Feedback Synthesizer

Feedback Synthesizer

Reference

• ID: S-PROD-04
• Category: Product
• Inputs: raw feedback (support tickets, CONV- artifacts, interviews, surveys, reviews, Slack messages)
• Outputs: themed feedback report with actionable insights → artifacts/product/
• Used by: Product Agent
• Tool scripts: ./tools/artifact/validate.sh

Purpose

Consolidates raw, unstructured feedback from multiple channels into a structured report of themes, each backed by evidence and linked to actionable product recommendations.

Procedure

1. Collect raw feedback from all available sources (support tickets, interviews, surveys, app reviews, internal notes).
2. Normalize feedback: strip duplicates, tag source channel, note user segment.
3. Perform open coding — assign initial labels to each piece of feedback.
4. Group labels into higher-level themes (e.g., "onboarding friction", "pricing confusion").
5. For each theme, count frequency and assess severity (blocker / pain / annoyance).
6. Rank themes by frequency * severity to surface the highest-impact issues.
7. For each top theme, provide 2-3 representative quotes as evidence.
8. Draft actionable recommendations tied to each theme.
9. Save the themed feedback report to artifacts/product/.
10. Validate the artifact using ./tools/artifact/validate.sh.

Small-Data Path

If fewer than 10 feedback items across all sources:

• Skip frequency scoring — insufficient sample size for statistical patterns
• Produce a Signal Log format: chronological list with severity ratings and direct action suggestions
• Do not generate PRD candidates from fewer than 5 items in the same theme
• Note the data limitation explicitly in the output
• Recommend: "Collect more feedback before synthesizing themes. Use customer-conversations skill to generate structured input."

Resume standard synthesis procedure when 10+ items are available.

Step 11: Produce PRD Candidates Section

After ranking themes (Step 6), for the top 3 themes by frequency × severity score:

1. Draft a PRD candidate entry for each:

• Theme name: from Step 4
• Problem statement draft: 2-sentence articulation of the user pain
• Target ICP: which user segment reported this most
• Evidence sources: CONV- artifact IDs + support theme artifact IDs
• Priority score: frequency × severity (1-25 scale)
• Suggested next action: /ship (score >15), validate first with /discovery-validation (8-15), monitor (<8)

1. Append a ## PRD Candidates section to the themed feedback report.

1. Do NOT auto-create PRD artifacts — present candidates for human review only.

1. Inform the user: "Run /ship '[theme name]' to initiate the PRD pipeline for any candidate you'd like to pursue."

Step 12: Produce Pain Matrix (when invoked by GTM Sprint)

When the feedback-synthesizer is called during a GTM Sprint (or explicitly requested), produce an additional Pain Matrix section after Step 6:

1. Build the Matrix Table

For each theme from Step 4, score:

| Theme | Frequency | $$ Impact / Occurrence | Urgency (1-5) | Total Score | |-------|-----------|----------------------|---------------|-------------| | {theme} | {count} | {estimated $$ or time cost} | {1=someday, 5=hair-on-fire} | frequency x urgency |

• $$ Impact: Extract from interview quotes. Use time-to-money conversion if needed (e.g., "wastes 3 hours/week" at $50/hr = $150/week = $7,800/year)
• Urgency: How soon prospects need a solution (1 = nice to have, 5 = actively searching right now)

1. Quadrant Classification

Classify each theme into a quadrant:

• Build Now (High $$ + High Urgency): immediate opportunity, prioritize in MVP
• Build Next (High $$ + Low Urgency): strong business case but prospects can wait
• Quick Win (Low $$ + High Urgency): small pain but prospects want it solved now
• Monitor (Low $$ + Low Urgency): track but don't prioritize

1. Quotable Quotes Section

For each "Build Now" theme, extract 2-3 direct quotes suitable for:

• Landing page social proof (above-the-fold pain quotes)
• Lead magnet content
• Pitch deck problem slides
• Sales outreach templates

Format: "[Quote]" — {Role}, {Company Size} (anonymize company name)

1. Append the Pain Matrix, Quadrant Classification, and Quotable Quotes as a ## Pain Matrix section in the feedback report artifact.

Quality Checklist

• [ ] Feedback from at least 2 distinct sources is included
• [ ] Themes are mutually exclusive and collectively exhaustive
• [ ] Each theme has frequency count and severity rating
• [ ] Representative quotes are included as evidence
• [ ] Recommendations are specific and actionable, not vague
• [ ] User segments are noted where relevant
• [ ] Pain Matrix includes $$ quantification (when requested)
• [ ] Quotable quotes are anonymized and landing-page-ready (when requested)
• [ ] Artifact passes validation

Cross-References

• gtm-sprint (S-GRO-09) — invokes Pain Matrix output during Day 2 synthesis
• customer-conversations (S-PROD-08) — provides CONV- artifacts as input
• landing-page-copy (S-GRO-02) — consumes quotable quotes for above-the-fold social proof
• prd-writer (S-PROD-02) — PRD candidates from top themes

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: vibbs
• Source: vibbs/company-os
• License: MIT

Install and usage instructions live in the source repository linked above.