Automated valuation model (AVM) info

An Automated Valuation Model (AVM) estimates list, sale, or rent prices for any address using machine learning. It's how lenders, insurers, and proptech apps value millions of properties at once.

Why Houski's AVMs

• $5.00 per valuation - Pay-as-you-go, no upfront commitment.
• Hierarchical fallback - City-specific model if available, then state or provincial, then country. Every address gets a prediction.
• All of Austria, Canada, Czechia, Estonia, Finland, France, Latvia, Netherlands, Norway, Poland, Slovenia, South Korea, Spain, Switzerland, the United Kingdom and the United States: every property scored using the 400+ fields that power our API.
• Transparent accuracy - Median MAPE, R-squared, and percent-within-10% published for every model below.

How to get a valuation

• Through the API - Hit our /avm endpoint with a property_id. Returns list, sale, and rent predictions plus a report. $5.00 is added to your API subscription as usage and billed on your next monthly invoice.
• Through the /predict endpoint - The same models behind every valuation on this page also answer our /predict endpoint, which returns the list price, sale price, or monthly rent estimate for a property on any date from January 2021 to today. It gives you the numbers without the report and each predicted value costs the same as reading that field.
• Through your dashboard - Sign up and order an AVM for any address, no integration required. Same $5.00, choose to bill it to your API subscription or pay it as a one-time charge to your card without a subscription.

What each report includes

A report is more than three numbers. Here is what you get and how each part is built.

• Three estimates with confidence scores - List price, sale price, and monthly rent each come from their own model, so one can be more certain than another. Each estimate carries a confidence score of 100 minus the model's typical percentage error, so a confidence of 95% means half of past estimates from that model landed within 5% of the real figure.
• Flags - Concrete things we noticed about the property or its area that are worth looking into before you rely on the estimate, such as elevated flood risk, monthly fees that run high for the value, a large gap from the assessed value, or older wiring. They describe the home and its surroundings, not our confidence in the number, and they appear only when something stands out.
• Property and area scores - We score the location from 1 to 10 across natural hazards such as flood, fire, and earthquake, and everyday livability such as walkability, transit, safety, and noise. Higher is always better, including the hazard scores, where a high score means lower exposure.
• Comparable properties - Nearby homes of the same type and similar size, ranked by how closely their price, size, and distance match this home, closest first. The price shown is the last asking price we saw before each one left the market, and the sale date is our own estimate rather than a recorded sale figure.
• A living estimate - The report is live, so the values update over time as the underlying data and the market change, and each report shows the date it was valued.