Fixing the Zestimate

 

Suggested Changes to Reduce Criticism and Improve Industry Relations

The Zestimate – Zillow’s AVM-calculated house price – has been a double-edged sword from the beginning. On the one hand, it gives property owners, buyers and sellers a comforting sense of empirical certainty – that an abstract, passionless computer algorithm, has declared from on high what a property should be worth. And it’s good entertainment for the general public. On the other hand, the Zestimate is not actually an empirical reflection of the value of the house. It is not an appraisal, a BPO, or even based on a casual walkthrough. It is simply a mathematical model that pulls together data such as the number of bedrooms, number of baths, comparable homes sold locally, and a dollop of Zillow’s special sauce. As such, it is simply not very accurate. Zillow explains just how inaccurate it is here:
http://www.zillow.com/zestimate/#acc (http://www.zillow.com/zestimate/#acc – but I don’t believe many people read and understand that document. And Spencer Rascoff has spoken many times about the context in which the Zestimate should be placed – but I don’t believe most people who see the Zestimate have ever heard him speak about it.

Instead, people believe in the Zestimate that they see on the listings on Zillow.com. One reason, which I’ve already mentioned, is that it gives people supposedly objective backup for what they would like to believe about the value of a house. Another reason, perhaps both more insidious and effective, is that it gives the impression of extraordinary precision and accuracy. In the Star Trek episode, Return to Tomorrow, McCoy asks Spock how many miles of rock they will have to tunnel through; Spock replies, “Approximately 112.37 miles.” Anyone who can give the answer to two decimal places, right off the top of his head, must be a genius. The rock layer must be between 112 and 113 miles thick, even if McCoy’s tricorder is only accurate to plus or minus twenty miles. Zillow has it both ways; it has its page of fine print indicating the wide range of values a Zestimate may represent, certainly not read by everyone who visits the site, and it has the minutely exact figures it displays on its many pages that everyone sees. If Zillow says a house is worth $182,379, one imagines the house is worth between $182,000 and $183,000. It’s the nature of seeing such a precise number. Now, if Zillow were to just display the range for the house ($170-$195K) and a “confidence” rating next to it – as most professional AVMs do – the confusion and controversy around the Zestimate might go away.

The way the Zestimate is presented currently, it causes a variety of issues. Some sellers end up bullying their agents into going along with using this beautiful, exact number they found online, and many agents will go along with this to get the listing. Arguing about why this number isn’t useful in pricing a listing wastes the time of professionals, and it may harm both sellers and agents who price their home too high or low after seeing the Zestimate. It also could harm buyers that don’t put a bid in on a house that is priced right but which they think is overpriced because of the Zestimate. We all want this number. But this number, at least as currently presented with its deceptive precision, is bad for everyone. Let’s just consider that.

I think I understand why Zillow likes the Zestimate. Pseudo-accurate Zestimates bind customers to Zillow’s services. Zestimates represent product differentiation at its finest, especially when it comes to the impression of accuracy that no other portal claims to have. When buyers and sellers engage real estate professionals, Zestimates give them a sense of information asymmetry — they can “turn the tables” on their professionals with the knowledge they’ve gained online. It gives some a false feeling of power, and others a misperception that the professional is less valuable.

If the Zestimate as it is today were scrapped, and a price range and a rating of confidence in that range were displayed either on its own or, as a compromise, in addition to the specific number, this would place the locus of knowledge and experience back with the Realtor®. The buyer or seller would be asking the Realtor®, “What does this range mean? Where do you think this home actually falls between the top and bottom numbers based on the condition of the house?” Changing the Zestimate to a price range and confidence rating would be good for buyers and sellers, because they would have more realistic if not more definite ideas about prices, and because it would cement rather than distance the relationship they have with their agents, whom Zillow and I both agree are still necessary to the home-buying process. Meanwhile, by modifying this feature, Zillow could eliminate one of the biggest criticism magnets regarding the site. To Zillow, are these benefits worth its taking its perceived omniscience down a peg, and somewhat lessening its product differentiation? I would argue that they are. I’m sure the people at Zillow have considered this approach, but I think it’s worth continuing to reevaluate.