How to Write a Literature Review

By Kevin Fernando · Sep 24, 2026 · 13 min read

A literature review is a methodical synthesis of existing research on a topic. Define a specific research question, search multiple databases systematically, screen sources against consistent criteria, extract data using a template, and organize findings to identify patterns and gaps. Document every step so your process is reproducible and defensible.

A literature review is not a summary of everything written on a topic. It is a systematic process of finding, evaluating, and organizing sources to establish what research has established, where evidence conflicts or is sparse, and what questions remain. The work is analytical before it is written. You gather sources, apply consistent standards to decide which ones belong, extract information in a structured way, and look for patterns that emerge only when you organize the data deliberately.

This guide treats literature review as a research method with concrete checkpoints. You will learn how to scope a question so it guides your search, conduct a reproducible search across the right databases, screen sources against criteria you set in advance, extract findings consistently, and synthesize what you find so gaps become visible. The process mirrors the research synthesis that IMBeacon.ai performs at scale: systematic source selection, rigorous evaluation, and organized synthesis that reveals what the evidence shows and what it does not.

What you will produce and what you need to start

A literature review is a written synthesis, but the product is only as strong as the research work that precedes it. Before you write, you will have conducted a systematic search, screened sources against explicit criteria, extracted data from each one in a consistent format, and organized your findings so patterns are visible.

You need access to academic databases (your institution, employer, or public library usually provides these), a way to track sources and notes (a spreadsheet, reference manager, or simple document works), and time to read carefully. You do not need to be an expert in your topic; you need to be willing to read primary sources and apply standards consistently. If your topic is very broad (artificial intelligence, climate change, mental health), you will need to narrow it first. If your topic is already specific (the effect of a particular drug on a particular outcome in a particular population), you are ready to begin.

Define your research question and scope

Start with a question, not a topic. "What do we know about leadership?" is too broad. "How does distributed decision-making affect team performance in remote software development?" is specific enough to guide your search and tell you which sources belong.

Your question should identify the population or context you care about, the intervention or exposure or phenomenon you are studying, and the outcome or finding you are interested in. Write it down. This becomes your filter: when you find a source, you ask whether it addresses your question. Sources that do not belong in your review, even if they are high-quality, because they study a different population, measure a different outcome, or examine a different phenomenon.

Scope also means deciding what types of sources you will include. Will you include only peer-reviewed journal articles, or also books, reports, and theses? Will you set a date range, or review everything ever published? Will you limit to one language? These decisions prevent endless searching and make your choices reproducible. Write them down alongside your question.

Search for sources systematically

A systematic search is reproducible and documented. Someone reading your review should be able to repeat your search and find the same studies, because you recorded which databases you used, which terms you searched, and what filters you applied. This is not how most people find sources on the internet, and it takes longer, but it is the only way to ensure you did not miss a large group of relevant studies by accident.

### Choose your databases

Different databases index different journals and materials. A study comparing systematic search qualities examined 28 widely used academic search systems including Google Scholar, PubMed, and Web of Science. The choice depends on your discipline and question.

For biomedical research, PubMed stores abstracts and bibliographic details of more than 30 million papers and is the standard. For broader social science and humanities work, Scopus covers approximately 90.6 million core records, while Web of Science covers approximately 100 million items. For education research, ERIC database hosts approximately 1.6 million items. Most institutions provide access to at least two or three of these. Check what your library or employer offers.

For a systematic review, you must search at least three databases, preferably including one large general database and two subject-focused ones. Searching two or more databases achieved greater than 95% coverage and at least 87.9% recall in systematic reviews where conclusions and certainty were unchanged. The combination of Embase, MEDLINE, Web of Science Core Collection, and Google Scholar achieved 98.3% overall recall and 100% recall in 72% of systematic reviews.

### Build and refine search strings

A search string is the phrase you type into the database. Start with the key concepts from your research question and list synonyms for each. If your question is about distributed decision-making and remote teams, your concepts might be: distributed decision-making (synonyms: decentralized decisions, shared decision-making, autonomous teams); remote work (synonyms: distributed teams, virtual teams, work-from-home); and team performance (synonyms: productivity, effectiveness, outcomes).

Combine these using Boolean operators. AND narrows your search: "distributed decisions AND remote teams" retrieves only sources that mention both. OR broadens it: "distributed decisions OR decentralized decisions" retrieves either term. NOT excludes: "remote teams NOT offshore" removes sources about offshore outsourcing. Most databases also allow you to search for phrases in quotation marks ("distributed decision-making") and to limit results by date, publication type, or language.

Run your first search and look at the results. Are they relevant? Too many? Too few? Refine your string. If you retrieve thousands of irrelevant results, your search is too broad; add AND and another concept. If you retrieve only a handful, your search is too narrow; try OR to include synonyms, or remove a concept. Document each version of your search string and how many results it returned. This record becomes part of your methods.

### Document your search process

Record the date you searched, the database name, the search string you used exactly as you typed it, and the number of results you retrieved. If the database allows, export the results to a file or take a screenshot showing the result count. This documentation serves two purposes: it lets you repeat the search later if you need to update your review, and it proves to a reader that your search was systematic and reproducible, not a casual hunt for sources that support your hypothesis.

Create a simple table with columns for database, search date, search terms, and results count. Update it as you search. If you search PubMed on January 15 with the string "(distributed decision-making) AND (remote teams)", record all of that. Later, when you write your methods section, this table becomes your reference.

Screen and evaluate sources for relevance and quality

You now have a list of sources. Most will not belong in your review. Screening is the process of deciding which ones do, using criteria you set in advance. This prevents you from cherry-picking sources that fit what you already believe.

### Apply inclusion and exclusion criteria

Before you read any source in detail, create a checklist of criteria that a source must meet to be included. Your criteria might be: published in a peer-reviewed journal, published between 2015 and present, studies a team-based context, measures an outcome related to decision speed or quality, and reports quantitative or qualitative findings. Any source that fails any criterion is excluded.

Start by screening titles and abstracts only. Read the title and abstract and ask: does this source address my research question? Does it meet my inclusion criteria? If the answer is clearly no, exclude it. If the answer is maybe or yes, keep it for full-text screening. This step eliminates sources that are off-topic without wasting time reading the full paper.

For sources that pass title and abstract screening, read the full text and apply your criteria again. Now you can see whether the study design is appropriate, whether the population matches what you are studying, and whether the outcomes are relevant. Again, exclude sources that do not meet your criteria. The sources that remain are your final set.

Use a simple form or spreadsheet to track your decisions. Record the source identifier, your inclusion decision (include or exclude), and the reason if you exclude it. This record shows that your screening was systematic and transparent.

### Assess methodological quality

Not all sources are equally strong. A well-designed randomized trial with a large sample and low risk of bias is more trustworthy than a small observational study with methodological limitations. You do not necessarily exclude lower-quality sources, but you need to know which ones they are so you can weight their findings appropriately when you synthesize.

Credible sources demonstrate reliability through verifiable accuracy, author expertise, transparent methodology, and appropriate evidence supporting claims. Credibility assessment involves evaluating trustworthiness, whether sources provide accurate, honest information, and expertise, whether authors possess relevant qualifications and knowledge.

Key criteria for evaluating sources include the author's expertise, the publication date, and the language's objectivity or bias. Peer review is generally applied to journals and books offered by recognized academic publications that are usually paid for by libraries, as well as by a growing number of high-quality open access educational materials. Reputable institutions typically adhere to rigorous publication standards, and if a source comes from a recognized university or research body, it is likely to have undergone thorough review.

For each source, note its strengths and limitations. Does it have a large sample or a small one? Does it use a comparison group? Are there potential conflicts of interest? Is the outcome measured objectively or by self-report? Record these observations in your extraction template so you remember them when you synthesize.

Extract and organize data from sources

You now have a set of sources that meet your criteria. The next step is to pull the information you need from each one in a consistent format. This allows you to compare across sources and spot patterns.

### Create an extraction template

Design a table or form that captures the information you will need to answer your research question. Data extraction typically includes study design, population characteristics, interventions or exposures, outcomes, key findings, and any indicators of quality or bias.

For your question about distributed decision-making and remote teams, your template might include columns for: study identifier (author, year), study design (experimental, observational, qualitative), team size, industry or context, decision-making structure (centralized vs. distributed), how team performance was measured, the main finding, sample size or number of participants, and any limitations the authors noted.

Keep your template simple. You want to extract enough information to answer your question and assess quality, but not so much that extraction becomes overwhelming. Plan early using your review question and inclusion criteria to shape the data extraction form. The form should be piloted with a few studies to uncover gaps or unclear fields. Piloting is the process of completing data extraction for a select number of studies so you can evaluate the process before extraction starts across all studies. Read through two or three sources using your template. Do you have a place to record everything you need? Are the column headings clear? Revise before you extract from all sources.

### Record source details and findings

For accuracy in data extraction, two or more people should extract data from each study. If you are working alone, extract from each source twice, with time between extractions, so you catch your own errors.

As you extract, be precise. If a study reports that distributed teams made decisions 20% faster, record "20% faster" not "faster". If a study does not report a specific finding, leave the cell blank rather than guessing. You are building a dataset that you will analyze; inaccuracy now compounds later.

When you finish extracting from all sources, review your completed template. Look for missing data. If several sources did not report a particular outcome, note that; it may be a gap in the literature. If one source reported something very different from all the others, flag it for closer inspection during synthesis.

Synthesize findings to identify patterns and gaps

Synthesis is where research becomes insight. You now have extracted data from all your sources. The task is to organize that data so patterns emerge and gaps become visible.

### Group sources by theme or finding

Look at your extraction template and group sources by what they found or how they approached the question. Some sources might show that distributed decision-making speeds up team response; others might show it reduces decision quality; others might show no effect. Some might study software teams; others might study manufacturing or healthcare. Some might use quantitative methods; others qualitative.

Create a summary table or narrative that organizes sources by these themes. This is not a list of every source; it is a synthesis that shows what the evidence collectively says. Suppose you found ten sources: four showed that distributed decision-making improved speed, three showed it improved quality, two showed mixed effects, and one found no effect. Organize your synthesis around these findings. Within each group, note which sources agree, which disagree, and what might explain the disagreement (different populations, different measures, different time frames).

This organization reveals what the research consensus is, where evidence conflicts, and which findings are supported by many studies versus only one or two.

### Identify gaps and limitations

Gaps are questions the literature does not yet answer. They emerge when you notice that most sources studied one population but not another, or measured one outcome but not another, or used one method but not another.

Limitations are weaknesses in the existing evidence that reduce how much you can trust it. They include small sample sizes, lack of comparison groups, potential conflicts of interest, or methodological choices that introduce bias. Errors such as poor organization and inappropriate selection of sources can compromise the rigour and trustworthiness of an investigation.

As you synthesize, note both. If all sources on your topic studied teams in tech companies, that is a gap: we do not know whether the findings apply to other industries. If most sources measured performance by self-report rather than objective metrics, that is a limitation: self-report is subject to bias. If no source studied the long-term effects of distributed decision-making, that is a gap. These observations become part of your review's conclusion and shape what future research should address.

Recognize common pitfalls and how to avoid them

Literature reviews fail for predictable reasons. Knowing them in advance helps you avoid them.

Scope creep: Your question starts narrow and gradually expands. You begin studying distributed decision-making in remote teams, then add office-based teams, then add decision quality, then add employee satisfaction. Before long, you are trying to answer five questions at once and your review becomes unfocused. Lack of clear research questions or objectives can result in an unfocused and overly broad literature review. Write your question down and refer to it every time you consider adding a source. If the source does not address your specific question, exclude it.

Cherry-picking: You read sources and unconsciously favor those that support what you already believe. A poorly defined search strategy can result in missing key studies or adding irrelevant ones. The antidote is a systematic search documented in advance. You commit to searching specific databases with specific terms before you know what you will find. This removes the temptation to search until you find what you want.

Using weak sources: Using incorrect or irrelevant sources is a frequent mistake that can undermine the credibility of a literature review, including relying on outdated, non-peer-reviewed, or non-scholarly sources. Set inclusion criteria before you screen and apply them consistently. If a source does not meet your criteria, exclude it even if it seems relevant.

Ignoring bias: Including biased studies in a systematic review can undermine its validity by biasing its overall result. Assess methodological quality for each source and note sources with high risk of bias. When you synthesize, weight high-quality evidence more heavily than low-quality evidence. Do not pretend all sources are equally trustworthy.

Verify your review is complete and defensible

Before you write, confirm that your review is systematic and reproducible. Use this checklist:

1. Do you have a written research question that is specific enough to guide source selection? Can someone else read it and know which sources belong in your review?

2. Did you search at least three databases? Did you document the search date, search terms, and result count for each database?

3. Did you create written inclusion and exclusion criteria before screening? Did you apply them consistently to every source?

4. Did you assess methodological quality for each source and note limitations? Can you explain why you weighted some sources more heavily than others?

5. Did you extract data from each source using the same template? Is your extraction consistent enough that someone else could verify it?

6. Did you organize your findings by theme or finding so patterns are visible? Can you show where evidence converges, conflicts, or is sparse?

7. Did you identify gaps in the literature and limitations in the existing evidence? Can you explain what questions remain unanswered?

8. Is your process documented so someone else could repeat it and find the same sources?

If you can answer yes to all eight questions, your review is ready to write. If you answer no to any, go back and complete that step. A review that passes this checklist is defensible because every choice is documented and reproducible.

Do this automatically

A researcher conducting a systematic literature review might spend 2-3 hours defining a precise research question, another 4-5 hours building and testing search strings across multiple databases, and then 1-2 hours per database documenting each search iteration, time that multiplies across the dozens of databases and search refinements a thorough review typically requires. Software can automate the systematic search, documentation, and initial synthesis of findings across multiple academic databases simultaneously, reducing the manual coordination and record-keeping that currently consumes the bulk of a literature review's preparation phase. Academia research surveys the literature, extracts the findings and methods described above, and attaches a citation to every claim it reports, across as many subjects as you need, with Mimir attaching a citation to every figure. It will not write your paper for you. It gathers and cites the evidence; the argument stays yours, and deliberately so. Every plan includes it, including the free one.

Start your first report and run it on something you already know well.

This article is general information and research commentary. It is not professional advice, and it does not account for your particular circumstances.