A bank using behavioral analytics and risk scoring to reduce false positives and fraudulent activity — Register for the INETCO payment fraud boot camp miniseries, **1-2-3 Detect: Tips to get your payment fraud strategy in top shape**. A link to a 5 minute video will be sent to your inbox once a week.

As efforts increase to protect customers from card-present and card-not-present fraud, there is an increasing risk lurking for banks and retailers: false positives. According to KPMG’s 2019 global fraud survey, 51% of respondents reported a significant number of false positives resulting from current technology solutions and decreasing efficiencies in fraud detection. With customers expediting their shift to digital channels, we see this being paralleled by an even sharper increase in both fraudulent activity and false positives.

There is no doubt that too many false positives becomes a significant point of friction for many banks and retailers. The approval or decline of a transaction is often the defining factor of customer experience, and a profitable card portfolio. This is why it is important that your high-risk transactions are being identified, scored and classified – as correctly and efficiently as possible.

So what is the secret to reducing false positives? Well, there are a few things to think about, as detailed below…

Nurturing your machine learning models with more data points

Predictive technology such as machine learning is commonly used to analyze several data points within a single transaction, producing a value that is used to score the transaction based on the level of risk. But it is important to note that your machine learning will only be as good and fast as the data it is fed. The more comprehensive, and timely, the data delivery, the more precision can be applied. There are over 190+ data elements contained in each ISO payment transaction. With access to all this information, you can configure a hybrid combination of rules-based alerts and machine learning models that continuously rebuild individual customer models every time an event occurs. With more data precision and faster modelling, comes more accuracy around which transactions need to be blocked.

Expanding attack vector and attack surface coverage across all links, third party transition points and payment channels

A payment transaction is made up of multiple links, networks and transition points. Most end-to-end transaction journeys include at least 4-8 links. Each time we add an API, service or device, we are essentially introducing a new place for a transaction to be hijacked. As payment related fraud attacks take on a new level of sophistication, it has never been more important to consider a holistic approach. More and more often, attacks are only detected by correlating multiple data points – such as a payment on a different device, from a different location and a different frequency than normal. The ability to screen every end-to-end transaction, across every link and every hop, is part of the secret to success.