01 · The Question
Are Your Research Data and Your Research Records the Same Thing?
Suppose your study produces 2,000 survey responses. Those responses are clearly data. But what about the questionnaire version, recruitment log, cleaning script, codebook, ethics amendment, notes explaining exclusions, preliminary analysis, and email documenting a consequential methodological decision?
They all relate to the research, but calling all of them “data” can obscure an important distinction.
Research data and research records overlap, and their formal definitions vary across policies. In many contexts, however, the research record is the broader concept.
03 · What You Need to Know
The Difference Is About Function and Scope, Not File Type
It is tempting to distinguish the concepts by format: perhaps spreadsheets are data while notebooks are records. That approach quickly fails.
A notebook may contain original observations that function as data. A spreadsheet may instead be an administrative tracking record. An audio file could contain research data, while another audio file could document a team meeting. The relevant distinction is what the material represents and how the governing policy defines it.
Research Data Are the Recorded Evidence Generated or Collected Through Inquiry
At a practical level, research data may include measurements, observations, survey responses, interview recordings or transcripts, images, sensor outputs, sequences, test results, computational outputs, coded observations, or other recorded material used as evidence in research.
Formal definitions are often narrower and policy-specific. Under the NIH Data Management and Sharing Policy, for example, “scientific data” means recorded factual material commonly accepted in the scientific community as being of sufficient quality to validate and replicate research findings, regardless of whether those data support a scholarly publication.
That definition matters specifically when determining what falls within NIH's Data Management and Sharing Policy. It should not automatically be substituted for definitions used by another organization.
Research Records Can Document the Larger Research Process
A research record can extend beyond the factual material being analyzed. It may include documentation showing what was proposed, authorized, performed, changed, analyzed, communicated, or reported.
The U.S. Public Health Service research-misconduct framework illustrates the breadth of the concept. Its definition of a research record includes the record of data or results embodying facts resulting from scientific inquiry and gives examples including research proposals, physical and electronic laboratory records, progress reports, abstracts, theses, oral presentations, internal reports, journal articles, and documents and materials needed to develop a complete record of relevant evidence.
That is substantially broader than a dataset.
| Material |
Could Be Research Data? |
Could Be Part of the Research Record? |
| Raw measurements |
Yes |
Yes |
| Survey responses |
Yes |
Yes |
| Interview recordings or transcripts |
Often |
Yes |
| Laboratory notebook |
May contain data |
Yes |
| Protocol or research proposal |
Usually not research data itself |
Yes |
| Data-cleaning script |
Usually documentation or analytical material rather than the observed data |
Yes |
| Preliminary analysis |
Depends on the definition |
Potentially |
| Published article |
Not ordinarily the underlying research data |
Can be part of the research record under some definitions |
| Communication documenting a consequential decision |
Usually not |
Potentially |
The repeated “could be” is intentional. Classification depends on context and on the definition being applied.
NIH Provides a Particularly Useful Example of Why the Terms Cannot Be Collapsed
The NIH Data Management and Sharing Policy explicitly excludes several materials from its definition of “scientific data,” including laboratory notebooks, preliminary analyses, completed case report forms, drafts of scientific papers, plans for future research, peer reviews, communications with colleagues, and physical objects such as laboratory specimens.
That does not mean these materials are scientifically worthless or should automatically be destroyed. It means they are not “scientific data” for the purposes of that particular policy.
This is a crucial distinction. A document can fall outside a data-sharing definition while still having substantial value as a research record.
Not classified as scientific data under a particular policy
The material falls outside that policy's definition of data.
Not worth preserving
A separate judgment governed by record-keeping requirements, evidentiary value, institutional policy, ethics, contracts, regulations, and the needs of the research.
Metadata Sit Close to the Boundary Because Data Need Context
Data rarely explain themselves. A column containing 0s and 1s is not useful unless someone knows what the variable represents, how it was constructed, and what those values mean.
NIH defines metadata as additional information intended to make scientific data interpretable and reusable, with examples including dates, variable construction and description, methodology, data provenance, transformations, and intermediate or descriptive observational variables.
Metadata therefore accompany and describe data. They also illustrate why a useful research record usually needs more than the bare observations themselves.
Raw Data, Processed Data, and Analytical Outputs May All Be Data
“Research data” should not automatically be equated with “raw data.” Research workflows commonly produce several data states.
Original observations may be cleaned, coded, transformed, aggregated, derived, or otherwise processed. Some derived material may itself become scientific data or an analytical output necessary to validate findings, depending on the study and applicable definition.
Preserving the relationships among these states is often more important than debating which folder deserves the label “data.” If a final analytical dataset cannot be connected to its source evidence or transformations, an important part of the research history has been lost.
A Physical Research Object Is Not Necessarily Research Data
Another common source of confusion is the difference between the object studied and the information recorded from it.
A biological specimen, archaeological artifact, material sample, or other physical object may be essential to research without itself being classified as data under a particular policy. NIH's definition, for example, excludes physical objects such as laboratory specimens from “scientific data.” Measurements, images, sequences, observations, or other recorded information obtained from such objects may nevertheless be data.
Research Records Matter Because Data Alone May Not Explain What Happened
Imagine finding a final dataset five years after a study. It contains 850 rows and 42 variables. Even if every value survives perfectly, you may still be unable to answer basic questions.
Which instrument version produced the variables? Were any cases removed? What does “99” mean? Which transformations created the derived scores? Which protocol amendment was in effect? Which analysis script generated the published table?
The broader record provides the documentary context needed to answer those questions. This is why preserving research records for research integrity cannot be reduced to preserving the final dataset.
Data Sharing and Record Retention Are Different Obligations
Researchers sometimes assume that everything they retain must eventually be shared. That is incorrect.
A data-sharing policy identifies material that should or must be made available under specified conditions. Record-retention requirements determine what must be preserved. Privacy, consent, confidentiality, intellectual property, contractual restrictions, security, and other considerations may further determine who can access retained material.
Consequently, a record may need to be retained but not publicly shared. Conversely, scientific data selected for sharing may require accompanying metadata and documentation so that other researchers can understand and use them.
Watch Out
Do not decide what to delete simply by asking whether a file qualifies as “research data.” Materials outside a data-sharing definition may still be required research records, evidence of compliance, or essential documentation of how the study was conducted.
04 · A Practical Example
One Survey Study Produces Both Data and a Larger Research Record
Hypothetical Example
A study of university students' use of generative AI
A research team administers an online questionnaire to 1,200 students and analyzes the relationship between AI use and several learning measures.
Research data
The participants' recorded questionnaire responses provide the principal factual material collected for analysis.
Metadata and documentation
A codebook defines variables, response codes, missing-value conventions, derived scales, and transformations needed to interpret the dataset.
Procedural records
The approved protocol, questionnaire versions, recruitment documentation, and amendments establish how the study was intended and authorized to operate.
Processing and analytical records
Cleaning scripts, exclusion logs, analysis code, and relevant outputs show how the source responses became the reported results.
Decision records
A dated note documents why the team changed the treatment of a problematic variable after discovering that one response option had been implemented incorrectly.
The 1,200 responses are research data. The broader collection documents the history of the study. Some items may themselves qualify as data under particular definitions, while others clearly serve a documentary rather than observational function.
Together, they make it possible to understand not only what evidence was collected but what happened to that evidence.
07 · A Quick Checklist
Before You Classify Something as “Data” or “Just Documentation”
Check the role and governing definition:
Identify which policy, institution, funder, sponsor, regulation, or other authority is using the term “research data” or “scientific data.”
Use the definition supplied by that authority rather than assuming that all organizations define data identically.
Distinguish the factual material collected or generated through the research from documentation describing how the research occurred.
Identify metadata and documentation necessary to make the data interpretable and reusable.
Preserve the relationship among source data, processed data, analytical datasets, code, and reported results where relevant.
Do not assume that materials excluded from a data-sharing definition can therefore be discarded.
Assess data-sharing obligations separately from record-retention and access requirements.
Verify privacy, confidentiality, consent, contractual, security, and other restrictions before sharing retained materials.
