Part 1 of How to Finish Your Research with Beeblio. This part turns an initial idea into a question and a checked map of the literature.
A literature search is easier once you know what the search must help you decide. The Everyday Habits & Well-Being demo begins with a question: How are sleep, physical activity, leisure screen time, and social connection associated with self-reported well-being in a synthetic cross-sectional adult survey? It names an outcome, four candidate predictors, a population, and a design. It also leaves room for a finding that does not match the hypotheses.
The point of departure: the demo project in Your Projects.
Start with a question you can investigate
Open Your Projects and choose Everyday Habits & Well-Being. The project opens its main report in 4-Reports/research-report.md. Read the “Research question and hypotheses” section, then open 2-Data/codebook.json. The report gives the plain-language question; the codebook shows what the dataset actually records, including units and ranges. That comparison is a useful first check: could each part of the question be answered with a defined variable?
Use the File Manager to move between those folders and keep related files open in tabs. The dedicated Literature, Data, Analysis, Reports, and Figures views help you find an artifact by its role; the folder tree keeps its exact path visible. When a project grows, Quick Open can jump to a file or reference without retracing the folder tree.
For your own project, create a short question document before collecting sources. Specify the population or material, the outcome or phenomenon, the comparison or predictors, and the kind of evidence you expect to use. Also write one sentence about the limit of the design. In this example, cross-sectional self-reports can describe associations, but cannot establish temporal order or causation.
Try asking Beeblio: “Which terms in this question need operational definitions, and which claims would this design be unable to support?” Treat the response as a critique to evaluate, not as a substitute for choosing your method.
A precise question at work: the plan and model it makes possible.
Search for evidence with a purpose
Open Literature → Search. Search for the concepts in the question, then vary the terms for the outcome and each candidate predictor. The goal is to understand how earlier researchers measured these concepts and what their designs can support. Inspect a result before saving it: check its title, authors, publication details, DOI when available, and relevance to the question. Save the citation and an accessible PDF when appropriate.
In Literature → Library, open the project bibliography, 1-References/references.bib. The demo includes papers and guidance used in its report. Match a citation key in 1-References/literature-notes.md to the corresponding entry in the bibliography. Before repeating a claim, open the underlying paper or authoritative record and verify that the summary reflects its population, method, and finding. A DOI is a way to locate a work; it does not itself verify a summary.
If you already have a bibliography, import it and check duplicates and metadata. Keep the library tidy now; it is much harder to repair citations while preparing the final manuscript.
Literature Search: evidence enters one verified item at a time.
A candidate examined in full before it earns a place in the library.
Build an evidence map, not a reading pile
The demo has two complementary views: 1-References/evidence-matrix.csv and 3-Analysis/literature-matrix.matrix. The interactive matrix compares study design, population, finding, and limitation. Open a row and ask what it can contribute to this research question. A scale-development paper can inform measurement; it cannot prove that sleep or activity changes well-being. A guideline can offer context; it is not the same as a primary empirical study.
For a new review, make columns that expose the decisions you need to make: sample, design, exposure, outcome measure, result relevant to your question, and a limitation. Add each saved study, then check the entries against the source. You can ask Beeblio AI to propose a comparison or draft cells, but review them before treating the matrix as evidence.
The Library’s Literature Map gives you another way to inspect relationships among saved references. Open it to explore citation links and clusters, then return to the underlying papers before treating a visual group as a research gap. The map helps you choose what to read next; the matrix records what you learned.
Read deeply with Knowledge
If a PDF is in the workspace, add it to Knowledge and wait for indexing to finish. Search a specific question, such as how a paper defined its outcome. Knowledge returns an answer with source files and, when available, page numbers and excerpts. Open the cited file and read the surrounding passage before you copy a detail into the matrix. You can also ask the research agent to search indexed project files in chat when the answer needs evidence from several documents. Record the source and page for a contested cell; a generated answer alone is not the evidence.
An answer with sources attached; the surrounding passage is the evidence.
Make a repeated review procedure reusable
If you will extract the same fields from many papers, open Agent Skills and create a custom skill. Give it a clear “When to Use” description and instructions such as: identify population, design, exposure, outcome measure, and one limitation; report a source passage or page for each field; mark any field that cannot be verified. Save it, then invoke it in chat with /skill-name while working through the next paper. Custom skills are available across your projects, so keep their instructions general enough to reuse. Check each output against the paper before adding it to the matrix.
Try asking Beeblio: “Compare the saved studies by population and well-being measure. Identify a possible gap, then list which sources support it and which parts still need verification.”
The evidence matrix: each saved study reduced to comparable fields.
Finish this stage with a decision
Return to the question document. What did the search change? Perhaps the outcome needs a more precise definition, an exposure needs a better measure, or the planned comparison is too broad. Revise the question and note why. Keep a search log if you are conducting a formal review, including databases, dates, terms, and screening decisions.
The handoff to Part 2 is concrete: a focused question, a checked reference library, and an evidence matrix. Those artifacts should tell you what the study must measure and what its design must avoid claiming.
Continue to Part 2: Design the study and collect data.