HOUSINGNEWS REPORT
HOUSING PRECOGS: BIG DATA PREDICTIONS BEYOND HEURISTIC HUNCHES
just based on Internet research,” said SmartZip’s Avi Gupta. “So, agents need to personally get in front of owners and build rapport over a period of time — which is hard to do for more than 300 to 400 owners. If an agent prospects with 400 owners that live next to each other, their chances of getting a listing in the next 12 months are the same as the organic turnover of that neighborhood, which averages 5 percent in the U.S. “Instead,” he continued, “predictive analytics affords us the opportunity to expand the aperture to about 2,000 homes, and then select the top 20 percent of homes with the highest likelihood to sell over the next year, i.e., the best 400 homes that are on average two to three times more likely to sell than average. This can double or triple the chances of winning listings.” Predicting Inventory While the bulk of all real estate inventory resides within local MLS systems, and while most properties
“With the most advanced version of AI, called deep learning, which is what we use, only 10 percent of the final prediction decision is determined by the data itself. That is because 90 percent of the predictive power comes from the extremely complicated interactions between the layers of neurons within the deep neural networks that we have created. We now live in a time where data availability is everywhere, but what you do with it is where the magic happens.”
BRAD MCDANIEL CO-FOUNDER & CEO, LIKELY.AI
neural networks that we have created. We now live in a time where data availability is everywhere, but what you do with it is where the magic happens.” Still the data is the foundation of the predictive analytics, and if the data is flawed the predictions based on that data will be flawed, notes HouseCanary’s Alex Villacorta. “When you have a lot of data and that data is ‘dirty,’ more data isn’t going to yield a better result,” he said. “You need to make sure that the data is
managed and refined in a way that serves your analysis.”
SmartZip’s Gupta pointed out that predictive analytics do not replace the important relationship-building aspect of the real estate business, but allow agents and brokers to better identify which relationships they should be building. “Real estate remains a relationship business, especially for home sellers, who typically choose agents based on trusted relationships and referrals, not
4
MAY 2018 | ATTOM DATA SOLUTIONS
Made with FlippingBook Online newsletter