Skip to content

borghei/Claude-Skills / product-team/research-summarizer

research-summarizer

Verified

Synthesize raw user research (interviews, surveys, tickets) into themed findings and decision-ready briefs. Use when synthesizing user interviews, building a findings brief, or communicating research to stakeholders.

references/research-synthesis-frameworks.md

View on GitHub
# Research Synthesis Frameworks

Practical reference for organizing raw qualitative research into themes
and insights.

## 1. The synthesis ladder

1. **Raw data** — transcripts, notes, recordings
2. **Observations** — atomic statements / quotes / behaviors
3. **Codes / tags** — categorical labels on observations
4. **Patterns** — clusters of related codes
5. **Themes** — interpreted groupings of patterns
6. **Insights** — explanatory statements about why
7. **Recommendations** — implied actions

Don't skip steps. Each one is verifiable; jumping breaks the trust chain.

## 2. Affinity diagramming

The most common synthesis technique. Steps:

1. Write each observation on a separate sticky note / row
2. Without grouping intent, place them on a wall / canvas
3. Move similar ones close together
4. Name emerging clusters
5. Refine; collapse, split, rename
6. Identify cross-cluster patterns

### When affinity works
- Smaller research (< 20 sessions)
- Co-located synthesis session with the team
- Need cross-functional buy-in

### When affinity struggles
- Remote teams without good digital tools
- Large data sets (use thematic analysis instead)
- Single-person synthesis (you'll bias)

## 3. Thematic analysis (Braun & Clarke)

Six phases:

1. **Familiarize** — read all data, re-read
2. **Generate codes** — descriptive labels per observation
3. **Search for themes** — cluster codes into candidate themes
4. **Review themes** — refine, merge, split
5. **Define and name themes** — write the theme statement
6. **Produce report** — final themes with evidence

### When thematic analysis works
- Larger data sets (20+ sessions)
- Mature research practice
- Replicable / auditable synthesis

### Key discipline
- Track which observations support which themes (always)
- Track which observations DON'T support (always — disconfirming evidence)
- Have a second coder for important studies

## 4. Jobs-to-be-Done (JTBD) framing

When the research goal is to understand motivations:

- **Job:** the progress the customer is trying to make
- **Outcome:** desired result from the job
- **Forces:** push / pull / anxiety / habit affecting adoption

Synthesis surfaces:
- Functional jobs (e.g., "submit my expense report")
- Emotional jobs (e.g., "look competent to my finance team")
- Social jobs (e.g., "not be the one who's always slow")

Useful when research informs product positioning, not just feature design.

## 5. Persona / segment synthesis

When research informs persona or segment definitions:

### Per persona / segment, capture
- Demographics + role + context
- Goals (jobs to be done)
- Pain points
- Behaviors / workarounds
- Quotes (representative)
- Sample size in your research

### Don't
- Define a persona from a single interview
- Smooth out conflicting evidence (note the variation)
- Confuse persona with audience segment (different uses)
- Forget the negative persona (who this is NOT for)

## 6. User journey synthesis

When research informs a journey map:

### Per stage of journey, capture
- What users are doing
- What users are thinking
- What users are feeling
- Pain points / friction
- Opportunities (gap between current and desired)

### Sources
- Interviews
- Support tickets (often best signal for pain points)
- Observational research
- Analytics (quantitative complement)

## 7. Themes — what makes a good one

A good theme:
- **Has a clear name** that captures the essence
- **Has a one-sentence statement** explaining the pattern
- **Has 3+ supporting observations** from different sessions
- **Cross-cuts** (not specific to one persona/segment unless that's the point)
- **Surprises** or **confirms strongly** (not "it depends")
- **Suggests action** (or at least the next question to ask)

A weak theme:
- "Users want a better experience"
- "Performance is important"
- "X is good for some and bad for others"

## 8. Tagging strategies

### Open coding (early)
- No predetermined tags
- Add tags as new patterns emerge
- Risk: lots of tags, hard to consolidate

### Closed coding (later)
- Predetermined tags from earlier rounds
- Risk: miss new patterns

### Hybrid (recommended)
- Start with framework tags (e.g., journey stage, sentiment)
- Add open tags as patterns emerge
- Review tag taxonomy periodically and consolidate

### Tools
- Manual: spreadsheet with row per observation
- Light tooling: Notion, Airtable, Coda
- Specialized: Dovetail, Condens, Lookback
- Spreadsheet + scripts works fine for most cases

## 9. Disconfirming evidence — actively seek it

For every emerging theme, ask:

- Are there sessions that contradict this?
- Are there segments where this doesn't hold?
- Did I sample enough to know if this is universal?

Reporting only confirming evidence misleads stakeholders. Report
disconfirmation explicitly (e.g., "5 of 8 users showed this; 2 actively
contradicted it; 1 was unclear").

## 10. Frequency vs intensity

Two dimensions of pattern strength:

- **Frequency:** how many users showed it
- **Intensity:** how strongly each user expressed it

A theme that 3 users mentioned with intense frustration may matter more
than one 8 users mentioned in passing.

Capture both.

## 11. Sample sizing reminders

For most product research:
- 5-8 users per segment surfaces 80%+ of usability issues
- 12-15 users surfaces most behavioral patterns
- 20+ users is needed to detect quantitative differences

Don't over-engineer N; you can run a follow-up round if the first is
inconclusive.

## 12. Common pitfalls

- **Synthesizing from memory.** Use the data; recall biases pattern detection.
- **Solo synthesis on important studies.** Bias amplifies.
- **Coding without re-reading.** First pass misses things.
- **Themes that are just feature requests.** Themes are about underlying user needs.
- **Equal weighting across all sessions.** Some sessions have more signal than others; note quality.
- **Skipping the cross-cut by segment.** Often the most interesting finding.
- **Burying disconfirming evidence.** Highlight it.
- **Insights without "so what."** If it doesn't change anything, it's data, not insight.
add-product-e2e-evalpaperclipai/paperclipAdd or extend a Paperclip full-stack runner E2E workflow, fixture, matcher, or report evidence path for local or Daytona execution.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
add-runner-evalpaperclipai/paperclipAdd or extend a Paperclip Runner protocol evaluation definition, roster, assertion, or report fixture with provenance and narrow validation.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
diagnose-why-work-stoppedpaperclipai/paperclipCompatibility entry point for failed, stalled, looping, or over-recovered Paperclip work. Use fix-failed-paperclip-run for the full diagnosis, regression, repair, real verification, and green-PR workflow; honor an explicitly requested diagnosis-only scope.Stars: 115kUpdated Claude CodeCodexVerifiedInstall
fix-failed-paperclip-runpaperclipai/paperclipDiagnose and fix failed, stuck, slow, or incorrectly blocked Paperclip tasks and runs. Use with a task/run URL or failure report to trace the root cause, prove a regression red then green, check second-order effects, verify the real workflow, and deliver a PR with passing checks.Stars: 115kUpdated Claude CodeCodexVerifiedInstall