Zillow Quit Home-Flipping Because AI Couldnt Model Real Estate Price Fluctuations

Houses are expensive. Buying and selling them is thusly complicated - as, perhaps, it should be.

Zillow, like so many tech companies before it, thought it could build an algorithm to model this process, scale it and make a ton of money by raising convenience while reducing price. What it forgot was that there are local factors, like too few craftspeople to renovate, or the closing/opening of new businesses or malls or warehouses that impact regional prices. It lost $381 million figuring that out and is now laying off 25% of its workforce. JL 

Alex Kirshner reports in Slate:

Zillow announced Tuesday it’s shutting that division down after it lost $381 million in the most recent quarter. Zillow is laying off 25% of its workforce in what sounds like a direct result of losing $381 million flipping homes in three months. Zillow has acknowledged that the company could not predict accurately how rapidly home prices would change: “We’ve determined the unpredictability in forecasting home prices far exceeds what we anticipated and continuing to scale would result in too much earnings and balance-sheet volatility.”

Some things are easy to sell. A pizza goes for whatever price the menu says it sells for, and the buyer does not spend several weeks calling in an inspector to make sure nothing is wrong with the olives or pepperoni. A lightly worn hockey jersey goes for whatever an eBay vendor can get someone to pay for it, and there are no lawyers or licenses, just seller ratings, required to make sure everyone is on the up-and-up. A piece of stock is even simpler; it goes for whatever a seller says it goes for, and goes up or down depending on how many people want it. You can learn the agreed-upon price with a two-second Google.

Some things are hard to sell. Like homes, as anyone who has ever tried to sell one would attest. That is probably how it should be, given that homes are generally the most expensive things people will ever buy or sell. But the entire process is still a drag, involving lots of intermediaries and preparations and verifications that can and do stretch over several months. Zillow, the website you visited to check out homes you could not afford during the first few months of the pandemic, tried something neat to help people get around that process (which worked!) and also make Zillow some money by flipping those homes (which did not). Zillow announced Tuesday it’s shutting that division down after it lost $381 million in the most recent quarter. Zillow is laying off 25 percent of its workforce, it says, in what sounds like a direct result of losing $381 million flipping homes in three months. The wind-down of the program, and the associated layoffs (about 2,000 out of 8,000 employees) will take a few quarters to wrap up. Zillow still owns thousands of homes from its buying spree.

On the one hand, Zillow’s failure is a typical infuriating business story. Some execs dreamed up something they could not execute, it blew up spectacularly, and the most obvious people to suffer from their decision are the many employees they’ll fire. (Zillow’s shareholders are also losing, which happens when companies release news of this kind on earnings calls with their investors.) On the other hand, I hope they try again? Or that another company does it better? The idea of selling houses like they’re something less complicated than houses is worth someone getting right, even if Zillow came nowhere close.

The concept of “iBuying” has been closing in on the mainstream for a few years now—at least since 2018, when Zillow spun up “Zillow Offers” in Phoenix and Las Vegas before expanding eastward. There are a handful of other companies in this game, and they tend to operate in markets where, as the New York Times explained, a lot of homes are relatively new and affordable and of similar sizes. There are more formalities involved than just this, but the conceit of iBuying is that a company makes an offer for a home without even going out to visit it, based on what an algorithm, in this case Zillow’s Zestimate, says it should cost—and the seller accepts or declines. The long inspection, appraisal, and closing processes in real estate do not apply. iBuying made up at least 1 percent of home sales in 18 markets in the third quarter of 2019, Redfin, another online real estate company, reported to the Times. It’s supposed to work for companies like Zillow, because the housing market is supposed to be somewhat predictable in the markets where iBuying happens. In 2019, maybe it was!

You may have noticed that the real estate market has changed a bit since then. Houses got very expensive, very quickly, during the pandemic. Zillow announced in October that it was halting the program for the rest of this year, and at the time, the company said the problem was “a labor- and supply-constrained economy” that held up its ability to flip homes competitively in a hot market. The COO called it “an operational backlog for renovations and closings” on the houses it had iBought and hoped to flip for a profit. A dearth of contractors to renovate all these homes’ kitchens was clearly not the whole story. Now Zillow has acknowledged that the company simply could not predict accurately how rapidly home prices would change: “We’ve determined the unpredictability in forecasting home prices far exceeds what we anticipated and continuing to scale Zillow Offers would result in too much earnings and balance-sheet volatility,” the CEO, Rich Barton, said in a statement to shareholders on Tuesday.

If a tech company wants to give people way too much money for their houses and be forced to sell those houses for a loss, then hey, sure!

The $381 million loss it took in this business over three months put the whole company (which was otherwise profitable) at a $169 million quarterly loss before taxes, interest, and the like. The company’s stock has taken a good thwacking as a result, going from around $100 last week to around $70 now. This all feels vaguely populist, sort of in the way that the brief run of MoviePass felt populist: If a tech company wants to give people way too much money for their houses and be forced to sell those houses for a loss, then hey, sure! If a tech company wants to pay for you to see a bunch of movies more or less for free, great! Unfortunately good things never last when companies are forced to explain them to their investors.

Zillow is easy to mock here. Losing almost $400 million in three months when you could’ve done almost anything else is extremely make-fun-able, not to mention retrospectively wasteful. It’s also possible that the machines will never be good enough to do this on their own in a way that works for both sellers and the companies trying to buy their houses.

Maybe, as Bloomberg’s Matt Levine speculated last month, the future lies in computers buying houses with a sprinkle of human verification: Instead of pressing a button and selling your house for X price, you press a button and sell your house for X price pending someone driving out there and making sure there’s nothing absurdly wrong with it. Maybe an actual real estate expert needs to spend 20 minutes looking at the final deal. It seems impossible for iBuying to accurately price in every little quirk about a home—it won’t know if it sometimes smells weird in the backyard or the neighbor plays music too loud—but the real estate market is so logistically daunting that I hope someone figures this out. If the market is fair and no one takes a bath on a deal, giant companies that don’t have to take out loans to buy houses are good candidates to move a sale along quickly in a way that makes a seller’s life a lot easier and saves lots of people lots of hassle. Zillow Offers is dead, but perhaps its death can be an inspiration for someone else to take a chunk out of conventional real estate.