Case 4 · Buying homes with an automated valuation

The goal. Buy homes directly from sellers, make light repairs, resell at a margin.

The system. An established home-valuation model — mature, widely used, and reasonably accurate as an estimate — was used to decide which homes to buy and at what price. Purchases were made at scale.

What happened. In Q3 2021 the company bought 9,680 homes and sold 3,032. In November 2021 it shut the unit down, took a write-down of more than $500M, and cut about 25% of its workforce (~2,000 people).

Your task

  1. The valuation model was reasonably accurate. So what failed?
  2. What changes when an estimate stops informing a browsing customer and starts triggering a purchase?
  3. They were buying at volume in the markets they were forecasting. What does that do?
  4. Would a more accurate valuation model have prevented this?

Sources are on the main case studies page — read them after the session.