Voice of Customer Synthesis Generator

# Task
Synthesize voice-of-customer data into actionable insights. Use ONLY the provided feedback.

# Inputs
- Feedback sources: {feedback_sources}
- Customer segments: {customer_segments}
- Goal: {goal}
- Known themes: {known_themes}
- Constraints: {constraints}

# Anti-hallucination rules
1. Every theme must be supported by multiple data points from the feedback. Single mentions are anecdotes, not themes.
2. Do not infer customer intent beyond what the feedback states.
3. If a theme contradicts another, present both without resolving the tension.

# Synthesis structure
1. **Executive summary** (3-5 sentences): Top 2-3 findings with business implications.
2. **Themes** (5-8): For each theme:
   - Theme name and description (1-2 sentences)
   - **Signal strength**: Strong (5+ mentions, consistent across segments) / Moderate (3-4 mentions or one segment) / Emerging (2 mentions, worth watching)
   - **Segments affected**: Which customer segments raised this?
   - **Representative quotes** (2-3, attributed to segment if possible)
   - **Actionability**: Quick win (fix in days) / Needs research (investigate further) / Strategic (requires roadmap decision)
3. **Segment view**: For each customer segment, what are their top 2-3 concerns? Where do segments agree or diverge?
4. **Requests vs. friction vs. sentiment**:
   - Feature requests (what they want built)
   - Friction points (what's broken or hard)
   - Sentiment signals (love, frustration, churn risk)
5. **Recommended follow-ups** (3-5): Specific actions tied to themes. Each with owner suggestion (product, CS, marketing, engineering).
6. **Gaps**: What questions does this feedback NOT answer? What would you want to investigate next?

# Output
Return all 6 sections.

<source>{feedback_sources}</source>
<context>
  <customer_segments>{customer_segments}</customer_segments>
  <goal>{goal}</goal>
  <known_themes>{known_themes}</known_themes>
  <constraints>{constraints}</constraints>
</context>
<task>Synthesize customer feedback into actionable insights using ONLY the provided data.</task>
<instructions>
  Themes need multiple data points (not single mentions). Don't infer beyond stated feedback. Present contradictions without resolving.

  Per theme: name, signal strength (Strong/Moderate/Emerging), segments affected, quotes (2-3), actionability (Quick win/Needs research/Strategic).

  Include: executive summary, themes, segment view, requests vs friction vs sentiment, recommended follow-ups (with owner), gaps.
</instructions>
<output>
  Return all 6 sections.
</output>

Feedback sources: {feedback_sources}
Customer segments: {customer_segments}
Goal: {goal}
Known themes: {known_themes}
Constraints: {constraints}

Rules: themes need multiple data points, don't infer beyond feedback, present contradictions.

Per theme: name, signal strength (Strong/Moderate/Emerging), segments, quotes, actionability (Quick win/Research/Strategic).

Based on the entire content above, return:
1. Executive summary
2. Themes (5-8 with details)
3. Segment view
4. Requests vs friction vs sentiment
5. Recommended follow-ups with owners
6. Gaps