Zillow lost money because they weren't willing to lose money (stevenbuccini.com)

Article summary

Zillow's iBuying business failed due to its inability to stomach the uncertainty and potential losses required to build a successful algorithmic underwriting system. The company oversimplified the problem and relied on existing data, rather than building a new organization with the necessary technical and cultural mindset. This led to a $500 million loss and the shutdown of the division. The article argues that Zillow's approach was flawed and that building a successful system requires rigor, a willingness to take risks, and a deep understanding of the data and the problem being solved.

Main themes

• algorithmic underwriting
• risk management
• data quality
• machine learning
• real estate investing
• business strategy
• adverse selection
• feedback cycle time

What commenters say

• Zillow's failure was due to its inability to build a good lasting algorithm in a dynamic world, where even seasoned Wall Street algorithms have failed.
• The company's data and methodology were flawed, and they weren't willing to pay the price to fix it.
• Zillow's management made a rational decision to cut losses and abandon the iBuying business, rather than continuing to throw money at a failing venture.
• The company's mistake was trying to de-risk their new venture, which is often incompatible with launching a new business strategy.
• Some argue that Zillow's failure was not due to risk aversion, but rather a lack of understanding of the market and the need for a more complex model.
• Others believe that the company's approach was not entirely flawed, and that they could have succeeded with more accurate market price estimations.
• The feedback cycle time for real estate investments is much longer than for other assets, such as advertising, which increases the risk and makes it harder to adjust the model.
• Adverse selection is a major issue in machine learning models for pricing, and it's not clear if training the model on this selection can actually lead to better outcomes.