Today, I’m pleased to announce the launch of Zillow Prize, a contest that allows data scientists from around the world to work alongside us to improve Zestimate accuracy and have the chance to win $1 million. This is the first time we’ve ever invited data scientists outside of Zillow to work on the algorithm.
The contest is staged in two rounds: The public qualifying round, which opens today and closes Jan. 17, 2018; and a private, invitation-only final round that kicks off Feb. 1, 2018, and closes Jan. 15, 2019. By the close of Zillow Prize, the person or team who improves the accuracy the most AND beats a Zillow benchmark model will walk away with $1 million.
This contest has long been a dream of mine, but in our initial years our resources were better spent refining the algorithm. Over these years, we have driven the median error rate from 14 percent down to around 5 percent today, thanks to investments in technology and cloud computing; our ever-expanding living database of all homes; and, of course, the tireless work of our ever-growing Zestimate team. In a recent benchmark study of 2016 transactions listed for sale on Zillow (a set of homes on which we observe higher accuracy than overall), the Zestimate error was 3.5 percent, pretty close to the 2.5 percent error rate of the actual listing price itself.
Our experience over the last 12 years has taught us a few things. We’ve come to believe that our strength in this area now allows us to be more open with respect to innovation because we are better positioned to take advantage of such innovation than anyone else. Business strategists call this “first-mover advantage” and we believe we have it in spades. Since our internal team has already built the most accurate home value estimate available, Zillow can rapidly incorporate and leverage new ideas coming in from external teams. Going forward, Zillow will be more focused on publicly demonstrating thought-leadership in this area and capitalizing on our ability to generate, gather and integrate new ideas faster and more effectively than anybody else.
This new posture is facilitated by the current scale and breadth of our living database of all homes, which allows us to better leverage many types of algorithmic improvements. We’re particularly excited about the exploration of more hyperlocal data and algorithms, a task well-suited to highly distributed, crowd-sourced efforts. Moreover, we are eager to tap into ideas from other industries utilizing machine-learning solutions that may translate well into the real estate domain – ideas from biotechnology, computer vision, driverless cars and financial markets, to name a few. While some of these discoveries may come from external teams, Zillow’s data and experience give us the ability to leverage them across all homes and utilize data from public records, listing information, user-contributed content and website.
The Zestimate was the first piece of content unique to Zillow, and it remains a single-feature microcosm of our overall mission: Power to the People. Before the Zestimate, consumers could only see one opinion of value, created by the seller, even though other opinions of value not seen by the consumer were being used by industry professionals to make all sorts of decisions that impacted them. The Zestimate brought this information out from behind the curtain, and Zillow Prize reaffirms our commitment to helping people make the best decisions about real estate.
We are very excited to see what this crowd-sourcing effort brings to light. A home is most often the largest financial investment a person makes in his or her lifetime, and we are committed to ensuring they have a clear picture of their home’s value.
