Zillow’s Zestimate home valuation mode has received a major upgrade to keep up with rapid market changes, according to a company announcement on Tuesday. Using machine learning and neural network technology, Zestimates has reached a median error rate of 6.9 percent for the more than 104 million off-market homes it tracks — a one percentage point improvement.
“Since we introduced the Zestimate in 2006, we have never stopped innovating in order to provide consumers with the most accurate home valuations,” Zillow Chief Analytics Offer and Zestimate creator Dr. Stan Humphries said in a statement. “The new architecture we’re debuting today represents another significant step forward in our efforts to harness big data to create more certainty for consumers, which leads to better decisions.”
Neural networks, Humphries explained, imitate how the human brain collects, organizes and interprets data. Neural network technology enables Zillow to “map hundreds of millions of data points efficiently, drawing connections among inputs and using the relationships formed to produce or predict an output.”
With the Zestimate home valuation model, neural networks are able to collect and correlate home facts from property data such as sales transactions, tax assessments and public records, along with housing market trends and home values to produce a more accurate and responsive Zestimate.
“As a result of this update, the Zestimate can now react more quickly to dynamic market conditions, providing homeowners with a more accurate estimate [prediction] of a home’s current value,” Humphries added. “In addition, transition to a neural network-based model will reduce Zestimate processing time.”
Throughout the pandemic, real estate agents have complained about a lag in the Zestimate AVM as home values skyrocketed due to booming buyer demand. In April, Sacramento-based team leader Tim Collom told Inman Zestimates’ margin of error had ballooned over the past 12 months.
“They’re not off by $10,000. They’re off by like $100,000 to $200,000,” he said. “In Sacramento and some other areas, Zillow estimates can be high by 10 percent and they also can be low by 10 percent, but very rarely do they hit the mark right now.”
“I think it’s because of the market volatility and how fast it’s going — even seasoned real estate agents are struggling to keep up with price trends right now,” he added.
However, Zillow said the latest Zestimate update has only increased its confidence in the model as evidenced by the platform’s February launch of live offer Zestimates. In 20 markets, homesellers could request Zillow Offers solely based on their Zestimate.
“We’ve long said, one of the beauties of Zillow Offers is that it’s the best first step for you to think about selling,” Jeremy Wacksman, Zillow’s chief operations officer told Inman in a previous article. “Nobody wants to think about selling, most sellers are thinking about buying, they want to be thinking about buying, but at some point they have to think about, ‘Oh my gosh I have to go through this process.’”
“The Zestimate is the starting point for that. If we can turn the Zestimate into a starting offer for certain customers, that’s going to get them to raise their hand and say, ‘Okay, I’m thinking about moving, help me understand, am I trading in? Can I talk to an agent about that?’”
Zillow estimates the neural network update will increase live offer amounts by 30 percent.