November 16, 2026

Year-End Research Cleanup: Files, Notes, Sources, and Decisions

A year-end research cleanup is the moment when a busy evidence base becomes reusable: the sources you can defend, the notes you can trust, and the decisions you made with your data. Many researchers enter the new year wi

Written byWisPaper TeamAI Research Workflow Team
Editorial illustration for Year-End Research Cleanup: Files, Notes, Sources, and Decisions

A year-end research cleanup is the moment when a busy evidence base becomes reusable: the sources you can defend, the notes you can trust, and the decisions you made with your data. Many researchers enter the new year with a library full of PDFs, a BibTeX file that may contain entries no one has checked, and scattered notes that later feel disconnected from the decisions they actually supported.

This guide diagnoses where your research base is strongest, where cleanup should start, and what a finished state looks like. It is built for clinicians and medical researchers who will soon be writing a protocol, updating a review, preparing a submission, or briefing a team. All examples use the same focused population and outcome throughout: an adult primary-care population for whom you are tracking the balance of benefit and harm of low-dose aspirin for primary prevention of cardiovascular events, with major bleeding as a safety outcome.

Start by identifying your current state in the table below, then follow the shortest route that moves your evidence base toward submission readiness.

What kind of year-end cleanup do you actually need?

Diagnosis matters more than order. Three researchers with the same messy desktop need different work. Use this matrix to classify your starting state.

Starting stateYou will recognize it byStart here
Source debtFiles are saved under many names; you cannot distinguish two versions of the same report; your BibTeX contains entries you have never verifiedHow do I make my source list defensible?
Note fragmentationYou have useful annotations, but they live in different tools, folders, or papers and do not track back to a claimHow do I turn scattered notes into a usable evidence base?
Decision driftYou remember the conclusion but not the records, rules, or screening choices that produced itHow do I reconstruct and record the decisions behind my synthesis?
Mixed overloadEverything is present but nothing is connected: files, notes, sources, and decisions are all entangledStart with source debt, then proceed through notes and decisions

If you last touched your evidence base more than six months ago, treat it as a new intake regardless of how organized it feels. The goal is not to preserve the old order but to make every retained item traceable to a purpose.

Where does the cleanup start for researchers preparing a submission?

Begin with identity, not with tidiness. The first question for every file is not “where does this belong?” but “which study or report is this, and is there another version of it?” Cochrane’s guidance on searching and selecting studies describes the same problem at scale: records must be linked to the same study so that you do not count one trial as two sources of evidence. DOI metadata is a practical tool here. A DOI resolves through the DOI system and can remain stable even when an object’s location or descriptive metadata changes, which makes it a reliable key for matching versions.

For a smaller year-end cleanup, work in three passes:

  1. Deduplicate records. Group files that represent the same underlying report. Do not merge before you are sure they are the same study: a protocol, a primary results paper, and a long-term follow-up are distinct reports that belong to one study. Keep the relationship, not one collapsed file.
  2. Verify reference fields. For each source you intend to cite, confirm title, author, publication, and DOI against registered metadata. Crossref’s REST API exposes scholarly metadata including DOI, title, author, publication, funder, license, and relation fields when deposited.
  3. Record what remains unverified. A source without a resolvable identifier is not automatically wrong, but it needs a note and a reason before it stays in a submission-ready set.

You can organize the same item under multiple project groupings without duplicating it. Collections in Zotero are designed for that: the same item can appear in a methods collection and a clinical-outcomes collection, while tags record topics, methods, or workflow status. This is a better model than nested folders, which force you to choose one primary location for a paper that actually serves several parts of your work.

When the aspirin reference set is processed, you should be able to answer three questions for every retained source: Which study does it belong to? Can I resolve its identifier? Why is it still in the set? A source that fails all three is cleanup output, not a research asset.

How do I make my source list defensible?

The defensible source list has four properties: every entry resolves, every retained record is linked to a study, every discarded record has a recorded reason, and the search can be repeated. Cochrane’s handbook frames this as search planning from eligibility criteria, using multiple sources, documenting the process, deduplicating records, screening in stages, and updating searches before publication. A year-end cleanup does not need to reproduce a full systematic review, but it should leave an audit trail that a colleague could follow.

The key distinction is between your eligibility criteria and your screening decision. Write the eligibility criteria once, as a short rule set, and apply it consistently. In the aspirin example, one reasonable rule set is:

  • Population: adults without established cardiovascular disease
  • Intervention: low-dose aspirin for primary prevention
  • Comparison: placebo or no aspirin
  • Outcomes: major adverse cardiovascular events and major bleeding
  • Study type: randomized trials for efficacy and safety; observational data only for long-term harms where clearly relevant

Then record, for each screened record, the decision that followed: included, excluded because of population, excluded because of duplicate reporting, or awaiting a field. This distinction between rules and decisions is what changes a pile of PDFs into a reusable evidence base.

BibTeX files deserve special treatment. A BibTeX entry can carry a reference that looks complete but does not correspond to a real academic record. TrueCite checks BibTeX entries against real academic databases to flag hallucinated references. At year-end, this kind of verification is worth doing early, before any team briefing or manuscript draft builds on the reference list.

Set a clear completion criterion: do not declare the source list clean until you have touched every entry and recorded its status. Partial cleanup is the most common source of false confidence in a pre-submission review.

How do I turn scattered notes into a usable evidence base?

Scattered notes are usually not a writing problem. They are a missing-index problem. The fix is to give every retained note one traceable relationship: to a source, to a decision, or to a question.

A useful note format for year-end medical research cleanup contains four fields: the claim or quotation, the source identifier, the outcome or section it informs, and the note’s status. “Status” is the field people most often skip. Mark each note as raw, contested, superseded, or synthesis-ready.

Here is the labeling discipline that makes later work fast:

  • Raw: copied or paraphrased, not yet connected to a decision
  • Contested: two sources disagree; keep both and record the disagreement
  • Superseded: later evidence has replaced it; do not cite alone
  • Synthesis-ready: linked to a source and an outcome, and phrased precisely enough to enter a draft

Do not rewrite all old notes. Triaging them is enough: keep notes that still map to the current question, mark those that have changed relevance, and archive the rest under the source they came from. The goal is coverage of your current decision set, not a complete history of everything you ever read.

For the aspirin brief, a note that reads only “aspirin reduces MI” is not usable because it lacks population, comparator, and magnitude. The synthesis-ready version is: In trial X, adults without CVD randomized to low-dose aspirin had a lower rate of nonfatal MI than control, with an accompanying increase in major bleeding; see source 14, outcomes table. That note can enter a draft. The original cannot.

When notes are spread across tools, consolidate by source rather than by topic first. Topics will later be reconstructed from the source-level index you build. Save or upload papers where they can be organized alongside your other materials; My Library provides that storage and organization workspace. Once papers are in one place, it becomes practical to ask questions against the uploaded set, which turns verification from a file-by-file hunt into a query-driven check.

How do I reconstruct and record the decisions behind my synthesis?

The most dangerous year-end debt is decision debt. A manuscript may be nearly finished, but the screening log, inclusion rules, and outcome prioritization may exist only in email threads or someone’s memory.

Reconstructing decisions means restoring the paper trail, not redoing the analysis. Work through three questions in order:

  1. What were the eligibility criteria, and where are they recorded? If the criteria changed mid-project, record the version history.
  2. What screening decisions followed? For each major exclusion, note the rule it applied to and the source it affected.
  3. What outcome hierarchy shaped the synthesis? In clinical work, distinguish the primary efficacy outcome from the primary safety outcome. In the aspirin example, cardiovascular events and major bleeding answer different questions and should never be folded into one “net outcome” without an explicit tradeoff decision.

Then perform a one-time reconciliation. Put your clean source list, your triaged notes, and your decision log side by side and ask: for every claim I intend to make, can I name the source and the decision path that supports it? A claim with no source is a gap. A source with no claim is clutter, not necessarily error. A decision with no record is a reproducibility risk.

If a full review of your materials is too large for manual reconstruction, use the Library QA workflow to ask questions based on papers uploaded to My Library or added from search results. This changes the cleanup task from “open every file” to “interrogate the set,” which is faster when the volume is high and the questions are well formed.

How do I know my year-end cleanup is finished?

A cleanup has a natural end state, and it is not the absence of mess. It is the presence of four artifacts:

  1. A deduplicated source inventory. Every retained source resolves to an identifier or carries an explicit note explaining its retention.
  2. A decision log. Eligibility rules, screening choices, and outcome priorities are recorded and dated.
  3. A triaged note set. Every retained note has a source link and a status marker.
  4. A claims-to-evidence map. Every claim you expect to make in the next quarter has a named source and decision path.

Any one of these can be audited in under fifteen minutes if it is clean. Use that as the test: could a colleague verify a single claim in your current manuscript without asking you for the underlying file? If yes, the specific artifact is ready. If no, keep that workflow open.

The distinction between evidence strength, applicability, indirectness, and limitations is where most medical year-end cleanups underperform. Researchers record the result but not the domain. For the aspirin example:

  • Evidence strength describes the design and conduct of the trials behind the claim.
  • Applicability asks whether the enrolled population reflects your clinic population and era.
  • Indirectness asks whether the outcome measured in the source is the outcome you care about, and whether the comparison is the one in your question.
  • Limitations are the remaining threats to a conclusion from any of the above.

Your cleanup is finished when each retained source has been sorted into these categories well enough that a later writer will not mistakenly treat a safety-adjacent observational report as a randomized efficacy result. Not every paper needs a full critical appraisal at year-end; every paper does need a category.

One supporting practice: update before you close. Cochrane recommends updating searches before publication, and the same logic applies to a year-end close: the set you declare clean in December should reflect the searches you actually ran most recently. Run or record your final search near the close date, and note the cutoff in your decision log. Run the search in the scholarly search interface, capture what was found, and then apply the same screening discipline to the new results.

What artifacts make this reusable next year?

A year-end cleanup should not require the same effort next December. The difference is whether this year’s work produced transferable artifacts or only a one-time tidy state.

The lowest-cost reusable artifacts are:

  • A short README in the project folder that names the population, outcome, and eligibility criteria
  • A decision log with dated entries and a search cutoff
  • A source inventory with identifier, status, and reason for retention
  • A tag convention for workflow status that you reuse across collections

The tag convention is the highest-leverage habit. If every future item is tagged with a status such as to-verify, contested, superseded, or synthesis-ready, next year’s cleanup becomes a status report rather than a re-archaeology. Record topics, methods, or workflow status with tags; do not rely on folder placement to carry meaning, because the same item will need different meanings in different project groupings.

Store the artifacts where the next researcher will actually look. For a team project, the README and decision log belong at the project level, not inside one person’s personal library. The Research Projects area is the workspace for creating a project that carries these materials. If the artifact is not visible from the project entry point, treat it as personal memory, not team infrastructure.

Finally, schedule a close-date checkpoint now. The single most useful year-end habit is not a new folder structure; it is a dated decision that says: the search cutoff is [date], the eligibility criteria are [version], and the evidence base is clean for the next [period]. Without that sentence, the cleanup has no boundary in time, and next year’s work will have to decide again what “current” means.

FAQs

No. Cleanup is a status assignment, not a purge. The goal is that every retained file is either linked to a source and a purpose, or explicitly marked as archived. Discard only records you can separate with confidence from the source they belong to. If you are unsure, keep the file and record that uncertainty in the source inventory.