United StatesReal estate2021 – ongoingZillow
Neural Zestimate home valuations
Zillow reported more than 15% lower relative error after launching the Neural Zestimate
Problem
Zillow's earlier valuation system depended on many regional models. Maintaining them was complex, and separately trained components limited joint optimization.
Approach
Zillow launched the Neural Zestimate in June 2021 for homes that were not actively listed. A national neural model combined historical property and transaction data while representing local geography.
Results
Takeaway
The reported improvement concerned estimated market values for off-market homes. It did not establish the accuracy of forecasts for future resale prices.
Categories
- Industry
- Real estate
- Business function
- IT and R&D
- Innovation type
- Product and service innovation
- AI technology
- Predictive analytics and ML
- Adoption
- Built in-house
- Goal
- Quality and risk
- Evidence
- Company's own claim
Sources
Source summaries
Zillow described replacing regional valuation pipelines with a national neural model. It reported lower relative error based on real-world sales data collected after launch.
Full details are in the original sources.
