Zestimate Takes Logical Step Into AI
Zillow recently announced accuracy improvements for off-market Zestimates, which will now utilize neural networks to generate valuations. We believe historically the majority of the Zestimate used algorithms that were designed by data scientists. These algorithms are static in nature and therefore, have difficulty keeping pace with a hot housing market. Neural networks are essentially algorithms that have the ability to dynamically create new algorithms on their own, hence the phrase machine learning. The bottom line is the new Zestimate should be more accurate because it will be more agile.
Initially, Zillow has reported modest improvements in accuracy from switching to neural network-based outcomes. What’s more important than the initial accuracy gains is that Zillow has just put AI on the clock, given we can now timestamp the Zestimate’s accuracy pre and post addition of the neural network. If history repeats itself, we should see measurable improvements in the Zestimate’s accuracy over the next two years.
Zestimate accuracy inches forward
As a reminder, there are two flavors of the Zestimate: on and off-market. Off-market Zestimates are the most important Zestimate for the success of Zillow 2.0 because the majority of live Zestimate cash offers will go to homeowners whose homes are not on the market. While sellers who have already listed their home could pivot and sell to Zillow Offers, they would likely have to navigate a listing agent contract, which we view as unlikely.
The off-market Zestimate median error rate is now 6.9%, compared to 7.8% previously. In plain English, this means 50% of the 104m off-market Zestimates are accurate within +/- 6.9% of a home’s eventual sale price. The other 50% fall outside of this +/- 6.9% range. In short, the off-market Zestimate is still unpredictable and has a long way to go. That said, the important thing is that it’s getting better, which we believe is important for three reasons:
- Expands live Zestimate buybox. The tangible result of the Zestimate improvement is that the pool of eligible homes for a live Zestimate offer will “likely increase by 30%” from 900k to roughly 1.2m (a fraction of the ~105m off-market homes in the US). Increasing the box of addressable homes is key for acquiring inventory, something Zillow struggled to do in the March quarter, which led them to guide Homes revenue for the June quarter about 10% lower than expectations. The more homeowners that see a live Zestimate and the more accurate it is should increase seller leads and conversion, leading to more inventory.
- Builds trust with consumers. The reason a more accurate Zestimate builds trust is because pricing transparency is core to the iBuying experience, given homeowners initially anchor their decision to sell based on price. If the live Zestimate undershoots, Zillow won’t build inventory. Regardless of whether it’s a machine or a human, if you bid too low, you’re not going to get the house. If the initial live Zestimate overshoots and the company must materially lower its final cash offer after an in-person inspection, consumer trust will erode.
- Maintaining its lead in search. The Zestimate helps Zillow control of the top of the real estate search and discovery funnel because the Zillow surfing phenomenon is rooted in people’s curiosity to know the value of their dream home or their neighbor’s. The more accurate the Zestimate, the more valuable it is to users. Maintaining control of the top of the search funnel is important for Zillow 2.0 because it provides a huge organic audience (220m monthly visitors) to which the company can advertise its iBuyer offerings at no added cost. Over time, this marketing cost advantage should translate into better margins for Zillow and better offers to home sellers, leading to more inventory and selection for home buyers, driving the flywheel.
Putting it together, Zestimate accuracy is a key piece in Zillow’s iBuying puzzle. We are long-term believers in the power of AI and believe that machines will eventually be capable of consistently appraising the value of a home better than humans.