What Is Data Minimization, and Why Does It Matter in Research?

Does Better Research Require Collecting More Data?

Researchers are often trained to anticipate future analyses, preserve potentially useful variables, and avoid discovering too late that an important measurement was never collected. Data protection introduces a seemingly opposing instruction: minimize personal data.

Does that mean researchers should deliberately build smaller, less informative datasets?

No. Data minimization is not an instruction to collect the fewest possible variables regardless of scientific consequences. It asks researchers to determine what personal data are actually necessary for a defined purpose and avoid processing more than that purpose requires.

Data Minimization Means Enough for the Purpose, but No More

Data Minimization Is a Purpose Test, Not a Data-Count Test

The principle begins with a defined purpose

You cannot determine whether data are excessive until you know what they are supposed to accomplish. A dataset that is excessive for one study may be entirely justified for another.

Under the GDPR formulation, personal data must be "adequate, relevant and limited to what is necessary" in relation to the purposes for which they are processed. European Commission guidance summarizes this as collecting and processing only the personal data necessary to fulfil the purpose.

The Philippine framework uses somewhat different language but reaches a related practical requirement through proportionality. The implementing rules of the Data Privacy Act state that processing should be adequate, relevant, suitable, necessary, and not excessive in relation to a declared and specified purpose. They further provide that only personal data necessary and compatible with that purpose should be collected.

These are jurisdiction-specific legal formulations, not a universal definition applicable everywhere. Researchers should identify the law governing their project. As a research-design principle, however, they illustrate why minimization begins by asking what the processing is actually for.

"Adequate" prevents minimization from becoming under-collection

The first word in the GDPR formulation is easy to overlook. Data must be adequate. The information must be sufficient to fulfil the stated purpose properly.

Suppose a study examines whether an intervention works differently across age groups. Age may be necessary. Removing it simply because it is personal data could make the planned analysis impossible or scientifically weaker. Minimization does not require that result.

Conversely, the same study may need age without needing a participant's full date of birth. The scientific requirement establishes what information must be available; minimization asks whether additional precision contributes enough to the purpose to justify processing it.

"Relevant" requires a rational connection to the purpose

A variable can be easy to collect and academically interesting while still having no meaningful relationship to the stated purpose. Researchers should be able to explain that relationship.

If religion, marital status, ethnicity, income, political views, exact location, or another personal characteristic is relevant to the research question, sampling design, confounding structure, equity analysis, eligibility criteria, or another defined function, its collection may be justifiable subject to applicable requirements. If it is included merely because demographic questions usually appear at the end of a survey, its relevance is much harder to demonstrate.

This is where deciding which personal-data variables the study actually needs becomes a methodological as well as a privacy decision.

"Limited to what is necessary" asks whether you can reasonably do with less

Necessity should not be reduced to whether information could conceivably be useful. ICO research guidance explains that, in contexts where necessity is required, processing must be more than useful or habitual. It should be a targeted and proportionate way of achieving the purpose, and researchers should consider whether the same purpose could reasonably be achieved by less intrusive means.

That can lead to several kinds of minimization. You might remove an unnecessary variable, reduce its precision, replace a direct identifier with a study code, separate identifiers from analytical data, limit access to particular team members, or disclose only the fields a collaborator needs.

Minimization is not only about the questionnaire

The most obvious application is deciding what to collect, but the principle concerns processing more broadly. Personal data can become excessive at later stages even when their original collection was justified.

Data minimization is therefore better understood as a continuing design constraint than as a one-time questionnaire cleanup.

Minimization can reduce the consequences of a breach

Every additional personal-data field creates another piece of information that may be exposed, misused, incorrectly shared, or accessed by someone who does not need it. The severity of that risk depends on the information and context, not simply the number of columns.

Imagine two stolen research files. One contains study IDs, age groups, and research scores. The other contains the same information plus names, exact dates of birth, home addresses, telephone numbers, personal email addresses, and several sensitive characteristics that the analysis never required. The security failure is serious in either case, but unnecessary information can enlarge the consequences of the incident.

Minimization therefore complements security rather than replacing it. You still need appropriate protection for the data you legitimately hold. Collecting less unnecessary information simply means there is less unnecessary information available to be compromised.

Anonymization and pseudonymization can support minimization, but they are different

Where research can be performed using genuinely anonymous information, personal-data obligations under GDPR-style frameworks may no longer apply to that anonymous dataset. Genuine anonymization, however, requires more than deleting names.

Pseudonymization replaces or separates identifying information so that data cannot be attributed to a person without additional information. It can substantially reduce risk and is explicitly recognized in research safeguards under GDPR-related provisions, but pseudonymized data remain personal data when reidentification remains possible.

The appropriate approach depends on the research. Longitudinal follow-up or record linkage may require a controlled ability to reconnect records to individuals. Other analyses may not need that capability at all.

Research provisions do not erase the minimization principle

Some data protection regimes provide particular provisions, safeguards, or exceptions for scientific, historical, statistical, or public-interest research. These should not be interpreted as permission to accumulate unlimited personal data.

For example, UK research guidance identifies measures respecting data minimization, including anonymization and pseudonymization where possible, as important safeguards for research-related processing. The Philippine implementing rules likewise tie research-related processing to safeguards, declared purposes, confidentiality, and proportionality in relevant circumstances.

The exact rules vary, so researchers should verify their own jurisdiction rather than borrowing an exemption from another country's law.

Minimization also applies when data move to someone else

A research team may legitimately hold a rich dataset while a statistician needs only selected variables, a transcription service needs only audio files, or a collaborator needs only de-identified measurements for a particular analysis.

Membership in the research team does not automatically establish a need for unrestricted access. Before giving another researcher participant-level information, determine which participant data that collaborator actually needs and what governance applies to the disclosure.

The same reasoning matters when transferring research data between institutions. Minimization can sometimes reduce the amount or sensitivity of information that must cross organizational boundaries.

Minimizing a Longitudinal Research Dataset Without Losing Follow-Up

The study still collects personal data because its design requires them. Minimization does not prevent longitudinal research. It asks the researchers to separate what is scientifically or operationally necessary from what was simply convenient to place in the same file.

Data Minimization Is Often Simpler Than Researchers Assume

Apply Minimization Every Time the Data Change Hands or Purpose

For practical research management, minimization can be turned into a recurring question: what is the least personal information that allows this particular task to be done properly?

Documenting these decisions can be valuable. If you decide that exact age, detailed location, or a particular sensitive characteristic is necessary, record why. Minimization does not require researchers to apologize for scientifically necessary data. It requires them to be able to distinguish necessary processing from accumulation by default.

That discipline becomes particularly important when selecting storage systems or sharing arrangements. Whether data can be placed in a commercial cloud service is a separate question from whether every field needs to be uploaded there in the first place.

How to Apply Data Minimization to a Research Project

Common Questions About Data Minimization in Research

Minimize What You Do Not Need, Not What the Research Does Need

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.