The role of machine learning in perfecting fraud management strategies

Register now

Now that we’ve entered the largest retain season of the year, it is the perfect time to turn attention to cyber security defenses and fraud protection.

Alarmingly, 96% of all ecommerce businesses have reported some form of fraud attack at their organization, with account takeover landing in the top three, according to Merchant Risk Council’s 2019 Global Fraud Survey Results.

Retailers need to prioritize proactive fraud management more than ever as the threat of fraudulent online activity continues to rise, especially as more consumers store payment, billing and shipping information with retailers directly. In order to catch fraudsters quickly and efficiently, retailers must consider the use of technology and machine learning – where it works, where it doesn’t and where it’s costing retailers.

From new automated technologies to manual review processes, ensuring that these methods apply across all channels is vital, here’s what retailers need to know in order to stay ahead of the fraud game.

Protecting customers across all channels

Omnichannel, omnichannel, omnichannel. It’s a blessing and convenience for today’s consumer, but it adds many layers of complexity for retailers especially when it comes to fraud management. Transactions are occurring instore, online and through mobile – sometimes all are used in a single buying journey – and consumers expect a seamless experience at every single point.

With so many purchasing environments available today, retailers must effectively manage fraud across more touchpoints. For example, on-site mobile payments are now common for smaller-scale retail transactions such as local coffee shops or boutiques who leverage tools like Square to make payment more immediate.

With a strong fraud management system in place, retailers can minimize cross-channel fraud and chargebacks while simultaneously avoiding checkout friction and rejected orders. Additionally, cross-channel data from fraud management systems can be leveraged to gather insights on fraudulent activity, protect revenue streams and deliver a seamless customer experience at every touchpoint.

Finding the balance between automation, machine learning and manual fraud review

To preserve the customer experience and provide fast and effective fraud prevention, many retailers are taking a multi-pronged approach to fraud prevention to save time and money. Nearly 30% of merchants believe controlling fraud is too expensive, which is why many are turning to this duet of strategies.

According to research, the average ecommerce merchant spends time manually reviewing 16% of orders, with each manual review taking at least 10 minutes. Manual reviews are a necessary component for fraud management but driving down the number of transactions that require a manual review should be a priority for retailers.

Depending on the industry segment, ideally retailers should aim to drive down their proportion of all transactions going to manual review to less than 2% - far lower than the average of 16%. This can be done by investing in technology that can automatically discern between “normal” and fraud patterns and base models off that activity to improve automatic approvals and reduce the need for manual reviews.

For example, TaylorMade, an American golf retailer, was able to free up valuable headcount that they were able to then use to support increased peak activity when they took advantage of automatic reviewing. This allowed for employees to focus on service and finance, both of which are vitally important throughout peak retail season.

With both automated fraud management as well as manual review processes, retailers are powering security with both the power of data and human touch and driving greater efficiencies as a result.

Fraud technology in 2020 and beyond

With multiple shopping channels and consumer expectations around delivery at an all-time high, effective systems that automate fraud checks are paramount.

Take for instance eGift cards. Retailers capitalize on these product offerings during the holidays because they are fast and convenient ways for consumers to send a friend or loved one a special gift. However, the instant delivery that shoppers expect makes them a ripe target for fraudsters.

According to a recent Radial CNP report, the attack rate for open-loop gift cards (those branded by Visa and Mastercard not connected with a specific retailer) is nearly 21%, while entertainment companies that offer digital gift cards see attack rates of 24%.

This is an area when machine wins out over manual reviews. While automation of the process is important for eGift orders, if a company decides manual review is necessary (for example: for an especially high-value transaction), it should not send any eGift card information to the customer until the transaction has been approved.

Looking at fraud management in the future, tools must rapidly innovate to keep up with emerging technologies and the fraudsters who find new ways to cheat the system. Behavioral analytics for instance, are becoming more of a focus as retailers look to quantify and understand more of the customer’s buying journey from online to in-store.

As shopping continues to be more personalized so must the tools, behavioral analytics systems can combine observed consumer behaviors with transactional patterns and history to validate a customer’s identity. With these additional layers of complexity and information, a more comprehensive view of the customer comes into view and their preferences are captured and utilized to help better protect them and their purchases.

As fraudsters’ methodologies are constantly changing, retailers must stay alert and consistently arm themselves with the information and tools to fight back. Those omnichannel retailers who are not implementing advanced technology like machine learning and device fingerprinting into their fraud management stack will find it difficult to keep their customers and profits safe in this ever-threatening ecommerce world we live in today.

For reprint and licensing requests for this article, click here.