Three Myths of Fraud Detection
October 25, 2016
As technology evolves, so does the sophistication of attack methods deployed by fraudsters.
Never before have there been so many tools, technologies and opinions about how to best to combat today’s threats. It can be both overwhelming and confusing for fraud managers to see beyond trendy buzzwords, outdated conventional wisdom and blanket statements about the effectiveness of new technologies. To help provide some clarity, we are dispelling three common myths.
1. Machine learning has replaced rules and negative lists
Machine learning scales risk scoring over large quantities of data. ML algorithms can uncover trends that may be overlooked by the human eye. Machine learning is a powerful way to tell you “what” happened, but its Achilles heel is being unable to explain “why.” By relying only on machine learning, you miss a chance to assess the data with fraud rules derived from human insights. For example, a new product added to your ecommerce website is data-less; it has no history, good or bad. Based on experience, however, a human may know the likelihood of how attractive it is to fraudsters from the price point and level of demand that increase its resell value. In a similar way, the sometimes-maligned negative list (when properly maintained) helps you find repeat fraudsters who otherwise evade detection by algorithms. Machine learning is thus complementary to, not a replacement for, rules and negative lists. By combining domain expertise with machine learning, you can achieve a balanced approach to fraud risk management.
2. Manual reviews are no longer necessary
Manual review for fraud is commonly dismissed as expensive and disruptive to the customer experience. While automation does help to reduce customer friction and operating expenses, there is still value in having a manual review process. In many cases, manual review is necessary to improve your fraud management strategy and customer experience. Manual review allows for Intelligence Gathering, where fraud analysts uncover trends before chargebacks start rolling in. The fraud operations team is on the front lines interacting with customers. This team receives important feedback around false positives which can be leveraged before a larger problem occurs. The Customer Service aspect of manual review is often overlooked; in some cases, a merchant may be the first to inform the card holder that their information has been compromised. This experience leaves a positive impact on your brand. As a Safety Net, manual review provides you with peace of mind that the riskiest of transactions can receive additional scrutiny for an added layer of protection. These benefits should be strongly considered when deciding whether or not to deploy a manual review team as part of your fraud management process.
3. Data types previously used for detection are obsolete
Data types previously used for fraud detection may seem to have lost value – but are not necessarily obsolete. For example, IP Address often looked at as outdated because it is so easily spoofed through VPNs and proxies. This myth of obsolescence ignores that a VPN itself is a unique and useful signal. It also discounts the value in a good customer’s repeated use of the same IP network. To use an analogy, when DNA technology emerged as useful in physical crime investigations, law enforcement did not discontinue the use of fingerprints. New technology and data types combined with existing data types help to paint a richer view of the transaction. Your strategy for fraud detection will achieve better results by using data types (even older ones) to complement existing tools and technology.
When evaluating new tools and technology we encourage you to approach bold claims with a sense of healthy skepticism. Don’t allow these and other myths to negatively impact your business.
About the Author
Coby is a Product Manager at Accertify focused on enhancing Accertify's fraud platform and reporting offering. When he’s not at work thinking of ways to help merchants combat fraud, he’s at home being bossed around by two tiny humans forcing him to play My Little Pony and Monster Trucks while fetching them Strawberry Kiwi Capri-Suns and Cheez-Its.