ChurnBase - Spot at-risk B2B SaaS accounts weeks before they churn

Hey makers, Every B2B SaaS I've worked with had the same broken loop: a "customer health score" sitting in a spreadsheet that only flagged accounts after they'd already mentally checked out. Quarterly NPS surveys. Color-coded dashboards. Nothing actually predictive. The frustrating part the signals were always there. Drops in admin logins, support ticket spikes, failed integrations, billing disputes but spread across five tools and never combined into one real risk score. By the time the CSM noticed, the renewal call was already a save-meeting. ChurnBase scores every B2B account on behavioral signals logins, feature depth, billing health, support patterns and combines them into one weighted risk score that updates daily. Not sentiment. Not NPS. Just what people actually do (or stop doing) in your product. A few design decisions I made deliberately: - No annual lock-in. €59/mo, cancel anytime. Selling 12-month contracts to fix retention is its own joke. - EU-built, GDPR-native. Customer data stays in Frankfurt. No "data processing addendum" theater. - Pre-order with refund guarantee. If we don't ship by Q3 2026, every month you paid gets refunded. I don't want anyone betting on a vaporware launch. This is a pre-order today, not a live product. I'm doing PH now because the founding €59/mo locks in forever, and the people who sign up early shape what we build first. Two questions I'd love your take on: 1. CS / RevOps folks what's the worst "health score" you've ever had to defend in a QBR? 2. Would you trust a model that flags risk weeks before it shows up in NPS or do you need it sentiment-confirmed? Happy to nerd out on methodology, signal weighting, or why I'm betting behavioral > sentiment.