Wired: Not everyone has been thrilled with their Zestimate. Tell us how it's done.
Barton: Our algorithm is a complex piece of AI that pores through a ton of data, looks for patterns, and creates predictive models. Then it goes through by zip code and identifies which models work best in each neighborhood. We're going to take that all the way down to the house level eventually. It's an interesting computer science and stats problem. A fun one.Wired: Why wait until now to upgrade your algorithm?
Barton: Zestimates are based on history; we plot house prices like stock prices. We are careful about major updates to the algorithm because we are literally rewriting history.Wired: So how does your users' data fit into the formula?
Barton: The Zestimate is only going to be as good as the information we have going in, and there are lots of holes and inaccuracies. We opened up Zillow so owners could correct facts about their homes, publish their own estimate of their house's value, and upload pictures. We can feed some of that information back into the algorithm. If you say, "No, there are four bathrooms, not three," we take that as reality. It makes the Zestimate significantly better.Wired: Can sellers really be trusted to give honest appraisals?
Barton: It's kind of like a warts-and-all corporate blog — stuff is going to come out that you wouldn't want, but everybody ends up better. It's a very modern concept. Smart homeowners get it: If there was a flood in the basement two years ago, it's going to be found out. So let's talk about the fact that it happened and what we did about it.