How to Conduct a Literature Review: A Practical Research Workflow
Follow a thesis question from eligibility rules and database searches through extraction, synthesis and writing.
Imagine that you are writing a thesis chapter about peer-led writing workshops. One paper measures students’ confidence, another scores their essays, and a third describes what happened in a workshop. Which papers can answer whether the workshops improved writing? That decision will shape your search, your extraction table and the conclusion you can defend.
1. Write the question your review needs to answer
Topics such as “AI in education” or “remote work” are too broad to guide a useful review. First write down what you need the review to help someone understand or decide. Are you mapping a developing field, explaining competing theories, estimating an intervention effect, or identifying a gap for a new study?
Turn that purpose into a question with a clear population or setting, concept or intervention, and outcome or phenomenon. For example, instead of “Does tutoring help?”, ask: “Among first-year university students, how do peer-led writing workshops relate to writing confidence and assessed writing quality during the first academic year?” This still needs scoping, but it names the population, intervention, and outcomes that determine which papers belong.
Check feasibility before committing. Search a few relevant databases, inspect recent reviews, and note whether the terms used by researchers match your wording. If thousands of records appear, narrow the setting, date range, or outcome for a reason tied to your purpose. If almost none appear, broaden a concept or consider whether the question should be exploratory. A well-framed question guides eligibility, searching, extraction, and synthesis; it should not be retrofitted to the studies that happen to be easiest to find. Cochrane’s guidance on review questions is written for intervention reviews, but its point about question scope shaping every later step is broadly useful.
2. Set boundaries you can apply to a paper
Write down the job of the review in one sentence. For the workshop chapter, it might be to explain which outcomes have been measured and how those measurements differ. If you are still choosing a design, see the companion guide to systematic and narrative reviews.
Write inclusion and exclusion criteria before screening in earnest. Specify the population or context, concepts or interventions, eligible evidence types, publication or date boundaries, languages, and outcomes or phenomena of interest. Avoid rules such as “relevant papers only”: they transfer the real decision to an unrecorded impression. Instead, make the reason operational: “Empirical studies of university writing courses, published from 2010 onward, reporting a writing outcome.”
For a narrative review, criteria can remain flexible enough to follow an important conceptual thread, but record how you selected sources and why particular literatures are central. For a systematic review, pre-specify the method in a protocol, ideally before outcome patterns are known. If a justified change becomes necessary, document what changed and when. This reduces the temptation—often unconscious—to let results shape eligibility or analysis.
3. Build a search that another person could repeat
Translate the question into concept groups and synonyms. For the writing-workshop example, one group might include “peer tutoring,” “writing centre,” and peer-led workshop; another might include first-year undergraduates or university students; a third might cover writing confidence, writing quality, or academic writing. Combine synonyms within a group with OR and combine groups with AND. Test the query against a few known relevant papers: a search that misses them may have a vocabulary or indexing problem.
Select sources that fit the subject and evidence type. A database is not a neutral window on all research: coverage, indexing, language, and document types differ. Record each database or platform, the date searched, the exact query, and any limits. For a systematic review, have someone other than the person who built the search check the strategy before you rely on it: a research librarian or information specialist where your institution has one, otherwise a supervisor or an experienced colleague. The PRESS 2015 guideline statement lists what that check should cover. Also consider supplementary routes such as reference-list checking, citation searching, trial registers, or grey literature when the question warrants them. A convenience search in one broad index is not evidence that “the literature” has been comprehensively covered.
Keep a search log as you work. Save the exported records and preserve the original search strings. Remove duplicates systematically, noting the tool or procedure. These records later let you explain how many references were identified, screened, excluded, and included. For a systematic review, these counts feed the flow diagram. PRISMA flow templates
4. Screen consistently and keep reasons
Apply the eligibility criteria in stages: first title and abstract, then full text for records that remain potentially eligible. Pilot the rules on a small batch before screening the whole set. For a Cochrane intervention review, at least two people must independently make the final eligibility decisions using full texts. Agree on a way to resolve disagreements before the team starts. For a solo thesis chapter, record who made each decision and ask a colleague to examine difficult cases where possible. Describe the actual process. Cochrane Handbook, study selection
At full text, record a concise reason for each exclusion—wrong population, no relevant outcome, ineligible design, and so on. This is useful quality control, not paperwork for its own sake: repeated “unclear” cases reveal a rule that needs clarification. Avoid deciding from an abstract alone when key eligibility details are missing; retrieve the full text or record that it could not be obtained.
5. Record what each study measured
Create an extraction table that reflects the question. Common fields include citation, setting, study design, sample, methods, outcome or concept, principal findings, limitations, and funding or conflicts where relevant. Add fields for what differs across studies—for instance, whether “writing quality” means rubric scores, grades, or self-assessment. Pilot the table on a few varied papers and revise it before using it consistently.
Separate what a study measured and found from what its authors infer. When a result matters to your argument, verify it in the paper rather than copying it from an abstract or another review. Note missing data and methodological concerns instead of silently filling gaps. For systematic reviews, use a risk-of-bias or quality-assessment tool suited to the included designs; there is no universal checklist that makes unlike studies comparable.
6. Group findings by the question they answer
Before drafting, group findings around the question: by outcome, theory, method, time, population, or mechanism. Ask what patterns hold, under what conditions results differ, and whether apparent disagreement reflects different definitions or study designs. A synthesis paragraph usually makes a claim, supports it with several studies, explains the relationship among them, and states the boundary of that claim.
Suppose two fictional reports describe different outcomes. In one, students report greater confidence after a workshop. In the other, independent assessors give essays similar scores before and after the workshop. You can report both observations, then explain that the confidence result alone cannot establish a change in writing quality. This is more informative than listing one paper after another and avoids treating unlike outcomes as if they were interchangeable.
Do not pool results simply because studies address a similar topic. Quantitative meta-analysis requires defensible comparability and appropriate statistical methods; qualitative evidence may call for thematic, framework, or other interpretive synthesis. When synthesis is not appropriate, explain why and present findings in a structured narrative. Be explicit about the limits created by publication bias, language restrictions, incomplete retrieval, or weak primary studies.
7. Write a transparent argument and audit it
Organize the review around the question and the logic of the synthesis, not a parade of citations. A useful structure often moves from the problem and review purpose to the method, the major patterns and differences, implications, limitations, and remaining uncertainty. In a thesis chapter, some methods detail may belong elsewhere; readers still need enough information to understand how the evidence was selected and interpreted.
Before sharing the draft, check that each major conclusion is supported by the evidence summarized; that contradictory results and meaningful limitations are visible; and that the methods reported match what you actually did. Ask a colleague to trace one conclusion back through the synthesis table to the original studies. For a narrative review, tools such as SANRA can prompt reflection on rationale, search description, referencing, scientific reasoning, and presentation, though it is an appraisal aid rather than a substitute for disciplinary judgment.
State what the evidence supports, what it cannot establish, and what next study or decision would reduce the most important uncertainty. If you are beginning with database discovery, the Field Guide chapter on search and saved evidence is a contextual starting point.
Sources & further reading
- Cochrane’s guidance on review questions
- PRESS 2015 guideline statement
- PRISMA flow templates
- Cochrane Handbook, study selection
- Field Guide chapter on search and saved evidence
- Lasserson, Thomas & Higgins, Cochrane Handbook, Chapter 1: Starting a review.
- Snyder, H. “Literature review as a research methodology: An overview and guidelines.” Journal of Business Research 104 (2019), 333–339.
- Grant, M. J. & Booth, A. “A typology of reviews: an analysis of 14 review types and associated methodologies.” Health Information & Libraries Journal 26 (2009), 91–108.
- Page et al. “The PRISMA 2020 statement: an updated guideline for reporting systematic reviews.” BMJ 372 (2021), n71.
- Baethge, C., Goldbeck-Wood, S. & Mertens, S. “SANRA—a scale for the quality assessment of narrative review articles.” Research Integrity and Peer Review 4, 5 (2019).