Manual fraud investigations plummet 66%, leading to $3.5 million in savings

Improving fraud filter accuracy reduces false positives for a better shopper and seller CX

Manual fraud investigations plummet 66%, leading to $3.5 million in savings

Improving fraud filter accuracy reduces false positives for a better shopper and seller CX

66% reduction in manual work

$3.5 million in cost savings

800% improvement in fraud accuracy

Though it was labeled the Wild West decades ago, online commerce remains a savage place to do business. Fraud threats continue to escalate, eating into profits, harming brand reputation, and compromising the customer experience.

Even when bad actors are not trying to swindle companies and their customers, flawed authentication processes threaten to undermine the experience, frustrating shoppers and those selling their goods online.

The challenge

An e-commerce marketplace prioritized authenticity, knowing it was crucial to build trust and loyalty among its customers as well as the sellers who relied on the platform for their livelihoods or just a little extra spending cash.

Our client wanted to ensure that goods offered for sale on its platform were authentic and available for sale when a shopper clicked “Add to Cart.”

Too many product listings were incorrectly flagged as fraudulent, however, leading to poor shopping/selling experiences and excessive costs of manual investigation.

Our solution

Our team conducted false positive analysis to identify the patterns and root causes that resulted in legitimate buyers and sellers to be incorrectly flagged as suspicious or identified as bots when in fact they were human. We refined the search algorithm to more accurately identify problem listings.

New, more accurate search strings with keywords such as “sty1e,” and “oak1ey” and “tiffanysty1e” were designed and implemented to help flag counterfeit goods and misrepresented logotypes.

We enabled associates to perform content moderation and account security across billions of active listings and millions of buyers. We deployed automated machine learning and AI algorithms to uncover and respond to account security issues including account takeovers (ATOs) and verification.

The results

The volume of transactions requiring manual investigation plummeted 66%. As a result of more accurate search outcomes, 1,079 hours were saved each month and full-time equivalents (FTEs) reduced by 60%, producing $3.5 million in savings over three years.

Fraud filter accuracy improved 800%.