user-research-synthesis

/user-research-synthesis - Turn Interview Data Into Insights

When the PM types /user-research-synthesis, transform raw user interview notes, transcripts, and observations into actionable product insights.

Context Routing Logic (Internal - for Claude)

Automatic Context Checks: When this skill is invoked, immediately check:

| Source | Files/Folders | Search Terms | What to Extract | | ------------------ | ----------------------------------------------- | ------------------------------ | -------------------------------------- | | Existing Research | thoughts/shared/pm/research/.md | topic from chat, user segments | Previous findings to avoid duplication | | Related PRDs | thoughts/shared/pm/prds/.md | problem related to interviews | Problem framing and hypothesis | | Strategy Context | thoughts/shared/pm/frameworks/*.md | user segment, strategic fit | How findings ladder to strategy | | Previous Synthesis | thoughts/shared/pm/interviews/ | topic name | Past research to build on | | Interview Guides | thoughts/shared/pm/interviews/ | topic | What questions were asked |

Context Priority:

1. Raw interview data FIRST (always use verbatim quotes)
2. Related PRDs and problem statements SECOND
3. Previous research on related topics THIRD
4. Strategic context FOURTH

Cross-Skill Links:

• After synthesis → Link to /prd-draft to turn insights into feature spec
• If about competitor mentions → Link to /competitor-analysis
• If about retention → Link to /retention-analysis for churn patterns
• If informing strategy → Link to /write-prod-strategy

Step 0: Understanding Your Research Context

Before we synthesize, let me understand what you've learned...

Checking:

• thoughts/shared/pm/research/ for previous findings on this topic
• thoughts/shared/pm/prds/ for the problem statement
• Interview guides used: what were you trying to validate?
• Previous synthesis on related topics

Based on what I find, I'll show you:

What You Were Researching

Topic:

• [What problem or feature were you exploring?]
• [Your original hypothesis: what did you expect to find?]

Sample:

• [Who did you interview? # of participants, roles, segments]
• [Notable characteristics: power users? churn risk? non-users?]

Context from Previous Research:

• [What do we already know about this problem?]
• [What did you expect research to confirm/disprove?]

PM-Specific Diagnosis Questions

1. Research Quality: Are these high-quality interviews with right users, or preliminary conversations?
2. Saturation: How many interviews did you do? (5-8 = good, 3-4 = preliminary, 10+ = very deep)
3. Bias Risk: Are you interviewing customers who love you, or a balanced sample?
4. Surprise Factor: What surprised you? (Best insights often contradict expectations)
5. Action Threshold: Do you have enough evidence to make decisions, or need more research?

When to Use

• After completing 5-8 user interviews
• Processing customer feedback from multiple sources
• Synthesizing usability test results
• Analyzing support tickets or sales call notes
• Converting qualitative data into product decisions

How It Works

This is a 4-step process:

Step 1: Upload Your Raw Data

Step 2: Extract Key Observations

Step 3: Cluster Into Themes (Affinity Mapping)

Step 4: Generate Actionable Recommendations

Step 1: Upload Your Raw Data

When the PM types /user-research-synthesis, start with:

Let's turn your user research into actionable insights.

**What data do you have?**

Upload or paste any combination of:
- Interview transcripts (from Grain, transcription tool, or manual notes)
- Usability test recordings or notes
- Customer support tickets
- Sales call summaries
- Survey responses (open-ended)
- Slack messages from customer channels

You can upload multiple files or just paste everything into this chat.

**How many interviews/data points?**
[Let me know so I can gauge the scope]

**What were you trying to learn?**
(e.g., "Why users churn after the first week" or "Pain points in the onboarding flow")

What to Look For in the Data

As you upload, I'll automatically start flagging:

• Direct quotes - Verbatim user language (the most powerful stuff)
• Behavioral patterns - What users actually did (not what they said they'd do)
• Pain points - Explicit frustrations or workarounds
• Jobs to be done - What users are trying to accomplish
• Unexpected use cases - How they're using the product in ways you didn't anticipate
• Emotion signals - Moments of frustration, delight, confusion

Step 2: Extract Key Observations

Once I have your data, I'll say:

Great, I've reviewed [X] interviews/data points.

I'm going to extract individual observations - each one gets its own "sticky note."

This will take a few minutes. I'll create:
- User quotes (in their exact words)
- Observed behaviors (what they actually did)
- Pain points (explicit problems they mentioned)
- Workarounds (clever hacks they've built)
- Context (their role, goals, environment)

Processing now...

Output Format

I'll create a structured list like this:

## Observation #1

**Type:** Pain Point  
**User:** Marcus (PM, Spotify)  
**Quote:** "I have 47 voice memos on my phone that are just 'remember to follow up with design about X.' I never convert them."  
**Context:** Uses multiple task managers, frustrated with manual entry  
**Emotion:** Frustration (high)

## Observation #2

**Type:** Behavior  
**User:** Priya (PM, Notion)  
**Quote:** "Half my tasks come from casual hallway conversations."  
**Behavior:** Captures ideas verbally but loses them before writing down  
**Context:** Fast-paced startup environment

[... continues for all observations]

Quality Checks

As I extract, I'll flag:

• ⚠️ Vague statements - "Users want a better experience" (not actionable)
• ⚠️ Future predictions - "I would definitely use this" (unreliable)
• ⚠️ Leading question responses - If I detect the interviewer asked a leading question
• ✅ High-quality signals - Specific stories, concrete examples, strong emotion

Step 3: Cluster Into Themes (Affinity Mapping)

After extraction, I'll say:

I've extracted [X] observations from your data.

Now I'm going to group these into themes using affinity mapping.

I'll look for:
- Patterns that appear across multiple users
- Contradictions (where users disagree)
- Underlying needs beneath surface-level requests
- Jobs-to-be-done that aren't being solved

Clustering now...

The Clustering Process

I'll create theme clusters like this:

## Theme 1: "Task Capture Friction"

**Frequency:** 6 out of 8 users mentioned this  
**Severity:** High (blocks daily workflow)

**Key Observations:**

- [Observation #1: Marcus voice memos]
- [Observation #2: Priya hallway conversations]
- [Observation #5: Jake loses tasks from Slack]
- [Observation #12: Sarah Excel spreadsheet workaround]

**Pattern:**
Users capture tasks in the moment (voice, text, conversation) but face friction converting them into their task manager. The "structuring" step is the bottleneck.

**Direct Quotes:**

- "I never convert them." - Marcus
- "Then 3 days later I'm like... wait, what API thing?" - Priya

**Jobs-to-be-done:**
When I have a spontaneous task idea, I want to capture it instantly without thinking about structure, so that I don't lose track of commitments.

---

## Theme 2: "Context Loss Between Tools"

**Frequency:** 5 out of 8 users  
**Severity:** Medium

[... continues for each theme]

Handling Contradictions

When users disagree, I'll explicitly c