What if you could know your home’s precise fair market value on any given day at any given second?
With artificial intelligence, you can, says Mike Shapiro, the managing director of the start-up Plunk, a real-time property valuation platform. Shapiro spoke last week about how, using machine learning, AI can take into account all factors affecting individual home value, broader market conditions and changes in buyer sentiment and purchasing power to arrive at a valuation that he says is more accurate than just comparing to recent sales data.
A Plunk app for homeowners is currently only serving the greater Seattle area, but as the company scales up, the intention is to price every dwelling in the United States, 144 million of them, with few exceptions. The company also has a version of Plunk geared toward professionals in the real estate, finance and insurance industries, including real estate investors.
Plunk will use AI to adjust pricing in real time based on market conditions, which Shapiro says will have a dramatic impact on moving sales along faster. Price readjustments will happen much more quickly, he says, increasing transaction volume, and more inventory will come on the market because of greater confidence in the accuracy of home valuation.
Shapiro also anticipates that AI will help banks and insurance providers make better decisions on loans and policies, and more quickly.
Shapiro, whose book “Read The Tape: Using Situational Awareness to Predict Business and Personal Probabilities,” was published this September, said his background is as a stock options trader and market maker in Chicago before entering the venture capital field and investing in a real estate company in Newport Beach, California, that he grew into the second-largest Sotheby’s in the world before selling it in 2018.
He noted that pricing houses is done by looking at comparable homes that have recently sold — called “comps.” In the Hamptons, value largely driven by location and square footage, he said, while in Manhattan, it can be super granular, down to the street, the apartment building, the condo complex or co-op.
“The problem with that is that housing is not isolated,” he said. “Housing is 100 percent impacted by geopolitical issues. It’s impacted by the price of lumber or the price of copper. It’s impacted by interest rates. It’s impacted by the stock market. It’s impacted by everything.”
As an asset class, real estate is highly integrated and correlated to almost everything else, he said.
“When you’re buying a house, and you’re out in the market, you’re instantly reactive to market mechanics,” he explained. “You’re instantly reactive to your stock portfolio or your business, where you work or your job or your aptitude. How much does it cost to build? You’re just reacting to the economic activity at large.”
Plunk works to identify correlated behaviors and correlations in other markets and then uses AI to create and maintain an accurate model of home value.
“If you can figure out what those correlated behaviors are, if you can figure out what those deltas are interactive with, then you can start getting a very highly accurate model with machine learning,” Shapiro said.
He explained that AI uses a multitude of models to be predictive and conclude what is most probable, quite like how the European weather forecasting method using 100 models to gauge the probability a hurricane will strike.
“It’s just like your brain,” he said of AI. “It’s absorbing as much information, as many models as possible, with machine learning to see if it’s making mistakes to make sure that you’re coming up with the formula for the correct answer. So that’s what we’re using in housing. I think it’s revolutionary, because I think that the entire system is flawed in just using comps because it’s already potentially 90 days old. It’s not reflective.”
Just last week, the National Association of Realtors reported that the number of pending homes sales nationally is the lowest it’s ever been since the organization began tracking in 2001.
In looking at the current state of the market, where homeowners are locked into their low interest rates and sales are restrained, Shapiro has a different take on why inventory is low: “There’s no inventory because people don’t feel like they can get the price that they want, and there’s low inventory because there’s a complete disconnect between the comp and the actual valuation,” he said.
He hypothesizes that if homeowners could see that prices are going down, they will be more likely to sell, just like when stock investors sell off a stock that was climbing but lost its momentum.
“You would see a significant increase in inventory if there was a price alignment on a national basis,” he said.
In a market like the Hamptons, where there are homes being used only two months a year, the motivation to sell would be stronger versus a primary home market, he noted.
Shapiro said sellers are just looking at the last comp but buyers are instantly reactive to their own portfolios, the geopolitical landscape, etc.
“It’s really a matter of, what is a buyer’s perception of the price of that house at that moment, simply put, based on all this other information,” he said.
And if the price is right, the home transacts, he said.
Plunk has hired photographers to shoot houses and has massive photographic archives to improve Plunk’s analysis, according to Shapiro.
It also uses data such as year built, home condition, finished square footage and number of bedrooms, bathrooms and stories.
“From a lingo point of view our data lake is massive,” Shapiro said.
When AI bridges the difference in price perception between buyer and seller, the time a home spends on the market before selling will be much shorter, he predicted, leading him to question whether agents’ commission rates will continue to be justified.
“Do you think the commissions would be as high as they are if people knew they were going to sell their houses in seconds and they didn’t have to spend all this time and money on marketing?” he wondered.
While “seconds” is hyperbole, he said a house could sell in a week rather than months.
