Automatic Analysis
How Lumifeed automatically categorizes and prioritizes feedback with Claude.
Automatic Analysis
Every feedback submission is analyzed automatically within seconds of arrival.
What gets analyzed
Sentiment
One of: positive, neutral, negative
Category
One of: bug, feature_request, question, praise, complaint, other
Priority score
A number from 0–100 reflecting the urgency and importance based on the feedback content, sentiment, and user context (plan, role, etc.).
| Score | Meaning |
|---|---|
| 80–100 | Critical — address immediately |
| 60–79 | High — plan for next sprint |
| 40–59 | Medium — backlog candidate |
| 0–39 | Low — track but deprioritize |
Summary
A 1–2 sentence neutral summary of the feedback.
Tags
Automatically extracted topic tags (e.g. export, safari, mobile).
Deduplication
Lumifeed uses pgvector to find semantically similar existing feedback. If a new submission is > 85% similar to an existing one, it's marked as a duplicate and linked to the original. Duplicates increase the original's vote count.
Viewing analysis
In the dashboard feedback detail panel, you'll see all AI analysis fields under AI Insights. Each field shows the model's confidence.
Plan limits
| Plan | AI analyses per month |
|---|---|
| Free | 25 |
| Starter | 500 |
| Pro | 2,000 |
| Agency | Unlimited |
When the limit is reached, new feedback is still stored but analysis is skipped until the next billing cycle.