Zillow’s decision to pull the plug on Zillow Offers, its tech-powered business of buying and selling homes, came as a surprise amid the still-hot housing market and prompted a question: Is the business model flawed?
As the third-quarter earnings reports from two of Zillow's competitors showed, it's too early to write off the business model.
Zillow Offers was one of a new breed of real estate fintech companies known as iBuyers. At its core, the business model is simple—buy a property, fix it and sell it—and it has been around as long as there has been property to sell.
Now, iBuyers have brought the deep pockets of institutional investors and high-powered algorithms to the equation, figuring they can build a national scale out of what has traditionally been a highly local business.
But when algorithms use flawed data or are not refined correctly, things can quickly go wrong. Zillow Offers is an example of this. When Zillow began, its main offering was listing properties. Those properties listed homes at values derived from algorithms that were then compared against actual real estate transactions and vetted by real estate experts. But this process was time-consuming and costly; Zillow decided to minimize the vetting of the prices to expedite the process.
Algorithms are only as good as the data they use. Zillow’s home valuations, known as Zestimates, were relatively accurate in some markets or when properties were already listed for sale. But when Zillow entered new markets or listed an off-market property, that proved to be problematic.
Those flaws in the algorithms, plus the rapid appreciation in home prices coupled with shortages of workers and materials to repair the homes, proved too much for Zillow. In the end, Zillow decided that scaling up Zillow Offers was too risky, too volatile and offered an insufficient return on equity, according to its third-quarter earnings call on Nov. 2.
Strength in the business
Other iBuyers have fared better. Opendoor and Offerpad, two of the largest companies in the market, reported encouraging earnings on Nov. 10 and see a bright future in the business. Overall, iBuyer purchases increased by 89% in the third quarter.
Opendoor, which holds more than 50% of the market share, increased its purchases in the third quarter by 79%, buying more homes than all major iBuyers combined. The company cited its ability to price homes, continuously adjusting its modeling and approach.
Offerpad, now the second-largest iBuyer, also cited its ability to achieve a variance of less than 1% between its aggregate estimated prices and actual sales prices.
Offerpad also highlighted its automated weekly market report, which summarizes real-time information from individual markets that is then reviewed and used to refine the company’s underwriting algorithm.
This feedback loop is important and has allowed Offerpad to decrease its risk. The company also said it had adjusted its home acquisition strategy to limit the number of homes purchased when certain supplies were unavailable. This adjustment, the company said, would ensure that the average time from acquisition to sale remained below its 100-day target.
This speaks to the company’s ability to factor other elements like supply chain constraints into its business model. Brian Bair, chief executive and founder of Offerpad, said that his company was “as much a logistics business as it was a real estate technology company.”
By the numbers
The results of both companies reflect their optimistic view of the market.
- Offerpad reported revenue of $540.3 million in the third quarter, up by 190% from $186.4 million a year earlier. The company also sold a record number of homes (1,673), up by 749 from the prior quarter. Its gross profit increased to $53.1 million, or 169%, from $19.8 million a year earlier. More than 99% of its inventory is owned for fewer than 180 days.
- Opendoor also reported favorable results, with revenue of $2.3 billion for the third quarter compared to $338 million a year earlier, up by 569%. Gross profit rose to $202 million compared to $35 million a year earlier for an increase of 465%. The company continues to expand, launching in five new markets in the third quarter to bring its total footprint to 44 markets in the United States.
Because house flipping is a low-margin business, building out ancillary services that increase the number of transactions per customer offers an opportunity for the companies to increase their margins.
Zillow said it plans to take write-downs of as much as $569 million and reduce its workforce by 25% as it winds down the business in the coming months. Zillow recently reached a deal to sell about 2,000 of the homes to Pretium Partners, a New York-based investment firm.
As with any business, companies operating in this segment must implement sound business practices and understand the risk. Some key considerations to keep in mind:
- Know the market: iBuyers have found success focusing on key markets, which gives them the ability to understand the data in the specific markets to more effectively predict sale prices.
- Mind the algorithms: Companies need to evaluate the rules that govern algorithms, maintain a continuous validation process, and always consider the accuracy and diversity of the data used to create the algorithms.
- Garbage in, garbage out: Companies need to have policies around the data inputs and have a rigorous process to evaluate how data is used. Moreover, data sources must have enough diversity to compensate for changing dynamics. Approaches like human-in-the-loop model training and testing can offset some gaps in the data inputs. This is all the more important in a low-margin business.
- Test, refine, repeat: Feedback loops are necessary. Developers of algorithms and their users are often far removed from each other, and without user feedback, it’s impossible to refine an algorithm. A continuous machine learning model is needed to determine if the models have what is known as model drift, or the reduction of the model’s predictive ability. That Zillow’s models were not compensating for changes in the supply chain caused the models to be less accurate in their predictions.
- Consider the life cycle: When algorithms break or perform in unpredictable ways (for example, offering a price below what the company paid), a framework is needed to check them as often as needed.
While a disruptive business model that depends on technology comes with risks, when the technology is used effectively, it has the potential to expand and change the way home purchases and sales are done.