Zillow burst onto the scene about 14 years ago with a way for both buyers and sellers to instantly get an estimate on a home’s value. Their computer model, drawing heavily on nearby sales and other data, caught the public’s imagination.
These automated valuation models, or AVMs, have been around long before Zillow and many new ones offer a wide range of practical uses. Mortgage originators should learn how these tools can help generate quality leads while making their business more efficient.
Zillow called their AVM a “Zestimate,” a name that received its share of ribbing but also was effective in catching the attention of consumers. Eventually, Zillow used the model to draw hundreds of millions of prospective homebuyer and seller eyeballs to the company’s website. This generated myriad leads for mortgage originators and other real estate professionals.
The Zestimate was the first widely successful consumer-facing version of this application. AVMs calculate the estimated market value of a property at a given point in time based on a mathematical model that typically relies on recent nearby sales of similar homes, among other factors.
From the outset, the Zestimate received plenty of flack for its often inaccurate calculation of property value. Even now, Zillow acknowledges that the Zestimate is just “a starting point in determining a home’s value and is not an official appraisal.”
Times have changed. Complex AVMs using increasingly sophisticated big-data tools such as machine learning and artificial intelligence are improving the accuracy of automated property valuations — empowering a wide variety of businesses to build innovative products that are disrupting different niches of the real estate and mortgage industries.
Mortgage originators should be aware of how both the development and application of AVMs has evolved. Here are some of the most innovative AVM applications that go far beyond just answering the question, “What’s your home worth?”
Value and leads
In the world of property development, one of the most innovative applications of property valuation comes from CityBldr, which answers the question, “What could your home be worth if developed at its highest and best use?” CityBldr’s AVM applies machine learning to examine public-record property data, as well as data from other sources, to determine the best use of any parcel of land and then estimate what the property would be worth given its highest and best use.
For instance, CityBldr CEO Bryan Copley said, take a look at a property in Capitol Hill, a hot Seattle neighborhood. The owner of the property learned through CityBldr that the property was worth $1 million more if it were to be redeveloped rather than sold outright. “That’s life-changing information that property owners should have before making the decision to sell,” Copley said.
Predictive-analytics companies such as Audantic and Likely.AI leverage an AVM to predict inventory hitting the market in the near future. This provides high-quality leads for real estate agents, mortgage brokers and real estate investors.
“In essence, we predict the likelihood of an owner’s desire to sell within 90 days. Once a property reaches a certain confidence threshold in our prediction models, it becomes a ‘Likely Seller,’ which are highly valued by lenders, investors and real estate professionals,” said Likely.AI CEO Brad McDaniel.
Audantic also scores properties, but specifically for a real estate investor audience, meaning its scoring is based on how likely a home is to sell at a discount below its after-repair property value.
“A lot of investors have used intuition to determine how to do their marketing, and we’re able to apply predictive analytics to the data and get much higher results,” said Audantic chief technology officer Franklin Sarkett, who worked as a data scientist at Facebook before joining Audantic. “We can eliminate 80 percent of the population, and we can double or triple their results.”
Cash and loans
So-called “iBuyer” companies such as Opendoor and Offerpad utilize AVMs to help determine how much they should offer to prospective home sellers visiting their websites. After entering their home address, owners receive a nearly instant cash offer that also comes with the convenience of not having to stage and show a home, along with the ability to select a prescheduled move-out date.
Offerpad combines AVMs to analyze market trends and ultimately make offers on individual homes, according to Dan Mayes, its chief real estate technology officer. “Offerpad receives data from many sources and, most importantly, knows how to use (the information) when buying and selling homes,” he said, noting that the company uses deep neighborhood data to help set a benchmark for offers, as well as providing property characteristics to better help determine the needs of its buyers.
Another way that AVMs are being used in the mortgage industry is to help people quickly apply for loans. As a younger generation enters prime homebuying years, the need for a low-friction, digital-mortgage experience is becoming more critical for lenders who want to compete.
Online mortgage marketplace LendingTree facilitates such an experience at the top of the loan-origination funnel, using an AVM and other property data to streamline the online application process — and also validate property value and other information about a home.
LendingTree’s streamlined loan-application process attracts quality mortgage leads even in market cycles when lenders are struggling to drum up business through more traditional lead-generation methods. “Our business model can operate efficiently in a rising-rate environment when lenders have difficulty in bringing in organic volume themselves,” said LendingTree chief economist Tendayi Kapfidze. “That’s when they’ll come to LendingTree to fill their pipeline and we can market into that demand, or pull back if the unit economics don’t make financial sense.”
Here’s an innovative AVM application that is more removed from the world of real estate — at least at first blush. Windfall Data utilizes data to help it identify, target and engage affluent consumers for nonprofits. Estimated property value and home equity represents a major source of wealth for many Americans.
“Windfall models out the net worth of affluent consumers with more than $1 million in wealth,” said Windfall Data CEO Arup Banerjee. “In addition to net worth, we also provide additional household-level data so that our customers can analyze their existing customer base and acquire net new customers through omnichannel marketing.”
Risk and insurance
Property values are being put at an increasing risk from natural disasters such as hurricanes, floods and wildfires, given the twin threats of climate change and a growing number of newly built homes that border wildlands. Risk Management Solutions (RMS) helps financial institutions and public agencies understand, quantify and manage that risk using vast troves of data.
RMS models — built on millions of data points — are used by insurers, reinsurers and other organizations to analyze the probability of economic loss from catastrophic events. RMS models also are able to quickly quantify estimated insured losses from an actual natural-disaster event.
Another innovative application of AVMs in the insurance space is em-ployed by Cape Analytics, which automates the insurance-underwriting process with the use of data. Cape Analytics delivers the AVM along with other key property attributes that may impact property value to insurers via a simple interface, allowing those insurers to better identify risk and streamline the underwriting process, while also improving the agent and customer experience.
“Whether during underwriting, rating or renewing, information about properties is critical for carriers to understand their exposure to growing risks from climate change-driven impacts,” said Kevin Van Leer, product manager at Cape Analytics. “At Cape Analytics, we generate new forms of actionable, risk-relevant data, using computer vision and machine learning to analyze aerial imagery of properties across the U.S.”
• • •
AVMs have come a long way in the past 14 years thanks to improvements in the modeling of property values, along with innovative applications like the ones outlined above. Expect to see exponential progress on both fronts over the next 14 years as the mortgage industry continues to move closer to a world of instant and accurate property valuations.
