The Zillow debacle: no algorithm is smart enough to overcome basic flaws

Market-making only works well for fungible goods with symmetric information

Don’t blame the algorithm

It’s been the promise of the past two decades that Big Data will supercharge business decision-making and liberate the CEO from having to rely on superstition and voodoo and actual business experience to make important decisions. Collect everything and save it for a day when machines are finally smart enough to understand your business better than you can. This is the promise that led Zillow to invest billions into their ill-fated Offers business.

Much of the reporting on this issue focuses on a runaway algorithm, or desire to win “market share”. These may be true, but this is not a story about Big Data. This is a story about fundamental risks that the entire “Instant Buyer” industry is exposing themselves to.

Zillow described a pilot of their “Instant Offers” program in their 2016 Annual Report, and expanded the program in 2018.

From Zillow’s 2018 10-K filing:

if Zillow Offers were available in the top 200 metro areas in the U.S., sellers of nearly half of the existing homes sold in 2018 across the entire nation, or approximately 2.7 million homes, would have been eligible to receive offers from Zillow to buy their home.

In the securities markets, an offer is called the bid, and the asking price is called the ask. The difference, where the bid in an orderly market is always lower than the ask, is called the bid/ask spread. Every purchase in the market occurs at the ask, and every sale occurs at the bid. Therefore, when you maintain a P&L of any trading activity, every purchase at the ask results in an immediate loss because you can only sell at the bid.

People who willingly take these immediate P&L losses are called market-makers. Market makers rarely earn profits from their trading activity in an orderly market, they make money from fees paid by exchanges and other market participants who benefit from the liquidity they provide to the market. For some reason, however, Zillow convinced themselves that they could both be a market-maker, and profit from trading activity through “flipping” houses. Big Data and machine learning, as wonderful as they may be, do not enable you to have your cake and eat it too.

iBuyer industry has fundamental flaws

Every share of stock in AAPL is the same, whereas every 1500 sqft house built in 1949 is not the same. Market-making only works well for fungible goods with symmetric information. A market-maker in the equity markets, for example, is quoting bids and asks for securities regulated market, and they are paid by the exchanges to provide this service.

Zillow is offering to every property owner — the person best situated to possess superior knowledge about the property and its underlying condition — a free binding quote on their property.

From Zillow’s 2018 10-K:

Zillow Offers provides homeowners in certain metropolitan areas with the opportunity to receive offers to purchase their home from Zillow.

It’s clear that Zillow did not understand how valuable this free service was to property owners.

By making these offers to anyone who asks, they re-structure the game to put themselves at a disadvantage. The seller is now enjoying a no-called-strikes game, where they can get a binding quote on their property at any time, and if they like the price, they will sell it to Zillow. If they don’t like the offer, they don’t have to do anything. Said another way, this amounts to offering a PUT option for the price of $0.00 to the owners of every property in their supported territory. When you give away, for free, options that have real value in the market, you’re at a disadvantage on day 1, before you’ve consummated a single transaction.

At any time, I can sell my least-favorite property to Zillow whenever I deem the price advantageous for me. No algorithm is smart enough to overcome this disadvantage.

The problem was not a lack of technical skill, or an algorithm that was given too much power. It was not even a desire to win market share. The problem at Zillow was a fundamental misunderstanding of core finance principles that underly the market apparatus: every option has value, and giving them away for free has has consequences.