Research Laptop Cleanup: Reference Managers and the PDF Hoarding Problem

Where the gigabytes actually hide

Zotero keeps every attachment under %USERPROFILE%\Zotero\storage, one subfolder per file with a random eight-character name — opaque to you, perfect for the tool. Sync downloads and locally imported copies both land there, and the folder only ever grows until you intervene. The sync settings page shows how much of your quota is used, but the local folder is the real number that matters for disk space.

Mendeley keeps its database and downloaded files under AppData\Local\Mendeley Ltd, and deleting that tree while it is your active manager deletes your library with it. EndNote hides attached PDFs inside the .Data folder that sits next to the .enl library file — invisible in normal browsing, huge in practice. And then there is Downloads: article.pdf, article (1).pdf, preprint_final_v3.pdf, the same text three times at different stages of review.

The cleanup that keeps your citations intact

One rule first: citations live in a small database, PDFs are the mass — never touch the database before the files are accounted for. Export a full backup before anything else: a Zotero RDF or BibTeX export from Zotero, a copy of the whole data folder for Mendeley. Only then start cutting, in this order:

• Empty Zotero's trash, then merge entries in the Duplicate Items collection — the merged item keeps the attachments of every copy, so check for twin PDFs afterward
• Check the real numbers: Zotero's sync settings show quota used, but what occupies the disk is the local storage folder
• Mendeley: never delete AppData\Local\Mendeley Ltd while it is your active manager — that tree IS the library
• Downloads: sort by name and kill the (1)…(7) copies of every paper in one pass
• EndNote: compress the library to a single .enlx file first — it is the backup format that carries the PDFs with it
• For future imports, attach files as linked attachments in one canonical folder outside the manager's storage — one copy, referenced everywhere

A policy that stops the regrowth

The laptop refills unless the workflow changes. Pick one canonical location for PDFs, let the manager point at it, and send the archive to institutional cloud storage rather than the sync quota. Purge Downloads monthly — a sorted-by-name view makes the duplicate families obvious in seconds. Cite-while-you-write keeps working as long as the database is intact, and a library whose backup fits in one folder is a library you will actually back up.

Questions and Answers

Read next