synthesis-frameworks

What this skill does

# Synthesis Frameworks

Frameworks for synthesizing qualitative research data to identify patterns, extract insights, and drive product decisions.

## Overview

Research synthesis transforms raw observations into actionable insights. Good synthesis reveals hidden patterns, validates assumptions, and informs strategic product decisions through structured analysis and collaborative sense-making.

## When to Use This Skill

**Auto-loaded by agents**:

- `research-ops` - For thematic analysis, insight generation, and research reports

**Use when you need**:

- Analyzing interview transcripts
- Synthesizing usability test findings
- Identifying user needs and pain points
- Creating research deliverables
- Documenting and communicating insights
- Running team synthesis workshops
- Pattern recognition across research data
- Translating research into actionable recommendations

## Core Synthesis Methods

### The Synthesis Process

Research synthesis follows a structured five-step process from raw data to actionable insights:

**The five steps:**

1. Immersion: Consume all raw data without judgment
2. Extraction: Identify individual observations atomically
3. Grouping: Find patterns through affinity mapping
4. Insight Generation: Ask "so what?" and connect to implications
5. Communication: Share findings in actionable formats

**Detailed template:** See `assets/synthesis-process-template.md` for step-by-step instructions, time allocation, and specific techniques for each phase.

**Time allocation:** Expect 8-16 hours total for a typical research project, with most time in grouping (30-40%) and insight generation phases.

### Affinity Mapping

Collaborative technique for organizing observations into themes through physical or digital clustering.

**When to use:**

- Team synthesis workshops
- Large amounts of qualitative data
- Need for shared understanding
- Pattern discovery (not testing hypothesis)

**Basic process:**

1. Silent note writing: Individual observation capture
2. Wall posting: Share all observations
3. Silent grouping: Cluster related notes
4. Name clusters: Identify themes
5. Generate insights: Discuss implications

**Complete guide:** See `assets/affinity-mapping-guide.md` for:

- Setup instructions (physical and digital)
- Step-by-step process with timings
- Color-coding strategies
- Tips for effective clustering
- Remote and solo variations

### Pattern Identification

Systematic approach to recognizing recurring themes across qualitative data.

**Five types of patterns to look for:**

1. **Frequency patterns:** How often something occurred (e.g., "8/10 users mentioned pricing")
2. **Behavioral patterns:** What users actually did vs. said (e.g., "All users skipped tutorial")
3. **Sentiment patterns:** Emotional responses (e.g., frustration, delight, confusion)
4. **Segment patterns:** Differences by user type (e.g., new users vs. power users)
5. **Journey patterns:** When issues occur (e.g., "70% dropped off at onboarding step 3")

**Deep dive:** See `references/pattern-identification-guide.md` for:

- Detailed examples of each pattern type
- How to interpret patterns
- Pattern validation techniques
- Common pitfalls to avoid
- Advanced pattern recognition methods

### Thematic Analysis

Formal method for coding qualitative data and identifying themes.

**Three coding levels:**

- **Level 1 - Descriptive codes:** What was said/done (e.g., "Mentioned pricing")
- **Level 2 - Interpretive codes:** What it means (e.g., "Price sensitivity")
- **Level 3 - Pattern codes:** Big picture themes (e.g., "Value perception issue")

**Four-phase process:**

1. Open coding: Generate initial codes liberally (30-100+)
2. Axial coding: Group related codes into categories (10-30)
3. Selective coding: Refine into core themes (5-10)
4. Validation: Ensure themes are supported and distinct

**Complete methodology:** See `references/thematic-analysis-guide.md` for:

- Detailed coding process with examples
- Tools and techniques (manual and digital)
- Quality checks and validation
- Common coding patterns
- Worked examples from transcripts

### Jobs-to-be-Done (JTBD) Synthesis

Framework for understanding customer motivations and the "jobs" users hire products to do.

**Job statement format:**

```
When [situation],
I want to [motivation],
So I can [expected outcome]
```

**The Forces Framework:**
Analyze four forces that drive or prevent adoption:

- **Push:** Why change from current solution? (pain, frustration)
- **Pull:** Why choose new solution? (benefits, aspirations)
- **Anxieties:** What holds users back? (fear, risk, learning curve)
- **Habits:** Why stay put? (familiarity, switching costs)

**Switching equation:** Push + Pull > Anxieties + Habits = Switch

**Full framework:** See `references/jtbd-synthesis-guide.md` for:

- How to identify jobs in research data
- Finding jobs through struggle and workarounds
- Complete forces analysis with examples
- Strategy recommendations for each force
- JTBD vs traditional research approaches

## Insight Quality

Not all findings are insights. Good insights are specific, surprising, actionable, evidence-based, and prioritized.

### The Five Characteristics

**1. Specific (not vague):**

- Bad: "Users don't like the UI"
- Good: "6/8 users couldn't find Settings in hamburger menu - move to top nav"

**2. Surprising (not obvious):**

- Bad: "Users want fast loading"
- Good: "Users prefer slow load with progress over fast load with no feedback"

**3. Actionable (clear implications):**

- Bad: "Users are frustrated"
- Good: "8-field signup causes 60% abandonment - reduce to 3 fields to match competitors"

**4. Evidence-based (not speculation):**

- Bad: "I think users would like dark mode"
- Good: "5/10 users requested dark mode unprompted, citing eye strain from all-day use"

**5. Prioritized (impact + feasibility):**

- Bad: List of 50 equal findings
- Good: Top 5 insights ranked by impact/effort with quick wins called out

### Insight Formula

**Template:** "[X%/number] of users [did/said specific thing] because [reason], suggesting we should [action] to [expected outcome]"

**Example:** "8/10 users abandoned onboarding at 'Invite Team' step because they didn't have emails ready, suggesting we should make this optional to reduce abandonment from 60% to 30%"

**Quality guide:** See `references/insight-quality-guide.md` for:

- Detailed breakdown of each characteristic
- Bad vs. good insight examples
- How to strengthen weak insights
- Insight quality checklist
- Teaching teams to recognize quality

## Synthesis Deliverables

### Executive Summary

One-page research summary for the team who need highlights without detail.

**Standard sections:**

- Research goals and methodology (brief)
- Top 3-5 key findings with evidence
- Prioritized recommendations (quick wins vs. strategic)
- Next steps and owners

**Template:** See `assets/executive-summary-template.md`

### Full Research Report

Comprehensive documentation (5-10 pages) for complete findings and recommendations.

**Structure:**

1. Executive summary
2. Background and research questions
3. Methodology and participants
4. Findings by theme (with quotes and data)
5. Recommendations (prioritized)
6. Appendix (full participant list, materials, extra quotes)

**Template:** See `assets/research-report-template.md`

### Insight Cards

One-page format for individual insights that can be shared independently.

**Includes:**

- Insight statement and evidence (quotes + data)
- Impact and why it matters
- Specific recommendation with expected outcome
- Priority, effort, and owner
- Visual (screenshot, diagram, or clip)

**Template:** See `assets/insight-card-template.md`

### Data-Driven Personas

User archetypes based on behavioral patterns discovered in research.

**Creation process:**

1. Identify behavioral patterns in research
2. Cluster users with similar behaviors
3. Define 3-5 distinct archetypes
4. Validate against real users

**Persona template includes:**

- Goals