How Should You Organize Collections, Folders, Tags, and Notes in a Reference Manager?

How Much Organization Does Your Reference Library Actually Need?

You begin with one collection for your research project. Then you add subcollections for theories, methods, variables, populations, and chapters. You create tags for “important,” “very important,” “read,” “to read,” “maybe,” and several topics that are almost, but not quite, the same thing.

Six months later, saving a paper requires several decisions before you can move on. Worse, you cannot remember whether a paper about student engagement was filed under “Engagement,” “Student Engagement,” “Learning Engagement,” or the dissertation chapter where you first used it.

Reference managers offer collections, folders, tags, notes, searches, and other organizational tools because research literature can become difficult to manage. The challenge is not using all of them. It is giving each one a clear job so that organization makes retrieval easier rather than creating another administrative system to maintain.

Organize for Retrieval, Not for Perfect Classification

Give Each Organizational Tool a Different Job

Reference managers differ in terminology and implementation. Zotero uses collections, subcollections, tags, notes, searches, and saved searches. Mendeley Reference Manager provides custom collections, smart collections, tags, search, filtering, and Notebook functionality. EndNote provides groups and tagging capabilities among its organizational features.

The underlying problem is nevertheless similar: one paper can have several relationships to your research at the same time.

A study might belong to your dissertation project, use qualitative interviews, address student engagement, provide a theoretical argument you want to revisit, and still be waiting to be read closely. Trying to represent all of those characteristics through one hierarchy quickly becomes awkward.

The solution is not a universally correct taxonomy. It is to use different organizational mechanisms for different kinds of information.

Collections or folders work well for stable contexts

Collections are useful when you want to see a recognizable body of literature together. Depending on your research, sensible collections might correspond to a project, course, manuscript, dissertation chapter, broad research area, or another relatively stable context.

Zotero's collection model illustrates an important principle. A collection is not equivalent to a physical folder on your computer. The same bibliographic item can belong to several collections without creating several copies of the item in the library. Zotero describes collections as functioning more like playlists: one item can appear in multiple collections while remaining a single library record.

This is particularly useful when research projects overlap. A paper about technology acceptance might belong simultaneously to a dissertation collection and a separate project on artificial intelligence adoption. You do not need to decide which project “owns” the paper.

Do not build a folder tree for every concept in your research

Hierarchies feel orderly because they resemble filing cabinets. Academic literature, unfortunately, does not behave like paper in a filing cabinet.

One study can address several concepts, use multiple methods, involve a particular population, support one argument, challenge another, and matter to several projects. If every characteristic becomes a nested folder, you eventually face arbitrary decisions about where a multidimensional paper belongs.

Use collections for groupings you repeatedly need to browse as units. If you rarely think, “Show me all papers belonging to this particular group,” that category may not need its own collection.

Keep the hierarchy relatively shallow unless deeper nesting genuinely reflects the way you work. A structure such as “Dissertation → Chapter 2 → Student Engagement → Behavioral Engagement → Online Learning” may look impressively systematic while making routine filing unnecessarily laborious.

A simpler project structure may be enough:

• Dissertation
• Journal Article: AI Feedback
• Systematic Review: Digital Assessment
• Teaching and Learning Analytics
• Future Projects

The precise labels are less important than whether you understand and consistently use them.

Tags are better for characteristics that cross collections

Tags provide another dimension of organization. Zotero documentation notes that tags can characterize items by topics, methods, status, ratings, or workflow categories. Mendeley similarly supports tagging references, while current EndNote versions provide customizable tags that can represent themes, methodologies, relevance, and other categories.

This makes tags useful for information that cuts across projects.

You do not need all of these categories. They illustrate different jobs tags can perform.

Use controlled tags rather than inventing a new label every time

The greatest weakness of tagging is also its greatest freedom: you can create almost anything.

Suppose you use “AI,” “artificial intelligence,” “Artificial Intelligence,” “generative AI,” “GenAI,” and “gen-ai” inconsistently. Some distinctions may be meaningful, but others simply fragment the same concept across several labels.

A small controlled vocabulary is usually easier to maintain. Before creating a new tag, check whether an existing one already expresses the idea adequately. Use consistent spelling and capitalization. Merge near-duplicates when they appear.

Do not attempt to predict every tag you might need for the next ten years. Let your vocabulary grow when repeated retrieval needs emerge.

Separate topic tags from workflow tags

A tag such as “qualitative-research” describes the paper. A tag such as “to-read” describes your current relationship with the paper.

Those are different kinds of information.

Keeping them conceptually distinct can make a tag system easier to understand. Topic and methodological tags may remain useful for years. Workflow tags are temporary and should change as the paper moves through your process.

For example:

This is only an example. You might need two states rather than three, or none at all. Do not create workflow tags merely because another researcher's system looks attractive in a screenshot.

Notes should preserve meaning that metadata cannot

Reference metadata tell you what a source is: title, authors, year, journal, DOI, and related bibliographic information.

Notes can record why the source matters to you.

A useful note might capture a methodological limitation you want to remember, the reason a paper matters to your argument, a connection to another study, a definition worth revisiting, an idea for how a method might transfer to your own work, or the reason you decided not to rely heavily on the paper.

The best notes are therefore not necessarily long summaries. A concise note such as “Useful operationalization of behavioral engagement; compare with Lee measure” may be more valuable six months later than a page of prose that merely repeats the abstract.

Do not use tags to reproduce information you can already search

Before creating a tag, ask whether the reference manager already stores the information in a searchable field.

If you can already retrieve all papers by a particular author through the author field, an author-name tag usually adds little. The same may be true for publication year, journal title, DOI, or other structured metadata.

Tags are most useful when they represent information that the bibliographic record does not already capture in the way you need.

This principle reduces maintenance. Every manually assigned label creates another piece of information that you may eventually need to update.

Search can replace surprisingly elaborate classification systems

Modern reference managers are searchable databases. That changes how much manual organization you actually need.

Zotero can search metadata, tags, and full-text content where available. Its Advanced Search supports combinations of criteria, and saved searches update automatically as items begin or cease to meet those criteria. For example, Zotero documents a workflow in which a saved search can identify items that do not carry a “read” tag.

Mendeley likewise provides library search, filtering, custom collections, and automatically generated smart collections.

This means you do not have to anticipate every future retrieval question when filing a source. If bibliographic metadata, tags, or searchable full text can recover the paper later, additional manual classification may be unnecessary.

Saved searches can turn metadata and tags into dynamic collections

A normal collection contains items you assign to it. A saved search instead contains a set of criteria.

In Zotero, saved searches continuously update. You might create one for recently added items, papers carrying a particular tag, records missing certain information, or combinations of conditions relevant to your workflow.

This can be useful when membership should change automatically. A static collection called “To Read” requires you to add and remove papers manually. A dynamic search based on a workflow tag can update as you change the status of each source.

Do not automate simply because you can, however. Saved searches are most useful for recurring questions that you actually ask of your library.

Automatic tags can create clutter

Some reference managers may import keywords or subject headings along with bibliographic records. Zotero, for example, can automatically add tags supplied by websites and library catalogs. These automatic tags behave like other tags but can be distinguished, hidden, removed, or prevented from being imported through Zotero's settings.

Imported subject terms can be useful, particularly when they complement your own vocabulary. They can also produce a long and inconsistent tag list because different databases and publishers use different terminology.

If your tag selector has become dominated by labels you never deliberately created, investigate whether they are being imported automatically before manually cleaning them one by one.

Your system should work across multiple projects

A common mistake is to rebuild the entire organizational structure for every new study. That can lead to parallel collections, duplicate tags, and multiple copies of the same literature.

A better approach may be to distinguish between project-specific organization and reusable knowledge. Collections can identify the projects to which a source currently contributes, while stable tags can describe characteristics that remain meaningful regardless of project.

A methodological paper, for example, might belong to three project collections while carrying the same “structural-equation-modeling” tag throughout. This becomes especially useful when you maintain one reference library across multiple research projects.

Your organizational system should not compensate for bad metadata

No collection hierarchy can fully rescue records titled “Full Text PDF” with missing authors and incorrect publication years.

Accurate bibliographic metadata remain the foundation of retrieval and citation. When an imported record is wrong, correct the citation metadata rather than creating tags or notes to work around the error.

Similarly, organizational confusion sometimes comes from duplicate records rather than insufficient classification. If several copies of the same paper appear throughout your library, address the underlying duplication rather than filing each copy more carefully.

One Paper Can Be Organized Along Several Dimensions

The paper has not been copied into several independent records. Instead, different organizational mechanisms describe different aspects of the same source.

That distinction is the key to a scalable system. Collections answer where the source belongs in your work. Tags describe characteristics that may cut across projects. Notes preserve your interpretation or intended use. Search retrieves information already represented elsewhere.

Common Mistakes When Organizing a Reference Library

Build the Smallest System That Reliably Finds What You Need

A reference library should reduce cognitive overhead. You should not need to remember an elaborate classification manual every time you save a source.

For many researchers, a sustainable starting system is surprisingly small. Create collections for active projects. Maintain a limited vocabulary of recurring topical, methodological, and workflow tags. Add notes when a source contains something you need to remember. Let search handle the rest.

Then adjust the system when friction appears. If you repeatedly search for papers using a particular method, that may justify a method tag. If one project becomes too large to browse comfortably, a few subcollections may help. If a tag has not been useful for a year, perhaps it does not deserve continued maintenance.

This also changes how you should decide which papers enter your reference library. You do not need to fully classify a source the moment you encounter it. Capture enough information to make the source retrievable and add richer organization when the research actually demands it.

Periodically check for structural problems rather than endlessly redesigning the system. Consolidate redundant tags, archive obsolete project collections where appropriate, correct obvious metadata problems, and prevent duplicate references from taking over the library.

Is Your Reference Library Organized Enough?

Frequently Asked Questions About Organizing References

Your System Should Make Literature Easier to Find, Not Harder to File

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.