Loading...
How the Model Works
Rent estimates triangulated from government, scraped, and learned sources
Every listing shows an estimated rent we do not take from a single source. We triangulate three independent signals — a government fair-market benchmark, scraped comparable listings, and a machine-learned model — then weight them so no single noisy input swings the number.
Area median rent set by HUD each year
Comparable radius for scraped rent comps
Re-estimated as new leases close
The single figure our deal score is built on
HUD publishes Small Area Fair Market Rents annually. It is a defensible, regulation-grade baseline — but it lags the market and is coarse at the ZIP level, so we treat it as one input, not the answer.
We pull active and recently-rented listings in the same submarket and compute a rent-per-square-foot distribution. This captures what tenants are actually paying right now, including amenities and condition.
A regression model blends beds, baths, square footage, age, and location into a predicted rent, trained on historical closed rentals. It generalizes where comps are thin.
The displayed estimate weights each source by confidence for that property. When comps are dense we lean on them; when a ZIP is sparse we lean on the model and the HUD floor. Outliers are down-weighted so one bad listing cannot distort a deal.