Domain in, traffic-intel out — on $0 of data
The core pipeline. Feed it a domain, it fans out to free/public signal sources (Tranco global rank, Cloudflare Radar rank-bucket + category), fuses them through a power-law model, and returns an estimated monthly-visits band with a confidence rating and tech-stack fingerprint. Proven live against known domains: github.com landed dead-on vs SimilarWeb's public figure.
The core pipeline. Feed it a domain, it fans out to free/public signal sources (Tranco global rank, Cloudflare Radar rank-bucket + category), fuses them through a power-law model, and returns an estimated monthly-visits band with a confidence rating and tech-stack fingerprint. Proven live against known domains: github.com landed dead-on vs SimilarWeb's public figure.
Working end-to-end prototype (projects/traffic-intel/scout.mjs). Tranco + Cloudflare Radar both authed and live on the Nuc. Generates rank, category, estimated monthly-visits band, confidence, and CDN/stack fingerprint. Verified: github.com estimate ~1.26B/mo vs SimilarWeb ~1.3B — dead on.
Estimate curve overshoots on brand/direct-heavy domains (stripe.com ~5x high, ahrefs.com ~3x high). Needs calibration against ~50 known SimilarWeb data points to tighten the coefficient. Currently CLI-only — no web front-end yet.
This module matters because it advances the Recom vision in a concrete, inspectable way.