Hey HN! I built Veltora (https://veltora-beta.vercel.app), an ML-powered customer churn risk predictor that benchmarks predictions against real published data from 35+ well-known companies across 9 industries.
The core problem I wanted to solve: most churn tools return a risk score with no grounding. A "72% churn risk" means nothing without knowing whether that's normal for your industry. Veltora answers that by calibrating a weighted logistic regression model against 2024–2025 benchmark research for companies like Netflix, Salesforce, AT&T, Chase, and many others.
Technical details:
– Weighted logistic regression with industry-specific feature weights
– Calibrated against published churn rate benchmarks (annual/monthly) from industry reports
– 9 verticals: SaaS, Telecom, Fintech, Media/Streaming, Retail, Healthcare, Hospitality, Insurance, Utilities
– Frontend: React + Vite + Recharts
– Auth: Clerk, Payments: Stripe, Hosting: Vercel
– Freemium model
The product has two main surfaces:
1. Custom Predictor — input your own customer metrics and get a risk score with benchmark context
2. Portfolio Dashboard — browse real churn benchmarks by industry and company
Happy to discuss the model methodology or calibration approach. Would love honest feedback on whether this is useful to anyone in customer success, rev ops, or subscription businesses.
tempoz•1h ago
The core problem I wanted to solve: most churn tools return a risk score with no grounding. A "72% churn risk" means nothing without knowing whether that's normal for your industry. Veltora answers that by calibrating a weighted logistic regression model against 2024–2025 benchmark research for companies like Netflix, Salesforce, AT&T, Chase, and many others.
Technical details: – Weighted logistic regression with industry-specific feature weights – Calibrated against published churn rate benchmarks (annual/monthly) from industry reports – 9 verticals: SaaS, Telecom, Fintech, Media/Streaming, Retail, Healthcare, Hospitality, Insurance, Utilities – Frontend: React + Vite + Recharts – Auth: Clerk, Payments: Stripe, Hosting: Vercel – Freemium model
The product has two main surfaces: 1. Custom Predictor — input your own customer metrics and get a risk score with benchmark context 2. Portfolio Dashboard — browse real churn benchmarks by industry and company
Happy to discuss the model methodology or calibration approach. Would love honest feedback on whether this is useful to anyone in customer success, rev ops, or subscription businesses.