By Evan SparksBanks are on the forefront of the growth of using “big data” to inform and even drive business decisions. And as computing power grows and the uses proliferate for the data that banks control, data science is rapidly moving to provide its users with prescriptive power and new tools to automate and advance business decisions.
For Cheryl Gurz, SVP for operations at the Bancorp in Wilmington, Del., “prescriptive analytics” is the fourth frontier in the evolution of data. “This is where tomorrow is going,” she says. This fourth phase follows basic reporting of data (such as through Bank Secrecy Act Suspicious Activity Reports), followed by descriptive analytics (digging into amassed data to answer specific questions) and predictive analytics, which Gurz notes involves “predicting the likely future outcome of events often leveraging structured and unstructured data from a variety of sources.”
The goal, she says, is for banks to answer the “what if” questions, constructing models of decisions that simultaneously address differing scenarios, provide insights that involve a vast array of relevant data and provide preferred solutions to thorny problems.
For example, Gurz recalled a bank that was exploring whether to offer insurance in its local market. A data scientist looked at all the ZIP codes in the bank’s footprint and incorporated a database of weather events. The answer: “Nope! We are making loans to areas that have high weather patterns with losses, and we don’t want to sell insurance because the payouts have a probability of being high.”
Banks have generally led in analytics, with 71 percent of banks using big data to improve operations, compared with 60 percent across all industries, according to an IBM survey. “Financial services are far ahead of the rest of the industry,” says IBM’s Michael Maxwell, who, like Gurz, was a featured speaker at the ABA/BAFT Global Payments Symposium in New York earlier this year. As banks continue to build on the promise of prescriptive analytics, they’re finding applications for these tools in three key areas: growing the bank’s business, gaining operational efficiencies and facilitating regulatory compliance.
Finding scarcer growth opportunities
“In this economic environment, how do banks obtain growth?” Gurz asks. “Growth isn’t easy anymore.” With margins squeezed and the best credits picked, banks are turning to prescriptive analytics to help them find every opportunity and make sure they don’t leave money on the table.
At Canada’s TD Cards—a division of the TD Bank Group, a global banking company based in Toronto with a large U.S. operation—managers were looking to grow their $9.5 billion (U.S.) credit card portfolio with 5 million active accounts, all without changing their risk appetite.
TD worked with the data and analytics company FICO to develop a solution. They created a sample customer and identified several “custom decision keys,” as TD’s Clifford King calls them, such as whether she maintains a stable balance and whether she has extra credit capacity. These questions were fed into the model, along with internal and external data like credit bureau reports. As King describes it at a conference in Washington earlier this year, the combination of custom keys and internal and external data yields the best customer-level decisions.
The result, King says: account authorizations rose 50 percent over two years while delinquencies stayed the same. The portfolio saw balances rise 7.76 percent while incremental utilization on the average account rose by 35 percent. “This means [the growth]was valuable,” King says. “We were getting a good rate of return on those decisions.”
Operational efficiency meets customer benefit
Through its Henderson, Nev.-based industrial loan company, Toyota Financial Services helps four million Americans buy and pay for their cars via retail lending, leasing and dealer financing. The company saw a default rate of 1.27 percent in its 2016 fiscal year and actively tries to keep that number low, recognizing the uniquely challenging effects of losing a car on a person’s livelihood and family.
So Toyota embarked on a major effort to understand and adapt its collections practices using data. As Jim Bander, national manager for decision science at Toyota Financial Services, says: “You buy what you can collect. By being more effective in collections, we can go lower in the credit profile.”
Toyota partnered with FICO to develop a data optimization solution that would incorporate predictive models of repayment, default and loss and choose one optimal scenario for different customers. The strategy brought together credit bureau data with Toyota’s internal data, especially how different “collections treatments” had worked.
Toyota now classifies each account according to its optimal collections treatment, generating a decision tree for each account nightly that goes into the collector’s queue for the following morning. With customized, data-driven, prescribed treatments, the program in its first year helped 1,600 borrowers avoid repossession, while protecting 10,000 from reaching a stage of delinquency that would ding their credit. Meanwhile, the program helped Toyota grow its loan portfolio by 9 percent without hiring additional collections staff.
Meanwhile, the data on how collections treatments perform gets constantly fed back into the system, resulting in improvements over time. “The leading edge around using big data, analytics and cognitive capability is to continually evaluate ‘how am I doing’ in real time and suggest rule changes,” Maxwell comments.
“Working with delinquent customers to keep them in their cars while working out payment options has helped Toyota avoid millions of dollars in losses,” says Bander. “It’s a win for our customers, and a win for Toyota.”
Filling gaps to improve compliance
Prescriptive analytics can help close small gaps in regulatory compliance—such as those formed because of siloes within the bank. Gurz gives an example of an institution that was reimbursing customers over a credit card fee inappropriate under Regulation E but that failed to remediate similar fees on the demand deposit account linked to the card.
Another area where banks are using powerful analytics for compliance is the Basel III liquidity rules, under which banks need to be able to predict where and when it will get access to the funds counting as liquid assets. Most of the problem is that many core banking systems don’t keep a timestamp, which makes it hard to do intraday liquidity reporting, says Maxwell. “By using big data, we’re able to help banks put that together.”
Moving toward prescriptive analytics is a challenge for any institution. While newer decision software options allow non-specialist bank employees to construct models, use of advanced analytics still requires a grounding in complex subject matter. Selecting, preparing, refining and revisiting data and the outcomes are big, tough projects that call for team efforts cutting across the organization.
But for banks that brave the process, the result can be greater efficiencies, stronger product performance and better compliance. Just let the data speak for themselves.