Zillow’s Zestimate improves by 20% in Phoenix

Zillow today introduces significant improvements to its Zestimate home valuation model. The changes allow the algorithm to respond more quickly to current market trends and improve the national median error rate to 6.9% – an improvement of almost a full percentage point for more than 104 million unmarked homes. These upgrades have improved the accuracy of the nearly 20,000 zestimates available for households in Phoenix by 20%.

The new Zestimate algorithm uses neural networks, the latest machine learning approach, and takes into account apartment details such as the number of square meters and location as well as a deeper history of real estate data such as sales transactions, tax assessments and public records.

ALSO READ: Phoenix # 1 in the nation for growth in property values ​​and rentals

Neural networks are artificial intelligence systems that mimic how the human brain works. You are able to efficiently map hundreds of millions of data points, draw connections between inputs, and use the relationships formed to generate or predict output. In the case of the Zestimate algorithm, the neural network model correlates housing facts, location, housing market trends and housing values.

As a result of this update, the Zestimate can now respond more quickly to dynamic market conditions and provide homeowners with a more accurate estimate [prediction] the current value of a house. Additionally, moving to a neural network based model will reduce Zestimate’s processing time.

“Since introducing the Zestimate in 2006, we’ve never stopped innovating to provide consumers with the most accurate property valuations,” said Dr. Stan Humphries, Chief Analytics Officer at Zillow and creator of Zestimate. “The new architecture that we are introducing today is another important step in our efforts to use big data to give consumers more security and thus better decisions.”

Fifteen years ago, the zestimate gave people instant and free access to an estimated value for millions of households across America for the first time. Over the past decade and a half, Zillow has released several major updates to the Zestimate algorithm, as well as incremental improvements between major upgrades, and now calculates ratings for more than 104 million homes across the country.

With the company’s increasing reliance on the accuracy of Zestimate, Zillow began taking the Zestimate as a live initial cash offer through its home purchase program, Zillow Offers, in February. The Zestimate is an initial cash offer for around 900,000 eligible homes in 23 markets. With this latest update and increased Zestimate accuracy, the number of homes eligible for a cash offer is likely to increase by 30%.

Applying a neural network model to a national real estate dataset was an innovation used by the winning team of the Zillow Prize, the biennial, $ 1 million data science competition that was attended by more than 3,800 teams from 91 countries working to improve the zestimate. One member of the team, Jordan Meyer, is now a Senior Applied Scientist at Zillow, working on home reviews for Zillow Offers.

Comments are closed.