Product-org-os interview-synthesis

Document Intelligence

This skill supports three modes: Create, Update, and Find.

Mode Detection

Signal	Mode	Confidence
"update", "add interviews", "incorporate" in input	UPDATE	100%
File path provided with "update"	UPDATE	100%
"create", "new", "synthesize" in input	CREATE	100%
"find", "search", "list syntheses"	FIND	100%
"the synthesis", "latest research"	UPDATE	85%
Interview notes/transcripts provided directly	CREATE	90%

Threshold: >=85% auto-proceed | 70-84% state assumption | <70% ask user

Mode Behaviors

CREATE: Synthesize provided interviews into a complete research findings document.

UPDATE:

1. Check document registry first, then search user's structure
2. Preserve existing themes and quotes
3. Integrate new interview data, update pattern counts
4. Show what changed: new themes, strengthened patterns, contradictions
5. Update participant count and coverage metrics

FIND: Check registry, then search user's folders for synthesis documents.

Search Locations

• {Product}/Product/research/

• {Product}/Product/interviews/

• context/documents/index.md

• context/feedback/

---

Gotchas

• Direct quotes are evidence; your interpretation is analysis — keep them separate
• Sample size matters — always state how many interviews inform each finding
• Don't over-generalize from passionate outliers — frequency matters more than volume of feedback

Vision to Value Phase

Phase 1: Strategic Foundation -- Synthesizing customer research into actionable insights that inform strategic decisions and opportunity identification.

Methodology

The Synthesis Pipeline

<!-- Source: Thematic analysis methodology from Virginia Braun & Victoria Clarke, "Using Thematic Analysis in Psychology" (2006). Six-phase process: familiarize, generate codes, search themes, review themes, define themes, produce report. Adapted for product management context. --> <!-- Source: Teresa Torres, "Continuous Discovery Habits" (2021). Weekly interview rhythm, opportunity identification through interview snapshots, and the experience map approach. Key insight: synthesis should happen continuously, not in batches. --> <!-- Source: Inspired by phuryn/pm-skills `summarize-interview` pattern and Anthropic's knowledge-work-plugins `user-research-synthesis` structure. Also draws from product-on-purpose/pm-skills interview synthesis approach. -->

Step 1: Capture Raw Data

For each interview, extract:

Element	What to Capture
Participant	Role, segment, tenure, context (anonymize as P1, P2...)
Verbatim quotes	Exact words, with context. These are gold.
Behaviors	What they actually DO (not what they say they do)
Pain points	Frustrations, workarounds, complaints
Goals	What they're trying to achieve
Surprises	Anything unexpected or contradicting assumptions

Rule: Preserve exact quotes. Never paraphrase at the capture stage.

Step 2: Code the Data

<!-- Source: Affinity diagramming / affinity mapping. Origin: KJ Method (Jiro Kawakita, 1960s). Adapted for product research by Jeff Gothelf & Josh Seiden, "Lean UX" (2013). Group observations by similarity, then name the groups. -->

Tag each observation with codes:

Code Type	Purpose	Example
Descriptive	What it's about	"onboarding", "pricing", "reporting"
Emotional	How they feel	"frustrated", "delighted", "confused"
Behavioral	What they do	"workaround", "avoidance", "habit"
Evaluative	Their judgment	"essential", "unnecessary", "broken"

Step 3: Identify Themes

Group coded observations into themes:

• Strong theme: 3+ participants mention it independently
• Emerging theme: 2 participants mention it
• Signal: 1 participant mentions it, but it's notable

<!-- Source: Pattern strength thresholds adapted from Erika Hall, "Just Enough Research" (2013). Hall's pragmatic approach: you don't need statistical significance in qualitative research, but you do need pattern recognition discipline. -->

For each theme, document:

• Theme name and description
• Supporting quotes (verbatim)
• Number of participants who raised it
• Sentiment (positive, negative, neutral, mixed)
• Strength (strong, emerging, signal)

Step 4: Extract Insights

<!-- Source: The distinction between observations, patterns, and insights from IDEO's human-centered design methodology. An observation is what you see. A pattern is what recurs. An insight is the "why" that creates design opportunity. Also: Vijay Kumar, "101 Design Methods" (2012). -->

Transform themes into actionable insights:

Level	Definition	Example
Observation	What we saw/heard	"3 users mentioned they export data to Excel"
Pattern	What recurs	"Users don't trust the built-in reporting"
Insight	Why it matters	"Users need to validate data against their own mental models before sharing with stakeholders"
Opportunity	What to do	"Build customizable report templates that match existing Excel workflows"

Step 5: Connect to Strategy

<!-- Source: Teresa Torres, "Continuous Discovery Habits" (2021). Opportunity Solution Tree — connect research findings to product opportunities, which connect to desired outcomes, which connect to business objectives. -->

Map insights to:

• Existing assumptions (validate/invalidate from

  /assumption-map

  )
• Active strategic bets (support/challenge from

  /strategic-bet

  )
• Opportunity tree branches (from

  /opportunity-tree

  )
• Known feedback themes (from

  /feedback-recall

  )

Output Structure

# Interview Synthesis: [Research Topic/Sprint]

**Date**: [YYYY-MM-DD]
**Researcher**: [Name/Role]
**Product**: [Product name]
**Related**: [SB-YYYY-NNN, A-NNN, or initiative name]

## Research Overview

**Interviews conducted**: [N]
**Date range**: [Start] to [End]
**Participant segments**: [Segment 1 (N), Segment 2 (N), ...]
**Research questions**:
1. [Primary question]
2. [Secondary question]
3. [Secondary question]

## Participant Summary

| ID | Segment | Role | Tenure | Key Context |
|----|---------|------|--------|-------------|
| P1 | [Segment] | [Role] | [Time using product/in role] | [Relevant context] |
| P2 | [Segment] | [Role] | [Time] | [Context] |

## Themes

### Theme 1: [Theme Name] -- [Strong/Emerging/Signal]

**Participants**: P1, P3, P5 ([N]/[Total])
**Sentiment**: [Positive / Negative / Neutral / Mixed]

**Key quotes**:
> "Exact quote from participant" -- P1
> "Another exact quote" -- P3

**Observation**: [What we saw]
**Pattern**: [What recurs across participants]
**Insight**: [Why this matters -- the deeper understanding]

### Theme 2: [Theme Name] -- [Strong/Emerging/Signal]

...

## Contradictions & Tensions

| Tension | Side A | Side B | Possible Explanation |
|---------|--------|--------|---------------------|
| [Topic] | [View from some participants] | [Opposing view] | [Why they might differ] |

## Opportunities Identified

| # | Opportunity | From Theme | Confidence | Impact |
|---|------------|------------|------------|--------|
| O-1 | [Opportunity statement] | Theme 1 | High/Med/Low | High/Med/Low |
| O-2 | [Opportunity statement] | Theme 2 | High/Med/Low | High/Med/Low |

## Assumption Validation

| Assumption | Status | Evidence |
|-----------|--------|----------|
| [A-NNN: Assumption text] | Validated / Invalidated / Inconclusive | [What interviews revealed] |

## Recommendations

### Act On (High confidence, high impact)
1. [Recommendation with supporting evidence]

### Investigate Further (Emerging signals)
1. [What needs more research and why]

### Monitor (Weak signals worth tracking)
1. [Signal and what to watch for]

## Raw Quote Bank

[All notable quotes organized by theme, for future reference and stakeholder presentations]

### [Theme 1]
> "Quote" -- P1, [context]
> "Quote" -- P3, [context]

### [Theme 2]
> "Quote" -- P2, [context]

## Methodology Notes

- Interview format: [Structured / Semi-structured / Unstructured]
- Duration: [Average length]
- Recruitment: [How participants were found]
- Known limitations: [Sample bias, missing segments, etc.]

Instructions

1. Preserve verbatim quotes. Never paraphrase participant language. Exact words carry meaning that summaries lose.
2. Accept input as: raw notes, transcripts,

   @file.md

   references, or pasted text. Multiple interviews can be provided at once.
3. Anonymize participants (P1, P2...) unless the user specifies otherwise. Include enough context (segment, role) to make quotes meaningful.
4. Be honest about theme strength. A single mention is a signal, not a theme. Don't over-interpret small samples.
5. The "Contradictions & Tensions" section is critical. Real research always has contradictions. If your synthesis has none, you're not looking hard enough.
6. Connect findings to existing strategic context via

   /context-recall

   and

   /feedback-recall

   before finalizing.
7. Run

   /feedback-capture

   for any strong quotes or insights that should be in organizational memory.
8. Save syntheses to

   {Product}/Product/research/

   or

   {Product}/Product/interviews/

   .
9. Offer to create an

   /opportunity-tree

   from the identified opportunities.
10. When updating an existing synthesis with new interviews, clearly mark what's new vs. what was previously identified.

Integration

• Feeds from: Raw interview notes/transcripts,

  /context-recall

  (existing knowledge)
• Feeds into:

  /opportunity-tree

  (opportunities),

  /assumption-map

  (validation),

  /feedback-capture

  (quotes into org memory),

  /context-save

  (key insights)
• Related to:

  /experiment-design

  (testing insights),

  /prd

  (requirements from research)
• Connects to:

  /feedback-recall

  (finding related existing feedback)

Vision to Value Operating Principle

"Customer interviews are the highest-fidelity input a product organization has. A good synthesis turns hours of conversation into minutes of actionable clarity."