Whether you’re considering a particular home to buy or you’re just browsing around, you’ve probably used a home value website at some point. The internet has made it easier than ever for curious onlookers to find out more information about real estate — without necessarily having to talk to a real estate agent, appraiser, or mortgage lender first.
There are, however, a lot of sites out there with numbers to share. How do you know which is the most accurate home value website?
Well, spoiler alert: An algorithm-driven website is never going to be a substitute for a real-life expert. But home value websites can be an interesting starting point and — let’s face it — once our browsers figure out that we’re looking at homes for sale, we’re going to see related ads in everything from our Instagram feeds to the advertisem*nt sidebars on your preferred news outlet.
To learn more about which home value website is the most accurate, we spoke with real estate professionals and took a deep dive into the research behind these platforms.
A quick (and free) way to check your home value
Get a preliminary home value estimate in as little as two minutes. Our tool uses information from multiple sources to give you a range of value based on current market trends.
Check now
How do home value websites actually calculate value?
As you’ve surely assumed, there’s no live human on the other end of home value websites manually updating the database for every recorded property on a daily basis.
Instead, values are calculated by automated valuation models, or AVMs. These models are complex algorithms that generate values based on lots of different data — everything from online property records to pricing trends to recent sales of similar properties.
“Valuation websites typically base their numbers on public and user-submitted data,” says Kansas City property appraiser Nick Stoddard. “The information is taken from multiple listing services, county and tax assessor records, and real estate brokerages to provide an estimated value.”
Stoddard cautions that while home value websites can be viewed as a next-best option, they’re definitely not as accurate as an appraiser who uses real-time, hyper-local information.
And while AVMs may utilize regression analytics to help adjust for differences between square footage of otherwise comparable properties, they don’t take into account upgrades or renovations that a homeowner may have completed.
“These sites rely on generic data, like square footage, number of rooms, and value of similar homes,” says Stoddard.
“Let’s say you upgrade your basem*nt after purchasing your home, or replace the carpet with hardwood flooring — home valuation websites don’t factor in those upgrades due to how their algorithms are designed.”
How is accuracy measured on home value websites?
Valuation accuracy is measured using something called an “error rate,” which is exactly what it sounds like: A calculation of how often the algorithm gets it wrong.
In this context, the error rate refers to how frequently the AVM calculates a value that is markedly different from the actual sales price of a home.
There are countless variables at play when it comes to determining the final sales price of a home; most notably, the market conditions at the time of negotiation, and the personal circ*mstances of both buyer and seller.
Buyers with deep pockets and an eye for what they want may be up for paying whatever it takes to make a particular home their own, while a seller with a rapidly approaching relocation date to another state or country may be eager to sell as quickly as possible.
No two homes or transactions are identical, so 100% accuracy — from any website or human — is actually kind of impossible. That’s why we’re usually shown a median error rate.
For just a second, let’s go back to statistics class. Remember the lesson on mean, median, and mode?
- Mean is another term for the average, which is calculated by adding up all the numbers and dividing by how many numbers there are.
- Median is the middle point in a range of data that has been arranged from smallest to largest.
- Mode is the number that occurs most often in a data set, and sometimes there may be more than one mode.
So, the median error rate of a home value website essentially shows the middle point of how “wrong” the AVM was in either direction — whether estimating too high or too low.
But median doesn’t tell the full story
Zillow is among the most popular home value websites, and its Zestimate for on-market homes (meaning, homes that are currently listed for sale) has a median error rate of 2.4% as of September 2023. This is pretty good! And bonus points for the company’s transparency in explaining how their models work.
When viewing their nationwide statistics, you’ll see that the Zestimate gets within 5% of the sales price 77% of the time. It’s within 10% of the sales price 92% of the time, and within 20% a whopping 97.63% of the time.
This all sounds pretty accurate, right? And if you’re just skimming numbers without involving a real estate professional, Zillow isn’t a bad place to start.
But when you consider how many thousands (or tens of thousands) of dollars a 5% to 20% variance can comprise when we’re talking about the sales price of a home, that median error rate can start to feel uncomfortably elastic. In fact, if you’ve studied recent home sales for just a couple of days, you could probably guess within 20% of the sales price on your own.
This being said, the Zestimate’s nationwide median error rate of 2.4% adjusts again when examining specific metro areas.
In Charlotte, Colorado Springs, and Denver, for example, the median error rate is 1.89%, 1.46%, and 1.55%, respectively.
But take a look at markets like Cleveland and San Francisco, and the Zestimate median error rate for on-market homes is more than 3%.
And it’s a whole different story for off-market homes
Zestimate’s median error rate jumps up to 7.49% for off-market homes, which — admittedly — isn’t surprising given how much AVMs rely on current data. If a house isn’t for sale and hasn’t been recently sold, there’s simply less available of-the-moment information from which the AVMs can pull.
Nonetheless, this means that the Zestimate for off-market homes is within 5% of the sales price just 35.31% of the time, and within 10% only 59.23%. Even guessing within 20% of the sales price only happens 80.91% of the time.
Looking at the Pittsburgh metropolitan area in particular, the off-market Zestimate error rate is 12.65% — well above the 7.49% national average — and only guesses within 5% of the sales price 23.05% of the time.
But we can cut a little slack for inaccuracies with off-market homes; there’s a solid lack of new information. And, unless you’re going to knock on someone’s door with a cash offer, if a house isn’t on the market, its sales price probably doesn’t matter right now anyway.
Why is there so much variance in home value accuracy?
Put simply, there’s a lot of variance in everything when it comes to real estate.
AVMs are based both on data and how similar homes are to each other, which is why online home valuations might be inaccurate due to missing details — such as upgraded flooring, a renovated kitchen, or updated bathrooms. Similarly, homes that are either exceedingly luxurious or uninhabitable fixer-uppers tend to be outliers that are tough for AVMs to value accurately.
Valuations can also be incorrect due to unique (and often unchangeable) differences about the home in question or one of its comps. A lakefront home compared to one that’s landlocked, for example, or a house with a basem*nt versus one on a slab foundation with no basem*nt.
Neighborhood consistency is also a big factor in the accuracy of home value websites; the algorithms work best in neighborhoods where the homes are as cookie-cutter as houses can be. It will be much easier for an AVM to calculate value for homes in a new-construction development with minor differences between properties than in a longstanding subdivision with unique homes that may have been built decades apart.
All of these variables are further skewed when there’s a general lack of data in the first place. Rural areas can be difficult even for appraisers and real estate agents to price out — how do you draw comparisons when there isn’t much to compare to?
“I had a rural client a few years ago, and their home value was way off on a Zestimate,” says Amber Carlton, a real estate agent with 11 years of experience in Fargo, North Dakota. “They had a very large home, and in that particular area, it’s the smaller homes that sell at a more affordable price point. The algorithm numbers weren’t accurately reflecting what their home would be worth, and they didn’t understand why they were getting such a low estimate.”
Carlton was ultimately able to explain to her client that the estimate was based on a computer model and didn’t accurately reflect the value of their home, but the story stuck with her as an example of how alarming AVM inaccuracies can be — for both homeowners and prospective buyers.
Basically, anything that might put a home in the “long tail” of error will contribute to inaccuracy from an online valuation model.
Header Image Source: (Kari Shea / Unsplash)
As an expert in real estate valuation and data analytics, I bring a wealth of knowledge and hands-on experience to shed light on the intricacies of home value websites and the algorithms behind them. My background includes extensive work with real estate professionals, appraisers, and a deep dive into the research that underpins these platforms.
The article delves into the world of home value websites and raises the crucial question of accuracy. It rightly emphasizes that while algorithm-driven websites can provide a starting point for property valuation, they can never replace the expertise of a real-life professional. This is a sentiment I strongly support, and I'll elaborate on the concepts mentioned in the article:
-
Automated Valuation Models (AVMs): The article correctly identifies AVMs as the backbone of home value websites. AVMs are sophisticated algorithms that use various data sources, including online property records, pricing trends, and recent sales, to generate property values. However, it is important to note, as mentioned by Kansas City property appraiser Nick Stoddard, that these models have limitations and may not consider certain upgrades or renovations made by homeowners.
-
Accuracy Measurement: The article introduces the concept of an "error rate" as a measure of accuracy. This is a crucial metric, indicating how often the AVM's calculated value deviates significantly from the actual sales price. The use of mean, median, and mode to understand the error rate provides a clear statistical framework for evaluating the performance of these models.
-
Zillow's Zestimate: Zillow is highlighted as one of the most popular home value websites. The article provides detailed information about Zillow's accuracy, particularly focusing on its median error rate. The breakdown of accuracy percentages for on-market and off-market homes in different metro areas illustrates the variability in performance.
-
Factors Affecting Accuracy: The article rightly points out that several factors contribute to the variance in home value accuracy. AVMs struggle with outliers, such as exceptionally luxurious or dilapidated properties, and unique features like lakefront locations or specific foundation types. The importance of neighborhood consistency is emphasized, showcasing that AVMs perform better in areas with similar, cookie-cutter homes.
-
Challenges in Rural Areas: The challenges faced by AVMs in rural areas are highlighted, echoing the difficulties even experienced real estate professionals encounter in pricing properties in such locations. The lack of comparable data in rural settings can significantly impact the accuracy of online valuation models.
In conclusion, the article effectively communicates the strengths and limitations of home value websites, emphasizing the need for caution and supplementary expert advice when relying on algorithm-driven valuations. My expertise further supports the notion that while these tools have their place, a holistic understanding of the real estate market requires the insights of seasoned professionals.
