Back to Results
Please enter a valid email address.
Stating a discriminatory preference in an advertisement for housing is illegal. If you think this content is discriminatory or otherwise inappropriate and feel it should be removed from Zillow, please let us know by completing the information above.
We will review this content. Thanks for helping make the site more useful to everyone. To learn more, read Zillow's Good Neighbor Policy.
Thanks for the thoughtful commentary here. I thought I'd wade in by first returning to the original issue about responding to feedback about Zestimates and changing individual Zestimate values.
What we desire to achieve with the Zestimate is that the models produce the same result given the same inputs (i.e., two identical houses right next door to each other, previously sold on the exact same date for the same price, and with the same tax assessment should get the same Zestimate) and that this result is as close to the actual sale price as possible. In some cases, however, this doesn't happen. Two identical homes may have different valuations or the valuations may be the same but they differ substantially from the real market value. These issues arise from defects in the overall algorithm or specific statistical models themselves (of which we create more than one million each time we create Zestimates three times each week), and it is these issues that we are constantly monitoring and attempting to fix through refinements in the algorithm and/or models.
Any valuation methodology will produce a difference (error) between the valuation and an ultimate sale price. We publish that data here. At Zillow, we endeavor that this error be only algorithmic in nature (and thus something that we will ultimately be able to correct with more sophisticated approaches) and not be subject to the vagaries of who has and has not contacted Zillow.
In essence, with this approach, we are betting that Zestimates will get progressively more accurate over time and that there will be fewer inconsistencies between identical properties. We have put this belief into practice as demonstrated by the fact that we have developed three different valuation algorithms since Zillow's launch in 2006, each getting significantly more accurate than the previous one.
Does this mean that we don't want to hear about issues with individual Zestimates? Absolutely not. We appreciate feedback on individual Zestimates, because that helps us improve the model so that all Zestimates are more accurate. These individual examples give us clues that help us find algorithmic issues that affect many more homes, and it is these algorithmic issues that we want to fix. Unfortunately, fixing these more general issues takes more time than just adjusting an individual Zestimate to some value closer to the current homeowner's estimate of home value. Sometimes it may be difficult to fix this new issue without negatively affecting many more Zestimates. And sometimes the issue is so subtle that it simply requires time to find a good solution.
So please rest assured, we are listening when consumers give us feedback about their Zestimates. We monitor this feedback to help us identify systemic issues that need to be addressed. Unfortunately, given only a handful of statisticians, we are not able to provide a response to every single piece of feedback or immediately identify the root cause for any specific inconsistency. We are trying to achieve the greatest amount of good for the greatest number of users, and we believe that the best way to do so is to constantly be improving the core algorithms in such a way that the overall accuracy improves and inconsistencies are minimized.
"As a final note, let me quickly address a few other issues that have come up on this thread:
On the accuracy of Zillow Zestimates: For 2011Q4, our median Zestimate error was 9.4% with 52.1% of sales in that quarter falling within 10% of the Zestimate. Since 2006 when we launched, our models have become substantially more accurate even while the overall market environment has become one in which pricing is more uncertain and volatile. We will continue to get more accurate. Information is better than no information in my mind, so I can't see the rationale for not having valuation information like the Zestimate. If we agree on that point, then the debate shifts to different sources of such information. I prefer lots of different sources. Accuracy conversations tend to become comparative in nature pretty quickly so I'd encourage Zestimate critics to be explicit about their preferred alternative valuation approach and its overall accuracy on a sufficiently large sample of sales."
"On whether Zestimates determine or simply track market prices: By design, Zestimate errors are generally symetrically distributed about zero meaning that there are just as many Zestimates that end up being above an ultimate sale price as those that end up being below the sale price. If Zestimates were, in fact, driving sale prices then it would be difficult to account for the fact that half of all buyers pay more for their house than the Zestimate.
On the suppression of individual Zestimates: I have to agree with hpvanc on this issue: Either Zestimates are useful or they are not. I personally believe a valuation of the accuracy we achieve is useful in a host of real estate decisions even though I value the additional input from a realtor or appraiser for certain of these decisions. If Zestimates are useful, then they are only useful if all are shown equally. They are considerably less useful if I'm only allowed to see the ones that are higher than the real market value."
On whether Zestimates determine or simply track market prices: By design, Zestimate errors are generally symetrically distributed about zero meaning that there are just as many Zestimates that end up being above an ultimate sale price as those that end up being below the sale price. If Zestimates were, in fact, driving sale prices then it would be difficult to account for the fact that half of all buyers pay more for their house than the Zestimate.
"When you say "Zestimates can and many times do show errors," do you mean by "error" that the Zestimate does not equal the sale price on the day of an actual sale? That's a high bar as I'm sure you are aware, and one that few appraisals and CMAs will meet. Zestimates, appraisals, and CMAs will all have some error. We publish our accuracy rates religiously, and I'm happy to discuss our accuracy in the context of any accuracy metrics for the other two.
"Regarding the FICO analogy: I would view the analog of an annual credit report (where you see the facts about your credit record and current FICO score) to be a home details page on Zillow where one sees the facts about the home and the current Zestimate. You can access that data anytime on Zillow, not just once a year as in the FICO analogy. Yes, while one can correct factual errors on a credit report, one can also correct factual errors about a property on Zillow. One cannot, however, request a different FICO score if the credit facts haven't changed.
Regarding credit reporting regulations and whether they should be applied to Zestimates: Zestimates are not appraisals and cannot be used in lending decisions so regulations designed for such decisions don't seem applicable. Here, I'd probably invoke another of the examples that I provided earlier: car valuations by Kelly Blue Book. Should these be subject to regulations as well?"
Please enter text in the "Enter the text to display" field.
Please enter text in the "Enter URL" field.
Please enter a valid URL.
Please insert a video embed only
Zillow Advice depends on each member to keep it a safe, fun, and positive place. If you see abuse, flag it. More on our Good Neighbor Policy.