What would you have done? Zillow let an algorithm buy 7,000 homes — a $569M write-down and 2,000 jobs cut

In 2021, Zillow let its pricing model drive the purchase of nearly 7,000 homes across 25 markets. The model could not forecast prices three to six months out in a fast-moving market: the company overpaid, then shut the business down with write-downs of up to $569M and 25% of its workforce cut. A forecasting model is only as good as the horizon it can actually see; at scale, in a volatile market, automation amplifies the error rather than the value. Before industrializing a forecast, bound its scope: let it adjust only a capped share of the plan, and have a human committee sign off on any decision that commits capital.

Source: Bloomberg, "Zillow Shuts Down Home-Flipping Business After Racking Up Losses," November 2, 2021.