Brief - The Hustle

What Redfin’s learned about building real estate algorithms

Written by Juliet Bennett Rylah | Jul 15, 2021 4:46:19 AM

AI may be better at finding your dream home than you are.

During VentureBeat’s Transform 2021 conference, Redfin CTO Bridget Frey explained how the real estate brokerage firm uses AI to help sellers net more money in less time.

Its AI learns how prospective buyers look for homes

That means which listings they click, if they schedule tours or make offers, which offers win, and what they buy. Agents across the country are also adding data.

Redfin found:

  • Customers are 2x as likely to click on homes Redfin’s AI recommends vs. ones that surface under their search criteria
  • A quarter of all Redfin website and app traffic is people browsing homes it’s suggested to them

So, what’s the key to Redfin’s AI?

  • Equity. The better the data, the better the algorithms — but not necessarily all data. Per the Fair Housing Act, agents can’t talk about neighborhoods’ demographics or socioeconomic status, so Redfin doesn’t use that data either.
  • Diversity. A team that “reflects the customer base” and can identify problems and solutions.
  • Practicality. Actually talk to the humans who will use the AI — or you might end up building the wrong thing.

While that all sounds reasonable, if Redfin’s AI doesn’t recommend a place with a pool or sauna… we’re questioning its tech.