borghei/Claude-Skills / project-management/discovery/interview-synthesis
interview-synthesis
VerifiedTurns customer interview notes into evidenced themes, an opportunity tree and follow-up questions. Use when you have just run interviews, need to show what customers said, or must find evidence gaps before the next round.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 8.9 KB |
| assets/interview_input_template.json | config | 1.8 KB |
| assets/opportunity_tree_template.md | markdown | 2.4 KB |
| examples/wayfinder-8-interview-synthesis.md | markdown | 15.1 KB |
| references/red-flags.md | markdown | 13.1 KB |
| references/synthesis-framework-and-tooling.md | markdown | 7.6 KB |
| references/synthesis-methodology-guide.md | markdown | 8.1 KB |
| scripts/interview_synthesizer.pyexecutable | script | 25.0 KB |
references/synthesis-methodology-guide.md
View on GitHub# Interview Synthesis Methodology Guide
This guide bundles three complementary methodologies into a single end-to-end synthesis playbook: Teresa Torres' Opportunity Solution Tree, Steve Portigal's interview-listening discipline, and Alan Klement's Jobs-To-Be-Done synthesis.
---
## 1. Why a structured synthesis matters
Raw interview notes have a half-life of about a week. Without a structured synthesis pass, teams remember the loudest quote rather than the strongest evidence, and they jump straight to solutions before mapping the opportunity space.
The synthesis discipline answers four questions:
1. **What did we actually hear?** (Snippets)
2. **What patterns repeat?** (Themes)
3. **What opportunities does this open?** (Opportunity tree)
4. **What evidence is still missing?** (Follow-ups)
A team that answers these four questions before any backlog work happens reduces feature-build-and-toss waste dramatically.
---
## 2. Snippet extraction (Portigal)
Steve Portigal, in *Interviewing Users*, argues that the interviewer's job is to listen for four signal types:
### Stories
A story has a time anchor ("Last Wednesday"), a specific context, an action sequence, and an outcome. Stories carry the highest evidence weight because the participant is reconstructing real behavior rather than imagining a future.
> Good story snippet: "Last Wednesday I was closing the month. I pulled the bank feed at 4pm, but two transactions were duplicates from the prior period, so I had to manually re-key the entire ledger. I left at 9pm and still was not sure the numbers were right."
### Contradictions
Contradictions appear when a participant claims one thing but their stories show another. They are the richest source of latent needs.
> Contradiction: "I said earlier that I trust the automation, but then I described checking every entry manually for the first three weeks. That's not trust -- that's verification."
### Surprises
Anything that violates the interviewer's mental model. Surprises are honesty signals: the participant said something the team did not script.
### Emotions
Frustration, delight, fear, and relief mark motivational intensity. Pay attention to the moments where the participant's tone changes.
### What to drop
Drop hypothetical-future statements ("I would use X if..."). They are low evidence. Drop solution suggestions ("You should add..."). Those belong in a separate ideation backlog, not the synthesis.
---
## 3. Coding (Klement-flavored JTBD)
Alan Klement, in *When Coffee and Kale Compete*, proposes that every meaningful customer behavior can be decomposed into:
- **Situation** -- The context that triggers the job
- **Motivation** -- The change the customer wants to make
- **Outcome** -- The improved state they expect
The coding pass tags each snippet with:
| Code Type | Question Answered | Example |
|-----------|-------------------|---------|
| Need | What underlying need does this point to? | "trust-in-automated-data" |
| Job | What job-to-be-done is invoked? | "When closing the month, I want to verify automated matches, so I can sign off without re-keying" |
| Pain | What friction or workaround appears? | "manual-reconciliation-of-duplicates" |
| Gain | What positive outcome would matter? | "audit-trail-per-rule" |
| Strength | How well-evidenced is this snippet? | 1 (single mention) / 2 (multi-participant) / 3 (multi-participant + behavioral) |
Coding is best done by 2 people independently, then reconciled. Single-coder synthesis is faster but biased.
---
## 4. Theme clustering
A theme is a coherent group of coded snippets. The discipline:
- **Threshold:** >=3 snippets, from >=2 participants
- **Headline:** 1 sentence that a stakeholder can read in 5 seconds
- **Evidence trail:** Each theme links back to its source snippets
### Worked example
Snippets coded with `pain = manual-reconciliation` and `need = trust-in-data`:
1. "I re-key every duplicate manually." (P1)
2. "I do not trust the matches until I check them." (P3)
3. "The automation flags conflicts but does not explain why." (P3)
4. "I export to Excel and rebuild." (P5)
**Theme headline:** *Finance leads do not trust automated matches because they cannot inspect the rule that produced each match.*
This headline is actionable: it points at an opportunity (visible rule trace) without prescribing a solution.
---
## 5. Opportunity Solution Tree (Torres)
Teresa Torres' tree has four mandatory layers:
```
Outcome
|
+-- Opportunity (customer need)
+-- Solution (candidate)
+-- Experiment (validation)
```
### Outcome rules
- One outcome per tree (you can build multiple trees for multiple outcomes)
- The outcome is **measurable**: "Reduce time to first import from 14 days to 3 days"
- Not a feature: "Add onboarding wizard" is wrong; "Reduce time to first import" is right
### Opportunity rules
- Customer-side framing only
- Traces to >=1 themed insight (no opportunities invented from thin air)
- Mutually exclusive when possible (overlapping opportunities suggest the outcome is too broad)
- 3-7 opportunities per outcome (more = re-scope; fewer = under-explored)
### Solution rules
- Multiple solutions per opportunity (the tree exists to compare alternatives)
- Each solution will need a validation experiment before commitment
- Solutions are de-prioritized at this stage; opportunity selection comes first
### Common anti-patterns
- **Pet-feature tree** -- The team puts their favorite solution at the top and works backward. The fix: start from outcome and themes; never start from solution.
- **Theme = opportunity** -- Themes describe what is true today; opportunities describe what the customer needs. They are related but not identical.
- **One solution per opportunity** -- Defeats the purpose of the tree. Force >=2 solutions per opportunity, even if one is "do nothing".
---
## 6. Follow-up question generation
After synthesis, identify:
- **Weak-evidence themes** (strength <= 2) that the team wants to commit on
- **Unmapped assumptions** -- things the team believes but no interview snippet supports
- **Surprises** -- moments that violated the team's model and need confirmation
For each, write 2-3 follow-up questions for the next interview round.
### Good follow-up patterns
- **Story-prompt:** "Tell me about the last time you [behavior]."
- **Contradiction-probe:** "Earlier you said X, but you also described Y. Can you walk me through that?"
- **Counterfactual:** "If [current workaround] disappeared tomorrow, what would you do?"
- **Day-in-the-life:** "Walk me through your morning yesterday from start to finish."
### Anti-patterns to avoid
- "Would you use this?" (leading, opinion-based)
- "Do you like X?" (closed, low signal)
- "How important is X on a scale of 1-10?" (false precision)
---
## 7. Synthesis ceremony (4-hour workshop)
A repeatable synthesis ceremony with the product trio (PM, Design, Engineering):
| Time | Activity | Output |
|------|----------|--------|
| 0:00-0:30 | Walk-through of interview tags and audio clips | Shared context |
| 0:30-1:30 | Independent snippet extraction (each person, every interview) | Snippet wall |
| 1:30-2:00 | Reconcile coding (resolve disagreements) | Coded snippet wall |
| 2:00-2:45 | Theme clustering (affinity diagram) | Themed clusters |
| 2:45-3:30 | Build opportunity solution tree | Draft tree |
| 3:30-4:00 | Generate follow-ups + next-step decisions | Action items |
Run this every 2 weeks during continuous discovery. The cadence keeps insights fresh and the tree alive.
---
## 8. Output and handoff
The synthesis produces three artifacts that flow into downstream PM skills:
| Artifact | Consumed by |
|----------|-------------|
| Themed insights | `create-prd/` (Background, Market Segments sections) |
| Opportunity solution tree | `outcome-roadmap/`, `prioritization-frameworks/` |
| Job-codes (Klement format) | `job-stories/` (When/Want/So format) |
| Follow-up questions | Next discovery round, `identify-assumptions/` |
---
## References
- Torres, T. (2021). *Continuous Discovery Habits*. Product Talk LLC.
- Portigal, S. (2013, 2nd ed. 2023). *Interviewing Users*. Rosenfeld Media.
- Klement, A. (2018). *When Coffee and Kale Compete*. NYC Publishing.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |