With more than 330,000 members and $4.5 billion in assets, Santa Rosa, California-based Redwood Credit Union is the eighth-largest CU in the state. The FI has used data analytics since its inception but began an initiative to centralize and expand its program just three years ago, according to Chief Information Officer Tony Hildesheim.
Leveraging financial and social data in accord
Like any other credit union of its size, Redwood processes thousands of financial transactions every single day. This provides the CU with a plethora of data insights on individuals’ spending habits, which then allows it to better personalize and market its products and services.
“We use information about external transfers and about people’s habits for moving money to predict solutions that will better meet their needs,” Hildesheim said. “By doing that, we can target-market those individuals and let them know we have a solution.”
Analyzing financial data may be crucial for building marketing campaigns, but it often only paints half the customer profile picture, he explained.
“There’s this idea out there that there’s a wealth of data that you have as a financial institution, that you should somehow [have enough information] to be able to make decisions based on [members’] needs,” Hildesheim said. “But the reality is that you’ve got to couple that with social information, because a lot of spending decisions are based on what’s happening with the individual from a social perspective.”
Social data takes priority over financial information in pre-approval for auto loans, for example. Redwood previously offered pre-approval approximately every 24 months. This interval was determined largely by each member’s credit score and the last time he or she financed a vehicle, but the formula did not produce a cohesive loan applicant profile.
“If you buy a car and you happen to have a partner in the household, then within nine months, that partner will [likely] want a new car as well,” Hildesheim said. “If the husband buys a truck, the wife’s going to want a new Volvo.”
Redwood is re-evaluating its preapproval frequency to resolve this problem, and is looking at partnering with social data aggregation company Acxiom to obtain the data necessary to explore similar options.
Using data analytics to detect fraud
As with many CUs, fraud detection is one of the biggest challenges Redwood faces today. The augment of faster payments and surge in digital transaction volumes have made it nearly impossible for human analysts to examine every transfer that gets red-flagged, a problem the credit union sought to resolve by turning to data analytics to recognize fraud patterns.
“As a consumer, you have a pattern about how you deposit, where you deposit and when you deposit,” Hildesheim explained. “But maybe, all of a sudden, you have deposits and withdrawals much more frequently. That heightens the likelihood that you might be falling victim to a scam.”
The company’s fraud detection system assigns the transaction a confidence score based on its likelihood of malice after a potentially fraudulent transaction is flagged, he added.
“Usually [the confidence score is] based on looking at hundreds of thousands of deposits … made by other people in the credit union, but also by that individual themselves,” Hildesheim said. “The more history we have with that individual, the better we are at doing this.”
The flagged transaction is then sent to a human analyst for review, a process he believes might be entirely handled by artificial intelligence (AI) in the future. There are some serious limitations to its current implementation, however.
“One of the challenges for credit unions and small banks is the amount of data that you really need to make AI effective,” Hildesheim noted. “Most people woefully underestimate the amount of data you need to actually get a reasonable response.”
The other major challenge with AI is bias, both within the system itself and from its creator. Redwood experienced this firsthand when it tested an AI system to identify people who needed financial counseling, as the solution predominantly singled out younger people.
“Theoretically, I guess you could say that the machine was right,” Hildesheim said. “Young people generally do need financial counseling, but it didn’t use what we thought was appropriate data to determine that.”
Will analytics take away CUs’ personal touch?
Data analytics can be a powerful tool, but it still has its limitations. Hildesheim feels the best way to determine CU members’ needs is often just to ask about them.
“At the end of the day, if you’re their trusted financial services partner, they’re ready to answer questions or provide information that allows us to help them better,” he said. “That’s been way more effective than trying to use data to figure it out, to be quite candid.”
In fact, Hildesheim has concerns that CUs are relying too heavily on data analytics for their daily operations. Consumers largely choose CUs over banks for their human touch, after all.
“If you don’t have people who are able to really understand the needs of that individual and meet them, then even if you have the best data program in the world, it doesn’t really matter,” he explained. “I think that’s the biggest fear I have for credit unions in general — that we forget who we are, where we came from and what our core values are.”
Striking that balance might be the key to successful implementation as data analytics becomes more prevalent in the CU space.