« Who you gonna call? | Main | Buddhism and the cartoon controversy »

Zillow

People with no real estate experience will often make incorrect assumptions about their properties' values.

Example: Someone might think "The guy down the street has had his property on the market for $100,000; my house is bigger, so it's worth more than 100K." However, the listed price of a property may have nothing to do with its actual market value. Anybody can list a house for any price; the relevant data point is not the listed price, but the sold price, which tells you what someone was actually willing to pay. If the home hasn't sold for the $100,000 asking price, that could just as easily mean that it was priced too high for the market.

Which brings me to Zillow. Zillow is supposed designed to help ordinary mortals see what a property's worth, the idea being that, if you have access to the same data realtors have at their disposal, and you a apply a smart algorithm to the numbers, that's all you need to make an informed estimate of a property's real value. However it could be that Zillow actually just legitimizes incorrect assumptions about value by associating them with real data.

I'm talking from (admittedly limited) experience. Marsha and I are selling a condo, and our own research suggested a value between 55K and 60K. (Marsha's a realtor, and she knows her way around a market analysis). Zillow shows a value around 72K.

Somebody's off by 20%.

Zillow allows you to look at the comparable properties it used, and we did. Zillow includes three sold properties from 2004. That's an issue off the bat - good comps would be more recent. But what really threw us was the sold data on the three comps: 294K, 5K, and 42K, with a price per square foot of 259, 4, and 71, respectively. The average sold price is 111 per square foot, and the recommended price is based on 79 per square foot, which suggests that Zillow appropriately considered other factors - but that price is still too high.

The sold price data alone suggested that the three comps were hardly comparable. Looking closer, we found that the house that sold for 294K was shown to have a value of 74K, so the 294K may be bad data. A realtor would have have had the good judgement to toss that property out of the mix... in fact, would probably have thrown all three properties out, because one was unusually low, and another was much older.

The local Multiple Listing Service has its own market analysis program that may be more effective, but no algorithm so far has successfully encoded human processing of visual data and judgement calls based on an analysis of property condition, relative size, modifications, condition of the neighborhood and surrounding houses, etc. Some of that is ambient data that isn't stored where a system like Zillow and find it. A good realtor will get beyond the data and spend a lot of time touring and evaluating comparable properties, applying an experienced business perspective to data and physical observation to get an accurate sense of market value.

Which is not to say that Zillow isn't a cool tool for gathering data. I just wouldn't use its results as the basis for real-world business decisions. And I think we have to acknowledge that there are very human aspects of business processes and decisons that you can't address with code alone.

jon posted this at 8:36 AM
Share on Facebook| email to a friend

 Subscribe to RSS feed for Weblogsky