Analytics & Big Data
May 22, 2015

Smart fraud detection techniques strengthen customer contact channels

Today's multi-device and omni-channel landscape promotes a number of ways to complete banking transactions. This has also created more opportunities for fraud to occur. When banking institutions beef up security in one area, fraudsters often shift their efforts to another, perhaps less protected, area. For many firms, call centers may be the likeliest avenue for fraud. Armed with some customer information gleaned from data breaches and then calling contact centers repeatedly to gain more, fraudsters can successfully impersonate a client. Once they have enough information, they may ask for a password reset to gain online access or request an additional card.

"According to Consumer Sentinel Network, U.S. Department of Justice, the total amount of credit card fraud worldwide in 2013 was USD 5.55 billion and on the rise."

The answer lies in smart data analytics
To combat this activity, developing fraud detection strategies that take into account all of the different touch points a customer has with his or her bank is important. One key solution is to harness multiple sources of data to understand the entities being compromised and gain a more holistic view of customers. This requires consolidating data sources such as transactional data, call center data, bureau, mobile data, along with social media data that are often trapped in siloed enterprise systems and using the data to distinguish fraudulent activity from normal activity.

For example, if a credit card customer fails to alert her card lender of her travel plans, a strategically implemented fraud detection system can enable the lender to automatically gain insight from mobile and social data that the customer is traveling and thus reduce the incidence of false positives. An example of fraud via customer contact channel is a customer requesting several password resets of his online account within a short time span and following this up with high-velocity card usage. In this case, if only transaction-level data were being studied, probably the second or third sale transaction attempt might be declined based on high-velocity triggers. However, if contact center patterns were also observed, multiple failed authentications/password resets would have already triggered suspicion before a monetary loss occurred.

Putting it all into context – a strategic solution
Fighting fraudsters that socially engineer call-center staff, the latter usually over-burdened with calls, is a challenge. However, fighting fraud via contact centers with additional touch points such as Interactive Voice Response (IVR) and web login is posing an even greater challenge for banks. Analytics-driven solutions use anomaly detection techniques, and those that address authentication are vital to curb account takeover and identity fraud, the two leading fraud types arising from such a fraud scenario.

At Genpact, we have constantly explored and leveraged new dimensions and data sources to give our solutions an edge and a holistic appeal. In the case of a U.S. credit card provider, account activities such as address change request, addition of an authorized user, and credit line increase requests were studied with the assumption that in an account compromise scenario fraudsters attempt to completely take over the customer account and carry out a large number of transactions. Over a period of three to four weeks from the time of compromise of the account information and subsequent changes such as address change and credit increase changes through contact center channels, sales increased by a whopping 96%.

A segmentation-based approach was followed to identify fraudsters using contact center interaction–based attributes such as the following:

  • Linguistic patterns (call reason, activity conducted, missed authentication, distress pattern)
  • Non-linguistic data (response time, authentication time)
  • Account contextual data (previous interaction data, caller phone number match, location match, transfer destination)

These findings open up a new dimension of data sources that can potentially be used to solve for fraud in a more holistic manner.

In the face of fast-changing trends in fraudulent activity, it is necessary to solve for fraud across channels. Fraudsters do not limit their attempts to any one channel, and so neither should fraud solutions. This is particularly important for solutions to mitigate fraud via contact centers since this kind of fraud, rampant in the customer's credit life cycle, entails activity across multiple channels. Thus, continuous efforts must be made to enhance existing solutions by ongoing solution monitoring and innovative sources of data.

To learn more, you can also read Integrated Risk Modeling and Analytics - guide to survive comply and grow or write to our Risk Management experts.

Author: Anisha Kumar - Senior Manager, Financial Services Analytics