How we estimate property value

Our Automated Valuation Model (AVM) is an XGBoost gradient-boosted regression trained on every recent sale in the county. It combines assessor-reported physical attributes, census tract demographics, permit activity, and flood risk to predict a market price.

Features we use

• • Square footage, lot size, beds, baths, year built, stories, garage.
• • Latitude/longitude and the surrounding census tract median income and home value.
• • Count and total value of permits pulled in the last 5 years.
• • FEMA flood zone risk score.

Accuracy

We target a median absolute percentage error below 7%. Every property page shows a confidence range (low and high) derived from that error.

Limitations

AVMs cannot see interior condition, renovations that were never permitted, or non-disclosed sales. Our estimate is not an appraisal and should not be relied on for loan or legal decisions.