At the Inman Connect conference last week Zillow announced that their Zestimates were getting better now achieving an accuracy of within 5% for 25% of the homes evaluated. They also said that they are within 10% of the sales price for 50% of the estimates and within 20% of the sales price for 75% of their calculated estimates.
I have a lot of respect for Zillow and their business model. However it seems that Zestimates will never result in an accuracy that one can rely upon. When there is a wide range of possible accuracies, without any determination of the range in which your property falls, your estimate becomes a shot in the dark. Which range does it fall within? The 5%, 10%, 20% or greater?
To see how valuable the Zestimates algorithms are, I took all sales for Wellesley, MA for 2007 and compared assessment to sales. This raw analysis came very close to beating Zillows algorithms' for the stated accuracy. However my analysis beats their performance if I added one data point; building permits - identifying those homes that were being torn down or heavily rehabbed. There are a number of other tweaks that one could do to improve this simple model but at the end of the day knowledge of the property and local market conditions will always prove to be the only way to obtain an accurate assessment of a particular property.
My results with building permit filter:
Estimated Value Within | Accuracy | Accumulated |
5% | 29% | 29% |
10% | 23% | 52% |
20% | 27% | 79% |
My estimate is only a small data point and may prove to be highly inaccurate with a larger sample.
;)
Links;)
Tags (Tag Cloud);)
Archives
I checked my home on Zillow about 2 weeks ago....$289k
As of today it shot up to $304k
I find it interesting....lol