Literature Review Synthesis

What This Does

This playbook provides a structured workflow for conducting literature reviews. Claude helps you search for relevant papers, extract key information from each, identify themes and gaps, and produce a synthesized review document.

Prerequisites

• Claude Code installed and configured
• Access to papers (PDFs, URLs, or databases)
• A research topic or question

Synthesis Matrix

Track themes across papers:

Paper	Theme A	Theme B	Theme C	Methods	Findings
Smith 2020	✓	-	✓	Survey	X > Y
Jones 2021	✓	✓	-	Experiment	X = Y
Lee 2022	-	✓	✓	Interview	X < Y

Commands

• /lit-review [topic] — Start a new literature review
• "Analyze this paper: [citation/pdf]" — Extract from single paper
• "Synthesize papers on [theme]" — Combine extractions
• "Find gaps in the literature" — Identify what's missing
• "Write the introduction section" — Draft based on synthesis

Output Formats

Annotated Bibliography

1. Smith (2020): [2-3 sentence summary focusing on contribution and relevance]

2. Jones (2021): [2-3 sentence summary...]

Thematic Review

## Theme 1: [Name]
[Paragraph synthesizing what multiple papers say about this theme]

## Theme 2: [Name]
[Paragraph synthesizing...]

Gap Analysis

## What We Know
[Summary of established findings]

## What We Don't Know
[Identified gaps and unanswered questions]

## Why It Matters
[Significance of the gaps]

## Step-by-Step Setup

### Step 1: Create the research folder

```bash
mkdir -p ~/research/{papers,extractions,synthesis}

Step 2: Add the CLAUDE.md template

Save the template above to ~/research/CLAUDE.md.

Step 3: Start a review

I'm conducting a literature review on [your topic]. Help me:
1. Define the search scope
2. Generate search queries
3. Create an extraction template specific to this topic

Example Usage

User:

"Start a literature review on remote work productivity"

Claude:

Literature Review: Remote Work Productivity

Phase 1: Scope Definition

Research Question: How does remote work affect employee productivity compared to in-office work?

Inclusion Criteria:

• Published 2019-2024 (post-COVID relevant)
• Empirical studies with productivity measures
• Knowledge workers / office-type roles

Exclusion Criteria:

• Opinion pieces without data
• Non-comparable job types (manufacturing, retail)
• Studies without clear productivity metrics

Search Terms:

• "remote work" AND "productivity"
• "work from home" AND "performance"
• "telework" AND "output"

Ready for Phase 2: Search & Collect. Do you have papers to analyze, or should we discuss search strategies?

Later - Analyzing a paper:

User: "Analyze this paper: Bloom et al. (2015) 'Does Working from Home Work?'"

Claude:

Paper: Bloom 2015 - WFH Experiment

Citation: Bloom, N., Liang, J., Roberts, J., & Ying, Z. J. (2015). Does working from home work? Evidence from a Chinese experiment. Quarterly Journal of Economics.

Type: Randomized experiment

Research Question

Does working from home improve or harm worker productivity?

Methodology

• 9-month RCT at Chinese call center (Ctrip)
• 249 employees randomized to WFH vs office
• Productivity = calls handled per minute

Key Findings

1. WFH increased productivity by 13%
2. 9% from working more minutes (fewer breaks, less sick leave)
3. 4% from more calls per minute (quieter environment)
4. WFH workers reported higher satisfaction

Limitations

• Single company in China
• Call center work (measurable output)
• Pre-COVID (voluntary WFH, not forced)

Relevance

Foundational experimental evidence for productivity gains, but limited to specific work type and context.

Later - Synthesis:

User: "I've analyzed 8 papers. Synthesize findings on productivity measurement."

Claude:

Synthesis: Productivity Measurement Approaches

The literature uses diverse productivity measures:

Objective Measures (4 papers):

• Bloom 2015: Calls per minute
• Emanuel 2023: Lines of code, commits
• Gibbs 2023: Tasks completed per hour

Self-Reported Measures (3 papers):

• Barrero 2021: Perceived productivity scale
• Aksoy 2022: Compared to office baseline

Manager Assessments (2 papers):

• Butler 2021: Performance reviews

Key Insight: Papers using objective measures show larger productivity gains (avg +8-13%) than self-reported (+5-8%). This may reflect measurement bias or different constructs.

Gap Identified: No papers combine multiple measurement types to validate findings.

Tips

• Use parallel analysis: If you have 5 papers, ask Claude to analyze all 5, then synthesize. Don't do one at a time.
• Create a synthesis matrix: Track themes across papers visually. This reveals patterns.
• Quote sparingly but strategically: A few key quotes > many quotes.
• Watch for publication bias: Studies showing effects are more likely published. Note null findings.

Troubleshooting

Problem: Claude's summaries are too generic

Solution: Provide the full paper or detailed excerpts. Abstracts don't have enough detail for good extraction.

Problem: Papers seem contradictory

Solution: Look for differences in methodology, sample, context. Create a table comparing these factors.

Problem: Can't find enough papers

Solution: Broaden search terms, extend date range, include related fields. Use citation chaining (papers that cite key papers).