Archiving Research Records: What to Keep and How

Archiving research records means freezing the finished, documented evidence behind your findings, then storing it in a trustworthy place with enough context that a stranger can understand it years later. That takes four decisions: what to keep, how to describe it, where to deposit it, and how long to hold it. This guide walks through each, then shows how the same records feed a transparency report that readers can actually check.

Hellenismwatch

Ground Research in Credible Sources

Hellenismwatch monitors developments affecting the Greek diaspora, with claims supported by credible sources and analysis independent of party or government influence.

Visit Hellenismwatch

TL;DR

Archive the records needed to verify and reproduce your published claims, frozen in a fixed state. A backup of working files is a different thing.
Document who, what, where, when, how and why, and keep an audit trail of every change back to the original.
Prefer repositories that issue persistent identifiers and are open about governance, retention and continuity.
Retention is commonly at least 5 to 10 years, but funder, legal and ethical rules can override it.
A transparency report should name sources, methods, funding, limitations and corrections.

What counts as an archived research record
Document context, provenance and integrity
Digital archiving standards and repository criteria
Retention, access, privacy and disposal
What a transparency report should disclose
Scrutinizing contested claims: diaspora and geopolitics
Our perspective
Contribute to a documented record
FAQ
Sources

What counts as an archived research record

Start with the distinction that trips up most teams. Backup protects files you are still changing. Archiving preserves what is finished. Ghent University’s guidance draws the line this way: preserved data, or milestone versions of it, are frozen and inactive, unlike the mutable data you handle during a live project. VU Amsterdam’s handbook says the same from another angle: archiving is putting data in a safe place after the research is finished, once the data are in a fixed state and no longer change.

Purdue University Libraries describes archived data as data no longer actively used but kept for future reference, compliance or institutional purposes. Preservation goes further. It covers keeping data accessible, usable and intact over time, including documentation and metadata, file-integrity checks, a plan for obsolete formats, and context about how the data were created.

A practical record set

VU Amsterdam describes a complete dataset as more than the numbers. It includes raw or cleaned data, project documentation, a codebook or protocol, a logbook or lab journal, and the software needed to open the files. For humanities, social science and policy research, translate that into your own materials:

Source documents or faithful copies: scans, photographs, transcripts, downloaded web pages, with the date and place you retrieved them.
Your working notes: coding schemes, interview guides, translation choices, search strings.
Cleaned datasets, tables and the scripts or spreadsheets that produced your figures.
The published output and each earlier version that readers may have cited.
Correspondence or permissions that affect how the material can be reused.

How much to keep, and for how long

Not everything deserves a permanent home. Ghent suggests keeping the data needed to verify and reproduce published claims for at least 5 years, and longer when reuse potential is high. The ERI Research Toolbox lists common selection criteria: data required by funder, publisher or institutional policy, legal or ethical obligations, unique data that cannot be easily regenerated, data likely to be reused, and data of great value for society. When unsure, keep what cannot be recreated.

Backups still matter, in their place

While you work, protect files with several copies. The University of Newcastle’s toolkit recommends three backup copies, so data survive even if one is damaged or lost. Once the project closes, the archived version is the one you freeze and deposit.

Document context, provenance and integrity

A file without context is a liability. A reader who finds a spreadsheet with no explanation cannot tell where it came from or whether it was altered. Good documentation answers that before anyone asks.

Document at two levels

MIT Libraries recommends documenting at the project level (who, what, where, when, how and why) and at the data or code level (the attributes in detail). Do it while you work, with readme files or an electronic lab notebook, recording file names and formats, how the files are organized, how the data were generated, and any alteration or processing.

The UK Data Service’s best-practice guide gives a useful checklist of what good documentation holds:

The context of collection: project history, aims, objectives and hypotheses.
Collection methods: sampling, instruments, scale, temporal and geographic coverage, and secondary sources.
Dataset structure.
Validation and quality assurance procedures.
Changes made to the data over time.
Conditions of access and use, including confidentiality.

Capture provenance and a chain of custody

Provenance is the history of a record: when it was created, by whom, with which tools and processes, and what has changed since. The CUrating for REproducibility guide notes that FAIR principles call for provenance as standardized, machine-readable metadata, and names the W3C PROV data model as one formal standard. It also gives honest advice for newcomers: formal capture can be daunting and slow for manually curated material, so begin by understanding which categories of provenance detail matter in your field and set up routines that keep capturing them.

The UK Data Archive’s preservation policy shows what rigor looks like. It documents the complete chain of custody of each collection through metadata. Only the original version counts as an integral copy of what was deposited. Preservation and dissemination versions are treated as authentic when an audit trail of all alterations leads back to that original. You can adopt the same principle at any scale: never overwrite the original, and log every derived copy.

Authenticity has three parts

The National Archives of the Netherlands defines authenticity as reliability, integrity and usability, and it works to guarantee authenticity from the moment of ingest rather than repairing doubts later. The International Council on Archives makes a related point about electronic records: the people who create records should be helped to produce and keep records that are authentic, reliable and preservable, across the whole life cycle. In practice, that means archiving decisions begin when you create the file, not when the project ends.

Choose formats that last

ETH Zurich’s research data guidelines recommend well-documented, non-proprietary file formats for long-term accessibility and reuse, and ask that the structure and processing steps of all research data be documented digitally. Plain text, CSV, PDF/A and open image formats are safer bets than obscure proprietary software.

Digital archiving standards and repository criteria

You rarely need to build an archive yourself. You need to choose a place to deposit that you can trust, and standards are how you judge that.

The reference model: OAIS

The Open Archival Information System model is the common vocabulary. According to the Dublin Core Metadata Initiative’s summary, it is published jointly as CCSDS 650.0-M-3 and ISO 14721:2025. It defines three package types (Submission, Archival, Dissemination) and six functional entities: Ingest, Archival Storage, Data Management, Administration, Preservation Planning and Access. The related ISO 16363 sets criteria for auditing and certifying trustworthy digital repositories on OAIS principles. The Dutch National Archives also builds its policy on ISO 14721 and ISO 16363, alongside the records-management standards NEN-ISO 23081 and NEN-ISO 15489-1.

What a trustworthy repository does

Two older and newer documents agree on the essentials:

Science Europe lists persistent and unique identifiers first among its minimum criteria, so data can be discovered, cited, retrieved and versioned. It also expects transparency about mission, scope, preservation policies, governance, financial sustainability, retention period and a continuity plan.
OCLC’s paper on trusted digital repositories says such a repository accepts responsibility for long-term maintenance, shows fiscal responsibility, follows accepted standards, and has policies and performance that can be audited. Its information must be independently understandable to the designated community, without needing the original experts to explain it.

That last test is worth applying to your own deposit. Ask whether a graduate student in your field, ten years from now, could understand the record using only what you deposited.

A quick selection checklist

Question	What you want to see
Does it issue persistent identifiers?	A DOI or similar for every deposit and version
Is governance public?	Stated mission, funding model and continuity plan
Are formats and retention stated?	Published preservation policy and retention period
Can access be controlled?	Embargo and restricted-access options
Is it auditable?	Alignment with OAIS or a certification scheme

Retention, access, privacy and disposal

Setting a retention period

There is no single global number. ETH Zurich says data deposited in a FAIR repository should, as a rule, be retained at least 10 years unless legal, funder or community rules require otherwise, with the same 10-year minimum for other data where technically and economically feasible. VU Amsterdam asks for at least 10 years after the last publication based on the dataset. Ghent points to at least 5 years for verification data. The ERI toolbox notes funder, publisher and institution policies usually require at least 5 or 10 years after project end.

The safe method: find the longest of your institution’s, funder’s, publisher’s and legal requirements, then add time if the material is unique. Write the period into the record’s metadata so the next steward does not have to guess.

Not everything can be open

ETH Zurich frames open research data as an ideal to strive for that may not be achievable in every case, and allows justified exceptions. Consent terms, personal data, security and cultural sensitivity are common reasons to restrict access. Record the reason and the conditions for access in the documentation, as the UK Data Service checklist above suggests. For records involving named individuals, check consent forms and your institutional review board’s rules before depositing.

Appraisal is a judgment call

Ghent acknowledges that appraisal of research data is still an evolving field. Its emerging criteria include legal and ethical requirements, funder or institutional policy, reuse value, scientific, historical or cultural significance, uniqueness, and whether the benefits outweigh the cost of preservation. Document your appraisal decisions, including what you chose to discard, so disposal is a recorded choice rather than a silent loss.

What a transparency report should disclose

Archived records are the raw material. A transparency report is how you tell readers what is in them and what to make of it. Whatever your discipline, the major reporting guidance converges on the same items.

Disclosure	What readers need	Guidance
Sources and design	How evidence was found, selected and assessed	STROBE checklist
Funding and support	All grants, equipment and funders’ roles, or a statement that none was received	COPE principles
Interests	Relevant interests and how conflicts are managed	NHMRC guide
Data and protocols	Whether they are shared, where, and how to reach them	APA TOP guidance
Limitations	Potential bias and imprecision, in direction and magnitude	STROBE
Corrections	A policy for corrections, retractions and expressions of concern	COPE

The NHMRC guide, published by Australia’s health research council, also asks that communications identify host institutions and all sources of support. A consensus-based transparency checklist from Aczel and colleagues frames the exercise as a chance for researchers to explain their choices, which is the right attitude: a transparency report is an argument for your credibility, not a confession.

For text-based sources, such as speeches, newspaper articles, institutional statements and letters, the 2022 principles in Perspectives on Politics ask researchers to explain the production context of each source, including who created it and why, and how it supports the claim.

Scrutinizing contested claims: diaspora and geopolitics

Research on diaspora communities, minority representation and foreign influence is exactly where records get contested. Claims are repeated, sources disappear, and a figure quoted once is soon quoted as settled. Good archiving counters that.

One research guide on archives and contested memory argues that durable records should include version history, correction notes and clear source links, so errors or omissions do not circulate as if unchanged. It suggests reporting can separate verified records, attributed claims, analytic inferences, disputed accounts and unknowns, and that an archive should preserve testimony, official records, court findings, scholarly interpretation and acknowledged uncertainty rather than choosing one convenient story.

A workable habit for analysts:

Save a dated copy of every web source and note the retrieval date.
Label each statement in your notes as verified, attributed, inferred or disputed.
Keep contradictory sources in the archive, with a note on why you weighed them as you did.
Log every correction publicly, with what changed and when.

Our perspective

We built Hellenism Watch around one conviction: claims about the Greek diaspora and the wider Hellenic world are only as strong as the records behind them. In a field full of rumor, advocacy and partisan framing, a source trail is the only defense an analyst has.

That is why we stay independent of governments and parties, and why every claim we publish is tied to credible sources. Archiving is not an administrative afterthought for us. It is how a reader, a rival scholar or a skeptical official can check our work without trusting our word.

We also believe no single researcher can document a region alone. Scholars who know a community, a language or an archive firsthand hold evidence that deserves to be preserved and cited properly. Open collaboration, with shared standards for provenance and corrections, is how a national perspective on Greek identity and strategy can rest on evidence rather than impression.

Alexios

Contribute to a documented record

Hellenism Watch is an independent, non-profit initiative that monitors the state, threats and opportunities of the Greek diaspora across the Balkans and Eastern Mediterranean. Our publications and articles are free, and each claim is tied to sources. You can read the work on the Hellenism Watch site.

Hellenismwatch

If you research diaspora institutions, foreign influence or Greek representation and want your data and analysis documented alongside a wider network of scholars, see our fellowship for researchers and experts.

FAQ

What is the difference between archiving and backing up research records?

A backup is a working safeguard for files that are still changing. Archiving preserves finished records in a fixed state, with documentation and metadata, for future reference and verification. Most guidance recommends doing both: multiple backups during the project, then a frozen archive afterward.

How long should I keep research records?

It depends on your institution, funder, publisher and applicable law. Published guidance commonly points to at least 5 years for verification data and at least 10 years for deposited datasets, with longer periods for unique or highly reusable material. Check your institution’s records policy and use the longest requirement that applies.

Which digital archiving standards should I know?

Start with OAIS (ISO 14721), the reference model for digital archives, and ISO 16363, which sets criteria for auditing trustworthy repositories. In practice, you mostly need to choose a repository that follows them, issues persistent identifiers and publishes its preservation policy.

What should a transparency report for research include?

It should state how sources were found and assessed, who funded the work, any relevant interests, whether data and protocols are available, the limitations and possible biases, and how corrections are handled. For text-based sources, explain who produced each source and why.

Do I have to make archived records public?

Not always. Open data is a goal with justified exceptions, such as consent limits, personal data and security concerns. Where access is restricted, record the reason and the conditions for access in the documentation. Ask your institution’s data steward or research ethics board when personal data are involved.

Archiving Research Records: What to Keep and How