R&I 2017 Meeting Report: Does Real Estate have the Data and Analytics Needed for a New Era?
(for a print version, click here)
“Grinding away” may be most people’s first instinct, but it doesn’t always get the outcomes desired – especially in a time of change and uncertainty. Real estate data today is not telling investors everything they need to know – and the ability to predict future outcomes is becoming more and more challenged by an era of surprises such as Brexit, the recent U.S. presidential election, technology, market and capital changes.
Change is a constant, but this may be a time of heightened transformation. A lot of change seems to have historically centered around the century mark – so that the 18th century seems remarkably different from the 19th, the 20th, and now the 21st. The most dramatic transformation doesn’t seem to take place until the second decade or so of a new century. For example, the 20th century didn’t really get going until the 1910’s & 20’s when mass adoption of the automobile, unprecedented migration to U.S. cities, electricity, new media, exploding global markets and the first of two world words forever marked the 20th century as apart from the 19th. How reliable would a real estate market forecast made in 1914 about 1920 have been?
Within the first 17 years of the 21st century, there is mass adoption of the Internet, migration into U.S. cities, alternative energy, new media, exploding global markets and global shifts of political and economic power. Is the real 21st century about to start?
Do real estate investment management firms have the data and analytics they need to thrive in this fast changing world?
Can data mining and machine learning improve forecasting?
Recently there has been a significant uptick in the amount of discussions around Artificial Intelligence. New products have hit the consumer market such as Amazon Echo, Google Home, and IBM’s Watson that show real promise for a future in which computers assist humans in an intuitive and thoughtful manner.
Douglas Downey, Associate Professor of Computer Science at Northwestern University, introduced to the group the concept of “deep learning” that is the foundation of much of today’s AI research. “This has been a major disrupting force in the field of AI recently and has shown wildly better performance on key tasks.”
According to Downey the basic theory behind AI has been: “Present the specific data in a format the machine can understand and then have it solve for a probable outcome given the current information.” But that is changing. Today, computers can be fed raw data in massive untranslated sets and, through deep learning, can arrive at outcomes and predictions. “The core methods in deep learning are not new, they’re decades old.” But with today’s data ands computational power deep learners, which are loosely inspired by the architecture of the human brain, can perform some tasks that were once the province of humans, such as recognizing which objects are present in a given image. By programming machines with this process, “we can now show them an image of two boats tethered to a dock and have them tell us what items the image contains. We can even go so far as to have the machine tell us in natural English.” So, the output would look like: Two boats at a dock on a lake. “Ten years ago, this was considered to be impossible.”
So, what does this mean for commercial real estate? If machines can identify and predict outcomes from unrefined data sets and even images, they could possibly replace the job of analysts or even data scientists. Computers will be much less likely to enter biased data or potentially forget to factor in a particular condition. Computers eliminate human error, and in the unending hunt for certainty in a world of uncertainty, replacing humans could become a compelling idea.
What is the right staff for research in a new era?
Until machines take over everyone’s job, investment firms are left with the task of finding and developing the most capable people around. Even more challenging, the skills needed by research teams today are different than they have been for decades.
“Many people find it hard to hire and retain people in data and research,” commented Doug Herzbrun, Global Head of Research or CBRE Global Investors. “There is a trend towards hiring clones of ourselves in the CRE world,” and the same kind of person may not be the best choice for a research or data science position. There is a human nature element to much decision making that includes prior experience and background.
“A proper data scientist actually doesn’t want any background because they want to take the data and glean from it what they deem to be important,” claimed Chris Happ, COO and founder of Goby. “If a person gets burned by a deal they’ll instinctively never want to touch anything like it again even though the data might tell them they should.” So, it’s important to hire someone that has the right mindset. But this raises another question: Should that person be hired outright or should they be outsourced? “It’s a question of competitive advantage, do you have people on your team that give you an edge? Or can anyone buy that expertise.”
“In my experience, it comes down to price vs. value,” added Jack Kern, Director of Research and Publications for Yardi Systems, “if you are going to be doing the same sort of tasks consistently it’s much more valuable to hire a person and train them in the nuances of your business.” They will then be able to perform better, faster, and with less instruction. “But if it’s just once in a while, find a local university and let them do it for you. The only caveat is that you have to give them extremely specific parameters to work with otherwise you might not get the information you are looking for.”
Finding someone to fill a data scientist role can be difficult. The ultimate decision makers in an investment management company are very likely to push back on hiring the right type of person. The best candidate may not be well versed in Real Estate at all which usually gets a veto from those who have been entrenched in it for decades. “We’ve got some people that know nothing about real estate but they really know how to structure the data and the validity of fields and how that all works together,” said Andrew Fein, Principal at Heidrick & Struggles. “If you bring that person in house and can train them in CRE you’ll get more out of a single analyst.”
Tiffany Gherlone, Head of Research and Strategy at UBS Realty Advisors, agreed. “We hired a data scientist with a background in economics who also had taken a couple data science courses and we were the only department willing to consider him. Now the guidance I’ve been getting is: ‘Just hire more people like that guy.’” At UBS, they didn’t have the luxury of hiring someone who only looked at data. “We needed someone who could also speak at committee meetings as well but he didn’t have a background in real estate.” That part he was able to learn on the job.
“You have to be willing to train the people you hire,” added Armel Traore Dit Nignan, Research Manager for Principal Financial Group. “Before starting at Principal I knew nothing about real estate. I learned it better from the models than I ever did listening to people talk about vacancy rates and the like. We as an industry would gain a lot by mentoring people rather than just saying ‘you’re a data scientist, here is our data, what can you come up with?’ If a company is willing to challenge their employees then you can do great things together, but if they are not, people will walk.”
What is the full cost and risk of data vendor agreements?
With respect to the price paid for the vendor provided data, “it’s not just the fee that you’re being charged,” warned Toni Fisher, Assistant General Counsel for USAA, “it’s also the risk you are incurring when you sign a license agreement.” Too often, for expediency sake, organizations will sign license agreements without serious consideration or meaningful negotiation of terms, and this may be a mistake. Too often, there are terms in these contracts that may prove onerous in the future, and don’t always pass the reasonableness test. “In looking through our stack of agreements I found that in many cases no changes were made to the contracts before signing even though they can always be changed.”
Since the data publishers create the contracts, they are by definition, “a little biased in their favor.” There are often items in the agreements that will increase costs and risk. “The basic objectives of the data providers are to make money and reduce risk, so naturally their contracts will tend to be written in such a way that pushes money their way, and risk yours.
For example, many agreements will state “You cannot use our data in an offering memorandum or an SEC filing.” This is put in place so that they aren’t liable if someone invests based on data that later turns out to be inaccurate, or if a company defends their decisions and practices based on the same set. Many agreements also contain these two opposing clauses, one: “If you use our data you have to cite us and give us attribution.” Two: “We don’t make any representations or warranties as to the data.” They’re not even telling you that its true. “So, the providers are taking mandatory credit for data they won’t even back up. This is patently unfair.”
However, in many cases the data providers are hamstrung as well. According to Jack Kern, “We realize that we have a relationship with the subscribers and we want them to use the information in ways that serve their best interests, but when you file a PMM, PPM, anything with the SEC, or any other public release you are sharing our information out of context and in a manner that we didn’t intend.” The data providers are not registered investment advisors so, “when that happens we immediately have a violation with the SEC for making statements that aren’t within the guidelines that they’ve published.” The data companies aren’t hiding behind legal documents because they aren’t confident in their product, they are doing it to avoid being dragged into lawsuits.
“As a lawyer, I like things to be clear. I don’t want to take my chances if we go into court,” says Fisher. “Make sure that all clauses match what you actually do. Then, go through all the restrictions and make sure that there is nothing that conflicts with what you do.”
The challenges of real estate research in a time of market transformation are considerable, but they are not impossible. This not a time for grinding away, but rather for intelligent and thoughtful work with new tools, new skills, and open eyes.← A&D 2017 Meeting Report: Investing in a New Era of Change Spring EO 2017 Meeting Report: How Can We Succeed in this Kind of Market? →