No one likes to have their expectations missed. Whether it’s a product that did not work as advertised, or a rebate with fine print that makes it unusable, it’s beyond frustrating when expectations don’t meet reality. This is especially true for automated valuation models (AVMs).
AVMs are software-based pricing models that often use public records to estimate the value of a home or other real estate. Commonly known AVMs used by consumers are Zillow’s Zestimate and Redfin’s Estimate, but there are many more of these sophisticated AVMs on the market.
“The difference in predicted home values that consumers see and the accurate values that often take them by surprise could be due to a disconnect between marketing and underwriting goals.”
Today, it’s an accepted norm to show homebuyers or sellers an automated home value that is accurate enough to get their attention. But these values often fail in the level of accuracy needed to be usable for a consumer’s financial transaction. This results in a rude awakening for a seller who expects to list their home at the value they’ve grown accustomed to — or a homeowner who is dreaming about the project they can finance with their current level of equity, only to be approved for a lower loan amount.
The difference in predicted home values that consumers see and the accurate values that often take them by surprise could be due to a disconnect between marketing and underwriting goals. There are many business reasons why marketing and underwriting operate in silos. Highly accurate AVMs built for underwriting purposes cost more to produce than a marketing-first approach.
Marketing-based AVMs are often displayed for free to consumers in order to generate leads. Some companies may even go as far as showing a less accurate value as clickbait, hoping the homeowner will become a lead just to fix what they know to be true about their home.
It’s time to rethink this current norm, especially in today’s market where affordability is a challenge for so many. AVM providers have a larger opportunity to reduce friction in housing finance and provide viable alternatives to labor-intensive processes.
“AVMs that harness large amounts of data can perform more calculations than a human in a matter of seconds, and they provide an objective approach to value.”
Known low-quality values ultimately erode consumer confidence, particularly if no context of value accuracy is provided. To make confident decisions involving real estate, both accuracy and transparency are needed to understand a property’s value. AVM solutions need to provide a precise indication of confidence that allows someone to act upon the value with an expected outcome. It matters not only to real estate professionals but to consumers, financial institutions and investors.
When consumers are shopping for homes or checking the value of their own home on their favorite site, an expectation has been created that an answer will always be available. This helps to create loyalty and confidence for return visits.
Imagine if you typed questions into Google and the searches frequently turned up zero results. Chances are, you would look for a different search option. It’s better to see something — even if it’s irrelevant to your original search — than to see nothing. The human brain wants to see results.
The problem comes when these results are not accompanied by some context of how accurate the model’s prediction actually is. AVMs are typically designed to predict a fair market sales price or an appraised value for a given property. Predicting the outcome before a sale or appraisal has happened can be a powerful tool for gauging the timing of getting a loan, listing a home or making a purchase. Knowing with certainty that a value is accurate, and that there’s a high likelihood it will be within a margin of error for a future sale or appraisal, is what actually empowers consumers to act upon the data.
Telling a friend you are 50% sure you will arrive at their house on time for dinner, versus being 98% sure, will probably change their expectations and actions. A homeowner equipped with the knowledge about the accuracy of an AVM can make informed decisions about their finances. They can better choose the timing to take advantage of their home equity, which might enable them to renovate their home, consolidate debt or send a child to college. Potential buyers with this context can ensure they are financially prepared to purchase a home. This reduces friction in the process and ultimately leads to fewer failed transactions.
With a transparent approach to communicating confidence, there will be an increased need for highly accurate AVMs to be used directly by consumers, instead of today’s two-tiered approach. Valuation accuracy can be the difference between a frustrated, discouraged homebuyer and a well-informed one. The combination of cloud-computing power, more available property data and modern technologies such as machine learning make it possible for AVM providers to increase accuracy while providing confidence in the value prediction.
AVMs are being increasingly used in home equity lending. With increased accuracy, instant results and lower costs compared to an appraisal, AVMs are especially suited for underwriting these loans.
When lenders market to consumers using a highly accurate AVM, they present consumers with realistic expectations. When an AVM is sufficient to satisfy underwriting requirements — typically on smaller loan sizes — it creates a more streamlined lending process and leads to better borrower outcomes.
For conforming mortgages, AVMs are not currently accepted as a replacement for an appraisal, but they can be used by underwriters in tandem with an appraisal to verify a home’s value and flag for overvaluation or undervaluation. AVMs that harness large amounts of data can perform more calculations than a human in a matter of seconds, and they provide an objective approach to value.
Historically, AVMs have been blind to a property’s current condition, which is why pairing them with a physical inspection has been key for using AVMs in lending decisions. Recent advances by innovative AVM providers and AI photo technologies have evolved the approach to include property condition as an input to the model, which produces a more accurate result. As AVMs become more accurate over time, underwriters will be able to rely on them further, and they will be used to determine whether an appraisal is required for the level of risk tied to a specific loan.
Accurate AVMs can also change the way mortgage servicers interact with borrowers. AVMs help to determine when a homeowner can remove their mortgage insurance, assuring that borrowers aren’t paying for it longer than necessary. Removing the insurance requirement can help a borrower reduce their monthly payment and better understand their current equity position.
The low cost of AVMs also means that the values of properties in a servicing portfolio can be updated more frequently. This provides better tools that enable servicers to deliver the right options to current borrowers and inform them of additional opportunities to make use of their equity.
Close to 90% of mortgage holders have interest rates lower than 6%. This has created a lock-in effect where homeowners are prone to focus on improving their current property using available equity rather than moving to a different home. Servicers that use accurate AVMs can play a big part in empowering homeowners to understand all of their options and make good decisions.
It’s understandable that some would see benefits to providing consumers with mediocre information all of the time, rather than great information some of the time. The recent evolution in AVM innovation means that there is no longer a need to compromise.
Machine learning, data availability and low-cost, massive computing power provide the ability to move past today’s two-tier system and focus on giving consumers direct access to underwriter-quality AVM values. This is an exciting development for homeowners and prospective buyers alike. Whether you’re a lender, originator, underwriter, servicer, investor or consumer, it’s OK to raise the expectation of accuracy rather than deal with the norm of missed expectations. And that is good news for everyone. ●