Manual fraud review isn’t as old-fashioned as you might think. It reduces false positives and customer churn. It also helps artificial intelligence-driven fraud-screening systems get smarter. Manual review can even work for merchants like digital content providers that need to make real-time order decisions.
In part one of this series, I wrote about how manual review can increase revenue and strengthen customer relationships. I also wrote about how false positives cost merchants far more than completed card-not-present fraud, and drive away up to 80 percent of rejected customers.
Now let’s look at what you need for a real-time manual review program to reduce false declines without slowing down decisions. Real-time manual review requires some resources and expertise that not all merchants may have in-house. Merchants that can’t implement the programs described here may want to outsource them.
First, a damage control program can reduce false positives by giving declined customers a second chance. This program allows customer service reps (CSRs) to temporarily add customers to a whitelist if they complain about a decline. Then, either they or the CSR resubmits the order, and it goes to the manual review team for further investigation.
This process reduces the damage false positives can have on merchants’ revenue and repeat business. However, damage control creates its own challenges. First, you can’t rely on your customer service team to handle these manual reviews. The skillsets and experience needed for customer support and manual review are different. Effective damage control requires two specialized teams, not one team trying to fill two roles that sometimes conflict.
Second, good customers aren’t the only ones who contact customer service to complain about declines. Fraudsters who are bold enough to do this can also fool manual reviewers who have only basic skills. That means damage control programs need experienced reviewers who keep up with current fraud tactics. Otherwise, your damage control program can create an easy path for fraudsters to exploit. For merchants in spaces like digital goods, with higher fraud costs than other online retailers, any increase in fraud is unacceptable.
The second program is control group review. It uses data instead of customer feedback to identify falsely declined orders. Analysts select and review random batches of orders that were automatically declined to figure out which ones were declined by mistake. The reason for using sample batches rather than reviewing all declined orders is cost. Most merchants don’t have the budget to review every order, and sampling yields good information more cost effectively. Accepted orders don’t need this kind of review because mistakenly approved orders turn into chargebacks. You’ll know about them.
Of course, you can’t go back and approve good orders that were mistakenly declined. Nor can you reach out to customers whose orders were rejected in error. That would be awkward, at best. Instead, the purpose of a control group is to find your false decline rate and reduce it, while still stopping fraud and delivering real-time decisions. For example, your data may show you need to change your automated rules to approve more good orders from shoppers who are buying while on vacation abroad.
Data from your damage control program, and especially from your control group program, can improve your AI fraud detection tools. Machines learn what we teach them. When you use your program data to update the status of past orders, you give your algorithms new information that changes the patterns they see as fraud.
For example, what if your control group found that dozens of declined orders from customers in China were legitimate orders? In very simple terms, retroactively labelling those orders as good will help your AI see a new pattern — one in which orders from China aren't automatically considered fraud.
Real-time order decisions will always carry more risk than orders that have more time for investigation. False declines will always be costly, in the moment and over the long term. However, with manual review programs that use customer feedback and data review, you can protect your store from fraud, reduce your false declines, and make your automated and AI-driven fraud tools more effective.
Rafael Lourenco is executive vice president at ClearSale, a card-not-present fraud prevention operation that helps retailers increase sales and eliminate chargebacks before they happen.
Original Article at: https://www.mytotalretail.com/article/why-manual-review-is-key-to-retailers-defense-against-fraud-part-2