Is iBuying Dead?

Written by Tyler Kastelberg

Share on facebook
Share on twitter
Share on linkedin
Share on email

Happy Sunday!

In this week’s letter, I’m diving into the carnage that is Zillow’s home buying division.

Grab a drink, hide from your kids, and let’s get to it!

Zillow stock is down 66% since its peak this year … what happened?

iBuying has been all the rage, as companies like Zillow, Redfin, and Opendoor use artificial intelligence to actively acquire, improve, and flip homes in hot markets of the US. In total, ibuying accounts for less than 1% of total home sales in the US. However, in Phoenix, it accounts for more than 5% of all transactions.

However, this week Zillow announced that they will be exiting the ibuying business, laying off more than 25% of its workforce and writing down the value of their home purchases by more than $300 million.

Side note – I can only imagine that Blackrock is giddy about this impending portfolio sale …

In a statement from their CEO, Zillow cited failures in their ibuying AI algorithm that drove the collapse of the division.

As a result, Cathie Wood’s Ark investment funds dumped $255 million of Zillow stock on Wednesday, which contributed to the collapse in the price this week. 

At the same time, Opendoor’s stock surged on the news that Zillow is leaving the market.

The problem with “AI” algorithms

Lots of companies claim they use artificial intelligence (a.k.a. machine learning) to solve hard problems. Truthfully, very few companies use AI effectively.

Why? AI algorithms require a massive amount of relatively clean data to provide meaningful insights.

I inquired about this with one of my buddies, Joe Veltkamp, who works for Aible, an AI software provider. Joe explained that insights generated from an AI model with less than 10,000 data points shouldn’t be trusted.

Moreover, AI models can be negatively impacted by their own feedback loops. For example – high-frequency stock trading firms have a cap as to how much money they can confidently invest at any one time … their own trades eventually degrade the strength of their models.

Zillow likely had different AI models for different cities (maybe even neighborhoods). Phoenix, Zillow’s largest market, has more than 2,000 home sales per month. At first glance, there might be enough data to create an effective algorithm …

The problem is that real estate transactions aren’t just about price and square feet. They’re about a complex set of facts (like size and condition) and human emotions (like how a neighborhood and property feels).

In other words, real estate is uniquely human …

Human behavior is driven by emotions, and the decision to buy a home is no different. This is one of the reasons why real estate agents are still very relevant in the home buying experience. 

Back to Zillow … my guess is that they didn’t have enough data to effectively execute their purchase strategy.  

Did Zillow’s engineering team know their models were flawed? Yes.

Was this avoidable? Probably.

What about Opendoor and others? I’m not bullish.

Send me a reply with your thoughts – especially if you disagree. I’d love to hear some counterpoints as to why ibuying is the future.

Hire team members, not freelancers

Bullpen connects commercial real estate experts and companies for on-demand work

Step1_with_arrow 4

Tell us about your job.

Step2_with_arrow 3

Meet your freelancer.

Step3_with_arrow 3

Get to work.

Subscribe to Bullpen

Analyst Resources. Financial Models. Growth Strategies

Join Bullpen today!

Get the real estate experts you need, when you need them.