Predictive Analytics and Complex Event Processing Technology Move to Cutting Edge of Financial Services Industry
Creating a Lifetime of Value
According to Thorpe, Wachovia has developed customer acquisition and target models, as well as models for attrition, using the SAS Analytics solution from the Cary, N.C.-based vendor. By combining those predictive models, the bank creates what it calls a customer lifetime value model, which helps it decide for which geographies and channels it should increase or decrease advertising spend.
"We're looking at long-term value to the bank rather than just pure acquisition," Thorpe says of the lifetime value concept. "We've worked really hard to develop the infrastructure to really be fact-based, quantitative-based and predictive analytics-based to understand return on our investment."
Wachovia also uses predictive analytics to increase the value of its existing customer relationships, having applied its models to develop customer lifetime values for the 13 million households in its retail bank, according to Thorpe. Rather than simply recommend new products to those households in a blanket fashion, Wachovia determines which households are most likely to respond to specific offers and which products offer the greatest profit opportunities. "We're taking that customer lifetime value knowledge into how we actually touch a customer," Thorpe relates.
Because predictive analytics enable institutions such as Wachovia to tailor products to specific customers' needs, similar innovations will become increasingly important in differentiating financial services firms as competition increases, according to John Lucker, a Hartford-based principal and national practice leader for advanced quantitative solutions with Deloitte Consulting. "It's not a one-size-fits-all world anymore," he says. "People demand customization." Further, Lucker adds, in the chase for higher investment returns, customers are more likely than ever to move their money from one institution to another, making identifying and retaining the most profitable customers more challenging and more important than in the past.
"Once you find these people, you have to do something to make them stay. It's not just finding them; it's reacting to their needs," Lucker comments. "Predictive analytics ties into finding the people and then using more traditional marketing efforts to get at, from a segmentation perspective, what it is they want and then catering to those needs."
Stopping Fraud
While predictive analytics can be used to identify the most-desired customers, the technology may be equally valuable when used to identify the least-desired customers -- those likely to commit fraud.
For insurers, predictive analytics can be leveraged during the claims process by finding commonalities among new, incoming claims and historical claims that may indicate fraud. When those common elements are uncovered, a carrier's special investigations unit (SIU) can trigger an investigation.
Wachovia's Thorpe says there's an opportunity to identify bank fraud using predictive analytics and pattern analysis as well. But particularly in banking, it's vital that fraud is identified immediately. "It has to be very dynamic," Thorpe explains, pointing out that someone with a stolen debit card will try it once to see if it works. If successful, they'll make as many transactions as quickly as possible to max out the card. "We have to be able to catch it the first time to prevent those repetitive ones. You have to be that quick," Thorpe continues. "To identify fraud, you have to do it instantaneously to prevent it from happening elsewhere."
While pattern analysis, or anomaly detection, has been used to detect fraud in the credit card industry for some time, lately it also has been garnering increased attention on Wall Street. Hacking against financial firms is at an all-time high, and perpetrators increasingly are hijacking customers' accounts to buy and sell stocks fraudulently online. According to security experts, pattern analysis can be a big help in detecting such anomalies.
The method involves running server-based processes in the background that authenticate users based on what types of transactions they are executing and/or from where they are logging on. The information is then compared with a profile of what is expected of each user. If an individual's behavior is out of range with what is expected, the transaction can be immediately flagged.
"If constantly kept running in the background, these systems can be highly effective at keeping criminals out," says Gartner VP and distinguished analyst Avivah Litan. "They are not intrusive to legitimate users unless the user's activity is suspect."
Tweet This