No Such Thing as “Moneyball” in Commercial Real Estate
September 11, 2019
The theme of the 2011 movie Moneyball has fueled countless Big Data initiatives. This adaptation of Michael Lewis’s 2003 book by the same name chronicles the Oakland Athletics’ quest to assemble a competitive team on a limited budget. The team’s GM Billy Beane (played by Brad Pitt), puts his faith in this nerdy assistant Peter Brand (played by Jonah Hill) who uses statistical analysis to outflank other MLB teams that erroneously rely on the traditional biases of seasoned baseball scouts. In this way, Moneyball crystalizes the power of Big Data and serves as a parable of analytical innovation.
At first glance, the commercial real estate industry appears ripe for just this kind of innovation. After all, adoption of tech solutions for CRE has been notably slow, data is decentralized and real estate professionals still largely rely on versions of the same tools and methodologies in use 20 years ago to do their work. Not surprisingly, the common lament that CRE is “behind the times” has spawned a myriad of CREtech start-ups, most with crisp, alluring narratives about how market participants can use their proprietary “Big Data” solutions to create more value and make better buy/sell decisions – in other words, play Moneyball with their CRE portfolios. But these start-ups are mostly just preying on the collective lust for a silver bullet. A more honest look reveals that the application of Big Data principles in CRE has some massive limitations.
The “Big Data” approach works best when the underlying data pertains to discrete outcomes that can be easily captured and tabulated. It is much easier to gather insights from Big Data when an abundance of standardized data points exists. Baseball is full of this kind of data (e.g. balls, strikes, hits, walks, outs, runs, errors, etc.) as are sectors like logistics and retailing. In contrast, CRE has a great deal of subjective data pertaining to unique assets that are measured on a continuum. Factors like “property condition” and “location” are still largely in the eye of the beholder and there is no agreement on how to definitively score them. At the margins, large swaths of geographic cell phone or credit card data can be used to make inferences about a specific location. But such data points only reflect a small fraction of the aspects of that location. Moreover, the quality of that location is itself dynamic and subject to countless externalities, influences and tastes that morph over time.
Not Enough Data
CRE assets are unique and operate in small competitive micro-markets. As such, there is typically not enough data to enable meaningful statistical analysis. Whereas data scientists tell us that an excess of 30 standardized data points are required to deem a result statistically significant, a typical CRE asset would have data points from maybe just five germane sources. The CRE market has a disturbing tendency to solve for the dearth of data by importing larger quantities of data either from similar assets that operate in different markets or from dissimilar assets that can claim homogeneity on the grounds of their nearby proximity. Clearly, however this type of extrapolation is seriously flawed. In truth, there is no way to accurately control for myriad locational factors that influence a particular micro-market. Put differently, on a broader geographic level the market preferences in, say, Seattle are vastly different from those in Houston. To draw meaningful inferences from a mixture multi-city data is dodgy business.
The Behavioral Component
The application of Big Data principles in CRE is further complicated by the difficulty of predicting behavior and motivations among market participants. In industries that produce a significant amount of standardized data, it is possible to make useful inferences about behaviors that surround transactions. In contrast, the transactions in CRE are almost always unique situations where the complex motivations of the parties are unknown. Negotiations and decisions are not made by algorithm, but by human beings with all manner of behavioral anomalies often sitting face-to-face across the table from one another. It is this plethora of unique behavioral dimensions that further hinders the application of Big Data principles in CRE.
Advantages Come from “Little Data”
It is easy to understand why so many in the CRE industry have been following the siren call of the Big Data cure-all: by throwing lots of resources at Big Data initiatives a platform can claim it has a material advantage over its competitors. However, to date these initiatives have yielded few concrete results. Such claims amount to little more than marketing hype aimed at dazzling potential users.
The focus of CREtech on the Moneyball-like aspects of “Big Data” is misguided and is mostly driven by academics or tech experts outside the industry who have little experience in the day-to-day CRE marketplace. The most reliable CRE professionals, on the other hand, have an innate understanding that true gains come from the painstaking work of capturing, curating and analyzing anomalous data relevant to specific assets and strategies. This “Little Data” often does not fit into neat buckets; it requires significant industry experience and judgement to interpret. For this reason, the technology efforts that focus on accelerating the organization of CRE’s “Little Data” are the ones destined to make the greatest impact.