Pain Point Weekly: turning real online complaints into evidence-backed pain signals, not guesses

Most startup advice says "talk to customers" or "check Google Trends" to validate an idea. But customer interviews are biased. People are polite to your face and trend data tells you a market exists, not whether it's actually underserved.

I wanted evidence that didn't depend on people being nice to me. So I built Pain Point Weekly: a pipeline that pulls real, unprompted complaints from Reddit, Hacker News, and Stack Exchange, classifies each one with an LLM. Complaint, feature request, or someone actively looking for a tool, then clusters the semantically similar ones together. When 15 independent people across different threads describe the exact same broken workflow in their own words, that's a real signal not a guess.

Right now it's tracking SME payment operations, reconciliation, payment integrations, account receivable. Last week it processed 99 posts about payment reconciliation and surfaced 56 real pain signals. The sharpest one: a small business owner had $24,000 held by AMEX for over a week with zero visibility into when it'd release. Another cluster showed people stuck manually reconciling Stripe and Paypal because QuickBooks' auto-reconciliation actively introduces errors.

I'm pre-revenue and still figuring out if this is valuable enough that people would pay for it. If you're building in B2B fintech, or just curious about evidence-based niche discovery, I'd love feedback, and you can follow the weekly findings here:
https://painpointreport.beehiiv.com