Feedback Analysis

Turn raw feedback into a structured voice-of-customer report.

Input

The user pastes or attaches:

• Support tickets / chat transcripts
• App store / review site comments
• Survey free-text responses
• Interview notes

If there are < 20 items, ask the user if they want to wait for more — small samples create false patterns.

Workflow

1. Code each item

For every feedback item, tag:

• Theme (1-3 word category, e.g. "onboarding", "pricing", "performance", "missing-feature").
• Sentiment (positive / neutral / negative).
• Severity (blocker / friction / annoyance / nice-to-have).
• Quote-worthy? (yes/no — tag the most representative quotes).

2. Roll up

Group by theme. For each theme produce:

• Count and % of total.
• Severity distribution.
• 1-3 representative quotes (verbatim, not paraphrased).
• Inferred root cause (if visible from the data).

3. Prioritize

Use a 2x2: Frequency (low/high) × Severity (low/high).

• High frequency + High severity → fix first.
• High frequency + Low severity → batch later.
• Low frequency + High severity → triage individually.
• Low frequency + Low severity → ignore for now.

4. Output report

# Feedback Analysis — N items, <date range>

## Top themes
1. <Theme> — <count> mentions (X%) — <severity mix>
   "<representative quote>"
   Likely cause: <inference>
   Recommended action: <specific>

## Quadrant
[2x2 of theme placement]

## Action list
1. <Action> — addresses themes A, B — owner: TBD

Anti-patterns to avoid

• Don't paraphrase quotes — verbatim only, even if grammar is bad. Paraphrasing strips signal.
• Don't over-weight the loudest user — frequency across users matters more than message length.
• Don't conflate "feature request" with "underlying problem". A user asking for X may have a different real need.
• Don't skip positive feedback — it identifies what NOT to break.