Three years after publication, a reader queries an odd data point in Figure 4. The lead author has moved institutions, the old laptop has been sold, and the only copy of the file sits on an external drive nobody can find. There was no fraud, but the group cannot demonstrate that they did things properly. In research integrity, being unable to show your work often has consequences uncomfortably close to having done it wrong.
Record-keeping and data retention are the least glamorous part of research. They are also the only part that can answer “what did you actually do?” years after memory has faded.
What a research notebook is for
- Reproducibility: you or someone else can repeat exactly what was done.
- Evidence: in disputes over ideas, authorship, intellectual property or suspected misconduct.
- Thinking: writing down what you plan and why, before you do it, protects you from retelling the story after seeing the results.
This applies to social research too. Field diaries, qualitative analysis decision logs and survey data-cleaning logs are all research notebooks.
What a good notebook entry contains
- Date, person, project code.
- Today’s aim in one sentence.
- Materials, equipment and software versions; for surveys, the questionnaire version.
- What was actually done, including deviations from the plan (“incubated 45 minutes, not 30, because the centrifuge was busy”).
- Raw results, or the exact path to the results file.
- Observations, problems and ideas for next steps.
The most common gap is not missing results but unrecorded failed experiments. Failures explain why the final protocol looks the way it does, and they show you did not simply pick the best-looking run.
Paper or electronic notebook?
| Criterion | Paper notebook | Electronic lab notebook (ELN) |
|---|---|---|
| Quick notes at the bench or in the field | Excellent | Depends on the device; gloves and connectivity get in the way |
| Searching years later | Hard | Good |
| Links to data files | Path written by hand | Direct attachments and links |
| Audit trail of changes | Single strike-through, initials, no torn pages | Automatic version history |
| Sharing and handover | Needs scanning or copying | Built-in permissions |
| Main risks | Loss, water, fire | Vendor lock-in, export formats |
Many groups combine the two: handwritten notes on the spot, photographed at the end of the day and filed in the ELN alongside the data. For paper, the old rules still hold: numbered pages, ink, no blank pages, and corrections made with a single line so the original remains legible.
Surveys and qualitative work
No test tubes does not mean no notebook. Three kinds of record are worth keeping:
- A data-cleaning log: how many responses were removed, why, and under which rule (for example, “completed a 60-item questionnaire in under four minutes”). Reviewers ask about this number constantly.
- A field diary: interview dates, settings, recording problems and first impressions written straight after each session.
- An analysis decision log: when two codes were merged, when a theme was renamed, and why.
File names and folder structure
A simple convention prevents most “which file is the real one?” moments:
- Start file names with an ISO date so they sort themselves: 2026-03-12_EXP05_sampleA_raw.csv.
- Keep three folders: raw (read-only, never edited), processed (generated by code) and output (tables and figures).
- Ban “final”, “final2” and “final-really”. Use version numbers or let version control handle it.
- Put a README at the project root explaining the structure and abbreviations.
Back up with the 3-2-1 rule
Three copies, on two different types of media, one of them off-site. In practice: the working copy on your machine, one on institutional storage or an institution-approved cloud service, and one on an external drive kept at home or in another office. Test a restore at least once a term; a backup that has never been restored is just a hope.
Data containing participants’ personal information must follow your approved ethics protocol: encryption, a separately stored key linking codes to identities, and no personal cloud accounts.
How long to keep it
There is no single figure that applies everywhere. Many universities and funders require raw data and research records to be kept for a number of years after publication, and clinical research usually for longer. Find your institution’s and funder’s data management policies and write the retention period into your data management plan at the start of the project. If you cannot find a rule, keep everything until all papers from the dataset are published and several years have passed, then consider disposal through a documented process.
Handing over when you leave
This is when data is most often lost. A minimum handover list:
- Notebooks (paper originals stay with the group; keep scans if permitted).
- Raw, processed and output folders plus analysis code, with a README.
- A list of samples and their physical locations: which freezer, which shelf.
- A map linking each paper, figure and table to its source files.
- A written agreement on which data you may continue to use after leaving.
Something to try this week: take your most recent paper and see whether you can locate the raw file behind every figure within 15 minutes. Any figure you cannot trace is the first thing to fix in how you record and store your work.
Câu hỏi thường gặp
What should be recorded in a research notebook?
Date, person, aim, materials and software versions, what was actually done including deviations, results or file paths, and observations. Failed experiments should be recorded too.
Is an electronic lab notebook better than paper?
Paper is quicker at the bench, while electronic notebooks excel at search, file links and audit trails. Many groups write by hand and photograph pages into an ELN at the end of the day.
How long should research data be kept?
It depends on institutional, funder and field requirements; clinical research usually has longer periods. Check the official policies and record the retention period in your data management plan from the start.
What is the 3-2-1 backup rule?
Keep three copies of your data on two different types of media, with one copy stored off-site. Test restoring from backup regularly to make sure it works.
Can I take my research data with me when I leave a university?
That depends on institutional data ownership policies and any agreements in place. Hand over everything to the group and get written agreement on which data you may keep using.
