Fighting fraud is a time-consuming task and most online retailers need help with it. Ecommerce platforms are stepping up to assist their merchants with AI-driven solutions, but do these AI-only tools provide enough fraud protection? Here are some factors to consider before opting in to a platform-provided fraud screening program.
First, it’s important to understand what AI screening programs can and can’t do. They can automatically apply the rules they’re given to each transaction, which provides a fast and easy way to spot orders with attributes that may indicate fraud. There are two important things to remember here. The first is that the AI program will only be as good as the data it receives, so regular updates with high-quality, validated information are necessary to keep it working well. The second is that AI can flag orders for possible fraud, but not all flagged orders are fraudulent, so it’s important to know what happens to orders that are flagged by the AI program.
Does the program prevent false-positive declines?
If the platform’s AI automatically declines flagged orders, that could end up costing you more money than it saves. That’s because some of those orders will be from good customers whose behavior overlaps with that of fraudsters: ordering from multiple devices in different cities or even countries, making high-value purchases and requesting rush shipping, and sending purchases to addresses in multiple locations. Each of these behaviors can be a fraud indicator—or a sign that a well-to-do customer who travels often for work or leisure is placing orders. The AI won’t know which it is and if it falsely declines the flagged orders that are actually good, your business will lose more than the value of those orders.
That’s because false declines disproportionately hit high-net-worth shoppers—research by MasterCard found that affluent consumers place more than half the e-commerce orders that are falsely declined. And nearly a third of shoppers surveyed said that after just one mistaken decline, they will never shop with that merchant again. Over the long term, false declines raise your cost to acquire customers, raise churn levels, and erode the average lifetime value of your customers.
This is why AI works best when it’s paired with human analysis to verify flagged orders, contact customers if necessary, and separate the good customers from fraud, rather than simply kicking all flagged orders out of the system.
How well are you protected against new types of fraud attacks?
Fraud prevention is not a set-it-and-forget-it business. CNP fraud is an organized criminal enterprise, which means it’s always looking for new weaknesses to exploit and new technology to use. AI excels at spotting known fraud indicators, but when new types of fraud arise that don’t match the AI’s parameters for flagging an order, they will simply pass through the system and be approved. Only when humans realize there’s a new type of fraud and give that information to the AI engine will the AI be able to detect those new indicators.
A more comprehensive approach includes proactive input from fraud analysts to narrow or eliminate the gap between the rise of new fraud and the system’s ability to screen for it. Screening every order—not just the ones that raise flags—is also required to build the most accurate and up-to-date dataset possible for the AI engine to use.
Is the chargeback protection comprehensive?
Some e-commerce platforms offer chargeback protection as part of their fraud-screening programs to spare merchants the cost of charged back orders and the related fees (which typically cost about $25 but can cost more, depending on the processor and their fee structure). This protection can be a great benefit, as long as you understand exactly what’s covered before you opt in and what the practical limitations are.
The first thing to clarify is whether the program covers all fraud-related chargeback costs or only those that aren’t associated with so-called friendly fraud. Friendly fraud occurs when an otherwise legitimate customer falsely claims that they never received an order or didn’t make the purchase so they can get a refund and keep the merchandise. Not every chargeback protection program covers these cases, but if they don’t you could be setting yourself up for more losses down the road. That’s because among customers who file a successful friendly fraud chargeback, 40% will do it again within 60 days.
The most effective anti-fraud programs are multi-layered and comprehensive enough to protect your business from false declines, lost customers, friendly fraud, and new types of fraud as well as known fraud methods. AI can be a powerful fraud-fighting tool, but it’s most effective when used as one tool among many. Knowing exactly what your e-commerce platform provides can help you decide whether to opt in and whether you’ll need additional fraud-prevention tools.
Rafael Lourenco is the Executive Vice President of ClearSale