Voice of customer prompt pack
Customer feedback is full of signal and impossible to read at scale by hand. The job of Voice of Customer analysis is to turn hundreds of reviews, survey verbatims, or interview transcripts into a short list of recurring themes, sentiment patterns, and prioritised actions — backed by real quotes, not vibes. This pack generates a prompt tuned to your specific data source and goal so an LLM does that synthesis reliably.
How it works
You choose the data source — product reviews, NPS comments, support tickets, or interview transcripts — and the synthesis goal: theme clustering, sentiment breakdown, prioritised pain points, or feature-request mining. The builder assembles a prompt that tells the model how that source is structured, what to extract, and how to present it: themes ranked by frequency, each supported by verbatim quotes, with sentiment tags and a short “so what” recommendation. It requires quotes as evidence so the synthesis stays grounded in what customers actually said.
Tips and notes
- Feed it volume. VoC synthesis is most valuable across dozens or hundreds of comments; a handful is better read directly.
- Demand quotes. The prompt requires a verbatim quote for every theme so you can sanity-check that the pattern is real before acting on it.
- Separate sentiment from frequency. A loud complaint from three users is not the same as a quiet one from three hundred — the prompt reports both.
- Close the loop. Each theme ends with a recommended action so the analysis drives a decision rather than sitting in a doc.