The property data provider ATTOM announced rollout of an updated automated valuation model (AVM) that is being promoted as a “a ground-up, AI-first rebuild designed to move beyond the limitations of traditional comparable-sales-based valuation models.”
According to the Irvine, California-based company, the new AVM is designed to analyze relationships between property characteristics, historical pricing patterns, and hyperlocal trends, with the goal of creating property-level valuations even in low-liquidity or data-sparse environments.
“We model how each neighborhood has evolved over the past 30 years and use that temporal context to translate historical sales into accurate present-day values,” said Aaron Wagner, vice president of data science at ATTOM. “This gives our valuation system a richer foundation to learn from, allowing each model to synthesize property characteristics, local market dynamics, and price history into a single, informed estimate. The result is consistent accuracy, even in markets where traditional comparable-sales approaches tend to break down.”
The company stated that in out-of-sample testing the new model generated a median absolute percentage error of 2.9%, with more than 80% of valuations falling within 10% of the actual sale price.
ATTOM added the new model is designed for enterprise use across mortgage, insurance, investment, and proptech applications. It is being made available through APIs, bulk delivery, and cloud platforms such as Snowflake and Databricks.
0 Comments