Fighting fraud with operational efficiency
Inefficiencies such as a bank’s disjointed data systems, manual processing of authentication for new account openings or outdated payment processes don’t just cause headaches for employees and customers alike — they also play directly into the hands of fraudsters.
Although these kinds of operational inefficiencies don’t rise to a level as egregious as leaving the vault unlocked, they can add up to serious losses for financial services organizations who are relentlessly buffeted by payment fraud, ATM skimming, counterfeiting, internal malfeasance and other scams.
Banks and credit unions can become more operationally efficient and, in turn, more resistant to fraud by first examining their upstream controls, says Mary Ann Miller, head of fraud strategy for Varo Money, a San Francisco–based direct bank. “Look at upstream controls such as your identity and authentications processes, and you can often stop fraud at an earlier stage of its life cycle during either log-in or during the payment process or the onboarding process.” Otherwise, she says, more routine work flows downstream and burdens operational functions.
Miller served on the Federal Reserve Board’s Secure Payments Task Force and pioneered the use of artificial intelligence (AI) and machine learning to help banks fight fraud when she was an executive with solutions provider NICE Actimize. Banks that fail to make the right investments in fraud control upstream will find that many of those processes revert to becoming manual controls, she says.
“The operations centers bear the brunt of those manual processes because of the lack of automation that could have been done upstream or because of the lack of efficiency of those processes,” Miller says.
Automated behavioral biometrics—which analyze patterns of human activity—can provide a quicker, more efficient and more foolproof method of digital authentication, according to Miller. “If there are early signals of an unusual pattern picked up early in the life cycle through behavioral biometrics, then you can stop a possibly fraudulent ACH or Zelle payment from being submitted in the first place.”
Miller offers this example: “A fraudster has an automated script in which they submit an unusually rapid series of payments. A behavioral biometrics system notes the velocity does not match the way the customer normally sends payments and interacts with the channel. That would trigger an alert in the operations center. That may require a follow-up investigation or review, but you instantly know what type of event may be occurring.”
The automated wizardry of AI and machine learning, however, does not spell the end of fraud, which is ever evolving. “We can’t lose [sight of] the importance of the human element when it comes to fraud prevention and detection, especially with some of the more complex frauds,” Miller says. “The best of both worlds is the best possible human fraud analyst and the best possible AI or machine learning or other types of automated fraud controls to complement each other.”
Ray Lundin spent 30 years solving murders for the Kansas Bureau of Investigation. These days, he’s fighting bloodless, financial crime as loss prevention officer for Envista, a nine-branch, 43,000-member credit union based in Topeka, Kansas. Envista is in the early stages of upgrading its core banking system, which will drive operational efficiencies in areas such as member services, mobile banking, accounting and fraud detection.
The new system will, for example, allow the credit union’s frontline staff to more efficiently vet checks by automatically determining if names on the account have ever been used in previous fraudulent activity. The system will also verify the account number, the routing number and the issuer of the check.
But Lundin does not feel threatened by the credit union’s big step toward greater operational efficiency. He and his team have plenty to do, and his detective’s keen sense of intuition can’t be replicated by a machine. “There are so many ways that fraud can come at you,” Lundin says. “You must continuously assess the risk because you can’t depend on the fraud staying the same. It’s a daily dance. You can’t simply be reactive to the fraudsters, but you must figure out a way to stay ahead of them.”
Not all the loss-prevention work can be automated. “We still have to do physical counts at the branches, as well as physical audits of internal control measures,” he says.
“It almost sounds contradictory that you can have operational efficiency and sufficient fraud prevention at the same time because fraud prevention on its own is cumbersome, requiring many redundant checks and balances,” Lundin says. But greater automation — in the lending process, for example — can reduce the redundancies and the organization can become more operationally efficient.
Christopher Gerda, risk and fraud prevention officer for Bottomline Technologies, a provider of business payment automation based in Portsmouth, New Hampshire, says one of the biggest operational challenges facing bank fraud investigators is having to work with siloed information rather than a centralized repository of data that would allow them to efficiently discover anomalous, possibly fraudulent activities.
“They don’t have all the data at their fingertips because it is in disjointed systems,” says Gerda, a former fraud investigator for a major bank. “But by getting all the data in one place, they can at least do needle-inthe-haystack type searches and look for IP addresses or do link analysis to draw out every bad guy who has used that IP address.”
The costs of creating a data lake can be rationalized, he says, “because the same data banks are using to detect fraud is often the same data they can use to justify a loan to a customer. It is a cross-corroboration of behaviors.”
“In a data lake, you can use machine learning to cluster up transactions in real time and kick out the anomalies. You now have some really good efficiencies. First, you’ve detected the fraud, and second, you’ve mitigated the losses. Data analytics, specifically behavioral analytics, play a huge role in identifying fraud.”
Edmund Lawler is a BAI Banking Strategies contributing writer who lives in New Buffalo, Michigan.