A research data management plan (DMP) explains how project data will be collected, organized, protected, documented, shared, and retained. This guide provides a reusable template, a storage and security checklist, a simple cost-estimation method, and worked examples you can adapt for a grant proposal, active study, dataset deposit, or project closeout.
Overview
A good DMP is both a planning document and a record of decisions. It should be specific enough for a research team to follow, but flexible enough to be updated when the methods, data volume, storage environment, or sharing requirements change.
Most plans should answer these questions:
- What data will the project create or reuse? Describe formats, approximate volume, collection methods, and whether the data are raw, processed, derived, or linked to other sources.
- How will the data be organized? Define file naming, folder structures, version control, metadata, and documentation practices.
- Where will data be stored and backed up? Identify approved institutional storage, working copies, backup arrangements, and recovery responsibilities.
- Who can access the data? Explain roles, permissions, authentication, encryption where appropriate, and procedures for removing access.
- What risks apply? Address human-subject information, confidential materials, controlled data, intellectual property, contractual restrictions, and re-identification risk.
- How will data be shared? Name a suitable repository or explain why access will be restricted, delayed, or limited to metadata or a controlled-access version.
- How long will data be retained? State the retention period or decision rule, who approves deletion, and how deletion will be documented.
- What will it cost? Include storage, transcription, digitization, specialist support, repository charges, preservation, and staff time when relevant.
A DMP does not replace an ethics protocol, information-security assessment, data-use agreement, or institutional records policy. Instead, it should point to those documents and keep their data-related decisions consistent.
How to estimate
Estimate the practical requirements of the project in five passes. Recording assumptions beside each estimate makes the plan easier to review and recalculate.
- Inventory the data. List each dataset, collection instrument, source, file type, expected volume, sensitivity, and owner. Separate original data from derived files and temporary working files.
- Estimate active storage. Calculate the volume needed while the project is running. Include raw data, cleaned data, analysis files, documentation, exports, and temporary processing space. A simple model is: active storage = raw data + working copies + derived outputs + documentation + processing allowance.
- Estimate backup and preservation storage. Apply the number of required copies or environments to the active volume, then add any preservation package or repository requirements. Do not assume that a synchronization folder is the same as an independent backup.
- Estimate people and services. Identify time for file management, quality checks, de-identification, metadata creation, repository preparation, access reviews, and closeout. Add specialist services only where the project actually needs them.
- Assign decisions. For every major activity, name the responsible role, approval point, system, and review date. A plan without ownership is difficult to implement.
Use the following decision table when selecting a management approach:
| Data situation | Planning decision | Evidence to record |
|---|---|---|
| Open, non-sensitive data | Use managed institutional storage during the project and a suitable repository for release. | Repository, file formats, license, metadata fields, and release date. |
| Human-subject or potentially identifiable data | Limit access, separate identifiers, document consent limits, and use approved secure systems. | Access roles, de-identification method, key location, review process, and sharing restrictions. |
| Large or frequently changing files | Define transfer, versioning, checksum, and backup procedures before collection begins. | Transfer method, naming convention, version rule, and recovery test. |
| Third-party or restricted data | Follow the data-use agreement and share only what the agreement permits. | Agreement owner, permitted uses, expiry date, and destruction or return procedure. |
| Data not suitable for public release | Share metadata, documentation, or a controlled-access version where permitted. | Reason for restriction, access request process, and responsible custodian. |
Inputs and assumptions
Use this reusable research data management plan template as a starting point. Replace bracketed text with project-specific information.
1. Data description
Project data: [Describe data to be collected, generated, or reused.]
Formats and volume: [List formats and estimated volume at collection, analysis, and closeout.]
Sources: [Identify participants, instruments, archives, databases, sensors, or collaborators.]
Data products: [List raw, cleaned, coded, derived, and final release files.]
2. Organization and documentation
File naming: Use a consistent pattern such as project_dataset_date_version_status. Use dates in one agreed format, avoid ambiguous names such as final_final, and document abbreviations.
Folder structure: Keep raw files separate from cleaned, analysis, output, and administrative files. Make raw data read-only when practical. Store a short README in each shared dataset folder.
Documentation: Record codebooks, questionnaires, calibration details, data dictionaries, processing steps, software versions, missing-data rules, and known limitations. For a literature review or evidence-synthesis project, preserve the search strategy, screening decisions, and extraction definitions alongside the dataset.
3. Storage, backup, and recovery
Working storage: [Approved system and project location.]
Backup: [Backup method, responsible role, and monitoring process.]
Recovery: [How a lost or corrupted file will be restored and how recovery will be tested.]
Prohibited locations: [Personal devices, unapproved cloud services, removable media, or other locations excluded by policy.]
4. Security and access
Classify each dataset before collection or transfer. State whether it contains direct identifiers, indirect identifiers, sensitive attributes, confidential business information, controlled data, or contractual restrictions. Describe role-based access, multi-factor authentication where required, encryption or secure transfer where appropriate, access reviews, incident reporting, and the process for removing a departing team member.
5. Sharing, licensing, and costs
Repository: [Institutional, disciplinary, general-purpose, or controlled-access repository.]
Release files: [Exact files, documentation, and metadata to be deposited.]
License: [License or access terms, subject to funder, institutional, consent, and third-party restrictions.]
Costs: [Storage, curation, de-identification, repository, preservation, and staff-time assumptions.]
6. Retention and deletion
State how long project records and research data will be retained, based on applicable institutional, funder, ethical, contractual, and legal requirements. Distinguish between the preservation copy, active working files, temporary exports, identifiers, and backups. Identify who authorizes deletion and how the decision will be recorded.
Worked examples
Example 1: Interview study
A project collects audio interviews, transcripts, consent records, coded analysis files, and a de-identified release dataset. The DMP should separate contact details and consent records from research data, restrict access to authorized staff, document transcription and redaction procedures, and explain whether the consent language permits sharing. The cost estimate should include secure storage, transcription time or fees, quality review, de-identification, documentation, and repository preparation. If the interviews cannot be shared, the project can still preserve a data dictionary, interview guide where permitted, analytic code, and a clear explanation of the access limitation.
Example 2: Computational research
A computational project creates source files, downloaded inputs, intermediate outputs, scripts, logs, model files, and final results. The plan should distinguish reproducible inputs from temporary caches, define software and dependency records, and specify where large files are stored. Estimate active volume by adding the input data, working outputs, and at least the copies needed for recovery. Record which files can be regenerated and which must be preserved. A repository deposit may contain code, a small demonstrative dataset, metadata, and instructions rather than restricted source data.
Example 3: Small student project
A student project uses a modest spreadsheet and survey responses. A proportionate DMP may be short, but it should still name the approved storage location, access permissions, backup method, file naming convention, sensitivity assessment, sharing decision, and deletion or retention date. Small scale does not remove the need to protect personal information or preserve enough documentation for another person to interpret the results.
When to recalculate
Revisit the plan at defined checkpoints rather than waiting for project closeout. Update it when the research question or collection method changes; new data sources or collaborators are added; volume exceeds the original estimate; a dataset becomes identifiable or more sensitive than expected; storage, repository, or preservation pricing changes; a repository changes its requirements; a funder, institution, ethics committee, or data-use agreement introduces a new condition; or the team begins preparing a deposit.
At each review, compare the estimate with actual storage use, staff time, access requests, incidents, and release work. Mark assumptions as confirmed, revised, or no longer applicable. Keep dated versions of the DMP and note who approved material changes.
For a practical final check, ask the project team to demonstrate four actions: find the current version of a file, restore a needed file from backup, explain who can access sensitive data, and prepare the intended sharing package without exposing restricted information. If any answer is unclear, revise the plan before the next collection or submission milestone. A concise, maintained DMP is more useful than a detailed document that no one consults.