The move, touted by a business executive at the time as “an exciting step forward”, was intended to streamline the process for homeowners considering selling to Zillow as part of his house flip business. Zillow promoted this option as a way to make it easier to sell a home while minimizing interactions with others during the pandemic. Barely eight months later, however, the company completely shut down that business, Zillow Offers.
Zillow declined a request for an interview with Krishna Rao, the company’s vice president of analysis. In a statement, Zillow spokesperson Viet Shelton told CNN Business that the company is using Zestimate for Zillow offerings “in the same way we encourage the public to use it: as a starting point.”
“The challenge we faced in Zillow Offers was the ability to accurately forecast the future price of inventory in three to six months, in a market where home value changes were larger and faster than ever,” he said. Shelton said.
Indeed, since Zillow entered the home turnaround industry in 2018, real estate markets have changed in extremely unpredictable ways. The pandemic resulted in a temporary freeze in the housing market, followed by an imbalance of supply and demand that caused housing prices to rise unprecedentedly. This may have only complicated the company’s decision to include Zestimate – which Zillow says is not an appraisal, but a “computer-generated estimate of the home’s value today,” taking into account available data “- as part of the Zillow bidding process. in more than 20 cities.
Artificial intelligence can examine much more information, much faster, than a single human could when considering a fair price for a home, weighing factors such as comparable home sales in a region. , the number of people searching in a specific neighborhood, etc. Still, “you can ask a realtor to look at a house and in a second choose a critical valuation factor that just doesn’t exist as ones and zeros in any database,” said Mike DelPrete, a real estate technology strategist. and Researcher-in-Residence at the University of Colorado Boulder.
A key element of Zillow
“Three times a week, we create over 500,000 unique valuation models, built on 3.2 terabytes of data, to generate current Z estimates on more than 70 million US homes,” the company wrote in a filing. of titles in 2011. More than 10 years later, the company publishes Zestimates for more than 100 million American households.
If you search for homes on the Zillow website or app, Zestimate is featured prominently in every listing whether the home is for sale or not. If the house is currently for sale, a red dot will appear next to the words “House for sale” and the Zestimate, if available for that house, will appear on the same line.
Zillow has spent years improving Zestimate, going so far as to host a multi-year data science competition to improve the accuracy of the algorithm behind it. The company awarded a team of three with the $ 1 million prize in early 2019.
Zestimate currently has a median error rate of 1.9% for homes that are on the market, Shelton said, which means Zillow’s estimates for half of the homes on the market are within 1. , 9% of the actual selling price. This error percentage is much higher – 6.9%, according to Shelton – for non-market homes. A spread of as little as 1.9% on a property with a Z estimate of $ 500,000 is still close to $ 10,000; this number is multiplying over many, many homes in different cities across the United States.
An art, not just a science
It’s one thing to create a template on a website that is often quite accurate. It’s another to then try to use this model in the real world to make very expensive bets – and do it on a large scale, according to Nima Shahbazi, a member of the winning team. Zestimate algorithm and CEO of Mindle.AI, which helps companies use AI to make predictions. For example, if one of the homes Zillow bought had some hidden issues – like a missed crack in the foundation – the Zestimate wouldn’t be able to predict those issues, he said.
“There are a lot of different parts between a very decent model and deploying the model to production that can go wrong,” he said.
Zillow used Zestimate to help him make buying decisions for homes he hoped to make a profit over time. But Nikhil Malik, an assistant professor of marketing at the University of Southern California, said algorithms tend to be good at making accurate short-term predictions, such as predicting stock prices a second in advance. But there just isn’t enough data for an algorithm to know more about longer busts and booms, according to Malik, who studies algorithmic pricing and has studied Zestimate in particular.
There are also many non-quantifiable aspects of giving a home a price tag, noted DelPrete, such as the value of living in the same neighborhood where you grew up or down the street from your parents. These can vary from person to person, making it even more difficult to outsource a home appraisal process to a computer.
“It’s a good tool for what it is,” DelPrete said of Zestimate, but it’s a mistake to think that it can be used to accurately predict house prices now or in the near future. to come up. He considers it “almost a toy,” intended more to pique your curiosity when searching for your or your neighbor’s house online.
“If you want to do iBuying and you’re going to be making thousands of deals every day, you’ve got to be really good at pricing homes, not just today but in three to six months,” he said. “And it is an art and a science.”
– CNN’s Anna Bahney contributed to this report.