Is your financial institution’s fraud case management system up to the challenges raised by today’s advanced predictive fraud detection models? Fraud case management functionality is vitally important to the successful deployment of predictive fraud detection models. You cannot be without it.
The importance of case management goes far beyond the obvious operational workflows and processes aligned with detecting fraud. Analyses of detected and missed frauds are both critical components of the analysis stage of the fraud management lifecycle. An effective case management system should support both of these critical activities as a starting point -- but there are also many other tasks it should be able to execute.
[For more on fraud detection and analytics, check out: Fraud Detection & the Customer Experience Continuum.]
Analysis of fraud data coming from your case management system provides a wide range of valuable information. Just ask yourself if your current approach can tell you what types of fraud scenarios were caught and how quickly they were caught. The system should also help you understand what types of fraud are decreasing and what types of fraud are increasing in severity. You can also learn what new cross-channel fraud activities are being used. And lastly, are those really “new” fraud types surfacing in your data, or do you have a tagging accuracy issue?
If you aren’t able to answer these questions, it could be you are losing the data you need to do so. Many important case management fields are corrupted or manipulated for operational reasons or just plain not fed into case management systems by core systems. Some of these “lost” fields include transaction date, date of first fraud, transaction time, transaction location, exact transaction amount, contiguous legitimate transactions, non-monetary transactions, and denied claims. These are straightforward fields to store within a case management system and facilitate accurate fraud analysis and predictive modeling.
Are you denying fraud claims regularly on the same account and customer? Was the account compromised at a point-of-compromise or was the card lost? Was it a mail theft? Was it an account takeover? All of these differences should be captured and fed into your fraud hub for inclusion in the next version of your predictive models. The system should also tag fraudulent transactions at the time of confirmation of the fraud and should include a fraud flag that specifically identifies the type of fraud identified or suspected.
Not all frauds that concern us are detected by current detection systems. The “best-practice” systems have an automated process to ingest externally identified fraud, closing the loop on false negatives. Similarly, not all transactions on a fraudulent account will be the same fraud type. For example, sometimes a card-not-present fraud type and a card-present fraud type will be seen on the same account. This contextual visibility can only be obtained from the details gathered by the case management system and stored in the fraud hub.
Keep in mind that fraudsters are channel agnostic -- is your analysis of fraud equally broad? If it isn’t, you risk missing important information that you could use to improve fraud detection and mitigate losses. If you use standardized industry fraud definitions, peer performance comparisons become possible. Employing a comprehensive case management system across the enterprise yields enhanced fraud analysis and detection performance.
[Learn more about the Internet of Things at Interop's Internet of Things Summit on Monday, September 29.]
Wesley Wilhelm (Wes) has more than 30 years of experience in banking and consulting to the financial services industry, with extensive knowledge of fraud management, payments, and retail banking technology and operations. He has held numerous management positions in risk and ... View Full Bio