Skip to content

afeezadebayo/research-learning-skills / literature-review

literature-review

Verified

Review research evidence for work and study questions: question framing and inclusion criteria, review types from quick scans to systematic reviews, database searching and search strings, screening, data extraction, quality assessment, synthesis by theme and write-ups with confidence and gaps. Use whenever someone needs to know what research says about a topic, run a rapid evidence review, find academic studies, summarise papers, check a claim about a study or write an evidence-based recommendation.

Literature Review

When a question has been studied before, the fastest route to a reliable answer is to find and synthesise that research rather than relying on a few articles or one memorable study. A literature review gathers relevant studies, assesses their quality and summarises what the evidence shows, where it agrees, where it conflicts and what is still unknown. Systematic reviews in medicine and social science follow formal protocols such as PRISMA for transparent reporting. [Reference, PRISMA statement] Most work and learning needs a lighter but still disciplined version: a rapid evidence review.

This skill runs literature reviews for work, study and policy questions: defining the question and inclusion criteria, searching academic databases, screening studies, extracting key data, assessing quality, synthesising findings and writing a clear summary with confidence levels and gaps.

Operating principles

  1. Question and criteria first. Decide what counts as a relevant study before searching.
  2. Search beyond one database. Combine sources to avoid missing important work.
  3. Prefer reviews and strong designs. Systematic reviews and meta-analyses summarise many studies.
  4. Judge quality, not just results. Sample size, design and bias matter.
  5. Synthesise, don't list. Explain patterns across studies, not each paper in turn.
  6. Be transparent. Record searches, criteria and decisions so others can check them.

Review types

TypeTimeUse for
Quick evidence scanHoursOrientation; low-stakes questions
Rapid evidence reviewDays to two weeksBusiness and policy decisions
Narrative reviewWeeksBroad topics, theory, teaching
Systematic reviewMonthsDefinitive answers; academic and clinical work

Where to search

SourceStrengths
Google ScholarBroad coverage, citations, "cited by" links
PubMedHealth and life sciences
Semantic ScholarAI-assisted discovery and summaries
SSRN, arXiv, RePEcWorking papers in social science, economics, computing (check peer review status)
Cochrane LibrarySystematic reviews in health
University and think-tank repositoriesReports and evaluations

When to use, and when not to

Use it when a decision depends on what research shows, for evidence-based policies and practices, academic work, writing authoritative content and checking popular claims about studies.

Don't use it for questions about your own customers or markets that published research cannot answer; use interviews, surveys and data instead.

Inputs

  • The question and the decision or purpose.
  • Population, setting, intervention or factor and outcomes of interest.
  • Time range and languages.
  • Depth and deadline.

Workflow

Step 1. Frame the question

Use a structure such as population, intervention or exposure, comparison and outcome: "In remote software teams (population), does pair programming (intervention) compared with solo work (comparison) improve code quality (outcome)?"

Step 2. Set inclusion criteria

Study types, dates, populations, settings and languages to include or exclude.

Step 3. Build search strings

Key terms and synonyms combined with AND and OR; test and refine.

Step 4. Search and record

Run searches in chosen databases; record date, string and number of results.

Step 5. Screen

Titles and abstracts first, then full texts against criteria; record reasons for exclusion.

Step 6. Extract data

For each included study: citation, design, sample, setting, measures, main findings, effect size and limitations.

Step 7. Assess quality

Design strength, sample size, risk of bias, funding and conflicts, replication.

Step 8. Synthesise

Group findings by theme or outcome; describe consistency, size of effects, conditions where results differ and overall strength of evidence.

Step 9. Write up

Summary answer, confidence level, key studies, gaps and implications for the decision.

Search string example

("remote work" OR "telework" OR "work from home") AND (productivity OR performance) AND (software OR engineering)

Extraction table

| Study | Design | Sample | Setting | Outcome | Finding | Effect size | Quality | Notes |
| Author 2023 | RCT | 1,600 employees | Travel company | Productivity, retention | Hybrid no worse on performance; lower attrition | Small | Medium-high | One company |

Synthesis template

Question: <...>
Bottom line: <what the evidence suggests, in two sentences>
Confidence: <high/medium/low and why>
What studies agree on: <...>
Where results differ and why: <settings, methods, populations>
Gaps: <what has not been studied well>
Implications for our decision: <...>
Key sources: <5 to 10 strongest studies>

Finding more studies

  • Backward search: check the reference lists of key papers.
  • Forward search: use "cited by" to find newer work that built on them.
  • Look for recent systematic reviews first; they point to the main studies.
  • Search author names of leading researchers in the field.

Quality questions for each study

QuestionWhy it matters
Was there a comparison group?Without one, effects are hard to attribute
How were participants chosen?Self-selection can bias results
How large was the sample?Small samples give unstable results
Were outcomes measured objectively?Self-reports can inflate effects
Who funded it?Interests can shape design and reporting
Has it been replicated?Single findings may not hold

Using AI tools for reviews

AI tools can suggest search terms, find candidate papers, summarise abstracts and fill extraction tables. Verify that each paper exists, check key numbers against the full text and make screening and quality judgements yourself for included studies. Record which steps used AI.

Presenting evidence to non-experts

Lead with the bottom line and confidence, explain in plain words what kinds of studies support it, give one or two concrete examples and say what is still uncertain. Avoid listing papers; decision makers need the pattern and its strength, with references available for anyone who wants detail.

Update the review when major new studies appear.

Share the search log with the summary.

Edge cases and judgement calls

SituationWhat to do
Very few studiesSay so; widen criteria carefully or use expert input
Many low-quality studiesWeight by quality; do not count papers as votes
Conflicting findingsLook for differences in populations, methods and measures
Only preprintsLabel clearly; check for later peer-reviewed versions
Industry studiesInclude with caution; note sponsors
Popular claim not supportedExplain what research actually shows

Handoffs to related skills

  • Planning: research-planning
  • Source quality: source-evaluation
  • AI tools for papers: ai-assisted-research
  • Note-taking while reading: active-reading-notes
  • Synthesis: research-synthesis
  • Writing the report: the Workplace Writing reports skill

Worked example

An HR lead needed evidence on whether hybrid work affects productivity before a policy change. A rapid evidence review searched three databases with defined criteria, screened 140 results to 18 studies and extracted design, setting and findings. Synthesis found that well-designed experiments generally showed hybrid work did not reduce measured performance and often improved retention, while results for fully remote work varied by role and measurement. Confidence was labelled medium, gaps were noted for creative and junior roles, and the policy kept hybrid work with support for new starters.

Common mistakes

  • Searching one database with one phrase.
  • Listing papers instead of synthesising.
  • Treating every study as equal.
  • Ignoring dates and settings.
  • Not recording search and screening decisions.

Output format

1. Question and criteria
2. Search log
3. Screening summary (numbers and exclusion reasons)
4. Extraction table
5. Quality assessment
6. Synthesis with confidence and gaps

Self-check before you answer

  • Were the question and inclusion criteria set before searching?
  • Were several databases searched and logged?
  • Was study quality assessed?
  • Does the synthesis explain patterns, confidence and gaps?

Created by Afeez Adebayo, Marketing, SEO & Growth Expert Website: https://afeezadebayo.com LinkedIn: https://www.linkedin.com/in/afeez-adebayo/ X: https://x.com/za_icon Source: https://afeezadebayo.com/skills-md/productivity/research-learning/literature-review/

Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). You may copy, share and adapt this skill, including for commercial use, as long as you credit Afeez Adebayo and link to the source URL above. Keep this notice in any copy or adapted version.

deprecation-and-migrationaddyosmani/agent-skillsManages deprecation and migration. Use when removing old systems, APIs, or features. Use when migrating users from one implementation to another. Use when migrating a database schema in production, such as renaming or dropping a column without downtime (expand/contract). Use when deciding whether to maintain or sunset existing code.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
performance-optimizationaddyosmani/agent-skillsOptimizes application performance across frontend, backend, queries, and databases. Use when performance requirements exist, when you suspect performance regressions, when Core Web Vitals or load times need improvement, when N+1 query patterns need fixing, or when profiling reveals bottlenecks.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
agent-cost-reportthedotmack/claude-memBelievable agent cost report for any period, default the last 7 full days PT, not counting today. Measured tokens from Claude Code transcripts priced at OpenRouter list prices (ESTIMATED), measured provider spend when a sanctioned source exists, note-taker cost separate, Timing-style HTML/PDF plus report.json, line-items.csv, evidence.json.Stars: 118kUpdated Claude CodeCodexVerifiedInstall