Rob Goldstein, chief operating officer and global head of BlackRock Solutions, and Jody Kochansky, head of the product group for Aladdin, the firm’s portfolio management system, can’t describe the future of data science at the $5.1 trillion money manager without comparing their ideas to the way things worked in the past.
Analyzing data is not new to the two executives. During a discussion Kochansky and Goldstein constantly interrupt each other in a lively back-and-forth that touches on robots, artificial intelligence, and machine learning. Both men started their careers at BlackRock a little more than two decades ago working on the iconic Green Package, the suite of risk analytics reports that took its name from the color of the only copy paper available in the building one night shortly after the firm’s founding in 1988. “Big data is not about breakthroughs with the math or methods,” says Goldstein, hired by Kochansky in 1994 as an analyst straight out of college. “It’s the data now available and the computing power to analyze that data. If you were to put us in a time machine and send us back 20 years, we would tell you that what you’re talking about being able to now do with data was not possible.”
Kochansky, who oversees more than 1,500 developers for Aladdin, BlackRock’s central nervous system, compares now with then in the mortgage market. In its early days the firm predicted how quickly homeowners would prepay their loans — an essential question for a money manager founded on its expertise in complex mortgage securities — based on information like the average credit score of borrowers in a bundle of loans. “Now Fannie and Freddie are giving you data that was unimaginable two decades ago about every mortgage,” says Kochansky. He stresses, though, that it’s not only about the breadth and volume of data: It’s about the science of models, including measuring their accuracy, fitting them within the current economy, and analyzing interrelated data sets. “You really need big data to wrestle this to the ground,” he adds. Kochansky’s 24 years at the firm have included a number of stints working closely with Goldstein, one of them building the Aladdin business from 1998 to 2007 and another rebuilding the platform after the 2009 acquisition of Barclays Global Investors.
Although Goldstein and Kochansky don’t say it out loud, their work finding new breakthroughs in the mature business of investing is essential to maintaining BlackRock’s premier spot in the industry. Other big money managers and hedge funds are gearing up data science efforts, but BlackRock’s edge is in Aladdin, which is a shared repository for all of the firm’s insights and ideas.
Of course the money management industry has always relied on patterns to pick stocks. But now that individuals and businesses have generated billions of pieces of data — 90 percent of all the data in the world has been created in the past three years — investors are frantically sifting through it to discover new leading indicators of the movement of stock prices. “We want to find the data sources that are not well known but are in fact predictive of future returns,” Kochansky says. Social media activity, satellite images of big retailers’ parking lots in developing markets, and company research reports are just a few examples of fertile sources of information. BlackRock’s job is to structure that unstructured data into information that can be played with by data scientists — by, say, creating an algorithm that counts cars in satellite images at certain dates and times. Data scientists can then analyze that information using typical statistical methods — all within Aladdin.
Goldstein cites a college friend from Binghamton University who became an equity research analyst and once spent three months, from November to January, counting cars, customers, and inventory at malls. In contrast, he says, Raffaele Savi — co-head of investments in BlackRock’s Scientific Active Equity group — told clients at a recent investor day that he reads every report every company analyst has ever produced. “Of course it’s not him. It’s the computer,” Goldstein stresses. And as BlackRock’s computers suck in all these analysts’ reports, regulatory filings, and other materials, they’re attempting to find other clues left behind. For example, they’re sorting through the language in transcripts of earnings calls, looking for changes in tone and counting positive and negative words. Naturally, once BlackRock is successful in decoding these reports, management at those companies will catch on and attempt to change the signals they’re inadvertently sending. Although clearly a more prosaic task, such parsing of clues from regulatory filings brings to mind Alan Turing and his team of mathematicians who cracked Nazi codes during World War II — which, once cracked, were used sparingly enough that Germany never caught on.
Data science is also part of BlackRock’s effort to make the sophisticated analytics and portfolio construction capabilities of Aladdin available not only to its historical base of institutional investors but to retail clients through their financial advisers. The firm offers Aladdin for Wealth and FutureAdvisor, a digital advice platform it acquired last year and provides through third parties like RBC Wealth Management. Goldstein says regulatory changes such as the Department of Labor’s fiduciary rule — which mandates that advisers for retirement accounts act in the best interests of their clients — and similar global requirements highlight the benefits of Aladdin capabilities like portfolio construction and transparency tools in the retail world. Big-data techniques allow BlackRock to scale Aladdin, offering its features cost-effectively to millions of retail accounts and arming its wholesalers, who focus on advisers, with relevant client information, model portfolios, and trade ideas.
Operations is another area rife with opportunities for data scientists. Aladdin carries out 250,000 trades a day and runs billions of economic forecasts and scenarios each night. The firm is using machine learning, as an example, to detect complex patterns in its huge amounts of data on trade activity in order to determine transactions most likely to fail.
Kochansky recounts another story from early in his career to explain one more application of machine learning: To find anomalies, he would put that day’s Green Package next to the previous day’s and manually compare the numbers. If a risk model had spit out a questionable figure, it meant tracking down the possible error — for instance, a bad price — and then rerunning the process. Although BlackRock has automated these activities, people are still essential to Aladdin’s exception-based work flow, where the computer identifies any problems that need human intervention.
BlackRock’s next goal is to teach the computer to fix the problem as well as identify it. The firm is building robots, or applications, to make judgments now decided by humans. “Every single time we detect an anomaly today and a human takes a corrective action, we are creating a giant data set so the robot can learn what actions are required,” says Kochansky.
Humans aren’t disappearing anytime soon. BlackRock, which has long proclaimed that it doesn’t solely rely on computer models, says this ethos will remain despite advancements in automation. “If you look at the research, it tells you that the ‘human plus computer’ will always lead to a better result than computer alone or a human alone,” Goldstein says.
BlackRock is well aware of the competition. Signals that seem promising today will stop working as others stumble onto them. If enough analysts can pick up on useful indicators based on the number of positive or negative words in an earnings transcript or whether the CEO and CFO were consistently upbeat, there’s no doubt companies will start changing their behavior. “Any edge has a half-life that is generally pretty short,” says Kochansky. “It really has always been an arms race.” Yep, everyone wants an Enigma machine.
Get more on asset management.
