Step-by-step guide
A seven-step guide to building a literature review matrix (synthesis matrix): define the research question, select 5–30 papers, set eight evidence columns, capture verbatim page-cited quotes, mark conflicts, synthesize, and export citations.
Updated 2026-07-22
What a synthesis matrix is, how many papers to include, which columns to use, and how to keep every cell verifiable.
A literature review matrix — also called a synthesis matrix — is a table that compares multiple papers against the same set of questions. Each row is a paper and each column is an evidence dimension such as sample, method, or result, so patterns, gaps, and contradictions across the literature become visible in one grid instead of staying buried in separate summaries.
Between 5 and 30 papers is a workable range for a hand-reviewed matrix, and it is the range WriteGo's Literature Review Matrix supports. Fewer than 5 rarely reveals cross-paper patterns; beyond 30, verbatim evidence capture stops being realistic and the matrix drifts toward unverifiable summaries. Narrow the research question before enlarging the matrix.
Eight columns cover most literature reviews: research question, sample, method, result, effect, limitation, conflict of interest, and DOI. WriteGo fixes exactly this column set so matrices stay comparable across projects. Fill cells with verbatim quotes carrying page locators rather than paraphrases, so every claim in the finished review can be traced back to a source page.
One row per paper, one column per evidence dimension, and every contentful cell holding a short verbatim quote with its page number — "n = 214, undergraduate sample (p. 4)" — rather than an unsourced summary. Disagreements between papers are flagged as conflicts instead of smoothed over, and each row ends with a DOI so citations can be generated without retyping metadata.
A matrix is only as sharp as the question it answers. Write the research question down first, and phrase it so a paper can agree, disagree, or be silent — that is what makes columns fillable. In the WriteGo matrix, research question is the first of the eight fixed columns, so each paper's own framing sits directly under yours and mismatches surface early.
Five papers is roughly where cross-paper patterns start to appear; past thirty, verbatim evidence capture stops being realistic. Include papers that disagree with your expected answer — conflicts are what a matrix is for. In WriteGo, upload PDFs up to 10 MB into the matrix's own paper library; server-side text extraction runs for every paper, with OCR for scanned pages, and each paper shows its processing status with retry and delete.
Use eight columns: research question, sample, method, result, effect, limitation, conflict of interest, and DOI. A fixed set forces the uncomfortable columns — limitations and conflicts of interest — to be captured for every paper, and keeps matrices comparable across projects. WriteGo fixes exactly this column set and auto-seeds candidate rows from extracted paper metadata checkpoints, so the grid starts from what the papers state.
Fill cells with verbatim quotes, not paraphrases, and record the page at capture time. A cell that reads "n = 214, undergraduate sample (p. 4)" can be verified in seconds; an unsourced summary cannot. In the WriteGo matrix, every contentful cell holds claims backed by captured quotes with paper-version, page, and character-offset locators, and cross-paper search stages useful passages as pending evidence instead of copy-pasted fragments.
When two papers disagree, mark the conflict — do not average it away. Conflicting evidence is usually the most interesting part of the synthesis and often becomes the discussion section. WriteGo keeps unreviewed, not-located, needs-review, and conflicting states visibly distinct and never layers a black-box quality score on top of them.
Read the matrix column-wise: group papers by method, compare effect directions, and let claims emerge from cells rather than memory. Write each synthesis claim so it can point back to specific cells, so every statement in the review can be traced to a verbatim, page-linked quote.
Export the finished matrix to Markdown or formula-safe CSV, and generate references in APA 7, MLA 9, Chicago, BibTeX, or RIS — resolved from each paper's DOI, arXiv ID, PubMed ID, or URL with per-field provenance. To draft from the matrix, hand selected evidence to Writer Agent or Integrity Review as a reference-only source package that carries page numbers and character offsets, never full text.
Yes — eight columns and one row per paper works in any spreadsheet, and the WriteGo matrix exports to formula-safe CSV if you want to finish there. What a spreadsheet cannot do is bind each cell to a verbatim quote with page and character locators or keep review states distinct.
No. An empty or not-located cell is honest data: a paper that never reports its limitations is itself a finding worth writing down. Never pad cells with guesses — the matrix's value is that every entry can be traced to a source page.
Currently yes. It requires sign-in, but uploading papers, building the matrix, cross-paper search, and reference export charge no credits today.
Not in WriteGo. The matrix deliberately has no AI chat and runs no AI detection — cells hold quotes captured from your uploaded papers, under review states you control. For AI question answering on a single paper, use ChatPDF, which keeps a separate paper library and verbatim-checks every page citation.