Details
The model considers 500+ data points per home, including physical attributes from MLS records (such as waterfront access, condition, and garage capacity), geographic factors (flood zones, proximity to amenities), user engagement signals, and public records. It uses an ensemble of machine learning techniques including random forest and gradient boosting models. As of 2024, the model achieves a median error of 1.99% for on-market homes and 7.64% for off-market homes. Redfin deliberately excludes crime statistics from its model inputs to avoid encoding racial or socioeconomic biases. The algorithm was last updated May 1, 2024. A 2017 independent study found the Redfin Estimate outperformed competitors on accuracy for homes actively listed for sale.
Have evidence about Redfin's AI practices? Submit a report.
Report a Sighting →