Define the question and the use of the answer

Begin with a project description that a colleague outside the team can understand. What question is being investigated, what information is needed and how will the result be used? Scientific research can be commercially or privately funded and can arise outside universities, but the activity must reasonably be described as scientific. The label is not established merely by purchasing an analytics platform or appointing a research manager.

Distinguish an aggregate finding about a population from a score that changes how a named customer or employee is treated. Statistical processing involves aggregate, non-personal results and is not a route for taking individual measures or decisions from the inputs or results. A project that contains both activities should describe and assess them separately rather than placing the whole workflow under one convenient heading.

Use the current safeguards framework

The Data (Use and Access) Act changed the UK research safeguards framework. The relevant changes commenced on 5 February 2026. Articles 84A to 84D now address research, public-interest archiving and statistical purposes; older explanations centred on Article 89 and section 19 of the Data Protection Act need to be read against those amendments.

The current provisions require appropriate safeguards and constrain continued processing of identifiable information. They require minimisation safeguards, and the processing must not be likely to cause substantial damage or substantial distress to a person. The safeguards are not satisfied where processing is used for measures or decisions about a particular person, apart from the specified approved-medical-research exception. The practical lesson is to examine the project’s real consequences, not simply its scientific vocabulary.

Keep the lawful basis and sensitive-data condition visible

Research is not itself an additional general lawful basis. Identify the applicable Article 6 basis for the activity. Public bodies and private organisations may be in different positions. Where health or other special-category information is involved, assess the additional condition and associated domestic-law requirements rather than assuming that a research purpose is sufficient.

In health research, consent to participate, consent for confidentiality purposes and consent as a UK GDPR lawful basis are different concepts. HRA guidance explains these distinctions. Do not copy a participant consent form into the data protection assessment as though it answers every question. Record which requirement each consent or authorisation addresses, and what happens if participation ends.

Ask whether identification is genuinely necessary

For project design, consider whether the question can be answered using fewer variables, less precise dates or non-identifying information. Where linking records is needed, separate the linkage function from general analytical access where practicable. Pseudonymisation can reduce exposure, but it does not automatically take information outside data protection law.

Our recommendation is to justify each identifying field and identify a point at which it can be removed or transformed. A team may need identifiers briefly to validate or link records without needing them throughout the study. Test the proposed dataset with the analysts before collecting everything by default. Keep the additional information needed for re-identification separately controlled and document who can use it.

Treat new uses as decisions, not assumptions

A dataset collected for one purpose may be attractive for another project. Research-related compatibility provisions can be relevant, but they do not eliminate the need to examine safeguards and other obligations. A claim that future research might be useful is not a sufficient operational plan for indefinite unrestricted access.

For each new use, record the purpose, recipients, information required, applicable grounds and whether the explanation to people remains adequate. Check contracts, confidentiality and other restrictions as well as data protection. A supplier’s technical ability to combine databases does not establish authority to do so. Where the new project changes risk materially, revisit the DPIA and approvals before releasing the records.

Make transparency understandable to participants

People should be able to understand who controls the information, the intended uses, sharing and relevant rights. In research, an ordinary corporate privacy notice may not explain the particular study well enough. Equally, a long participant information sheet can obscure the essential information if it is written only for specialists.

We recommend testing the explanation with people who resemble the intended audience. Can they describe what participation involves, where information goes and whom to contact? Keep a record of the version supplied. If a specific exception to providing information is considered, assess its conditions and safeguards rather than assuming research always removes the obligation. Clear explanations support both reliable participation and accountable information handling.

Apply rights provisions individually

Research can engage specific exceptions or exemptions, but they are not a blanket refusal of access, correction, objection or deletion requests. The grounds and safeguards depend on the right and the circumstances. A general statement that research data cannot be deleted may misrepresent the actual position.

Create a request-handling route that involves the study team and the person responsible for data protection. They need to understand what can be identified, what has become genuinely anonymous and what effect the requested action would have. Document the applicable reasoning and communicate it clearly. Do not promise withdrawal effects that the technical design cannot deliver, or retrospectively change the explanation when a participant asks a difficult question.

Control release, retention and reuse

Before sharing findings, check whether small groups, unusual combinations or quotations could identify people. A table without names is not necessarily anonymous. Decide whether detailed records belong in a controlled environment rather than a public download, and review access arrangements for collaborators and external analysts.

Set retention and review decisions for source records, working copies, linkage material and outputs separately. Some research may justify longer retention, but the reason and safeguards still need to be explicit. Record what will happen when funding ends or the principal researcher leaves. A well-governed project should remain understandable after the original team has moved on, with a clear account of what was collected, why it was retained and what uses remain permitted.

Sources and further reading

Primary guidance behind this briefing. Requirements depend on the facts; check the current guidance before making a decision.

This is general guidance, not a legal opinion on your circumstances or a guarantee of compliance. An individual assessment may be needed, particularly where sensitive information, disputed rights or regulatory action is involved.