November 16, 2026

Post-Publication Research Archive: Preserve the Evidence Trail

Finishing a paper is rarely the end of the work. The week after submission is when the evidence trail actually starts to matter: an examiner asks which database you searched on which date, a supervisor returns comments t

Written byWisPaper TeamAI Research Workflow Team
Editorial illustration for Post-Publication Research Archive: Preserve the Evidence Trail

Finishing a paper is rarely the end of the work. The week after submission is when the evidence trail actually starts to matter: an examiner asks which database you searched on which date, a supervisor returns comments that reference a source you can no longer find, or you open the folder from last term and find five versions of the same PDF with no record of which one you cited. A post-publication research archive is the answer to all of those moments. It is the organized set of search records, deduplicated source lists, screening decisions, and stable links that let you reconstruct exactly how a claim entered your manuscript.

This guide gives you a practical workflow for building that archive, whether you are closing out a thesis chapter, responding to supervisor feedback over a break, or setting up the next project so that January does not become another archaeology dig through Downloads.

Where Is Your Evidence Trail Breaking?

Before adding another folder or spreadsheet, identify the specific failure point in your current system. Most researchers have one of three starting states.

The finished paper with no process record. You have the manuscript and a reference list, but no record of the searches, databases, or screening decisions that produced it. If an examiner asks how a source was selected, you cannot answer from your files.

The working mess. You have the sources, but they are scattered across email attachments, browser downloads, and handwritten notes. The same paper exists in four places with four different filenames.

The organized-but-fragile setup. You use folders or a reference manager, but every record connects only to a local file path. If you move machines, switch computers, or lose a hard drive, the archive disappears with no persistent trail to the underlying records.

Pick the description that matches. The fix is different for each, but all three converge on the same principle: separate the record of what you did from the record of what you found, and make the found items point to stable identifiers rather than local file paths.

What Belongs in a Research Evidence Archive?

An evidence archive is not the same as a bibliography. A bibliography lists what you cited. An evidence archive records how you arrived at that list and keeps the trail reproducible.

A complete archive for a single paper typically has five layers:

LayerWhat it containsWhy it matters
Search recordSources searched, dates, query strings, any filters appliedLets you show when and where a search happened
Retrieved recordsDeduplicated list of items found, with identifiersSeparates what you found from what you read
Screening logWhich items advanced at each stage and whyDocuments selection decisions for title, abstract, and full-text stages
Linked study groupsWhich records refer to the same underlying studyPrevents counting one study twice
Final inclusion setThe subset that directly supports the manuscriptConnects claims to their evidence

The Cochrane guidance for systematic searching describes exactly this workflow: plan from eligibility criteria, search multiple sources, document the process, deduplicate records, screen in stages, link reports of the same study, and update searches before publication. That structure is designed for review-level rigor, but the same five layers apply to a thesis chapter or first manuscript at a lighter weight. You do not need a full systematic review protocol; you do need enough documentation that you can answer "where did this come from?" for every source.

The search record is the layer most people skip because it feels like interruption. A practical version takes about one minute per database or search interface.

For each search session, record four fields: the source or interface, the date, the exact query or filter configuration, and the number of records returned or saved at that moment. A plain text file or table works. The point is not to be exhaustive but to create a dated trail. Searching in WisPaper search gives you a single interface where you can run scholarly searches and then move the relevant results into a saved collection rather than copying citations across another tool.

Store each search record immediately. Retroactive reconstruction is where archives fail: by the time you realize you need the date you ran a query, the browser history has rolled over.

How Do You Deduplicate Without Losing the Trail?

The same study often arrives through multiple routes: you found it in one database, downloaded it from a preprint server, and received it again attached to an email from your supervisor. Keeping all three copies as separate items creates double counting and citation confusion.

The standard practice is to deduplicate at the record level, not the file level. Group records that refer to the same underlying study, keep one canonical record, and note the alternates. This is the practice Cochrane describes as linking reports of the same study. A tool such as My Library lets you keep a single saved copy of a paper instead of scattering duplicates across devices, which is the file-management half of the problem. A reference structure that supports organizing the same item in multiple project groupings without duplicating it handles the collection half.

Where a stable identifier exists, use it as the deduplication key. Two records with the same DOI are the same registered object, even if the titles are formatted differently or a preprint and a published version differ in small ways. The DOI system is designed so that an identifier can remain stable even as the object's location or descriptive metadata changes, which makes it the most reliable anchor for a long-term archive.

What Makes the Archive Persistent Instead of Fragile?

A folder on a single laptop is not a research archive; it is a working draft with an expiry date. Persistence requires connecting local records to stable identifiers and to a platform that survives device changes.

For each included source, capture the persistent identifier where one exists. Scholarly metadata resources such as the Crossref API expose registered fields including DOI, title, author, publication, funder, license, and relation information when deposited. If you have the DOI, you have a route back to the registered record even if your local file is lost.

Create a project home for each paper or chapter and keep the archive layers inside it. For readers starting a new project or converting an old one, Research Projects provides a defined place to create and hold the working record. This is also where the January reset pattern works well: spend an afternoon moving the previous term's sources into one project with a single search record, a deduplicated list, and stable identifiers.

How Do You Reconnect Citations After a Break?

Returning to a project after even a few weeks creates a specific failure mode: the reference list says one thing, and your working files suggest another. Before you continue writing, run a citation-to-source reconciliation.

The goal is to verify that every reference in the manuscript corresponds to a real record in your archive, and that the identifier in your citation matches the identifier on the source. Where your bibliography was produced from BibTeX, a screening pass against an academic database can catch entries that point to references that do not actually exist. TrueCite is designed for exactly this check: it screens BibTeX against real academic databases to flag hallucinated references. Run this before sending a revision back to a supervisor, because that is the moment when a phantom citation becomes visible.

This reconciliation is also the right time to update the search record. Cochrane notes that searches should be updated before publication, and a thesis or article submission benefits from the same habit. Rather than redoing the original hunt, search once more with the same recorded queries and add the dates to the archive.

What Decision Does the Reader Need to See Next?

The archive is not finished when all the files are in place. It is finished when a reader can follow the decision trail without asking you questions. Run this short diagnostic against your archive.

Can you answer all four questions from the record alone?

  1. Which sources were searched, on which dates, with which queries?
  2. Which records were removed as duplicates, and which canonical record replaced them?
  3. Which records were excluded at screening, and under what criterion?
  4. Which final records support which claims in the manuscript?

If any question forces you to rely on memory, the archive is not ready. If all four are answerable from the files, you have reached submission readiness on the evidence side.

When Do You Hand the Archive to Long-Term Storage?

The strongest archives are the ones you stop editing. Once the manuscript is submitted or the chapter is finalized, freeze the project record: export a clean copy of the source list, keep the persistent identifiers, and label the archive with the submission version.

A frozen archive is the thing you will open months later to answer a query or to start the next project. It should contain the search record for that version, the deduplicated source list, the screening notes, and the citation-to-source reconciliation result. If an examiner asks a question in three months, the answer should be a file, not a memory.

Is There a Faster Route for the Next Project?

The archive you built retrospectively for this paper becomes the template for the next one. Instead of starting from an empty folder, start from the five-layer structure and fill it as you go. Searching happens inside one interface, results move into a saved collection, deduplication happens by persistent identifier, and linking happens before you start writing rather than after.

For researchers who want to turn an archived project into a new line of inquiry, Idea Discovery offers a workflow that starts from research topics, background, and references to generate new research ideas. If you have just finished a chapter during a break and are deciding what to work on next, the archived trail doubles as the input to that process.

FAQs

A reference manager stores and formats sources. A research archive additionally captures the process: the search dates and queries, the deduplication decisions, the screening log, and the link between each claim and its supporting record. You can build an archive around a reference manager, but the manager alone does not document how sources were found and selected.