The Data Behind Online Appraisals: How AVM Tools Pump Out Home Values

Whether you’re getting ready to sell or wondering where your equity stands, there are a plethora of online appraisal tools that can give you an estimate of your home’s value in minutes.

While the output may be swift, the input isn’t so simple. These online appraisal tools, known in real estate as “automated valuation model,” or “AVMs” for short, use a combination of public records and homeowner-inputted data to calculate a property’s fair market value.

For a deep dive into the algorithms behind AVMs, we consulted Mischa Fisher, Chief Economist at Angi, Inc. (parent company of HomeAdvisor and Angie’s List) and former economic policy advisor for the governor of Illinois.

“AVMs are beneficial for reaching large audiences by scaling simple solutions. The process of sending an individual to your home to manually estimate its worth is a time-consuming task,” says Fisher, who has spent over a decade dissecting real estate and home improvement data.

It is worth noting that while the convenience of an instant home value estimate is appealing, the accuracy of an AVM is not always on par with that of a traditional home appraisal. Online appraisals are there to help you get a ballpark estimate of your home’s value — you shouldn’t rely on them to price your home for sale.

We’ll explore the elements behind the computer-based curtain, breaking down how these algorithms spit out your home’s estimated value in seconds.

Public data

Fisher shares that AVMs use a variety of public data, including “records of demographics, housing attributes, tax assessment, and county records on sales to train their predictive models.”

Public records contain detailed information about individual properties, such as lot size and square footage. AVMs analyze this data to discover your home’s past nine lives and deepest, darkest (publicly accessible) secrets.

AVMs may gather public data from the following:

• Government databases
• County assessors
• Recorder’s offices
• Real estate property listing websites
• Title companies

Types of user-generated data

Common types of public data used in AVMs include:

• Original sale price: Whether your home was first sold in the roaring twenties or the recent decade, AVMs analyze how your home’s value has appreciated (or, in some cases, depreciated) in value over time.
• County tax data and property tax records: An online appraisal reviews the property’s assessed value, the value that the county or local government assigns to your home for tax purposes.
• Liens associated with the house: AVMs review your property’s title to determine if there are any liens (e.g., a mortgage) that could impact the value of your home.
• History of deed transactions: Records of the property’s real estate transactions include sales deeds, mortgages, refinancing, price reductions, foreclosures, and beyond, which may influence an online appraisal’s value.
• Local market data: Recent home sale data of comparable properties in the area, or “comps,” is a crucial piece of the AVM puzzle.

Public data gaps

Fisher explains that “while public data is extremely valuable in determining a home’s value, it can be limited in scope.” One of the greatest discrepancies with public records is that many have incomplete, inaccurate, or outdated data.

Since public data sources are often run by varying government entities (i.e., city, county, and state) with different policies and procedures, the input of data can be inconsistent from county to county, leading to conflicting, outdated, and otherwise inaccurate data.

The lag-time in public data entry can present even more challenges in a hot market, where home values are rising rapidly.

Amber Taufen, a Managing Editor at HomeLight who has written about real estate for almost a decade and spent more than two of those years specializing in appraisal software, sheds light on public data discrepancies:

“You have to wait for the county to record sales information while home prices are going up weekly or monthly. So if you’re pulling data from sales that are three months old, your AVM might not be right.”

For instance, say you purchased your home 10 years ago, and your last tax assessment was five years ago. The property value indicated on those public records may be drastically higher or lower than your home’s current value.

User-generated information

User-generated information entails just what the name describes: data that is shared by the user of the online appraisal tool (yes, you!). Many tools allow users to input specific details about their home, such as renovations, upgrades, or additional bathrooms to more accurately assess the property’s value.

If you spent the last year renovating your kitchen, you want your return on investment to be reflected in your estimate, right?

Fisher explains how user-generated data can “capture attributes about more preferential, less immediately empirical sources of information about a home’s value, such as the desirability of the neighborhood, walking score, school district quality, and proximity to bars and restaurants, among other things.”

Not all online valuation tools collect the same user-generated information. If you use multiple online appraisal tools, you may notice that some ask for more or different information about your home than others. For example, HomeLight’s Home Value Estimator asks homeowners seven questions about their property to boost the estimate’s accuracy.

Types of user-generated data

Online valuation tools may ask for details on remodeling, renovations, or updates to structural elements, such as your sturdy new roof.

Many tools also inquire about your home’s surrounding area to gauge the desirability and accessibility of local landmarks, hot spots, and happy hours since location influences how much buyers are willing to pay for your home.

User-generated data gaps

Human error is the primary downfall of user-generated data. There’s no guarantee that what the homeowner is sharing is 100% accurate. Whether intentional or otherwise, homeowners may embellish or omit certain features of their home, property, or neighborhood that could impact the estimate.

Furthermore, it’s challenging for an algorithm to evaluate qualitative data, such as property condition and desirability of features. Compared to quantitative data like the number of bedrooms and square footage, qualitative data is more subjective, and therefore, more challenging to measure.

While many AVMs have categorical formulas designed to measure qualitative data, the subjective nature of this information limits the algorithm’s precision.

Sales data

AVMs gather the sale prices of recently sold comparable homes. Comps are recently sold homes in your area with similar features to yours. An AVM compares your home to these properties, adding and subtracting values based on similarities and differences.

Types of sales data

Online appraisal tools collect sales data from a variety of sources, including public records, listing websites, and private home sale platforms.

Sales data gaps

AVM’s cannot generate a comparable market analysis (CMA) as accurately as a real estate agent can to pinpoint your home’s value relative to recently sold properties. Here are three reasons why:

• Location: The recent sale of comparable homes is a major factor in the accuracy of AVMs. If your home is in a rural location, the AVM’s algorithm may struggle to find the recent home sale data in the surrounding area.

• Uniqueness: Neighborhoods with unique homes or unusual features present their own set of challenges for AVMs. The algorithm may struggle to determine an accurate value estimation if there is a lack of comps available.

• Non-disclosure: There are 12 “non-disclosure states” that do not disclose real estate home sale prices as public records. A real estate agent can access this sales data through the MLS (multiple listing service); however, AVMs cannot.

Factors that online appraisals do not take into account

While online appraisal tools are an efficient and easy way to jumpstart your journey of understanding your home’s value, their home value estimates are not entirely accurate. For an AVM to be accurate, it needs to understand the details of the house and neighboring houses, which is trickier for machines than mortals.

Fisher explains how one of the drawbacks to AVMs is the inability to account for all factors:

“[AVMs] only consider what they were created to consider. So, if there’s something novel or new about a certain home, neighborhood, or metro area, or something is changing faster than expected, AVMs can lose accuracy and present worse results than a trained professional could in the field.”

Taufen agrees, commenting, “Houses are not cars. You can’t put a Blue Book value on them.” While you may have two houses on the same street that appear to be identical to the algorithmic eye, they may, in fact, have striking inequalities once you step inside. “The interior of one might have Viking appliances and granite countertops, while the other may have Whirlpool and Formica.”

Here’s an overview of some details that an AVM might miss:

• Local area knowledge: The quality of your local school district and walkability to hipster havens are all tough for an algorithm to account for but can significantly influence your property value.
• Buyer preferences: Real estate agents understand what features and finishes local homebuyers value. An AVM cannot assess how a local buyer pool’s preferences for features like countertop material, storage, and flooring impact home values.
• Home condition, character, and quirks: An online appraisal cannot conduct a thorough property inspection of your home, which can leave home improvements and problems unaccounted for.

Get an online appraisal from HomeLight

As we noted earlier, not all AVMs are equal. For the most accurate instant home value estimate, choose an AVM that requests substantial user-generated data. HomeLight’s free online estimator tool uses a unique algorithm alongside a short questionnaire to generate the most accurate online home value estimate possible. Input your address and answer seven simple questions for a detailed analysis of your home value estimate.

A desktop appraisal works differently than an AVM

On a final note: Don’t confuse an online property value estimate with an appraisal a licensed appraiser conducts online, known as a “desktop appraisal.” Desktop appraisals grew in popularity during the COVID-19 pandemic when the real estate industry strove to reduce contact for home sale transactions.

A desktop appraisal is different from an AVM since a licensed appraiser is involved in the process. Even if an appraiser never sets foot on your property, they can access up-to-date MLS data and assess your home’s qualitative features far better than a batch of code.

Taufen explains how this process is a good alternative to having “boots on the ground” while still having an appraiser doing the actual evaluation. With this approach, a third-party data collector typically takes pictures of the property and shares them with the appraiser, allowing the appraiser to evaluate a home’s value from their office.

Some appraisers also work with image recognition programs that automatically rate the condition of a home by detecting issues like water or structural damage.

Educate yourself before embarking on your AVM journey

To summarize, free online home value calculators are a great way to get a glimpse into the value of your home without ever leaving it. But you should take your online estimate with a grain of salt, followed by a healthy dose of guidance from a top real estate agent in your area. Your agent can advise how local market trends and your home’s qualitative features shape your property’s value.

Header Image Source: (Markus Spiske / Unsplash)