Independent Study Finds Redfin Estimate to be Most Accurate Among Top Automated Home-Value Estimates

SEATTLE -- Redfin, the next-generation real estate brokerage, today announced the findings of a comprehensive, independent study that found that the Redfin Estimate more accurately predicted the value of thousands of homes for sale than estimates from other leading providers of public home-value estimates. Of the more than 5,000 home sales evaluated in the study, 64 percent sold within 3 percent of the price predicted by Redfin, compared to 29 percent for Zillow and 16 percent for Homes.com. The median error rate of the home-value estimate was 2.06 percent for Redfin, 5.95 percent for Zillow and 10.26 percent for Homes.com. Redfin commissioned independent research company SSRS to conduct the study.

SSRS is an award-winning social science and market research firm with widely cited research on topics including the presidential election, government-provided health insurance, mobile e-commerce, immigration, opioid abuse, perceptions of disabled veterans and racial attitudes among children. The firm's clients and partners include the Harvard T.H. Chan School of Public Health, the Pew Research Center, the George W. Bush Presidential Center, the Sesame Street Workshop, the Atlantic and The New York Times.

"Our findings are clear: Redfin performed significantly better than Zillow and Homes.com at predicting home-sale prices," said Dr. Aniruddha Banerjee, SSRS's senior vice president of advanced analytics. "We hope that others attempt to replicate our results. Our methodology for coming to this conclusion was consistent with the rigorous approaches taken at academic institutions."

Comparison of Redfin, Zillow and Homes.com Home-Value Estimates

"Artificial intelligence is now being applied to the most fundamental questions in commerce, government and society, to make startlingly accurate predictions about global temperatures, stock-market prices, political movements and, for individual Americans, the value of their homes," said Adam Wiener, Redfin's executive in charge of data science. "And this is not like earlier victories for artificial intelligence, like beating a chess champion or winning at Jeopardy; it has supreme practical value for someone trying to make big decisions about what she can afford and where to live. What's most compelling about these predictions is not just their practical value, but that they can be proven right or wrong. When Redfin or another application says a listing is worth $325,000, we don't have to argue about who was right, we just have to wait for the final sale."

Study Methodology

Each Sunday through Thursday between Oct. 19 and Nov. 30, 2016, for the 194 census-defined U.S. metropolitan areas where Redfin has listing and estimate data, Redfin sent SSRS a file including all residential homes that had a new sale pending. From this file, SSRS randomly selected 500 to 1,000 homes each day, until by Nov. 30 a total of 24,789 pending sales were under evaluation. The homes in this sample spanned 7,531 postal codes and 189 census-defined metropolitan areas, and included single-family homes, townhouses, co-ops and condominiums. Within 36 hours of being notified about the pending sale of a home, SSRS noted the home-value estimates from Redfin, Zillow and Homes.com. When the home sold and its sale price was published, SSRS then compared the estimated values with the actual sales price. 6,338 had closed by November 30. 5,661, or 89 percent of the closed sales, had home-value estimates from both Zillow and Redfin. 5,074, or 80 percent, had home-value estimates from both Homes.com and Redfin.

Why the Redfin Estimate is Accurate

To calculate the Redfin Estimate, Redfin evaluates billions of data points using proprietary machine-learning software running on next-generation cloud technology. Because Redfin is a broker, the company has complete access to the Multiple Listing Services used by real estate agents to describe properties in extensive detail, with each MLS tracking different attributes of a home that Redfin can use to calculate a more accurate estimate, such as whether a home is located on the water, has a view or faces a busy street.

The Redfin Estimate error rate is dynamic and fluctuates depending on the data available at a given time. Redfin posts the most current national and local error rates on the Redfin Estimate page. The error rate updates daily for on-market homes and weekly for off-market homes. The Redfin Estimate is available for more than 65 million homes in major U.S. metropolitan areas. Zillow estimates the value of more than 100 million homes, covering areas beyond those major metros. Homes.com does not publish its estimate coverage.

To view the full study, click here.

About Redfin
Redfin is the next-generation real estate brokerage, combining its own full-service agents with modern technology to redefine real estate in the consumer's favor. Founded by software engineers, Redfin has the country's #1 brokerage website and offers a host of online tools to consumers, including the Redfin Estimate, the automated home-value estimate with the industry's lowest published error rate. Homebuyers and sellers enjoy a full-service, technology-powered experience from Redfin real estate agents, while saving thousands in commissions. Redfin serves more than 80 major metro areas across the U.S. The company has closed more than $40 billion in home sales through 2016.