These Main Order Variables Indicate Fraud-But How Can We Tell Fraud From Good Customer Behavior?

Fighting fraud while keeping good customers happy is one of the key determinants of ecommerce success. Allow too much fraud and your bank may close your merchant account. Turn away too many good orders and your customer acquisition costs will skyrocket. Technology—especially AI and data analytics—have made it easier to find the right balance. However, whether it’s a person or an algorithm looking at the orders, they’re assessing order variables for clues—and the nature of those clues is changing.

Here’s a look at which variables are the most useful fraud indicators, and why the way merchants look at them may need to evolve.

What are main order variables and why do they matter?

In an order, variables are any pieces of data that can change from one order to the next, such as the items purchased, the day of the week the purchase is made, the type of device the order was placed from, and so on. The main order variables—the ones that we find most useful for indicating potential fraud—are:

Order amount. In general, higher-value orders are more likely to be fraudulent than lower-value orders. That’s because fraudsters are shopping with stolen payment data, so they’re not worried about budget constraints, and because they often target high-end items or large lots of items for resale.

Product category. Designer sneakers, smartphones and notebook computers are popular targets for fraudsters because these items are easy to resell in the underground economy.

Delivery address. The product destination is the only physical link that ecommerce orders have to the buyer. Because fraudsters are usually part of an organized crime ring, they tend to ship orders to collection points near ports and other distribution infrastructure. As a result, destinations in some postal codes may be riskier than others.

What’s tricky about evaluating main order variables?

At first glance, it seems like the simplest solution is to err on the side of caution with these variables, perhaps automatically rejecting orders that surpass a certain value threshold, contain too many sought-after items, or ship to specific locations. However, that approach was not a good idea even before the pandemic, because automatically rejecting orders risks damaging a merchant’s relationship with good customers.

For example, in 2018, 30% of all consumers said they’d had an online order declined, and that percentage doubled for shoppers with income of $1 million or more. Why? Wealthy consumers may spend more per order, buy sought-after items like designer sneakers and bags, and ship purchases to the hotel, resort or secondary home where they’re staying.

Now, consumers of all income levels have changed their behavior in response to covid-19, lockdowns and economic uncertainty. They’re more likely to buy in bulk when they find items they need, due to concerns about shortages and the supply chain. Many people who kept their jobs and started working from home have saved enough money on commuting and office expenses to indulge in trendy items. And many people have moved, temporarily or permanently, since the start of lockdowns.

According to data we collected from more than 1,000 online shoppers in each of five countries (US, UK, Mexico, Australia and Canada) for our State of Consumers Attitudes, Fraud & CX 2021 Survey, 78% of shoppers spent more online and/or shopped more often online than before the pandemic, and 53% now buy products in categories they hadn’t before March 2020. In US and Mexico, 17% made their first online purchase ever after the pandemic began.

While consumer behavior was changing in ways that muddied main order variable data, fraudsters increased their attacks on ecommerce merchants. In 2021, account takeover fraud, synthetic identity fraud and third-party fraud increased compared to the year before—and 2020 was a big year for ecommerce fraud. So, the likelihood of fraud is higher now than ever.

At the same time, many consumers have run out of patience for bad experiences. Forty percent of our survey respondents won’t shop again with a merchant that declines their order, and 34% will complain about the merchant on social media. Merchants have to balance the risk of fraud losses against the risk of customer churn and brand damage—and they have to do so while adapting the way they assess main order variables.

Best practices now for assessing main order profile variables

One way to assess main order variables accurately is to screen all orders against a database that holds information, like:

• How many times people have ordered deliveries to the shipping address they included. A high number of orders to one address can be a fraud flag, especially if they come from different customers.
• How many times the customer placing the order has filed a chargeback. A higher number can be an indicator of friendly fraud.
• Whether their orders have been approved before. This can help merchants avoid declining loyal customers.
• Whether multiple shoppers are using the same address, phone number and email address. This might indicate fraud—or a household where an adult is helping teens or elders shop online.

This kind of database should be continuously updated with fresh information, because static data degrades too quickly to be useful for very long. In addition, the database and order variable assessment shouldn’t lead to automatic order declines. To reduce the risk of rejecting good customers, merchants should manually review all flagged orders and then feed the results of those reviews into their fraud control machine-learning, so the AI gets better at sorting good orders from fraud.

We’ve seen a tremendous amount of change in consumer behavior, ecommerce growth, and ecommerce fraud over the past year and a half. With good data, smart screening methods and multiple layers of fraud control, ecommerce merchants can avoid fraud and retain their good customers, even in times of major change.

Original Article: https://www.merchantfraudjournal.com/main-order-variables-indicate-fraud/