Credit card fraud is ever-growing and ever-present. According to the 2016 LexisNexis True Cost of Fraud study, the total cost per dollar of fraud losses has reached $2.40. The total cost of fraud as a percentage of revenue has continued to rise as well, growing 11% from 2015 to 2016. Given status quo, this will only increase at pace — or even accelerate. There are a number of reasons for this increase. With the rapid increase in swipes/minute, especially out on the edge of networks, it makes sense that more fraudulent transactions fall through the cracks. In addition, more transactions are taking place via e-commerce and m-commerce portals. Current automated fraud detection methods were not designed for this volume of e/m-commerce transactions, despite the fact that more merchants are considering mobile expansions.
Of course, if the problem with current fraud detection was just volume, it would be easy to solve: add more resources. However, fraudsters constantly innovate and find ways to beat the system. Fraud detection methods need to stay ahead of fraudsters, but are currently failing to do so, despite the increased resources being dedicated to fraud countermeasures. As such, the number of successful fraud attempts increases every year. Changes in fraud detection must be implemented to get — and stay — ahead of fraud.
As if fraud wasn’t enough of a problem, any detection and prevention method must meet the demands of customers. With post-transaction fraud detection, any false positive that holds the card or affects the customer will cause them to look at other options. And any in-transaction system needs to be milliseconds fast to cause the least amount of delay possible; a wait of seconds could cause a customer to leave the service.
Post-transaction fraud detection, used by numerous banks and institutions, is not prevention but rather a fix after-the-fact. Manual post-transaction detection is slow, expensive, and leads to many false-positives. In-transaction fraud detection and prevention works, but comes with its own set of challenges. As mentioned before, the seconds added by a poorly-implemented in-transaction system will cause you to lose customers. But optimizing solely for speed is foolhardy; using an eventually-consistent solution may increase false-positives. If your data is not consistent by the time there’s another transaction (which could be mere milliseconds), there’s a real chance that the incorrect data causes a false positive. The inconsistency of eventual consistency behavior makes it a risky choice for real-time fraud detection. For a successful in-transaction fraud countermeasure, you need a real real-time transactional database.
VoltDB enables financial institutions to leverage real-time data for their fraud countermeasures. With VoltDB’s serializable ACID consistency, you can be sure that your data is always correct while processing hundreds of thousands — or more — card swipes per second. With a VoltDB powered fraud solution, you can detect and stop fraud in-transaction in milliseconds, instead of just fixing it after the fact.