Record research exclusions with clear reasons, supporting evidence and reconsideration conditions, so you avoid repeat work without hiding uncertainty.
Direct answer: Keep a short record containing the source, the question you were investigating, the specific reason you excluded it, and the evidence behind that reason. Distinguish a source that fails your criteria from one you cannot yet assess, and write down what would justify reconsidering it. Record consequential exclusions, not every irrelevant search result you pass.
A rejected source can disappear from your thinking while remaining in your search results. When you encounter it again, you either repeat the assessment or rely on a vague memory that it was somehow untrustworthy. Neither gives you a defensible reason for leaving it out.
The stronger approach is task-specific. A document may be excellent evidence about an earlier software release and unusable for instructions about your current one. Excluding it from this answer does not mean declaring its author unreliable about everything.
Use the source exclusion record
The source exclusion record is an editorial method for ordinary research, not a substitute for a formal systematic-review protocol. Its purpose is to preserve a decision you might otherwise have to reconstruct.
There is a relevant distinction in formal research: Cochrane asks review authors to document explicit reasons for excluding studies that readers might reasonably expect to see included. Its requirements apply to Cochrane reviews; you need not copy that entire process for a short technology briefing. The transferable principle is making important selection decisions inspectable. Cochrane handbook: searching for and selecting studies.
Start with the question written at the top of your notes. For example: “Can our current managed account export the attachments we need?” Your exclusion record should explain why a source cannot answer that question, rather than why it failed to impress you.
This is a narrower addition to keeping useful research notes: preserving the reasoning behind what you decide not to use.
Capture enough identity to recognise the source again
Before assessing a source, retain its title, author or issuing organisation, URL or document identifier, and the version or date shown. Add the date you assessed it. If the document has several editions, include the exact edition rather than the publication's general name.
You can do this in an existing document or spreadsheet. No new app is necessary. Keep a stable source identifier such as S14 so that your exclusion note and any quotations refer to the same item.
Avoid retaining unnecessary personal information in shared research notes. A public article's author and publication details may be relevant; a private email address copied from a downloaded file usually is not. Use the approved storage location for work material, and keep restricted documents out of public note collections.
Success means that a colleague can find the same source without rerunning your original search. If you have only a screenshot of a search result, the identity is incomplete. Locate the actual document before making a firm judgement about it.
Replace labels with observable reasons
“Biased”, “old” and “not relevant” are weak exclusion notes. They hide the test you applied. Instead, name the failed condition and point to the evidence.
| Weak note | Useful replacement | What could change the decision |
|---|---|---|
| Too old | Describes release A; our question concerns release C | Evidence that the relevant behaviour is unchanged |
| Wrong audience | Instructions require administrator access; readers are ordinary account members | A member-accessible alternative documented by the provider |
| Unreliable claim | Numerical claim has no identifiable dataset or method | Original data and an inspectable method |
| Not relevant | Measures attendance, whereas our question concerns task completion | Evidence connecting the reported measure to the required outcome |
The table is an editorial aid, not a universal scoring system. Different questions need different criteria. Historical research should not exclude old sources merely because they are old, and an account of personal experience should not be discarded because it is not a controlled experiment.
Also separate an author's commercial interest from the status of a particular claim. A vendor can be the correct primary source for a documented export format. The same vendor's enthusiastic description is weaker evidence of how much time your team will save.
Keep uncertainty out of the rejection bin
Use distinct statuses: excluded, awaiting evidence, and useful for another question. A paywall, inaccessible attachment or ambiguous product edition may leave you unable to assess a source. It does not prove that the source fails your criteria.
For an awaiting-evidence item, add a concrete next step: request the report through your library, find the official archived version, or ask the publisher which account tier the instructions describe. Assign a review date only when you genuinely expect to act on it.
For an excluded item, preserve a short evidence locator. “Eligibility paragraph, page 4” is better than a long copied passage when that paragraph contains the relevant condition. If a quotation is necessary, mark it as quoted and retain its source context.
Do not delete the source from a shared evidence set simply to make the remaining collection look cleaner. Keep your assessment reversible. If your question changes, review the old reason and append a new decision instead of silently replacing the history.
A worked screening example
Consider a student examining whether a particular digital submission process suits part-time learners. All quantities and times in this example are illustrative assumptions, not observed research results.
The student screens 28 sources and excludes nine: three concern a different process, two use an incompatible outcome, two describe a superseded configuration, and two repeat material already represented by an identifiable original. The arithmetic is 3 + 2 + 2 + 2 = 9 exclusions, leaving 19 sources for further assessment. This does not mean all 19 deserve inclusion.
Suppose writing each exclusion note takes 45 seconds. Nine notes take 9 × 45 = 405 seconds, or 6 minutes 45 seconds. Setting up the small record takes another four minutes. Initial overhead is therefore 10 minutes 45 seconds.
During the following week, six excluded sources reappear. Reassessing each from scratch would take an assumed three minutes: 6 × 3 = 18 minutes. Recognising the item and reading its note takes an assumed 20 seconds each: 6 × 20 = 120 seconds, or two minutes.
Avoided repeat work is 18 − 2 = 16 minutes. After the initial overhead, net time released is 16 minutes − 10 minutes 45 seconds = 5 minutes 15 seconds. No cash has automatically been saved. If none of the sources reappears, the record has cost time while still providing an audit trail.
The example supports selective recording, not elaborate administration. Keep a fuller record when an omission could change your recommendation or when several people are likely to encounter the same source.
Challenge convenient exclusions
My recommendation is to record a source's strongest usable contribution before excluding it when it challenges your preferred answer. This extra sentence makes it harder to dismiss inconvenient evidence through an overly broad label.
For example, a small study may not justify a general claim about all users, yet still identify a failure mode worth testing. Record both judgements. “Insufficient for prevalence estimate; useful example of recovery failure” is more accurate than “sample too small”.
Another reasonable approach is to keep only sources that pass a predefined protocol. That is appropriate when the protocol genuinely governs the work. For exploratory technology research, however, use separate roles for background, direct evidence and unresolved challenge. Do not pretend every useful source serves the same purpose.
Set up the record in your next 20 minutes
- Spend five minutes writing the current research question and the criteria that actually affect inclusion.
- Take ten minutes to document three consequential exclusions, each with identity, evidence and reconsideration condition.
- Use the remaining five minutes to ask whether someone else could reproduce those decisions from your notes.
If they cannot, improve the evidence locator before screening more material. Stop adding fields when they no longer change a decision. Revisit an exclusion when its stated condition changes, not simply because the source has appeared in another search.
Related guides
Frequently asked questions
Should I record every search result I ignore?
No. For ordinary research, record exclusions that required meaningful judgement or that another reader might reasonably question. A search result about an unrelated subject does not usually deserve a full entry. You can record the overall search query and date separately if you need to reproduce the search. The exception is work governed by a formal review protocol, where screening records and reporting requirements may be more extensive. Follow that protocol rather than this lightweight approach. The purpose here is to retain useful reasoning, not to create an administrative record of every click.
Can I reject a source because it is sponsored?
Sponsorship alone is not a sufficient explanation of why a specific claim is unusable. Identify what the source is being asked to establish and examine the supporting evidence. A sponsored tutorial may accurately describe a setting, while an unsupported claim about universal time savings remains unsuitable for a purchasing recommendation. Record the declared interest and the specific evidential limitation separately. If independence is an explicit requirement of your project, a conflict may make the source ineligible for that role. It may still help you locate the underlying documentation or an assertion that needs checking.
What should I do when a source becomes unavailable?
Change its status to unavailable or awaiting verification rather than concluding that it was false. Retain the bibliographic details, your access date, and the evidence you previously recorded. Look for an authorised copy through the publisher, library or archive, checking that it is the same edition. If you cannot recover a passage central to your argument, avoid treating your memory as a quotation or as fresh verification. You may need to narrow the claim or find independently accessible evidence. Unavailability is a limitation on checking the source, not a verdict on its content.
How do I handle disagreement with another researcher?
Keep both assessments until you have identified the exact criterion producing the disagreement. One person may be judging factual accuracy while the other is judging relevance to a particular population. Ask each to identify the passage that supports their decision, then compare those passages against the written question. If the disagreement remains consequential, seek another qualified reviewer or mark the item unresolved. Do not hide the disagreement by averaging confidence scores. When the project has a formal adjudication process, use it and preserve the final decision together with the reason that resolved the difference.
Can an AI assistant maintain the exclusion record?
It can help format decisions you have already made, provided the material is appropriate for the approved service. Give it the source identifiers, your reasons and the evidence locators, then check that it has not strengthened uncertainty into rejection or added unsupported credibility claims. Do not ask it to invent reasons for excluding sources that weaken your argument. For private or confidential research, use permitted storage and avoid uploading documents until their handling is authorised. A small manually maintained table may be quicker than reviewing a generated record, particularly when only a few sources are involved.
When should I delete old exclusion records?
Keep them while they explain decisions you still rely on, subject to your project's retention rules and any restrictions on the underlying material. An exclusion tied to a finished informal search may eventually have little value, whereas a published report's evidence trail may remain useful for corrections. Review the record before removing it and retain a minimal decision history if that is permitted and necessary. Do not keep confidential files indefinitely merely because storage is available. Equally, do not delete consequential notes to remove a visible disagreement or make a past recommendation look more certain than it was.
Sources and verification
- Cochrane handbook, chapter 4: searching for and selecting studies, consulted 11 September 2026 for the distinction between explicit exclusion reasons and formal review requirements.
- The supplied parent guide was read in the publication's local source. Its public route could not be retrieved during verification. The source exclusion record and numerical example are editorial constructions, not reported experimental results.
This article is practical guidance. Apply it in proportion to your tools, evidence, risks, and responsibilities.
