AI and tossing out the manual fraud review manual

Time is money, as the old saying goes, and that idea applies to few places more keenly than commerce.

The presence or absence of speed, particularly around transactions themselves, can be the difference between a conversion happening or not. Optimizing for smoothness and speed is increasingly a non-negotiable competency among merchants, especially for those working in a multi-channel context.

That’s where the persistence of manual fraud review process can be a problem, said Carl Tucker, vice president of risk solutions at Cybersource, a Visa solution, in a recent conversation with PYMNTS. Manual fraud review, he said, is not so much a fraud prevention tool; it’s a revenue saving activity for transactions that could have otherwise been rejected by the fraud strategy.

In places like pure-play eCommerce, quick-service restaurants (QSRs) or physical merchants with buy online, pickup in store (BOPIS) programs that run the same day, he noted, these processes are less common because they are so time-consuming that they are simply not suited to the near real-time fulfillment environment. There is no such thing as manual fraud review on a burrito sale because there is no context in which it could ever be usefully applied.

But in other spaces, particularly where merchants started in the physical channel and are branching out, manual fraud review persists and acts as a drag on commerce. The problem, Tucker said, is in realizing a fraud-fighting strategy that made sense in one channel doesn’t translate to others, and trying to insert it is costly both in terms of resources expended and in terms of sales lost.

“What we see is that it leaves too many customers at risk for false positives and unnecessary manual review,” he said. “And it just means that the overall fraud strategy is not as dialed in as it could actually be. We really consider manual review to be the last resort in a fraud strategy, and across the board we’re working with all clients to reduce or virtually eliminate manual review because it leads to faster fulfillment and a better customer experience.”

Getting rid of what doesn’t work, however, isn’t enough—to remove a fraud strategy that is underperforming, he said, the key is developing a fraud strategy that aligns with your customers’ expectations in each channel.

Improving approvals without increasing risk

The problem with manual review, he said, is that it is slow and cost intensive—two things that not only fail to add to the customer experience, but can lead to cart abandonment. And given that it is an expensive mechanism by which to fight fraud, he said, the goal for Cybersource is pretty simple—put customers on the path to using more sophisticated fraud-fighting technologies and strategies that actually reduce fraud rates while maintaining or improving acceptance rates.

There may still be a place for manual review in that process. When Cybersource’s artificial intelligence (AI) flags certain types of transactions as risky, there may be value in sending those transactions on to a manual fraud reviewer to see if there is still a way to accept these transactions and pass them through—or definitively reject them. That, however, is a last resort and one that a merchant with a sophisticated fraud detection strategy in place hopefully won’t have to use all that often.

What Cybersource has seen, Tucker said, is that as the market is increasingly digitized, and eCommerce players have entered and changed the field, the firm has really been able to push the industry forward in terms of using technology more efficiently to detect fraud instead of relying at all on manual processes.

“Necessity is the mother of invention, and merchants that are unable to have manual review due to customer expectations, like digital merchants, are the ones that are really leading the way when it comes to reducing or eliminating manual review,” he said.

And once manual review is out of the picture, or at least seen its footprint greatly reduced, Tucker said, it makes room for a more interesting conversation to begin.

Building the next-generation fraud detection tools

When the time and expense of manual review is off a merchant’s plate, Tucker noted, they get to start thinking more creatively and extensively about how they want to redeploy that time and treasure. What sells merchants on moving away from manual review and more toward automated detection, he said, is the fact they can demonstrate over time that their acceptance rates will go up, their manual review needs will go down, and their losses to fraud will still fall.

And when those changes are in place, he said, they move onto the next level of conversations about how to optimize the payment flow structure in a manner that optimizes their overall acceptance and authorization rates.

Thus far, he said, eCommerce merchants have had an advantage in this regard because they were mostly single-channel operators who leapfrogged manual review in the development. But the landscape has changed drastically in the last decade, or more accurately, even in the last five years. Merchants that were married to old processes then, are now looking at the data Cybersource shows them, and seeing the writing on the wall when it comes to consumer shopping habits.

“We can show clients how merchants that have invested in an eCommerce and an omni-eCommerce strategy how they’re outperforming those that have not,” he said.

And it’s made merchants more willing to leave their manual processes in the past, or in a much-diminished role, he said, as most would prefer to be among the outperforming than the outperformed.