AI-first company

Company case

Zillow Offers: home-price forecasts missed, and Zillow shut its home-buying business

Zillow bought homes to resell, at prices set with its own algorithms and in-person checks. In 2021 its forecasts of resale prices missed by far more than it had planned for. It wrote down $407.9 million of homes, shut the business and cut about a quarter of its staff. Zillow named labour and supply problems as well as the forecasts.

By
Last reviewed
Corrections
Report an error
Outcome Went wrong

Something failed in work where AI was used, and there is a public record of it.

At a glance

Company
Zillow GroupOnline real estate marketplace; Zillow Offers bought and resold homes
Size
Enterprise (1,000+ employees) Zillow Group had 8,005 full-time employees on 31 December 2021, by its annual report.
Where and when
United States. Started 2018, wind-down announced 2 November 2021.
Who produced the figures
The company itselfAll figures come from Zillow Group’s own filings with the US Securities and Exchange Commission: its results for the third and fourth quarters of 2021 and its annual report for 2021. The annual report’s accounts are audited. The reasons given for the failure are the company’s own.

What happened

Zillow runs the most visited real estate website in the United States. In 2018 it started Zillow Offers. The business bought homes directly from sellers, renovated them and resold them. Zillow says the business depended on forecasting the price of homes accurately three to six months ahead.

Zillow says it priced the homes “using in-person evaluations and data science and proprietary algorithms”. Its annual report says the same data that feeds the Zestimate, its public home-value estimate, also powers “the pricing algorithms that we use for our Zillow Offers business”. The filings do not say which methods the algorithms used.

Zillow set itself a limit in public. The result on each home was to stay within 200 basis points of break-even, that is within 2 percentage points of the price either way.

In the third quarter of 2021, more sellers than before accepted Zillow’s offers. Zillow says this happened because it was “unintentionally purchasing homes at higher prices than our current estimates of future selling prices”. On 18 October 2021 it stopped signing new contracts to buy homes for the rest of the year. On 2 November 2021 its board decided to shut Zillow Offers.

The numbers

From Zillow’s results for the third quarter of 2021 (July to September):

  • 9,680 homes bought and 3,032 sold. Zillow ended the quarter with 9,790 homes in stock and 8,172 more under contract to buy.
  • About $304 million written down. Zillow reduced the value of the homes it held, because it had paid more than it now expected to sell them for.
  • A loss of $80,771 on each home sold, on average, before overheads. That is nearly 2,100 basis points (21%) of the selling price.
  • A loss of $422 million before tax in the part of the company that bought and sold homes. The whole company lost $328 million.
  • About 1,200 basis points (12 percentage points). That is how far the result on each home swung between the second quarter and Zillow’s forecast for the fourth. The limit was 200 either way.

From Zillow’s annual report and results for 2021:

  • $407.9 million of write-downs on homes for the year. Before 2021 Zillow had recorded no large write-downs.
  • $93 million of the write-downs came in the fourth quarter. In November Zillow had expected losses of $240 million to $265 million there. Homes resold for more than it had expected.
  • About 10,000 homes still in stock at the end of 2021.
  • A loss of $881.5 million before tax for the year in the part of the company that bought and sold homes.
  • About 25% of staff to go during the wind-down. Apart from the write-downs, Zillow expected the wind-down to cost $175 million to $205 million in all, including $99 million to $112 million for staff who were let go.

Why Zillow says it failed

Rich Barton, Zillow’s co-founder and chief executive, gave the main reason: “We’ve determined the unpredictability in forecasting home prices far exceeds what we anticipated”.

The letter to shareholders says Zillow had been “unable to accurately forecast future home prices at different times in both directions by much more than we modeled as possible”. It names the events behind this: a pandemic, a short freeze in the housing market, and then home prices rising at an unprecedented rate.

The forecasts were not the only cause Zillow gave.

  • Not enough renovation workers. A shortage of renovation workers slowed the resale of homes. The homes Zillow held waited longer, and the risk from price changes grew.
  • Labour and supplies. The board’s decision names “home pricing unpredictability, capacity constraints and other operational challenges”, made worse by “a difficult labor and supply chain environment”.
  • Too few sellers. Zillow bought from only about 10% of the serious sellers who asked for an offer.
  • Too much capital. Running the business at the size needed to make a steady profit would have tied up too much money and made earnings swing too much.

The annual report also names a limit of the pricing model: it “may not account for submarket nuances – for example, the location of a home on a hill or in a building”.

What happened next

Zillow sold the homes it held faster than it had expected. By February 2022 it had sold, or agreed to sell, more than 85% of the homes it planned to resell during the wind-down. Zillow kept the Zestimate and its advertising business. It said it would look for ways to help people sell a home without buying it.

What the case shows

  • Compare the forecast error with the margin. Zillow aimed to keep the result on each home within 2 percentage points of break-even. Its forecasts missed by far more, so each mistake cost more than a sale could earn.
  • Forecasts can fail when the market moves fast. Zillow says its forecasts went wrong in both directions as prices first froze and then rose at an unprecedented rate.
  • A limit on the result does not limit how much you buy. Zillow’s limit was on the result per home. In one quarter it bought 9,680 homes, more than it could renovate and resell on time.
  • People checking each home did not prevent the losses. Zillow used in-person evaluations as well as its algorithms. The error Zillow names is in forecasting where prices would go next.
  • Watch who says yes. More sellers accepted Zillow’s offers than before. That was a sign that the offers were too high.

What this says about the practices

Practices the company applied

Sources

  1. Zillow Group (2 November 2021). Q3 2021 shareholder letter. Form 8-K, Exhibit 99.3, US Securities and Exchange Commission. sec.gov
  2. Zillow Group (2 November 2021). Zillow Group Reports Third-Quarter 2021 Financial Results & Shares Plan to Wind Down Zillow Offers Operations. Form 8-K, Exhibit 99.1. sec.gov
  3. Zillow Group (10 February 2022). Q4 2021 shareholder letter. Form 8-K, Exhibit 99.3. sec.gov
  4. Zillow Group (10 February 2022). Annual report on Form 10-K for the year ended 31 December 2021. sec.gov