A walk-through of Redfin’s powerful AI-based recommendation engines

For real estate company Redfin, leveraging AI and data is vital to ensuring its mission of netting property sellers more money in less time. In a conversation at VentureBeat’s Transform 2021 conference with VentureBeat founder and CEO Matt Marshall, Redfin chief technology officer Bridget Frey said the company relies on tracking analytics to see what homes users are clicking on and contacting real estate agents to develop its powerful recommendation algorithm. By looking at people’s browsing patterns, the company can break out data to determine what attracts users to certain properties.

Real estate has exploded during the COVID-19 pandemic as an increasingly remote workforce searches for properties with more space to handle work-from-home needs. However, the homebuying process can take months, providing a unique purchasing challenge for Redfin’s data teams.

As a national real estate brokerage, Redfin can also look at data real estate agents input about their buyers’ habits. To measure the results of its AI-driven enterprise, the company analyzes how its property recommendations perform compared to properties that pop up under users’ selected criteria.

“A lot of AI has turned into these science projects where you hire some really smart people and say ‘Where is the data correlated? Can you come up with something that matches?’ But if you can’t turn it into having an impact on the business, it doesn’t matter,” Frey said.

In the past, AI-based algorithms have been cost-prohibitive to build. Relying on Amazon Web Services to set up solutions for Redfin has helped speed up the transformation of its business.

Frey said real estate agents are working to reconcile the history of structural racism and segregation in housing with present-day diversity initiatives. Building a diverse team is crucial to Redfin’s strategy for growth: 36% of Frey’s team is made up of women, and 10% are Black and/or Latinx. Those numbers will increase, she said, to reflect the country’s demographic makeup.

“More diverse teams have gotten better results,” Frey said. Redfin’s vision is to make property ownership more equitable, although diversity is not the company’s No. 1 priority, she said.

Redfin deployed an algorithm to construct automated tour schedules for homebuyers and real estate agents but found that many were not engaging with the product. After bringing in more gender diversity to the engineering team, the company deployed a listening tour in various markets and updated its algorithms accordingly.

The lesson? Putting AI hand-in-hand with user feedback, particularly from a diverse group of voices, creates better solutions. It may not be the only factor in AI’s improvements, Frey said, but it’s a trend playing out time and again in her work.