A computer algorithm beat three veteran brokers when it came to picking the listings someone searching for a home preferred, according to a study conducted last month. Researcher Creed Smith pitted his algorithm, or bot, against three human agents to see who could do the best job of picking home listings for a client.
Real estate editor John Rebchook, served as the test subject. Each day for three days, he picked a home listing that appealed to him. In response, the bot offered up three listings and the brokers each picked one they thought he would most like.
Rebchook then ranked the recommendations from one to six. He didn’t know who was offering what, but on all three days he ranked the bot’s picks No. 1. The bot managed to get two of its picks in the top three each day.
“I couldn’t figure out any pattern on why a robot was picking something versus a broker,” Rebchook said. “If a robot can beat humans at chess and Go, they can beat us at picking houses.”
In March, a Google computer managed to beat the grand master of Go, a 3,000-year-old Chinese game considered more complex than chess. So maybe selling a home isn’t as complex as Go, but there is strategy involved in matching a client to the perfect house that Smith thinks could be improved.
Many brokers start out by asking buyers what they want in a home, Smith said. But after years in the business, he learned buyers often don’t understand what they really desire until they start looking.
“Actually, it’s amazing how often a buyer ends up buying something very much different from what they tell you and what you show them for six months,” he said. “This allows buyers to play and learn what they really like.”
By letting buyers pick out favorite listings and then making recommendations based on that, the bot in some ways mimics what Netflix and Amazon Video do when they recommend movies based on movies already watched. Smith’s algorithm starts with the standard criteria that home search engines use, such as neighborhood, home price and number of rooms, but then goes deeper.
Home listings are known for being full of fluff and puff and wasted words. But a few select terms carry emotional weight and meaning for buyers, Smith said. The bot has a dictionary that categorizes those terms and links them to other listings.
“We work every aspect of the home down into a mathematical score and offer the list back to the buyer with the best homes near the top of the list,” Smith said.
Given that close to half of buyers now find the home they want to buy online, the bot’s advantage isn’t necessarily in replacing human brokers. But it can save brokers and their clients time slogging through listing after listing. Smith sees the bot and related website as incremental technologies brokers could adopt to become more efficient and competitive with online sites, such as Zillow, Trulia and Redfin.
Smith said the brokerage industry remains technophobic and hugely inefficient, with 1.3 million people employed to sell 5 million homes a year. “Brokers are in such denial that anything could change the way things are done. It could never happen to me,” he said.
Unless the brokerage industry develops technologies that bring value to clients and boost efficiency, it will be overtaken by outsiders, he warned.