The recent financial crisis took some bank executives by surprise; they didn't fully understand how their banks were making money and how the crisis was impacting their business, says Aite Group research director Gwenn Bezard. "Part of fixing that is to have better data, better access to data and better information about the business," he says. "Across institutions, we see a drive toward better access to data to better understand their business, which means more investment in analytics."
Even community banks, whose CEOs may be neighbors with their customers, are gaining an appetite for analytics tools that used to be reserved for the large banks, Bezard notes. And while business intelligence software used to be relegated to a few specific areas of the bank -- risk management, customer analytics, financial performance dashboards -- it's now creeping into new territory. Today banks are using analytics to predict bankruptcies, forecast long-term risks and even catch bank robbers.
Modern business intelligence (BI) software enables bankers to sift through raw information and extract nuggets of knowledge that can be displayed in reports and dashboards. Beyond examining just the current state of affairs, banks are embracing predictive and, most recently, prescriptive analytics -- that is, "If I can predict the future, what do I do about it?" explains John McHugh, managing director in Accenture's global financial services group.
1. Predicting Unexpected Bankruptcies
In the recent economic downturn, some bankers realized that a FICO score is not enough to predict consumer defaults. "It's nice and good, but it's not enough," Aite's Bezard comments. "The ability of the score to predict has been weakened by the recession. How do you get a better handle on the risk of default? How do you make a better lending decision?"
According to Bezard, Kabbage, a start-up founded by a former banker that plans to lend to merchants that sell through eBay, is pulling account aggregation data and transaction data from merchants' eBay and PayPal accounts to aid in lending decisions. "They're getting access to data banks don't have because [the banks] don't ask for it," Bezard says. "That gives [Kabbage] live, direct insight into the cash flow of the business."
But banks finally are heading down this path, too. In a survey of Bank Systems & Technology readers, 55 percent said they use risk analysis and assessment tools to gain insight into customer behavior. Some of these banks are combining sophisticated credit scoring with risk analytics to determine the future solvency of their borrowers, according to Bill Spinard, a senior executive in Accenture's global risk management practice.
"Banks are beginning to see that with the right analytics, new statistical techniques and new databases, they can obtain marginal improvements on their credit scores," Spinard says. This allows them to adjust their reactions to the scores, he explains, whether by expanding or shrinking a credit line, by adjusting pricing, by calling a customer more often or by not calling at all.
"I may miss a payment, but the analytics might suggest that I have a habit of missing a payment once a year," Spinard relates. "This kind of analytics has the potential to significantly improve the risk management practices of banks, allowing them to make finer distinctions as to the kinds of customers they want to keep and the pricing they want to offer."
Analytics are even helping banks spot signs of unexpected bankruptcy. In the past, banks could easily predict an impending bankruptcy -- it would follow three to five months of late payments. In the current economic climate, however, a customer who has been making credit card and loan payments every month suddenly could declare bankruptcy. As a result, banks are applying predictive analytics to credit card databases, mortgage data, deposit data and even social networks to find more subtle clues. A bank might see a customer's credit card spending habits shift from Brooks Brothers to Walmart, for instance, Spinard says.
"Does that indicate a change in his potential solvency? It might," he observes. "If I start going to the casinos a lot, does that suggest that perhaps on my mortgage, the bank should be a little bit careful?"
2. Planning for the Far Future
According to the recent BS&T reader survey, 36 percent of respondents use forecasting analytics. A case in point is Chicago-based Alliant Credit Union (the eighth largest credit union in the U.S., with $6.3 billion in assets), which is using predictive risk analytics to plan for the future. Mona Leung, the credit union's CFO, says she likes to look at risk both from a long-term (five to seven years) and medium-term (three to five years) perspective.
"Analytics are not just for looking at where I am," Leung says. "They play in the creative element of, 'Where do I want to be?' "
Leung took all Alliant employees through a two-year process of identifying risks and creating standard definitions and processes. "Our CEO has always wanted a transparent, standardized way to understand risk across the organization; he thinks there are risks we may not be identifying," she relates.
Using analytics, the credit union began putting its balance sheet through predictive scenarios (that it purchased from consultancy Decision Strategies International) that envision the next five years in terms of the economy, the political environment, consumer behavior and loan demand, according to Leung. "You want to understand how bad and how good it can be, and what you are willing to do when analytics suggest that a risk is coming," she says. "The real creative conversation is what mitigations you can use when you predict the future."
The credit union's analytics, SAS OpRisk Management, are like "Excel on steroids," Leung reports, adding, "Doing analytics without a tool like this is like manually weaving cloth." In addition to minimizing business losses, she says, she hopes the software also will help the financial institution align with the Basel framework, quantifying true operational costs and risks, and streamline operations.
3. Catching Bank Robbers
While employing analytics to predict bankruptcies and conduct long-term planning are innovative uses for analytics, Union Savings Bank ($2.5 billion in assets) is using analytics in perhaps an even more cutting-edge way. The Danbury, Conn.-based bank uses facial recognition technology, video analytics and customer data analytics to instantly identify robbers.
"Our video surveillance system used to work like a VCR," recalls Bill McNamara, Union Savings Bank's SVP, technology and security. "If you wanted to find something, you had to toggle your way around."
The bank's new surveillance and facial recognition technology from San Francisco-based 3VR Security, however, detects movement and creates short, searchable video clips, he explains. The system counts the number of pixels between the eyes and different facial points to find a match, McNamara relates. Once a face is identified, it can be searched for across all locations, and the system will produce all relevant video.
The facial recognition has been surprisingly accurate, according to McNamara. For instance, a branch manager had long blonde hair one day and short black hair the next; the system identified her with no problem, he says.
The bank worked with its core processor, Fiserv, to write an interface to the surveillance system, allowing video clips at the teller line to be indexed to transactions, McNamara adds. A staff member or investigator can type in an account number and view video clips of all transactions that took place on that account.
"That's helpful when we're doing a fraud case," McNamara says. "If we identify an account that's being defrauded, we search that account, we see who's been transacting on it, then we can search those faces to see if they're transacting on any other accounts. It takes guesswork out of an investigation and reduces the amount of time needed." He estimates that incidents can be researched in about 10 percent of the time required previously.
Although video can be storage-intensive, McNamara acknowledges, the 3VR system uses strong compression, minimizing the bandwidth required to stream video and the storage needed to archive it, he reports. Each location has its own storage device; a central server queries all the devices.
In addition, the surveillance system can alert staff every time somebody walks toward a vault or other high-security area. And the bank also has installed license plate recognition software and cameras trained on its drive-up ATMs so it can capture the license plate of each user to help catch any perpetrator of ATM fraud.
In future uses of the system, McNamara says, he would like to identify the movement of branch traffic to modify branch staffing models. He also would like to use the facial recognition to trigger an alert when key people enter a branch.