High quality customer interviews are a superpower for figuring out which features to build and which to kill. I've conducted hundreds over the years, working toward product-market fit.
tl;dr When you ask the right questions, body language tells you what matters most. Transcripts tell you the rest.
Here's a complete LLM prompt framework to get started, created from my real-world usage, including the exact interview script I follow.
https://github.com/shawnyeager/customer-interview-metaanalysis
The bare-bones MVP is that you clone this repo, point your LLM at a directory of transcripts, and tell it to go.
===
Free idea: build a lightweight version of this, but on freedom tech.
https://www.aha.io/discovery/overview
Potential components:
+ jitsi for video
+ blossom for video storage (encrypted?)
+ nostr:nprofile1qyxhwumn8ghj7mn0wvhxcmmvqydhwumn8ghj7mn0wd68ytnzd96xxmmfdecxcetzwvhxgegqyz9lv2dn65v6p79g8yqn0fz9cr4j7hetf28dwy23m6ycq50gqph3xc9yvfs or nostr:nprofile1qyjhwumn8ghj7en9v4j8xtnwdaehgu3wvfskuep0dakku62ltamx2mn5w4ex2ucpxpmhxue69uhkjarrdpuj6em0d3jx2mnjdajz6en4wf3k7mn5dphhq6rpva6hxtnnvdshyctz9e5k6tcqyp7u8zl8y8yfa87nstgj2405t2shal4rez0fzvxgrseq7k60gsrx6zeuh5t for post-interview transcription and—most importantly—patterns and insights from a campaign of interviews
