Kount lets merchants be less picky about fraud detection

Listen to this article

Kount Inc. has announced a new version of its fraud detection software that can be configured to be less stringent on fraud prevention.

While this might seem counterintuitive, there are business reasons behind it. Kount’s software, Kount Complete, detects potential fraud in electronic commerce. But being too careful about fraud can eliminate legitimate customers acting in an unusual way, such as using a different credit card, ordering merchandise to be shipped to a different location or buying a larger amount than usual.

That fraud detection process can also add to the challenges on a site for customers to purchase goods, which is known as “friction.” When there’s too much friction, customers can give up on a site and go to one of its competitors instead.

With the new version of the Kount software, businesses can decide to relax certain criteria, such as being less concerned about people shipping to different locations or ordering a large amount of merchandise during the holiday shopping season, or logging in from a different location during the summer when they might be on vacation.

“Our whole platform puts control in the hands of our merchants,” said Tricia Phillips, senior vice president of product and strategy for Kount, in Boise. “Some of them may have really small margins, and their tolerance for fraud is extremely low.”

On the other hand, a merchant of digital goods with larger margins may be more concerned about friction, she said. “The business can determine the risk they’re willing to take.”

While there are other fraud detection products on the market, they don’t offer this feature, said Steven D’Alfonso, research director for IDC Financial Insights, in Framingham, Massachusetts.

“Other fraud solutions don’t have the capability to be able to control how much fraud they accept, instead of simply going after fraud in and of itself,” D’Alfonso said. “There are a lot of firms that can do that, but that sacrifices the customer experience, and introduces more friction into the process when you might not need to.”

The new version of the software has created a new way of scoring potential fraud. Kount has always used artificial intelligence, or machine learning, to analyze the potential of fraud in a given transaction. The software does this in two ways: Supervised and unsupervised. In supervised machine learning, a human teaches the software how to recognize fraud. In unsupervised machine learning, the software goes back and analyzes Kount’s 12 years of transactions to learn how to recognize fraud on its own.

Previously, these two scores were used separately. The new version of Kount’s software now includes Omniscore, which combines the two scores for a result that is up to twice as accurate as previous scores, Phillips said. Because the score is more accurate, it means that merchants can hire fewer fraud analysts, she said.

“The key here is that it’s taking the two different types of machine learning, supervised and unsupervised, and putting them together in a way that’s achieving better business outcomes,” D’Alfonso said.

In the next five or 10 years, merchants may be able to automate up to 80% of their fraud detection, which would free their fraud analysts to focus on more complex cases instead, D’Alfonso said.

“Omniscore takes a big step toward that vision,” he said. “I’m interested to see how the industry adopts it.”

IDC has not yet done a market analysis on fraud detection, but Kount would likely rank in one of the analyst firm’s top two vendor tiers, D’Alfonso said. While the market is getting more crowded, Kount is well established and could be more viable, long-term, than its newer competitors, he said.

Kount Complete has 6,500 merchant customers, Phillips said.

The company received $80 million of venture capital in 2015 from CVC Capital Partners.

The new version of the software, sold as a service, is available now. Existing users will be upgraded.