What happened
Zillow Offers bought houses directly from their owners, using the company's own computer valuations to decide what to pay, then renovated and resold them. It grew fast: Zillow sold 5,337 homes in 2020 and 15,436 in 2021 [2].
On 2 November 2021 Zillow's board decided to wind the business down. The company told the SEC — the US financial regulator — that the decision was made "in light of home pricing unpredictability, capacity constraints and other operational challenges". The wind-down would take several quarters and cut roughly 25% of Zillow Group's workforce [1].
The same filing disclosed a $304.4 million write-down — a loss booked because the homes were now worth less than Zillow had paid — for the third quarter alone. The cause, in the company's words, was "purchasing homes during the third quarter at higher prices than Zillow Group's current estimates of future selling prices after selling costs". A further $240–265 million of losses was expected on homes it was still contractually bound to buy [1].
The full-year figure was worse. In its annual report for 2021 (the FY2021 10-K), Zillow recorded write-downs "totaling $407.9 million" against its stock of homes, describing the cause as "unintentionally purchasing homes at higher prices" than its own estimates of future selling prices [2].
The 10-K is unusually candid about the mechanism. Among the risks it lists: "Our pricing model may not account for submarket nuances — for example, the location of a home on a hill or in a building — which could have a significant impact on price" [2].
Where the same model helped
The valuation model behind this was not a failure as an estimate. As a free public number that helps millions of people get their bearings in a housing market, it was and remains genuinely useful — being roughly right about a lot of houses is exactly the job. What changed was not the model's accuracy but the consequence attached to it. The same estimate that is helpful as a suggestion becomes a purchase order when a company buys the house. Nothing about the number had to get worse for the outcome to [1][2].
Where it burned
An estimate that is right on average can still be ruinous, because the mistakes do not cost the same in both directions. When the model bid too low, Zillow simply did not buy the house and lost nothing. When it bid too high, Zillow bought it — every one of them. So the business systematically bought the houses its model was most wrong about, and the write-down is the sum of those mistakes. Nothing in the model's average accuracy would have revealed this. Only asking what happens on the wrong side would have [1][2].
The tell
Before acting on a model's number, ask two questions: what happens when it is wrong, and how quickly would I find out?
Accuracy is the wrong first question for anything that drives a decision. What matters is the shape of the mistakes: whether errors in one direction cost more than the other, whether the system ends up selecting its own worst guesses, and how long the feedback takes. It is like a fruit stall that lets customers pick their own — you are left holding every bruised apple. Zillow's feedback loop was months long: the model bid, the house was bought, and the error only showed at resale. A number you cannot check for six months is a number you are committing to, not consulting.
The check is a habit, and habits are trained. Statistics, understood is about exactly this — averages, spread, and why the tail of a distribution is usually where the money is.
Sources
Every source below was opened and read. Last verified 15 August 2026.
- [1] Zillow Group, Inc. — Form 8-K, Item 2.05 (wind-down of Zillow Offers) — U.S. Securities and Exchange Commission (EDGAR), 2 November 2021
- [2] Zillow Group, Inc. — Form 10-K for the year ended December 31, 2021 — U.S. Securities and Exchange Commission (EDGAR), February 2022