ACADEMIC WRITING GUIDE
How to Write a Research Data Management Plan for a Dissertation (With Worked Examples)
A dissertation can have a clear research question, a carefully designed questionnaire, and a suitable analysis method—and still run into trouble when files are misplaced, participant identities become visible, or nobody remembers what a spreadsheet column means. These are research-data management problems. They are easier to prevent before data collection than to repair after analysis.
A research data management plan (DMP) explains what data a study will use, how they will be organised and documented, who can access them, how they will be protected, and what will happen after the dissertation is submitted. This guide shows how to write a practical plan, with complete hypothetical examples for a quantitative survey, qualitative interviews, and retrospective clinic records.
Important: A student dissertation is not automatically subject to the same DMP template as a funded research project. Always follow your university, ethics committee, data custodian, funder, and applicable legal requirements. The examples below are models of writing and decision-making—not approved protocols or completed studies.
What is a research data management plan?
A DMP is a structured description of decisions about research information throughout its lifecycle: acquisition, documentation, active use, protection, preservation, responsible sharing, and disposal. It makes the work understandable to the researcher now and, where permitted, to authorized future users. The Digital Curation Centre's DMP checklist addresses data types, documentation, ethics, storage, access, sharing, preservation, responsibilities, and resources; the UK Data Service emphasizes practical controls while research is underway.
A DMP is not a list of software products, a consent form, or an analysis plan. “I will use Excel and save my files” tells a supervisor almost nothing about what will be protected, how errors will be traced, or how records will be retained. Nor should the plan promise open access to data that cannot safely or legally be shared.
How a DMP differs from related dissertation sections
| Section | Main question | What a DMP contributes |
|---|---|---|
| Data collection procedure | How will evidence be obtained? | Where files go, how identifiers are separated, and how transfers are recorded |
| Ethical considerations | How are participants and other people protected? | Practical access controls, storage, retention and sharing arrangements |
| Data analysis plan | How will the evidence answer research questions? | Stable source files, documented transformations, metadata and reproducible outputs |
| Data management plan | How will information be governed across the project? | One connected account of formats, roles, file organisation, security and end-of-project decisions |
These sections should agree. For the sequence of recruitment and instrument use, consult our data collection procedure guide. Your DMP should refer to that procedure rather than reproduce it.
Nine steps to write a defensible dissertation data management plan
Step 1. Identify every type of data and how it enters the project
Begin with a simple inventory. Distinguish raw survey responses, audio recordings, interview transcripts, administrative extracts, scanned field notes, consent records, linkage keys, data dictionaries, analysis scripts and tables. Record the approximate expected scale and whether files contain direct or indirect identifiers. Include third-party datasets and their access restrictions; using a dataset does not automatically give permission to redistribute it.
For example, an interview project might produce 15 recordings, 15 transcripts, a separate contact list, field memos and a coding framework. These files have different risks and need not share the same access rules. If the study is still a proposal, estimate rather than present your inventory as a completed fact.
Step 2. Choose formats that the research team can actually use and preserve
State the expected working formats and, when appropriate, a stable export format. A survey platform may export CSV alongside a codebook; a transcript might be edited in DOCX and retained as an approved text-based copy; analysis scripts could be stored as plain-text files. Explain specialized software when a format cannot be interpreted without it.
Do not claim CSV preserves every feature of a spreadsheet: formulas, styling and multiple sheets may be lost. Keep the original instrument and any essential context. Name formats that suit your actual data, not an impressive-looking list. When defining survey codes and missing-value labels, align the plan with our operational definitions guide.
Step 3. Design filenames, folders, version control and a codebook
Use a predictable structure such as raw/, working/, documentation/, outputs/ and a restricted identifiers/ area if identifiers must be held. The principle matters more than the exact folder names: distinguish source material from derived products, keep restricted information separate, and document changes.
A useful filename encodes the project, content, date or version: feedback_survey_clean_v02.csv is clearer than final_latest_2.csv. Avoid participant names in filenames. Define who may replace a file, how a new version is recorded, and how an earlier dataset can be recovered. Keep a data dictionary describing each field, allowed values, units, missing codes, and derivations.
| File or document | What it contains | Why retain it |
|---|---|---|
survey_raw_2026-11-10.csv | Unaltered export from an approved survey platform | Trace the origin of analysed values |
survey_clean_v02.csv | Versioned working dataset with corrections | Reproduce the analysis from documented inputs |
data_dictionary_v02.csv | Variable names, labels, allowed values and missing codes | Understand and validate each column |
cleaning_log.csv | Dates, decisions, affected fields and reasons | Explain what changed and why |
analysis_script_v01.R | Reproducible commands, if used | Connect the analysed dataset to reported outputs |
All filenames and dates in this table are illustrative. The working data and keys must live only in locations permitted by the study's approval and institutional policies.
Step 4. Specify approved storage, backups and secure transfers
Identify the actual storage provider or institutional service to be used, who controls it, and how data will be backed up and restored. “The cloud” is not a meaningful security description. State whether files are encrypted in transit and at rest where required, whether multi-factor authentication is used, and whether backups follow institutional controls. Do not presume a personal email account, removable drive, public share link or consumer AI service is approved for identifiable information.
Describe the data path: how survey exports move from the collection platform to approved storage, who can export them, and how temporary local copies are removed. If an offline collection device is authorized, explain how it is secured, transferred and cleared. A backup strategy is only credible if someone can recover a needed earlier version; explain who would test that ability where required.
Step 5. Allocate access and protect identities realistically
List roles rather than saying “only researchers.” For example, the student researcher may access pseudonymised responses, the approved supervisor may access an agreed restricted subset, a designated custodian may hold a re-identification key, and a transcription contractor may receive only the material authorized under an agreement. Do not imply that a pseudonymised dataset is anonymous: a separate key or distinctive combination of details may still make re-identification possible.
For human-participant projects, the ethical considerations guide provides the participant-rights and review context. The DMP must turn those commitments into operations: locations, access decisions, redaction rules, audit logs, and responses to accidental disclosure. The UK Data Service stresses that communication, technical protections, and access governance work together; names alone are not the only identifiers.
Step 6. Plan quality checks without rewriting the evidence
Record how the team will detect missing records, inconsistent codes, damaged files, duplicate exports, incorrect dates and unexpected values. Preserve a raw copy and record each justified correction in a change log. Plan version checks and codebook updates so that a chart can be traced back to its source dataset.
A blank response should not silently become zero; a large value should not be removed merely because it changes the result. Our research data-cleaning guide explains the detailed decisions. The DMP states how those decisions and resulting versions will be documented and protected.
Step 7. Address consent, licenses and any possible sharing restrictions
Describe what participants were or will be told about future reuse, what the ethics decision permits, and whether third-party terms restrict redistribution. If the project could share a responsibly de-identified dataset, say what would be considered for release, through which institution-approved route, and when permission and risk review would occur. If the data cannot be openly shared, explain the specific reason: for example, indirect identifiability, contractual access conditions, or limited consent.
FAIR does not mean every dataset must be public. Findable, Accessible, Interoperable and Reusable practices can coexist with restricted access and appropriate metadata. An open codebook, search instrument or aggregate statistics may be shareable even when person-level responses are not. The NIH's Data Management and Sharing guidance is a useful example of how sharing must be weighed against ethical, legal and technical constraints, but its funding rules do not automatically apply to every dissertation.
Step 8. State retention, archiving and disposal decisions
Explain which raw and derived materials must be retained, who owns them, where the approved archive will be, and who can authorize deletion. Never invent a universal retention period. University, funder, ethics and data-custodian rules may differ; applicable laws or contracts may require longer or shorter periods. Write the actual period and trigger only when confirmed, otherwise state whose policy will determine them before collection.
Deletion must include working copies, authorized local exports and backups where the system's retention rules permit; some institutional backups cannot be immediately erased on demand. Also state when identifiers or linkage keys will be destroyed or retained separately, and how that affects withdrawal requests. Do not promise participants instantaneous erasure if the approved technical process cannot deliver it.
Step 9. Assign ownership, review points and a realistic contingency
Name the role responsible for maintaining the file inventory, checking storage permissions, reviewing data-quality logs, documenting amendments and transferring the final approved archive. Plan review points—such as before collection, after a pilot, before analysis and before submission—without turning the DMP into a second project timeline.
Consider plausible failures: accidental shared-folder access, a corrupted export, a lost field recorder or an unexpectedly identifiable quotation. State who must be informed and which institutional reporting process applies. Update the DMP when actual collection procedures change; a proposal plan and the final dissertation account should not contradict what happened.
Worked example 1 — Undergraduate feedback survey
Hypothetical project: A dissertation investigates the association between perceived usefulness of written instructor feedback and academic engagement among second-year undergraduates at a selected university. A cross-sectional online questionnaire is proposed. The study does not test whether feedback causes engagement.
| Decision | Proposed, study-specific management arrangement |
|---|---|
| Files | Survey export, consent or information-page record if required, questionnaire, codebook, cleaning log, analysis outputs |
| Identifiers | No names or student numbers collected unless approved and necessary; platform metadata settings reviewed |
| Storage and access | University-approved restricted storage; student researcher and approved supervisor access only as authorized |
| Quality and versioning | Preserve original export; record duplicate checks, scale scoring, missing values and v01/v02 changes |
| Preservation and sharing | Retain under confirmed institutional policy; consider sharing questionnaire, codebook and aggregate tables; no automatic release of row-level data |
Complete model DMP paragraph — proposed research
“The study will generate structured questionnaire responses on perceived feedback usefulness and self-reported academic engagement, together with the final instrument, a variable dictionary, processing logs and analysis outputs. Before recruitment, the researcher will verify the university's required ethics and data-storage arrangements and confirm which survey-platform metadata can be disabled. Original exports will be retained unchanged in an institution-approved access-controlled location, while analysis will use versioned copies. A codebook will specify question wording, response categories, missing-value conventions and any scale-scoring procedures. The researcher will log all exclusions and corrections and will not substitute zero for unanswered questions. Access will be limited to roles authorized under the approved protocol. Results will be reported in aggregated form, with small categories reviewed for potential identifiability. Retention and secure deletion will follow the confirmed institutional schedule; no individual-level dataset will be made public without a separate assessment of consent, risk and permission. The final questionnaire, codebook and non-identifying summaries may be considered for approved sharing.”
Why this works: It names actual outputs, distinguishes originals from working copies, explains which data require special controls, and connects sharing to permission. It does not falsely claim the survey platform was configured or the ethics decision already obtained. If a study uses identifiable login tokens, the plan must describe how these are separated from response data, not call the survey completely anonymous.
Worked example 2 — Employee interviews about workplace communication
Hypothetical project: A master's dissertation proposes 15 semi-structured employee interviews at one organisation. The researcher will examine workplace communication experiences; identifiable workplace narratives create risks that are not solved by replacing names alone.
Data-flow example
| Stage | Example artifact | Distinct safeguard |
|---|---|---|
| Recruitment | Restricted contact/consent list | Kept separately from interviews; access limited to authorized recruitment roles |
| Recording | Encrypted audio in approved storage | Uploaded through approved method; no casual messaging or unrestricted cloud transfer |
| Transcription | Versioned transcript and corrections | Authorized transcriber, if used; confidentiality terms and controlled access |
| Analysis | Pseudonymised transcripts and reflexive memos | Contextual identifiers reviewed; approved researcher access |
| Reporting or sharing | Redacted quotations, methods description | Quotation-level disclosure review; no automatic publication of full transcripts |
Complete model DMP paragraph — proposed research
“The project will create interview audio, transcripts, a separate recruitment and consent record, interview-guide versions, reflexive memos, a coding record and dissertation excerpts. Subject to institutional approval and organisational permission, recordings will be transferred through an approved secure route into restricted university storage. Contact information and any participant-code key will be stored separately from the interview content, with access limited to personnel whose roles require it. Transcripts will be checked against audio and assigned study codes; names, workplaces, distinctive roles and other potentially identifying details will be reviewed in context before extracts are used. Changes will be recorded in a version history and a redaction log. Original audio, derived transcripts and analysis documents will be retained or deleted according to the approved institutional schedule and participants' stated consent conditions. Full transcripts will not be released publicly merely because direct names have been removed; any proposed reuse will require a separate disclosure-risk, consent and permission assessment. The final dissertation will include only approved, appropriately protected illustrative quotations.”
Why this works: The plan covers more than the storage of recordings. It protects the linkage between people and accounts, acknowledges indirect identification, and makes anonymisation a reviewed process. The UK Data Service warns that distinctive life events, job roles and locations can identify someone even when names are omitted. A qualitative project also needs an audit trail of interpretive changes, but those methodological details belong mainly in its analysis or trustworthiness sections.
Worked example 3 — Secondary clinic appointment records
Hypothetical project: A student proposes investigating patterns in missed clinic appointments using retrospective records. No patients will be contacted, but the data could still contain sensitive and indirectly identifiable information.
Risk-to-control matrix
| Risk or constraint | Proposed response | Evidence the student should obtain |
|---|---|---|
| Access to patient-level records | No access until ethics/governance determination and custodian permission | Written determination and data-use agreement where required |
| Excess collection | Agree a minimum necessary extract: date period, visit category and relevant appointment outcome fields | Approved data specification and dictionary |
| Re-identification | Exclude direct identifiers when possible; assess combinations of rare categories | Disclosure review and agreed cell-suppression rules |
| Unauthorized sharing | Restrict files to approved systems and users; prohibit unapproved transfer | Access list and permitted processing location |
| Data retention and publication | Follow custodian and university terms; release only approved aggregates or other permitted outputs | Retention agreement and publication review decision |
Complete model DMP paragraph — proposed research
“Before any access, the researcher will obtain the ethics or governance determination required by the university and the clinic's written authorization for secondary use. The data custodian and researcher will agree on the minimum appointment-level extract needed to investigate attendance patterns, including the eligible period, variable definitions, linkage permissions and prohibited uses. Records will be provided only through the custodian-approved transfer route to an approved restricted workspace; identifiable raw clinic files will not be copied to a personal device. Where feasible, direct identifiers will be excluded at source, and combinations of dates, services and rare appointment categories will be assessed for residual disclosure risk. The original approved extract will be preserved without editing; cleaning rules, derived fields, exclusions and output tables will be recorded in a separate processing log and data dictionary. Access will be limited to authorized personnel. Record retention, disposal, disclosure review and any external sharing will follow the university and clinic agreements; individual-level records will not be released without explicit authorization.”
Why this works: The plan recognises that ownership of a database does not automatically authorize research access. It also avoids an unrealistic promise of complete anonymity. If the data custodian prohibits exports or public sharing, the DMP must follow that restriction rather than claiming openness as a research-quality badge. A one-clinic dataset may also limit generalisability; explain that interpretive issue in the limitations-of-a-study guide, not by adding a duplicate limitations essay to the DMP.
Turn a vague promise into a usable plan
Weak version: “Data will be kept confidential on a computer and deleted after the research. The dataset will also be available to future researchers.”
This is internally unclear. It does not identify the data, permitted users, approved system, retention authority, deletion trigger or sharing conditions. It may simultaneously promise deletion and reuse without explaining which version remains.
Stronger version: “Original survey exports will be retained unchanged in the university-approved restricted workspace. The researcher will create dated working datasets and maintain a codebook and cleaning log. Access will be limited to roles approved under the study protocol. The applicable institutional retention period and disposal procedure will be recorded before collection. No participant-level file will be shared outside the approved team unless consent, disclosure-risk review and institutional permission support that use; non-identifying summary tables and the questionnaire may be shared separately if permitted.”
The stronger version is useful because a supervisor can check its claims and a researcher can follow the steps. Replace generic terms with named approved services and the confirmed retention rule when the institution supplies them. Never copy an illustrative procedure as though it has been approved or carried out.
An adaptable dissertation data management plan template
Use the following structure as a starting point, not a universal institutional form:
“This study will generate or use [data types, source and approximate scale] to answer [research questions]. Original data will be obtained through [approved collection or acquisition procedure] and stored in [named approved system] with [specified access controls]. Research identifiers and any linkage keys will be [excluded/separated and governed how]. Working files will follow [folder, naming and version scheme], with [codebook, metadata and correction log] documenting definitions and changes. Backups, transfers and incident responses will follow [institutional arrangements]. Data quality will be monitored through [relevant checks]. Retention, secure disposal, sharing and reuse will be determined by [actual university, ethics, agreement and funder requirements]. Responsibility for checking and updating this plan will rest with [role], including review at [relevant project milestones].”
Write only arrangements that are real, approved or explicitly proposed pending approval. Do not promise an exact retention period, unlimited confidentiality or public release without a verified basis.
Common mistakes and how to correct them
- Writing only “data will be secure.” Specify the approved location, access restrictions and handling procedure.
- Calling coded data anonymous. A code key or combination of attributes may still permit identification; describe residual risk accurately.
- Treating the DMP as the data-analysis plan. Document data governance rather than repeating each statistical test.
- Editing the only copy of raw data. Preserve an authoritative original and create traceable working versions.
- Using personal email or consumer cloud tools without approval. Confirm permitted services and data-transfer routes first.
- Inventing a retention period. Use the actual applicable policy, consent terms and data-use agreement.
- Assuming all datasets must be open. Responsible sharing may involve restricted access or only non-identifying outputs.
- Ignoring indirect identifiers. Roles, dates, rare events and location combinations can expose participants.
- Failing to document decisions. Without a codebook and processing log, a second researcher may not understand your results.
- Leaving responsibility unspecified. Say who manages permissions, backups, amendments and end-of-project disposition.
Frequently asked questions
Is a data management plan required for every dissertation?
Not necessarily as a separate document. Universities, funders and disciplines vary. Even where no standalone DMP is required, the methodology and ethics materials still need credible arrangements for handling data. Check the dissertation handbook and approved research process.
Is a DMP the same as a data analysis plan?
No. A DMP governs the files, documentation, protection, access and lifecycle of research information. A data analysis plan specifies how that information will be used to answer research questions. They should be consistent but not duplicate one another.
Should I write my DMP before collecting data?
Yes, usually during research planning, with updates when approved procedures or practical needs change. A completed dissertation should distinguish original intentions from what actually happened; do not silently turn future-tense promises into past-tense claims.
Can I keep research files on my personal laptop?
Only if your institution, approval conditions and data-owner agreement permit it, with required safeguards in place. Sensitive records often require approved managed storage rather than personal devices. Do not infer approval from having a password on the laptop.
Does removing names make interviews or clinic data anonymous?
Not necessarily. Linkage keys, unusual roles, small groups, combinations of dates or distinctive quotations can allow identification. Assess the realistic disclosure risk and permitted access in context.
Should I share my entire dataset after finishing my dissertation?
Only when the study's consent, ethics decision, applicable law, licenses and institutional policies permit it. Some projects can share anonymised or controlled-access data; others can share only a codebook, instrument, aggregated results or a description of restricted access.
Final data management plan checklist
Before submitting, confirm that you can answer yes to each relevant question:
- Have I identified all important raw, processed, identifier and documentation files?
- Is the planned data collection or acquisition authority clear?
- Have I selected workable and appropriately preservable formats?
- Are file naming, versioning and source-data preservation explained?
- Will a codebook, metadata and a decision log accompany the data?
- Are storage, backups, secure transfers and recovery procedures specific?
- Are permitted users and access controls matched to the actual risks?
- Have I addressed direct and indirect identification rather than names alone?
- Are quality checks and transformations auditable?
- Have I checked consent, licenses, data custodian and institutional restrictions?
- Are retention and destruction decisions supported by genuine requirements?
- Is any possible sharing lawful, ethical and accurately described?
- Is someone responsible for updating and monitoring the plan?
- Have I used proposal or completed-study tense truthfully throughout?
Final takeaway
A good dissertation DMP makes research manageable and defensible. It explains what information exists, how it moves through the project, how decisions are documented, who may access it, and what happens when the study is finished. It does not need a long catalogue of software. It needs a small number of specific arrangements that genuinely match the research design and approval conditions.
If your university requires a special template, use that template first. If you need help aligning your data management, methodology and dissertation writing, you can request academic guidance from Acadelyra while remaining responsible for your study's actual procedures, approvals and evidence.
References and further reading
- Digital Curation Centre. Checklist for a Data Management Plan. https://www.dcc.ac.uk/dmps
- UK Data Service. Managing data during the project. https://ukdataservice.ac.uk/learning-hub/data-producer-support/managing-data-during-the-project/
- UK Data Service. Documenting and describing data. https://ukdataservice.ac.uk/learning-hub/data-producer-support/preparing-data-for-sharing-and-reuse/documenting-and-describing-data/
- UK Data Service. Anonymising data and protecting participants. https://ukdataservice.ac.uk/learning-hub/data-producer-support/preparing-data-for-sharing-and-reuse/anonymising-data-and-protecting-participants/
- National Institutes of Health. Writing a Data Management and Sharing Plan. https://grants.nih.gov/policy-and-compliance/policy-topics/sharing-policies/dms/writing-dms-plan
- UK Research and Innovation. Guidance on best practice in the management of research data. https://www.ukri.org/publications/guidance-on-best-practice-in-the-management-of-research-data/