← Back to docs
Recipe: NPS comment analyzer
Build a pipeline that ingests raw NPS survey comments, classifies sentiment, extracts feature requests, and routes actionable items to your product backlog — all inside Meridian.
Pipeline overview
1Ingest — CSV or API webhook drops raw comments into a Meridian table.
2Classify — LLM step tags each comment: promoter / passive / detractor, plus topic labels.
3Extract — second LLM pass pulls out concrete feature requests and bug mentions.
4Route — webhook sends high-priority items to Linear, Jira, or Slack.
Sample output
| Comment | Sentiment | Action |
|---|---|---|
| “Love the speed but need dark mode” | promoter | Feature request |
| “Export keeps timing out on large files” | detractor | Bug report |