My daughter Liz sensed three years ago that she might be pregnant based on the ads that popped up next to her Google searches, which rendered promos about diapers, car seats and other baby stuff.
What the heck? It was not accidental, Google deploys a scientific method called machine learning, which can predict all sorts of things based on your search history.
Machine learning is a marriage of data and algorithms. Algorithms are built on a model based on data inputs (my daughter’s searches) and then use that data to make predictions (Liz was pregnant). If the algorithm is sound, the machine gets smarter and smarter as more and more data is fed into it.
Machine learning is being used in driverless cars, robotics and Nest’s thermostats. Recommendation engines such as Netflix with movies and Amazon with books are powered by machine learning.
While machine learning is a big term, it wraps up several technical ideas such as data mining, artificial intelligence, pattern recognition and predictive analytics.
A simple example is how my iPhone predicts my next word when I type a text or when I fill out a form by starting with the letter “B” and the form is instantly populated with my name, address and related information.
Real estate is full of examples of machine learning. Have you completed a Zestimate or filled out the digital documents in a real estate purchase agreement lately? Most of these digital services remember your data, preferences and signature, and save you time when you fill out a doc again. And the company gets more and more information on you and can predict what you want, where you are going and what you might buy.
In the future, for real estate, an all-purpose algorithm will be at work, integrating the matrix of independent real estate steps currently handled by people, including offers, closing docs, mortgage approval, the move-in process and utility hookups. Each piece of this puzzle is being implemented now, but they will be integrated in the coming years and be instantly performed with little or no data input by the homebuyer and seller.
In a few years, neither the lender nor the agent will have their hands in the current mess we now call a closing.
The applications for real estate are endless.
Inman reporter Paul Hagey is working on a story on predictive home search on how many innovative companies are pushing the envelope using data and algorithms to help people spot the home and neighborhood that fits their needs.
Redfin offers “collaborative filtering” — surfacing both homes and neighborhoods to users that others with similar searches have liked. The Seattle company has a machine learning algorithm that collects listings that a consumer might like based on the search behavior on the site and the firm’s mobile apps, but it’s sent only to consumers who are currently working with a Redfin agent.
Realtor.com operator Move Inc. is the sole investor in Ylopo, a search-focused startup co-founded by a Move exec that uses machine learning and data mining “to curate and recommend homes that (users) will be most attracted to.”
Zillow is mum on its plans with predictive search but that is probably because it is working on something that it does not want its competitors to know about. And consider that Zillow with Trulia has more behavioral data on consumers than almost anyone in real estate.
Machine learning can also be a way to predict the future, such as where home prices are headed.
In their research study “Housing Value Forecasting Based on Machine Learning Methods,” researchers Jing Mu, Fang Wu and Aihua Zhang showed how machine learning could be used to more scientifically predict future swings in home values in the Boston area.
The advent of powerful machine learning applications has scared some scientific experts such as physicist Stephen Hawking, recently predicting a dark future for humanity. He wrote in The Independent, “One can imagine such technology outsmarting financial markets, out-inventing human researchers, outmanipulating human leaders, and developing weapons we cannot even understand. Whereas the short-term impact of artificial intelligence depends on who controls it, the long-term impact depends on whether it can be controlled at all.”
In this future of machine learning, real estate could be fundamentally altered. Smart machines are predicting buyer and seller behavior before agents have a moment to catch up; the closing and mortgage process will be collapsed with integrated digital documents; and maps and imaging technology make seeing a home an easy process. Even the negotiating process could be taken over by machines, without representation.
Like the Hawking scenario, that may sound like a dark outcome for the industry. But not one that is around the corner.
The best course is to act, not fret.
Smart real estate firms will become more like software companies enabling machines to perform much of their work and increasing their margins along the way as they improve productivity.
“Powering companies by software allows them to be responsive and data-driven and, hence, able to react to changes quickly,” wrote Erik Meijer and Vikram Kapoor in their recent article, “The Responsive Enterprise: Embracing the Hacker Way: Soon every company will be a software company,” in ACM Queue.
That is a tall order for real estate companies. One alternative is partnering with companies that are investing hundreds of millions of dollars in machine learning technology such as Zillow, Trulia, Move, Redfin and many others. Another strategy is acquiring smaller technology companies as Realogy did with ZipRealty.
Another example is Windermere’s investment in its own technology offshoot, Moxi Works, which launched a new software venture that merges a customer relationship management platform, email marketing, listing presentation tool, intranet and agent websites into one broker platform.
The worst alternative may be letting the machines own the future.
