate proposals include an Open Science and Data Management Plan as a 2-page section in the proposal PDF, right after the references, and it doesn't count against the science/technical/management page limit. Astrophysics Data Analysis is an exception that puts the plan inside its 15-page section, and instrument and technology development programs are generally exempt. The plan covers data, software (including an open license), and publications. Under NASA's science information policy (opens in a new tab), research data and software are shared at the time of publication or the end of the award, and publications are open with no embargo.

DOD and Other Agencies

Department of Defense requirements vary by component (Army, Navy, Air Force, DARPA) and usually appear in the solicitation or award terms. Expect stricter storage and security language when controlled unclassified information is involved. Grantsights didn't find a single DOD-wide plan format to cite, so read the specific announcement.

DMS Plan ElementNIH RequirementNSF Requirement
Data typesDescribe types and amountsDescribe data, samples and materials produced
StandardsMetadata and data standards; common data elementsData and metadata standards
RepositoryName the repositoryName the repository or archive
Access timelineNo later than publication or end of performance period, whichever is firstData supporting publications at publication
PreservationPer repository and planDescribe archiving and preservation of access
BudgetDMS justification, half a page or lessInclude in budget and justification
OversightOversight element requiredAccountability item in the Research.gov form

Grantsights Analysis: How 4 Agencies Review the Plan

Grantsights analysis of the current policy pages for NIH, NSF, DOE (department-wide and its Office of Science) and NASA's Science Mission Directorate, checked October 18, 2026, found that NIH is the only one of the 4 agencies that keeps the plan out of scored peer review. Method: we read each agency's official policy notice or guidance page and recorded the page limit, required elements, who reviews the plan, and whether it affects the merit score.

AgencyPage limitRequired elementsWho reviews itAffects merit score?Source
NIH2 pages or less (recommended)6 elementsProgram staffNo, unless the funding opportunity says soNOT-OD-21-014; DMS policy overview
NSFResearch.gov tool; ENG cites 2 pages5 content areas; 8 items per product in ENG formPeer reviewersYes (ENG: subject to peer review)NSF 26-202; ENG guidance
DOE Office of ScienceNone stated5 requirementsMerit reviewers comment on itReviewers assess appropriatenessOffice of Science data management page
NASA SMD2 pagesData, software, publicationsPeer reviewersYes, as intrinsic meritROSES OSDMP FAQ
DOE (department-wide)None stated5 requirementsSponsoring officeSet by solicitationDOE data management policy

What this changes for you: if you recycle an NIH plan for an NSF or NASA proposal, you're moving a document that nobody scored into a place where it's scored. That's worth an extra hour of rewriting. The reverse also holds. An NIH plan doesn't need persuasive framing, but it does need dates and repository names that program staff can approve without a back-and-forth before the Just-in-Time request. The NIH Just-in-Time guide covers where a plan revision lands in that step.


The 6 Sections Every DMP Should Cover

Regardless of the funder, every plan answers the same 6 questions. They're ordered here to match NIH's elements, and they map onto NSF's 5 content areas and DOE's 5 requirements.

Section 1: Data Types and Amounts

Describe exactly what data your project will generate. Be specific about formats, volume, and how data will be collected.

Weak: "This project will generate survey data and interview transcripts."

Strong: "This project will generate three primary data types: (1) structured survey data from approximately 800 participants collected via Qualtrics, exported as CSV files, estimated at 2 GB total; (2) semi-structured interview recordings (approximately 60 interviews, 45-60 minutes each) stored as WAV audio files, estimated at 30 GB total; and (3) de-identified interview transcripts produced via professional transcription and stored as UTF-8 plain text files, estimated at 500 MB total."

Include:

  • All data types (quantitative, qualitative, observational, computational, imaging)
  • File formats for each data type
  • Estimated data volume
  • Collection methods and instruments
  • Whether data is generated from human subjects, animal subjects, or other sources

Section 2: Standards and Metadata

Describe the standards you'll use to format, document, and describe your data so others can understand and reuse it.

Metadata is the documentation that makes data findable and interpretable. It answers: what does this variable mean, how was it measured, when was it collected, and under what conditions?

Address:

  • Metadata standards relevant to your discipline (Dublin Core, DDI, EML, FGDC, etc.)
  • Data documentation practices (README files, codebooks, data dictionaries)
  • Naming conventions for files and variables
  • Quality assurance procedures
  • Version control practices

Example: "All survey data will be documented using the Data Documentation Initiative (DDI) metadata standard. Each dataset will include a machine-readable codebook with variable names, labels, value labels, missing data codes, and skip patterns. Interview data will be documented with a README file describing the sampling strategy, interview protocol, de-identification procedures, and transcription conventions."

Section 3: Access and Sharing

This is where most researchers get nervous, but the expectations are clearer than you might think. Describe who can access the data, when, and under what conditions.

Address:

  • Which data will be shared publicly
  • Which data have access restrictions (and why)
  • Sharing timeline (when data becomes available)
  • Data use agreements or access request procedures
  • Whether data will be embargoed during the project period

NIH-specific: Share scientific data as soon as possible and no later than an associated publication or the end of the performance period, whichever comes first.

NSF-specific: Under NSF 26-202 (opens in a new tab), data supporting NSF-funded publications must be shared at the time of publication. Justify any exception in the plan.

NASA-specific: Research data and software are shared at the time of publication or the end of the award.

Section 4: Preservation and Repositories

Describe where your data will be stored long-term and for how long.

Repositories to consider:

  • Discipline-specific repositories (ICPSR for social science, GenBank for genomics, PDB for protein structures)
  • Generalist repositories (Dryad, Figshare, Zenodo)
  • Institutional repositories maintained by your university library
  • NIH-supported repositories (dbGaP for controlled-access genomic data, ClinicalTrials.gov for clinical trial data)

Address:

  • Repository name and URL
  • Whether the repository assigns persistent identifiers (DOIs or accession numbers)
  • Data retention period (minimum: duration of award plus funder-required retention)
  • Who pays for long-term storage
  • Backup procedures during the active project

Example: "De-identified survey data and codebooks will be deposited in ICPSR (Inter-university Consortium for Political and Social Research) within 12 months of data collection. ICPSR assigns DOIs to deposited datasets. Interview transcripts will be deposited in the Qualitative Data Repository (QDR) with controlled access requiring a signed data use agreement."

Section 5: FAIR Principles

The FAIR principles (Findable, Accessible, Interoperable, Reusable) are the framework federal funders increasingly expect. NIAID's guidance on FAIR data (opens in a new tab) describes how they apply to NIH-funded research, and NOT-OD-21-013 added a prompt about how data will be findable, such as through a persistent unique identifier.

Findable: Use repositories that assign persistent identifiers. Include full metadata. Register your dataset in a searchable catalog.

Accessible: Deposit data in repositories with clear access protocols. If access is restricted, provide a clear process for requesting data.

Interoperable: Use standard file formats (CSV, XML, JSON rather than proprietary formats). Use controlled vocabularies and discipline-standard ontologies.

Reusable: Include thorough documentation. Attach clear licensing terms (Creative Commons licenses are common for research data). Describe the conditions under which data can be reused.

You don't need a separate FAIR section, but weaving these principles into your answers shows reviewers and program officers that you know current expectations.

Section 6: Budget and Resources

Data management costs money, and all 4 science funders above let you request it. According to NIH's budgeting page for data management and sharing (opens in a new tab), applicants with detailed budgets request these costs in the matching cost categories and add a "Data Management and Sharing Justification" of half a page or less to the budget justification. Modular applicants put the same text in the Additional Narrative Justification, which the NIH modular budget guide explains.

Allowable at NIH, per that page:

  • Curating data and developing documentation
  • Formatting data to community standards and transmitting it to a repository
  • De-identifying data
  • Preparing metadata for discovery and reuse
  • Local data management infrastructure needed before deposit
  • Repository deposit fees, including more than one repository if the plan names several

Not allowable at NIH: infrastructure already covered by indirect costs, routine costs of collecting or accessing data, and anything charged twice. Costs must be incurred during the performance period, so a 10-year repository fee has to be paid before the award ends.

Repository fees vary widely. According to Dryad's costs page (opens in a new tab), an unsponsored researcher pays a $150 data publishing charge for a dataset up to 5 GB, effective May 6, 2025, and Dryad partner institutions sponsor submissions up to 10 GB. Check whether your university is a partner before you budget a fee you won't pay.

Example budget line: "Data Management and Sharing: $4,200/year. Research coordinator time for data cleaning, metadata creation, and repository deposits (5% effort, $3,200); repository deposit fees ($500); encrypted cloud backup ($500)."


Using DMPTool to Build Your Plan

DMPTool is free, maintained by the California Digital Library, and holds templates for NSF, NIH, DOE, NEH, USDA and other funders. A 5-step workflow gets most plans from blank to final. Many institutions have registered, and many university libraries give their researchers access.

What DMPTool does well:

  • Pre-loaded funder templates
  • Funder guidance embedded in each section
  • Sample plans from funded proposals
  • Collaboration features so co-investigators and your sponsored research office can comment
  • Export to Word, PDF, or HTML

How to use it in 5 steps:

  1. Create an account through your institution (if available) or with any email
  2. Start a new plan and select your funder and directorate
  3. Work through each section using the prompts and guidance text
  4. Share the draft with your sponsored research office for review
  5. Export the final version. For NSF proposals submitted on or after April 27, 2026, paste each section into the Research.gov DMSP tool instead of uploading a PDF, and keep a copy, since university library guidance warns the tool may not save drafts.

DMPTool won't write the plan for you, but the templates stop you from missing a required element.


Discipline-Specific Considerations

Your field changes the details more than your funder does. Here are 5 points to add for each of 4 common discipline groups.

Biomedical and Clinical Research

  • Address human subjects protections, HIPAA compliance, and IRB requirements
  • Specify de-identification methods (Safe Harbor or Expert Determination under HIPAA)
  • Identify controlled-access repositories (dbGaP, BioData Catalyst)
  • Address informed consent provisions for data sharing (prospective consent language matters)
  • Include ClinicalTrials.gov registration and results reporting for clinical trials

Social and Behavioral Sciences

  • Address confidentiality for qualitative data (names, locations, identifying details)
  • Specify de-identification procedures for interview transcripts
  • Use DDI metadata standards for survey data
  • Consider controlled-access options (ICPSR restricted-use data, QDR)
  • Address data from vulnerable or hard-to-reach populations

Physical Sciences and Engineering

  • Specify data formats and standards for instrument-generated data
  • Address large data volume management (computational outputs can generate terabytes)
  • Identify discipline-specific repositories (Materials Data Facility, Protein Data Bank)
  • Include software and code sharing plans (GitHub, Zenodo for code DOIs)
  • Document computational environments for reproducibility

Environmental and Earth Sciences

  • Use FGDC or ISO 19115 metadata standards for geospatial data
  • Address sensor data management and quality control
  • Identify repositories (PANGAEA, NCAR, USGS ScienceBase)
  • Include field sample management and physical specimen preservation
  • Address spatial and temporal resolution of datasets

NIH DMS Policy: What Changed in 2023

The 2023 DMS Policy replaced NIH's 2003 Data Sharing Policy and widened it in 5 ways.

Applies to all research generating scientific data, not just large awards. The 2003 policy required a sharing plan only for applications seeking $500,000 or more in direct costs in any year. The 2023 policy applies regardless of award size.

Plan is reviewed and approved. Program staff review the plan and may ask for changes before award. The approved plan becomes a term and condition of the Notice of Award.

Sharing timeline is defined. Data should be shared no later than an associated publication or the end of the performance period, whichever comes first.

Costs are explicitly allowable. Curation, metadata, de-identification and repository fees are allowable direct costs if incurred during the performance period.

Monitoring is ongoing. Investigators report data sharing in RPPRs. According to NIH's policy overview, NIH can take various actions in cases of noncompliance.

Public access now runs alongside it. The August 25, 2022 OSTP memo told agencies to update their public access and data sharing plans by mid-2023 and fully implement them no later than December 31, 2025, according to the OSTP announcement (opens in a new tab). NIH moved its 2024 Public Access Policy up to July 1, 2025, according to NOT-OD-25-101 (opens in a new tab). Under NOT-OD-25-047 (opens in a new tab), the author accepted manuscript goes to PubMed Central on acceptance and is public without embargo on the official publication date. That covers papers; your DMS plan covers the data behind them.


The RPPR Trap Built Into NIH's October 2024 Reporting Update

RPPRs submitted on or after October 1, 2024 ask 6 questions about data management and sharing. According to NOT-OD-24-175 (opens in a new tab), they ask whether data has been generated and what type, whether it's been shared, the repository and unique identifiers for shared data, the status of anything not yet shared, corrective actions if you haven't followed the approved plan, and any significant plan changes expected in the next year.

The trap is a plan written at a high level of generality, like "survey data and interview transcripts." You'll still have to report what was generated, where it went, and under which identifiers, and compare that with what the approved plan promised. A vague plan makes that comparison harder. For example, a hypothetical team whose plan said only "participant questionnaires and biomarker assays" could struggle to show which datasets were shared, and might need to submit a revised plan if its approach has changed. Writing a vague plan at proposal time is a mistake you'll pay for at year-one reporting.

NIH notice NOT-OD-24-176 (opens in a new tab), released September 23, 2024, set out 3 pathways for plan revisions: the Just-in-Time screen before award, Section C.5.c of the RPPR, or the Prior Approval Module for off-cycle changes. A Signing Official must submit the request at least 30 days before the change, and the approved plan stays in effect until the funding institute approves the revision. A team that switched repositories mid-award without that request is out of compliance even if the new repository is a reasonable choice.

NIH's Generalist Repository Ecosystem Initiative (opens in a new tab) includes 7 repositories: Dataverse, Dryad, Figshare, Mendeley Data, Open Science Framework, Vivli and Zenodo. Discipline-specific repositories, which NIH encourages when they exist, issue identifiers differently. ICPSR assigns DOIs at the dataset level, while controlled-access repositories like dbGaP issue accession numbers. Know which identifier type your repository issues before you reach the reporting stage.

The practical rule: write data type categories at the granularity the RPPR will ask about. If you'll generate 3 distinct data types, name all 3 with formats, name the repository for each, and confirm each one issues a persistent identifier. The grant progress report guide covers the rest of the RPPR.


Real Examples: What NIH Pays for Data Sharing

NIH doesn't just require sharing; it funds the infrastructure. According to NIH RePORTER (opens in a new tab), the University of Michigan received $565,461 in FY2025 for a supplement to "Data Sharing and Archiving for Demographic Research" (3R24HD048404-20S1, NICHD), a social science archive of the kind you'd name in a demography plan. In the same year, according to NIH RePORTER (opens in a new tab), the University of California, San Diego received $704,894 for "Global proteomics mass spectrometry data sharing infrastructure" (5R24GM148372-03, NIGMS).

Training is funded too. For example, Vanderbilt received $204,807 in FY2025, according to NIH RePORTER (opens in a new tab), for "Data Management for Data Sharing: A Short Course for Researchers in Children's Learning and Development" (5R25HD114368-02, NICHD).

Two lessons follow. First, if your field has an NIH-funded resource like these, name it in your plan; program staff recognize it. Second, if your field has no repository, that's a gap NIH has funded through resource awards, and it's better than writing "available on request." You can search federal grant programs to see which resource and infrastructure programs fit your field.


Common Mistakes That Get Flagged

Each of these 6 mistakes leaves a gap that NSF and NASA peer reviewers can score against or that NIH's 6-question RPPR check will expose later.

Mistake 1: Generic Plan With No Project-Specific Details

"Data will be stored securely and shared with the scientific community" is not a plan. Name the repository. Specify the file format. Quantify the data volume. Give a deposit date. Reviewers can tell when a plan was copied from an earlier proposal.

Mistake 2: Ignoring Directorate or Division-Specific Requirements

NSF's SBE directorate expects different things than BIO or GEO, and the Research.gov form changes with your primary directorate. Writing a generic NSF plan without the directorate guidance is a common mistake. The NSF review process guide explains where the plan fits in merit review.

Mistake 3: No Repository Identified

"Data will be made available upon request" is a red flag for most funders. Name a specific repository, confirm it accepts your data type, and describe the deposit process. If your discipline has no standard repository, use a generalist option like Dryad, Figshare, or Zenodo.

Mistake 4: Not Addressing Restricted Data

If your project involves human subjects, protected health information, or proprietary data, you can't say "all data will be shared openly." Describe what can be shared, what restrictions apply, and how restricted data can be accessed (controlled-access repositories, data use agreements, IRB approval for secondary analysis).

Mistake 5: Budget Doesn't Match the Plan

If the plan describes curation, repository deposits and long-term storage but the budget has zero dollars for data management, reviewers will doubt you can carry it out. If you've budgeted $15,000 for data management behind a vague paragraph, the budget item looks unjustified.

Mistake 6: Treating the Plan as Static

At NIH, a plan change needs a formal revision request at least 30 days ahead. Build in a review at each progress report so the plan still matches what you're doing.


DMP Template for NSF Proposals

This 5-part structure follows NSF's content areas. For proposals submitted on or after April 27, 2026, paste each part into the matching Research.gov field for your directorate.


Data Management and Sharing Plan

1. Types of Data

[Describe all data types, formats, and estimated volumes. Include how data will be generated or collected.]

2. Data and Metadata Standards

[Name the metadata standards you'll use. Describe documentation practices, codebooks, and quality assurance procedures.]

3. Policies for Access and Sharing

[Specify when data will be shared (data supporting publications at the time of publication), any restrictions and their justification, and how access requests will be handled. Name the repository or repositories.]

4. Policies for Reuse, Redistribution, and Derivatives

[Describe the licensing terms for shared data, such as Creative Commons CC BY 4.0. Note any restrictions on commercial use or redistribution.]

5. Plans for Archiving and Preservation

[Name the repository for long-term preservation. Describe how long data will stay accessible, following your directorate's and repository's guidance. Describe backup procedures during the active project.]


DMP Template for NIH Proposals

This 6-element structure follows NOT-OD-21-014. Keep it to 2 pages or less.


Data Management and Sharing Plan

1. Data Type

[Describe the scientific data to be generated or collected, with types, amounts, and formats. Say which data will be preserved and shared.]

2. Related Tools, Software, and/or Code

[List tools needed to access or use the data. Include software versions and whether tools are open-source or proprietary.]

3. Standards

[Describe metadata standards, data quality standards, and common data elements you'll use.]

4. Data Preservation, Access, and Associated Timelines

[Name the repository and how data will be findable, such as a persistent identifier. Give the sharing date: no later than publication or the end of the performance period, whichever is first.]

5. Access, Distribution, and Reuse Considerations

[Describe any limits on access, distribution, or reuse. Address privacy, informed consent, and data use agreements. Say what's open and what's controlled access.]

6. Oversight of Data Management and Sharing

[Name who carries out and monitors the plan, how often it's checked, and who submits revisions.]


What Agency Data Policies Don't Tell You

The 2022 OSTP memo set a December 31, 2025 deadline for agencies to fully implement public access, and the agencies didn't converge on one format. NIH asks for 6 elements, NSF uses a directorate form, DOE lists 5 requirements, and NASA wants data, software and publications in 2 pages. A plan that fits one funder is rarely the right shape for another. If you apply to several agencies, keep one master description of your data and rewrite the plan for each.

Reviewers check whether your data budget matches your plan. If you're depositing in a repository that charges fees, the fee should appear in the budget. If you're using an institutional repository, include staff time for curation. An unfunded plan signals that you won't follow through.

Plans that cover only final datasets miss what agencies ask about. NIH's second element asks for the tools, software and code needed to access or manipulate the data, and NASA's plan requires an open license for software. Document analysis code and software versions, not just the final spreadsheet.

Finally, the plan isn't the same as recordkeeping. Retention rules under 2 CFR 200 apply to financial and programmatic records; the grant recordkeeping guide covers those. Your data plan covers the research data itself.


These 6 guides cover the documents that sit next to a data plan in a research proposal:


Sources

Frequently Asked Questions

Can I budget data management costs in my grant?

Yes, at NIH, NSF, DOE and NASA. NIH's budgeting page lists allowable costs such as curating data, preparing metadata, de-identifying data, local data management infrastructure needed before deposit, and repository fees. NIH asks for a 'Data Management and Sharing Justification' of half a page or less in the budget justification (or the Additional Narrative Justification for modular budgets). Costs must be paid during the performance period, so a multi-year repository fee has to be paid before the award ends. Infrastructure already covered by indirect costs isn't allowable as a direct cost. DOE's policy page also lets proposals include plan implementation costs in the budget.

What happens if my project doesn't generate data?

You still file something. NSF says a project that won't produce data must include a document justifying that in place of the plan, and its Engineering directorate says proposals without a plan are returned without review. At NIH, the plan requirement applies to research that generates scientific data, so check the funding opportunity and explain the nature of the project if no scientific data will result. Don't leave the section blank.

How long should a data management plan be?

NIH recommends 2 pages or less (NOT-OD-21-014). NASA's ROSES guidance asks for a 2-page Open Science and Data Management Plan section in the proposal PDF. NSF is in transition: since April 27, 2026, plans are built in a Research.gov tool, NSF 26-202 struck the old two-page PDF limit, and NSF's Engineering directorate page still refers to a two-page limit that solicitations can change. DOE's policy page sets no page limit, so follow the solicitation. Shorter and specific beats long and generic everywhere.

What is DMPTool and should I use it?

DMPTool is a free, web-based tool from the California Digital Library that holds funder templates for NSF, NIH, DOE, NEH, USDA and others, with guidance text in each section. Many university libraries offer institutional accounts. It's worth using for a first plan or a new funder. For NSF proposals submitted on or after April 27, 2026, draft in DMPTool if you like, but paste the final text into the Research.gov DMSP tool, because NSF no longer accepts an uploaded PDF.

Do I need to share all my data publicly?

No. NIH, NSF, DOE and NASA all allow limits on sharing for privacy, confidentiality, security, intellectual property, and consent restrictions. NSF 26-202 says any exception to sharing must be described and justified in the plan. Say what will be restricted, why, and what can still be shared, such as de-identified files through a controlled-access repository like dbGaP or ICPSR.

Does NIH's public access policy change my data plan?

It sits beside it. The 2024 NIH Public Access Policy, effective July 1, 2025 under NOT-OD-25-101, requires the author accepted manuscript to go to PubMed Central on acceptance and be public without embargo on the official publication date. The DMS Policy covers the underlying scientific data, which must be shared no later than publication or the end of the performance period, whichever comes first.

Last updated October 18, 2026. Corrected October 18, 2026: what changed. Published by Grantsights from public sources. See our editorial policy, how we build and check guides, corrections and how we make money.