Redfin has improved the accuracy of the Redfin Estimate and expanded coverage to five new metro areas and several additional property types, including land and multi-family properties. The Redfin Estimate is a calculation of a home's worth, built on Redfin's propriety machine learning algorithm that uses billions of data points about properties across the country, the company announced in a release.

Redfin improved the accuracy of its estimate for off-market homes by 70 basis points, as the median error improved from 6.3 percent to 5.6 percent. An error rate of 5.6 percent means that half of all Redfin Estimates fall within 5.6 percent of the home's eventual sale price. For a typical $300,000 home, the updated Redfin Estimate is now $2,100 closer to the market value and will be within $16,800 of the final sale price half of the time.

Among Redfin's most popular features, the Redfin Estimate gives homeowners, buyers, sellers and real estate agents insight into the value of more than 85 million homes. Homeowners can claim a home on Redfin to track its Redfin Estimate over time and edit facts about their property to further improve the accuracy. When it's time to sell, the Redfin Estimate is often the starting point in a conversation between a seller and her Redfin agent, the company stated.

"Everyone wants to know what their home is worth, which is why the Redfin Estimate drives so many homeowners to visit our site and start a conversation with a Redfin agent," Redfin Chief Technology Officer Bridget Frey said in the release. "In real estate, every tenth of a percentage point matters and can represent hundreds or thousands of dollars in a home's value. Agents know that a smart, data-driven pricing strategy is often the difference between a home that sells quickly with multiple offers and one that sits on the market.

“The accuracy of the Redfin Estimate speaks to Redfin's data advantage as a brokerage that employs thousands of real estate agents across the country who works hand-in-hand with our engineers, providing uniquely human insights as we train our machine learning algorithms. With the best data, top tech talent and deep agent expertise, Redfin has the necessary ingredients to transform the real estate industry through AI and machine learning," Frey added.

A highly accurate estimate is also critical for RedfinNow, Redfin's instant-offer business. RedfinNow buys homes directly from homeowners for cash, fixes them up, if needed, and then quickly resells them.

 "The new release not only improves the accuracy of the estimate, it includes back-office tools that show the confidence of the estimate and project the future sales price, so we can better assess the risk and opportunity of buying a particular home," Quinn Hawkins, head of RedfinNow, said. "With continued advancements in AI and machine learning, we can envision a future where the Redfin Estimate becomes a live offer for a subset of properties."

As to more accuracy, Redfin, according to the release, has improved the Redfin Estimate for both off-market and on-market properties. For homes actively listed for sale, the Redfin Estimate error rate improved by 5 basis points, improving from 1.62 percent to 1.57 percent. These accuracy gains for both on- and off-market properties were measured by comparing the performance of the previous algorithm to the updated algorithm between February and August 2019. The Redfin Estimate error rate is dynamic and updates daily for on-market homes and weekly for off-market homes. The most current national and local error rates are listed on the Redfin Estimate page.

“Redfin's engineers and data scientists improved the Redfin Estimate's accuracy by conducting hundreds of experiments, testing and retesting new models and incorporating new data,” the company said. “Because Redfin is a brokerage, it has complete access to the Multiple Listing Services (MLS) 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.”

Redfin also expanded the Redfin Estimate to five new markets: Connecticut, Boise, Des Moines, Spokane and Chattanooga. Previously available for three property types: single family homes, condos/coops and townhouses; Redfin can now provide estimates for several new types of properties including land, mobile homes and multi-family buildings.

Redfin is a tech-powered real estate brokerage, today announced it has improved the accuracy of the Redfin Estimate and expanded coverage to five new metro areas and several additional property types, including land and multi-family properties.