Blog: Tackling Fraud with Data
As the dust settles on the 2014 retail holiday season, it isn’t surprising to learn that e-commerce was once again the winner. ComScore reported that online holiday spending through Dec. 21 was $48.3 billion, a 15 percent increase over 2013. And there’s nothing to suggest that this growth trajectory will flatten. While these trends are encouraging for online retailers’’ sales departments, they must be challenging for their fraud and loss prevention teams. According to the 2013 Federal Reserve Payments Study, card-not-present fraud rates were approximately three times higher than card-present fraud rates in 2012.
Just before the holiday shopping season, CyberSource released its 15th Annual Online Fraud Management Benchmark Study. This 2014 study reveals that merchants improved order conversion through lower rejection rates, while keeping their fraud losses stable. Naturally, I was curious about the tools that yielded these results and wondered to what extent they might have changed. Using CyberSource’’s 2012 study to compare, I found some surprises.
In 2012, validation tools were used the most—79 percent of merchants used a card verification number and 77 percent used address verification. Of the merchants that did not use these tools, 81 percent indicated they planned to implement a card verification number and 61 percent planned to use address verification. While merchants can implement these tools with little cost, their effectiveness, according to the surveyed merchants, is limited.
“No single tool is a panacea for fraud management. A layered approach using multiple tools and data elements is critical for success.”
Given the 2014 report’s positive findings, coupled with the expected very high use of card verification numbers and address verification reported in 2012, I was expecting merchants to rate the effectiveness of these tools higher. Interestingly, even though these validation tools remained the most prominent, their usage did not increase as expected, despite the number of merchants that planned to implement them following the 2012 study. And there was not a significant increase in their reported effectiveness.
Here’s what did change: the use of proprietary data tools, such as customer order history, in-house positive and negative lists, and company-specific fraud scoring models. Purchase device tracking tools, such as fingerprinting, also saw increased usage, though not as large of an increase as the proprietary data tools. And it’s these tools that, generally speaking, are rated as the most effective fraud management tools by the merchants surveyed.
The 2014 study highlighted improved fraud management. I have several of my own highlights. Merchants appear to be more apt and capable of leveraging their own data today than during the preceding several years. And, they are finding that using these data is more effective in combating fraud than traditional validation services. I think it’s important to note that only two tools (device fingerprinting and a fraud scoring model) were selected by more than 50 percent of merchants as most effective. Even though traditional validation services are still highly used and useful, no single tool is a panacea for fraud management. A layered approach using multiple tools and data elements is critical for success. I suspect this trend of merchants using their own customer data to manage CNP fraud will continue. I also expect that data-centric tools will become more effective as merchants become more sophisticated with data analysis.
What’s your view on the future role of proprietary data in CNP fraud management?
Douglas A. King is a payments risk expert in the Retail Payments Risk Forum, Federal Reserve Bank of Atlanta. He can be reached at [email protected]. This article originally appeared on the Atlanta Fed’s Portals and Rails blog.
In Blogs & Viewpoints, prepaid and emerging payments professionals share their perspectives on the industry. Paybefore endeavors to present many points of view to offer readers new insights and information. The opinions expressed are not necessarily those of Paybefore.