1.9 percent. In February 2021, that was the median error of the Zestimate, Zillow’s automated guess at what a home is worth, on homes listed for sale. Zillow had been publishing Zestimates for 15 years, on nearly 100 million homes. Now it would put money behind them. For eligible homes in more than 20 cities, the Zestimate became an initial cash offer, shown at the top of the property page.
“This exciting advancement demonstrates the confidence we have in the Zestimate,” said Jeremy Wacksman, Zillow’s chief operating officer.
The buying arm, Zillow Offers, was about three years old by then. It bought houses straight from their owners and sold them on. The aim, as the company later put it, was to be “a market maker, not a market risk taker.” That rested on one skill above all: forecasting what each home would sell for three to six months later. Zillow publicly set itself a target: results per home within 2 percentage points of breaking even, either way.
That spring the machine sped up. Zillow bought 3,805 homes in the second quarter, a record and more than double the quarter before. Its letter to shareholders credited, among other things, “the progress we have made in strengthening our pricing models and automation.” On each home it sold that quarter, it made an average gross profit of $33,849.
In the third quarter it bought 9,680.
More sellers were taking Zillow’s offers than ever before. In November the company said why: it had been “unintentionally purchasing homes at higher prices than our current estimates of future selling prices.” Renovation crews, meanwhile, were short of workers, and Zillow sold only 3,032 homes in the quarter. It ended September owning 9,790, with 8,172 more under contract. In October it stopped signing new contracts for the rest of the year.
On November 2, 2021, Zillow said it would wind Zillow Offers down. It wrote down $304 million on the homes it held and expected $240 million to $265 million more in losses, mostly on homes it was already committed to buy. About a quarter of the company’s workforce would go.
“We’ve determined the unpredictability in forecasting home prices far exceeds what we anticipated,” said the CEO, Rich Barton. The guardrail had been 2 percentage points either way. Between the second quarter and the outlook for the fourth, the swing was about 12.
Insiders had a nickname for the pricing push, Bloomberg Businessweek reported: Zprices. Current and former employees, competitors and counterparties told the magazine the trouble lay less in the data or the code than in how Zillow used them.
The Zestimate stayed. In the same letter that ended the buying business, Zillow called improving it “fundamental and foundational” to its audience and brand. It just wasn’t an offer anymore.