By Sondra Campanelli
April 22, 2018
Printable version of this article
Data-driven real estate lending has the potential to overtake traditional lending practices in the middle market space thanks to advances in data science, transaction speed, and an increase in borrowers’ comfort levels with the lenders, according to Tingting Zhang, founder and president of Terracotta Group, a non-bank lender. “We think there’s something in the CRE credit underwriting universe that can be fundamentally changed through enhancement of data and a data-driven approach,” she said. “I don’t think it’ll be very long before we say, ‘Remember how we used to make credit decisions in [a non-data-focused] way?’”
One aspect of the lending process that has been revolutionized by data is comparable sales. “When we first started [in 2004], we would go to a retail site to count the cars,” Zhang said. Now, the group gathers data from companies that can track foot traffic in retail centers remotely.
The dimensions for measuring comps have increased significantly along with new trackable data points, which is leading lenders to believe that data could overtake one of the central tenants of real estate lending—touching and feeling the underlying real estate. “We still [go out and look at the properties in person], but I do think that data has become very powerful” she said. “There are so many ways of measuring foot traffic nowadays.”
It’s taken time for data-focused lending to grab hold in the CRE world. One reason is that it’s hard to get data scientists and traditional lenders in the same room, speaking the same language. But Zhang also points to CRE as a cottage industry with lots of variables and real-world complexity. “It takes an understanding of how the CRE assets behave, but it also takes time as well.”
Advances in data science, which have made it more likely that an analyst can work with a “messy” data set that’s missing key data points, are helping to move that process along.
One common omission Zhang pointed to is a lack of recorded cap rates during the Great Recession due to a paucity of sales in many markets. It can also be difficult to find sales comps for larger transactions in small submarkets where a 50,000-square-foot property trades infrequently. “We have various data techniques we can use to fill in the blanks and create understandings of lease rates, so it’s not a battle that’s lost,” Zhang explained.
Due to its focus on data, the group, which lends to the middle-market space, has a very clear definition of the risks it’s willing to take. Primarily, it’s looking for a strong sponsor that has been executing the same business plan within the same geography through multiple market downturns.
But more importantly, the firm runs 55 unique investment scenarios within a 200-variable regression that not only factor in sponsor risk, but also property risk and submarket risk. “We look at the interaction of stress scenarios and highlight compound risk,” Zhang said. “Our one hard criterion is that no investment [we undertake] will cause us to lose our principal.”
Terracotta does use widely accepted measurements of risk, like loan-to-value or debt service coverage ratios, but doesn’t think they paint the full picture. “LTV is one measurement of risk, but there are deficiencies in it,” Zhang said. “Like any ratio, there are some things it measures and some things it misses.”
As an example, she points to two hypothetical scenarios: lending on a Houston office building at 65% LTV or a Santa Monica multifamily property at 80% LTV. High incomes and strong demand drivers for multifamily properties in Santa Monica might make that a safer lending opportunity than the property in Houston, which is a more volatile market with property prices higher than pre-recession levels.
“The reason why we’re using a 200 variable-driven analysis is that we’re stabbing at the credit at every possible angle to see interactions the human mind might miss.”