AI Glossary / Data engineering and architecture / Beginner
Research data archiving
Definition
Research data archiving is the managed preservation of research data and the information needed to understand and reuse them over time. It involves selecting, documenting and depositing material in an appropriate archive or repository, with clear access and stewardship arrangements. Archiving is more than keeping a spare copy: it aims to preserve an interpretable research resource beyond the immediate project.
What it means
A spreadsheet without variable definitions, methods or version information may survive physically while becoming scientifically unusable. An archive connects data with context, such as how observations were collected, what codes mean and which transformations were applied. Some archived data are openly available, while others require controlled access. The intended audience and permitted uses influence the deposit arrangement. A repository's preservation commitment and acceptance criteria must be checked rather than inferred from the word archive.
How it works
Researchers identify the material worth preserving and review ownership, permissions and sensitivity. They prepare data, documentation, metadata and any supporting code or instruments needed for interpretation. The repository checks the submission according to its policies and assigns or maintains references for discovery and citation. Preservation activities can include integrity monitoring, format management and documented changes. Researchers and repository staff clarify how corrections, new versions and access requests will be handled after deposit.
Three practical examples
- A survey team archives data with questionnaires, variable labels and an explanation of missing-value codes.
- An AI research group preserves an evaluation dataset, preprocessing code and version references needed to interpret published results.
- A historical research project deposits digitized material with provenance and access conditions so later users can assess its context.
Think of it this way
Think of leaving a laboratory notebook with labeled specimens rather than a box of unlabeled samples. Future users need both the material and the explanation of what it represents. The analogy is limited because digital data can depend on software, file formats and linked resources that change. Long-term usability therefore requires ongoing stewardship, not just a carefully packed deposit on one day.
Worked example
Suppose a dataset has a field Status with values 0, 1 and 9. The archived file contains 100 records: 70 coded 1, 20 coded 0 and 10 coded 9. Without documentation, a later user might treat 9 as a numerical status or a positive response. A codebook explaining that 1 means completed, 0 means not completed and 9 means missing changes the interpretation: completion is 70 of 90 known responses, about 77.8%, if that denominator is appropriate. The archive should preserve both the codes and the stated analysis choices, not silently replace them with a guessed interpretation.
Illustrative case study
In a fictional study, Pine Research prepares a deposit after publishing an analysis. A colleague attempts to reproduce one table from the proposed package and finds that a recoding step is missing. The team adds the transformation script, documents software requirements and labels the deposited version. Some identifying information remains outside the shared dataset under the approved access arrangement. The case illustrates testing reuse before deposit, without claiming that archiving alone guarantees reproducibility or permits unrestricted sharing.
Common misconceptions and mistakes
Archiving is not the same as routine backup. Backups help recover working systems, while an archive needs enough context and management for future use. Open access is not appropriate for every dataset. A persistent identifier does not itself preserve the underlying files. File conversion can affect values or functionality and should be checked. Keeping every intermediate file without selection can also obscure the authoritative material. Finally, a deposited dataset may support verification while still requiring specialized knowledge or unavailable external resources.
Business use
Plan archiving early so consent, documentation and file organization support the intended deposit. Choose a repository suited to the discipline, data type and access needs, then review its requirements. Preserve the relationship between raw data, derived datasets and published results. For AI projects, record dataset versions, labels, exclusions and processing steps, while respecting restrictions on source material. Budget time for curation and validation. A complete deposit package should allow a qualified future user to understand what can and cannot be concluded from the data.
People also ask
Must archived research data be public?
No. Controlled-access arrangements may be necessary, and the appropriate repository should support the required conditions.
Should code be archived too?
Often it is important for interpreting transformations and analyses, subject to rights and practical requirements.
Does deposit prove the research is correct?
No. Archiving preserves evidence and supports scrutiny; it does not validate every method or conclusion.
Check your understanding
A dataset is backed up in three places but has no explanation of its variables. Is that sufficient research archiving?
Answer: No. Redundant copies help protect the bits, but missing documentation can prevent meaningful interpretation and reuse.
Related terms
Learn next
Ask a colleague unfamiliar with a small dataset to interpret it using only the proposed archive package. Record every missing definition or dependency. Then compare the package with a suitable repository's deposit checklist and access requirements before finalizing the submission.
More in Data engineering and architecture
- Relaxed sequential - Relaxed sequential describes a parallel-programming approach in which parallel execution is an opportunity for acceleration rather than a requirement for...
- Relvar - A relvar, short for relation variable, is a variable whose value is a relation.
- Remote procedure call - A remote procedure call (RPC) is a way for a program to request execution of an operation in another process, often on another computer, through an...
- Retention period - A retention period is the length of time information or records must or should be kept under an applicable retention rule.
- Retrievability - Retrievability describes how readily a particular document can appear in the results produced by an information retrieval system.
- Reverse data management - Reverse data management starts with a desired result or condition and asks what changes to underlying data would produce it.
Sources
Sources: UK Data Service, “Depositing data: FAQs,” https://ukdataservice.ac.uk/help/deposit-data/faqs-help-depositing-data-uk-data-service/; UK Data Service ReShare, “Help guidance,” https://reshare.ukdataservice.ac.uk/help/index.html; UK Data Service, “Policy documents A–Z,” https://ukdataservice.ac.uk/help/general/policy-documents-a-z/; Library of Congress, “Data Integrity Management,” https://www.loc.gov/programs/digital-collections-management/inventory-and-custody/data-integrity-management/. Primary repository and institutional search excerpts consulted; full policy and editorial review remain pending. Repository examples do not establish universal deposit or legal requirements.
Previous: Resampling Next: Research hypothesis