feedback-synthesis

---
name: feedback-synthesis
description: When the user needs to analyze, categorize, or extract actionable insights from customer feedback across multiple sources, especially feature requests.
related: [user-research-synthesis, churn-analysis, prd-writing]
reads: [startup-context]

tags: [nontechnical, startup-founder-skills, feedback-synthesis]
----|-----------|-------------------|-------------------|---------------------|----------|
| [Theme] | [Count] | [Segments] | [Score] | [High/Med/Low] | [H/M/L] |

## Top 3 Opportunities (Deep Dives)
### Opportunity 1: [Theme Name]
- **Rationale:** Customer needs and strategic alignment
- **Representative quotes:** Direct user language
- **Alternative solutions:** Other ways to address this need
- **High-risk assumptions:** What must be true
- **Cheapest test:** How to validate with minimal effort

## Quick Wins
Actions that address frequent feedback with low implementation effort.

## Not Prioritized (and Why)
Themes explicitly deprioritized with reasoning.

## Appendix: Raw Data Summary
Breakdown by source, segment, and time period.
```

## Frameworks & Best Practices

### Opportunities Over Features
Never let customers design solutions. Prioritize opportunities (problems), not features. When a user says "I want a Gantt chart," the underlying opportunity might be "I need to visualize project timelines and communicate status to stakeholders." Always dig for the job-to-be-done behind the request.

### Opportunity Score (Dan Olsen)
Score each theme: Opportunity Score = Importance x (1 - Satisfaction), normalized to 0-1. This surfaces problems that are both important and underserved. A high-importance, high-satisfaction area is already well-served and should not be prioritized over a high-importance, low-satisfaction gap.

### Signal vs. Noise Rules
- **One customer saying it is not a pattern.** Require 3+ independent mentions of a theme before treating it as a signal. Exception: if the one customer is a whale account citing it as a churn risk.
- **Recency bias check.** A flood of recent feedback about one issue can overshadow a persistent problem. Always compare against the prior period.
- **Loudest does not equal most important.** Power users and vocal customers generate disproportionate feedback. Weight by segment size and revenue contribution, not volume alone.
- **Praise is data too.** Track what users love. Knowing your strengths prevents you from accidentally breaking them during a redesign.

### Assumption Testing
For each top-priority opportunity, identify the highest-risk assumption and design the cheapest possible test. Do not build the full solution to validate an assumption that could be tested with a prototype, survey, or Wizard of Oz experiment.

### Source-Specific Guidance
| Source | Strengths | Watch Out For |
|--------|-----------|---------------|
| Support tickets | High signal, specific problems | Skews toward bugs, misses satisfied users |
| NPS/surveys | Broad coverage, quantifiable | Low response rates can bias results |
| Feature request boards | Organized, vote counts available | Power users dominate voting |
| Sales call notes | Revenue-adjacent, prospect perspective | Prospects request features they may never use |
| App store reviews | Public, includes competitor comparisons | Skews negative, vague complaints |
| Social media | Unfiltered, real-time | Noisy, hard to segment |

### Avoiding Common Mistakes
- **Cherry-picking quotes** that support a pre-existing hypothesis. Present the full distribution, including contradictory feedback.
- **Conflating frequency with importance.** A low-frequency issue that causes churn matters more than a high-frequency annoyance users tolerate.
- **Delivering data without recommendations.** A theme map without action items is a report, not a synthesis. Always end with what to do next.
- **Ignoring the silent majority.** Users who never complain may be happy or disengaged. Segment analysis helps distinguish the two.

## Related Skills
- `user-research-synthesis` -- Chain when feedback analysis reveals gaps that need dedicated user research (interviews, usability tests).
- `churn-analysis` -- Chain when feedback themes correlate with churn patterns and need deeper retention analysis.
- `prd-writing` -- Chain when a clear opportunity emerges from the synthesis and needs to be specced into a PRD.

## Examples

### Example 1: Feature request prioritization
**User:** "Our feature request board has 150 items. Help me figure out what to build next quarter."

**Good output excerpt:**
> **Executive Summary:** 150 requests cluster into 9 themes. The top opportunity is not the most-requested feature (SSO, 34 votes) but the most underserved need: "real-time collaboration on shared documents" (Opportunity Score: 0.82). SSO scores lower (0.45) because existing workarounds satisfy most users adequately.
>
> **Opportunity 1: Real-time collaboration**
> - **Rationale:** 22 requests across 4 segments. Cited as expansion blocker in 3 enterprise deals worth $85K ARR. Current satisfaction: 2/10.
> - **Alternative solutions:** (a) Full real-time editing, (b) Lightweight commenting and presence indicators, (c) Async review workflow with notifications
> - **High-risk assumption:** Users want simultaneous editing, not just awareness of others' changes
> - **Cheapest test:** Add presence indicators only (show who is viewing a document) and measure whether collaboration-related tickets decrease

### Example 2: Multi-source synthesis
**User:** "We have 200 support tickets, 50 NPS responses, and notes from 10 customer interviews from last month. What are customers telling us?"

**Good output excerpt:**
> **Theme 1: CSV export broken for large datasets** (Opportunity Score: 0.91)
> - 47 support tickets, 8 NPS detractors, 3 interviews. Users hitting the 10K row limit work around it by splitting exports manually.
> - **Strategic alignment:** High -- data export is core to our "open platform" positioning.
> - **Cheapest test:** Not needed; this is a clear bug/limitation. Fix directly.
> - **Quick win:** Increase CSV export limit to 100K rows (engineering estimate: 2 days).