Have you ever wondered how accurate the Zillow “Zestimate” really is? Whenever I am doing appraisals I have homeowners tell me what their Zillow “Zestimate” is. I am sure they are hoping that my appraisal is similar to that number (if it is high), however the two could be close, but more times that not they are pretty far apart.
I actually just found out recently how you can determine the accuracy of Zillow in your area. Take a look at the video to learn how. If you don’t see the video then click HERE to go to my Youtube channel.
Zillow has gone back in time to compare their historic Zestimate with what the property actually sold for. They claim that their “estimated market value” is not an appraisal, but in reality an estimate of market value is an appraisal, but it’s one that cannot be used by a bank and one that you may want to think twice about before using to determine a sales price for your home. They provide a range in value which can vary widely. This information is based on public and user submitted data, both of which can vary in accuracy. County records in many states do not reflect newly finished areas or additions where a building permit was not obtained. Zillow explains that the further apart the spread from high to low the less accurate their estimate is. The counties in which various sources of information are available are typically the ones that are the most accurate.
I believe that you have to see Zillow for what it is and don’t expect much more. Without actually looking inside a property, and using the human touch to sift through sales information, you are only going to attain a certain level of accuracy. Another thing I would like to point out is that computers do not recognize similar market areas. An example of this is school districts. In many cities property values are driven by what school district the home is located in, and I have seen Zillow comparables pulled from areas in different school districts, which can give an inaccurate indication of market value. One other very obvious area where inaccuracies can occur is with property condition. Appraisers are trained to research this information so that adjustments can be made. If a comparable is in inferior condition an upward adjustment will be made to reflect this.
The bottom line is that an appraisal made by a person is going to be much more accurate than a value estimate provided by Zillow or other similar website. Have you had a “Zillow” experience where the value was close? far off? I would like to hear your opinion so just leave me a message below.
If you have any real estate appraisal related questions you can call me at 205.243.9304, email me, or connect with me on Facebook., Twitter, or Youtube.
Six weeks ago Zillow estimated the market of my home as $425,000. Today they show the market value as $229,000. I purchased this home in 2014 for $299,000 and the real aporaisal was $335,000. That is only one example of how Zillow is hurting sellers. They may have a disclaimer about not being an appraisal, but few people read it. They simply look at the Zestimate compared to the selling price and either are interested or move on. Zillow should remove this fake information or get ready for more class action lawsuits.
You are right Michael, and other homeowners feel the same way. I’m not sure if you have read about it but an Illinois home owner has filed a lawsuit because the Zestimate on their home is low and it is causing them to have a hard time selling it. Most consumers do place a lot of weight on the Zestimate which is why it should be ended, especially considering how inaccurate it is.
What’s happening with Zillow is that they appear to have cooked their historical data in order to justify their algorithm, which puts the identical house sitting next each other as far as north and south – approximately $100K difference in the valuation. I bought my house back in January, and the historical data I saw in January and those I am seeing now are NOT the same anymore. The Zestimate on my house back in January indicated the range of $650K to $780K from 2010 through 2017. Now the same Zestimate on my house shows $530K to $680K from 2010 through 2017. Basically, the graph had moved down by about $100K. The funny thing is, my next door neighbor has the exact same floor plan as mine except that his house doesn’t have a detached casida (so a little smaller in the overall square footage). As of this writing, the Zestimate on his house says $577,567 for 2,887 SQFT while on my house $525,792 for 3,170 SQFT. Does anyone understand this? Can anyone please explain this to me?
You are making perfect point! They are adjusting historical data so they algorithm looks correct. I bought house in 2006 for $450K and seen its Zestimate at the time being $550K. During next 10 years I have seen price go up as much as $670K and down to $330K. These highs and lows are gone now from the graph and showing high $540K where it was $670 and $370K where it was $330K. HISTORICAL DATA CANNOT CHANGE IF YOU ARE NOT A LIAR!
As of now the price is so skewed, you wouldn’t believe. The house next to mine is a little more than 1200 sq feet in size, is “zestimated” at $359K, never renovated, 12000 sf lot.
I have 3,493 sf living area, new kitchen, 2 new bathrooms, granite counters everyhwhere, Anderson windows + 65000 sf lot.
Guess how much is “zestimate”: $475K. In 11 years after all the renovations, and real estate in the town up almost 50% compared to 2006, my zestimate went down $75K!
Come on, Zillow, get real! Fire you programmers and think what you do to people!
Luckily, I love my house and don’t want to sell it!
Totally agree with you on this one. I have heard it a million times too, where the homeowner thinks the price of their home is the Zestimate. It’s almost frustrating sometimes. The problem is, sometime it actually is really close to the value of the property, and sometimes its the furthest from the truth. Thanks for the insight.
Yeah, I have seen the Zestimate be close to the value of a home but it is usually the exception rather than the rule. I think with appraisers educating the public on the weaknesses of the Zestimate they will begin to know they should not rely on them.
I understand Zillow doesn’t walk into homes, and also that they don’t realize what upgrades have been done in many circumstances, but how can they be so far off when an actual traditional sale has been made, and a market has a certain average appreciation after that. Buying our home in Sept 2012, we paid (with conforming appraisal, and with several offers) $530k. At the time, Zillow valued the home at $470k. Do you think the actual sale had any impact on the increase in ‘zestimate’ over the next 18 months? None. Based on average appreciation in the area, the current zestimate is $560k, which is almost exactly the increase expected from average appreciation IF they were right about their valuation in Sept 2012. Realtors and apprasiers put it at $610k.
Thanks for sharing John. You have to realize that Zillow is not in the business of giving accurate and reliable value estimates because that really cannot be done with computer models, especially in areas where homes vary in age, style, design, and quality. I think they use the value estimate feature to attract people to the site so they can get them to use the other portions of their site where they make the most income for their company. That’s just my two cents. 🙂
Tom,
Great article and video. I recently spoke with a realtor that indicated that home owners can go into zillow and modify their properties (ie. bedrooms, baths, sq ft). This can manipulate the already erroneous values. Have you seen anything like this?
Thank you
Jason, I have not heard of this. As a matter of fact I have heard of some homeowners that were very upset because they could not change the information to make it more accurate. I would be interested in finding out how you modify the Zillow information. Thanks for sharing.
I was able to change my information on zillow and was shocked when my zestimate went DOWN! My home was appraised for 165k at the end of 2009 and i paid 150k for it. The home had old nast carpet, cracked and scuffed base boards, old parkay flooring in the dining and bonus room and the walls were a million different colors. Since i bought it i put in all new carpet, new base boards and fresh paint throughout the house, put dowkn new wood in the dining room, layed new tile in the small bonus room between the garage and house and turned it into a large laundry room/ mud room, along with a new security system and ac unit….remember it was appriased at 165k less then 5 years ago, before the updates the zestimate was 88k and after updating with all this information it went down to 72k!!!! Makes no since to me! (Fyi this house had a brand new roof and kitchen when i bought it so the unly things that arent brand new and the bathrooms)
That website is a JOKE!
It does make me scratch my head most of the time Stephanie.
Zillow estimate is so inaccurate that some of the condo in my building the same features varies so differently.. Some are over 200k difference which does not make sense..
It doesn’t make sense to me either. That is the problem with Zillow because it gives very unreliable results. People need to realize they cannot make decisions based on what they see in Zillow. An appraisal by a live person who sees the property is always the best way to find out the value.
Thanks for the video! What do you think the main cause of the inaccuracies are? Do you think that a home that has been sold within the last year would have a more accurate Zestimate? We bought our home for 600,000 and now our Zestimate shows 617,000 after eight months with no work done to it (almost a 3% increase!). I understand that if a house has not been on the market in a long time, the Zestimate could be way off.
Great question Bryan. I think the main problem that exists with Zillow is their data sources for square footage information. It is my understanding that they take the recorded sale price of a home that sells and divide it by the square footage noted in the public records. Many times this square footage information is wrong, so the sold price per square foot is incorrect. They then take the sold price per square foot and multiply it by the square footage of the home in question. Again, this square footage can be wrong also, so the estimate of value is then wrong as well.
Sometimes the living area noted by the county records can be accurate, depending on what area you live in, as well as what type of house you have. I say this because in my area the square footage information on a one level home with no basement is more accurate than a 1.5 story home with basement.
Look, Zillow is not stepping into the homes. All it has are property parameters such as square footage, lot size, number of rooms etc. I tried to create my own statistical home valuation tool and was able to come within 2% of Zillow estimates most of the time in my neighborhood in So Cal. So, understand that when they give an accuracy parameter, most likely, it is the “standard deviation” from a statistical analysis. Now, most of us probably have heard of “six-sigmas” in manufacturing. Basically, a standard deviation means that if you have a bell-shaped curve, 70% of the values lie within 1 standard deviation on either side of the mean, 95% lie within 2 standard deviations and 99.9% lie within 3 standard deviations. So, obviously, saying that you can only value the home to near 100% accuracy by assigning a spread of +/-25% to the Zestimate is not very useful. This would also address Surbiton’s comment regarding the interview with the Zillow CEO. This is not a knock on Zillow, but it is the best you can do statistically. What Zillow cannot say is what the condition of the home is, what kind of upgrades were performed, what the neighborhood looks like, how motivated past buyers and sellers were in particular instances etc. Those are the limitations. Yet, constructing this valuation was useful for me. It was useful in seeing whether a home merited above-average valuation or whether it was priced lower than average for its condition. It did help me close on a home that was fairly priced with respect to its attributes. Statistics properly used can be useful. YMMV.
You have a very interesting story. Are you in the valuation business? Just curious since creating your own valuation model similar to Zillows was probably not easy. I do realize the limitations of their model but a lot of people seem to put a lot of faith in the “Zestimate” when maybe they shoudl not given their low level of accuracy.
Hi Tom,
I’m not in the valuation business, but I was looking to buy my first-home and as an engineer, I like to play with numbers. Creating a valuation model is not easy, and as you say, school districts do play a role in the valuation. Surprisingly though, in my neighborhood, I came closer to selling prices and some of the zillow estimates by using the school (in my case, the elementary school since all of the sub-divisions had the same high school) also as a parameter. One can end up seeing how elementary schools do play a role in the valuation. It was fun trying to do this. And as you note, the price per square foot is not linear. That is, larger homes sell at a lower price per square foot and you can create a model that creates this into account. It is slow work to create a model, but it is not really that hard (well, my day job I think is harder). Look up “hedonic regression” if you are more interested in this sort of thing (I think Zillow’s approach must be similar).
It’s good to see support on a larger scale (your regression model) for what we find on a more isolated and local level. Thanks for sharing your experience, it has really been interesting.
BTW, I think the Zillow Home Value index has value in reporting average trend in house price changes for a neighborhood. Less noise when you use a lot of homes in the analysis, and more accurate as you increase the coverage area (i.e., city instead of zip code). Would be interested in your thoughts regarding it.
Yeah, I agree that the data probably increases in accuracy when the amount of data increases. I think that “evens out” some of the imperfections in data.
Only a few days ago Zillow CEO Spencer Rascoff accepted in an Bloomberg interview that 17% of zestimates are more than 25% incorrect. Here is the link to the interview
http://www.bloomberg.com/video/zillow-sales-model-change-boosted-results-ceo-says-0qGTD~MvSvCjZ8dqUAFetA.html?cmpid=yhoo
It is outrageous that Zillow refuse ALL requests from homeowners to correct or delete obvious nonsense zestimates when the algorithm is so fundamentally flawed. Zillow causes more misery and financial damage to 20 million homeowners with garbage zestimates than the Taliban could ever hope to inflict, but Zillow is allowed to get away with this BS with no regulation whatsoever.
Thanks for sharing the link. Zillow could definitely learn better how to handle these problems with inaccurate Zestimates. I have heard many horror stories but I don’t think that most people take these values very seriously because they’ve seen how wrong they can be, if that is any consolation.
About 5 years ago had been getting lots of questions from lenders, borrowers/owners, etc about Zillow and their Zestimates so I did a quick analysis. I pulled 10 recent appraisals that I had performed and looked at the Zestimates for those 10 appraisals. I’m in the Denver, CO market which per the Zestimate Accuracy they give themselves 4 stars (apparently their highest ranking). In this area we have very good MLS and also very good Assessor records online so if their approach would work it seems that it should definitely work here.
However, the results were disastrous! Of the 10 properties, only 2 were within about a 10% variance of my appraised value. The other 8 Zestimates were wildly and inconsistently off. When I looked at their “comps” used with those 8 properties if an appraiser were to use their comps and method and value conclusions that appraiser would lose their license and likely end up in jail for mortgage fraud. The most surprising thing to me was how inconsistent they were.
I strongly suggest if you as an appraiser are being asked about Zestimates, or Trulio value or any other site that you perform your own local comparison to some of your recent appraisals to see for yourself how good or bad these sites are in your area. Here in Denver, where they feel they are most accurate, they were running at 20% or 80% go-to-jail accuracy based upon my analysis.
Wow Dave,that’s crazy. I have done this in my area as well and found them to be very inconsistent. It’s interesting that it is that far off even in their “Four Star” areas. At least maybe this give us appraisers a little more relief that we may not be replaced any time soon with computers, huh?
Humans > Machines.
I agree Ryan, Thanks.
Dear zestimate,
2 weeks ago you reported your estimate for one of my houses – 1104 D st, Antioch, CA, on your chart in amount of $135,000. However, today there is no $135,000 value on your chart at the end of August – beginning of September of 2014. What happened to the historical data?
Are you frequently changing your historical zestimates?
Thank you,