SimilarWeb has a billion-dollar moat: a browsing panel that cost ~$100M+ to assemble. You can't out-software that. Recom's bet is different — fuse the free, real, and public traffic signals nobody has bothered to productize (Chrome UX Report, Cloudflare Radar, Tranco, Common Crawl), wrap them in an AI analyst that does the interpretation SimilarWeb makes you do by hand, and win the wedge they ignore: small sites, real-time freshness, per-vertical benchmarks, and a price that isn't a $2k/yr wall.
Shipped, deployed, and running in production today.
Real working code shipped to production, but phases still on the build queue. Each card lists what's done and what's left.
Designed, documented, and on the build roadmap. No production code yet.
Actual measured Chrome traffic, free
Google publishes real Chrome-user experience distributions for millions of origins via the Chrome UX Report API — measured traffic behavior, not modeled estimates, at zero cost. Almost nobody builds consumer products on it. Recom folds CrUX in as the accuracy backbone the moment the API is enabled on our Google Cloud key.
Tighten the bands to reality
The power-law fusion nails ordering but overshoots absolute numbers on brand/direct-heavy sites. Calibration anchors the model against ~50 domains with known SimilarWeb figures, fitting per-category coefficients so the estimate band lands inside a believable range instead of 3-5x high.
Ask 'who's eating my traffic and why'
SimilarWeb hands you dashboards and makes you interpret them. Recom's AI-native layer lets you just ask — 'who's stealing my traffic, and how' — and get a narrated, sourced answer that reads the fused signals, spots the movers, and explains the shift in plain language.
Search vs social vs referral vs direct
Beyond a single visits number: estimate the channel mix driving a domain's traffic by fusing SERP-visibility signals, backlink/referral graphs, and social-embed footprints. Answers 'is this competitor winning on SEO, paid, or brand?' — the question that actually informs strategy.
Purpose-built for one industry at a time
SimilarWeb is a generalist. Recom ships per-vertical instances with industry-specific benchmarks, competitor sets, and the metrics that matter to that niche — so a board-game publisher or a SaaS founder sees their world, not a generic dashboard they have to translate.
Live signal, not weeks-old data
SimilarWeb's panel data lags weeks. Recom's signal sources (Radar, CrUX, live SERP checks) can refresh continuously, surfacing movement — a competitor's traffic spiking or cratering — while it's happening, with alerting on the domains you watch.
Domain in, report out — in the browser
The consumer surface: a clean web app where you type a domain and get the full fused Recom report — estimate band, source breakdown, stack fingerprint, and the AI analyst's narrated read — with shareable report links and a watchlist. Wraps the engine the CLI prototype already proves out.
Known issues and cleanup work, tracked so nothing gets forgotten.
What we know is broken, half-done, or worth cleaning
Anything flagged as worth cleaning up later lives here. Items are filed the moment they're found so they do not evaporate into chat history.
Recom is built on the honest premise that you don't need a $100M panel to give someone a directionally-right, AI-explained read on who's eating their traffic. If it's on this page it's on the roadmap. If it's not here, it doesn't exist yet.