The Cost of Ignorance: How a Flawed Algorithm Caused Almost a Billion Dollar Loss in Real Estate.

Does not knowing what you don't know make you liable? Perhaps yes, especially when it costs investors millions of dollars.

This is the situation the proptech giant Opendoor is currently facing.

If we learned something from the 2008 crash is that the real estate market is far more vulnerable to market shifts than what assumptions and abstract models may have made us think. Clear rules and relations are great when you want to create a tech solution for a problem, but in real estate, risks are exponentially higher, and absolutes can yield losses in the millions in the blink of an eye. 

The first casualty of this belief was Zillow's home-flipping business, Offers. According to Zillow's CEO, Rich Barton, the company "(... )determined the unpredictability in forecasting home prices far exceeds what we anticipated.", realizing that they were not in a position to accurately predict where home prices would be in six months “within a narrow margin of error.”

Zillow was unintentionally purchasing homes at higher prices than current estimates of future selling prices, reflected in a net loss of $422 million in Q3 2021, where 72% of that loss was in write-downs, so the shutdown Offers. On the other hand, Opendoor, whose revenue streams come mainly from their IBuying business, made the same mistake almost a year later, incurring an almost billion-dollar net loss by Q3 2022, losing money on 42% of its transactions in August, and triggering the filing of a class action securities fraud lawsuit in October of that year.

However, why would this loss trigger a security fraud lawsuit, of all things? In its SEC filing, Opendoor stated that its operations were dependent on its ability to accurately price and portfolio manage inventory and that an ineffective pricing or portfolio management strategy may have a material adverse effect on its business, sales, and results of operations. To achieve this, they claim to use price algorithms that use machine learning to drive pricing decisions and that, over time, they have improved the pricing accuracy of their models as they added new data inputs and refined model logic, improvements that compound with experience and scale. 

These statements place significant emphasis on how critical Opendoor's algorithms are for their business model, how they can adapt to changing market conditions, and how experience and scale further refine their models. But, a year removed from Zillow's withdrawal, the evidence shows how this critical component couldn't adapt to changes, leading plaintiffs to allege that the offering documents contained untrue statements, overstating the benefits and competitive advantages of the algorithms, increasing the risk of sustaining significant and repeated losses due to residential real estate pricing fluctuations.

From an all-time high stock price of $34.59 in February 2021 to an adjusted price target of $2.70 in October 2023, Opendoor's experience is a cautionary tale on how constant business oversight over algorithms is essential to avoid the pitfalls of growth, and how magnitude differentiates proptech from other startups. As algorithms have become increasingly prevalent in real estate, it is important to examine what Opendoor's algorithms fail to assess.

First, Opendoor's algorithms failed to understand how people behave when it comes to real estate. In a business where negotiations are expected and, at least, one party is deeply invested, the fact that they accept your first offer regularly should be treated with a healthy dose of skepticism. Real estate transactions often involve emotional considerations and may be challenging to quantify accurately through algorithms, potentially limiting their effectiveness.

Second, they relied on historical data and did not receive appropriate input from current experience, failing to capture sudden market shifts or unique circumstances that deviate from historical patterns. Zillow's decision should've been the canary on the coal mine for Opendoor.

And finally, they failed to address magnitude. A $100,000 house could be considered a rounding error for a $1.6 billion company like Opendoor, but when you operate at scale, this will become a problem faster and in a more impactful way, as mistakes in real estate are easily five-figure mistakes per transaction.

From an all-time high stock price of $34.59 in February 2021 to an adjusted price target of $2.70 in October 2023, Opendoor's experience is a cautionary tale on how constant business oversight over algorithms is essential to avoid the pitfalls of growth, and how magnitude differentiates proptech from other startups.