Using Customer Data To Help Prevent False Declines

As more commerce moves online, customers expect merchants to recognize them. McKinsey found that in November, 71 percent of customers expect personalized interactions with companies, and 76 percent “get frustrated when this does not happen.” Personalization typically involves marketing messages, email campaigns, and product suggestions, but what could be more personal — and more crucial to a personalized shopping experience — than recognizing good customers when they place an order?

Unfortunately, even merchants that invest in personalized marketing sometimes trip on this final step before the sale. Instead of recognizing a good order, their fraud screening tools flag it as possible fraud and the order gets declined. According to more than 5,000 adults who took part in the most recent ClearSale State of Consumer Attitudes, Fraud and CX report, these false positives, or false declines, happened more often since the start of the pandemic.

Rising Rates of False Declines in Ecommerce

Fifteen percent of survey respondents in the U.S., Canada, Mexico, the U.K., and Australia said they experienced online payment fraud between March 2020 and March 2021. However, 25 percent said they had an order declined while trying to buy online, and almost half of those consumers said they experienced more declines during the survey period than in 2019.

Gen Z and Millennial shoppers were far more likely to be declined by mistake: 48 percent of 18- to 24-year-olds and 34 percent of 25- to 39-year-olds reported at least one decline during the survey period. Among Gen X and baby boomer respondents, only 16 percent experienced a decline. Why was there a generational difference in decline rates? It could be that older consumers have more historical data to draw on, or it could be that they are less likely to make purchases in categories with higher fraud risks, like online games and event tickets. 

Whatever the reason, many customers are not forgiving when these declines happen. Sixty-five percent of all online shoppers in the survey said that if a merchant declined their order and then asked for more information to approve it, they would not provide it. In other words, if the merchant does not recognize them right away, most customers will simply opt out. In the same survey, almost half said they will never come back to a merchant after a decline. That represents a loss of the customer’s lifetime value and of the marketing budget that the merchant spent to bring them to the store in the first place. 

Understanding Why False Declines Happen

With the stakes for customer experience and retention so high, it is worth understanding why false declines happen even though there is so much data available to screen orders for fraud. There are two key issues: the type of data that a merchant’s fraud control program uses, and what happens after that data is used to screen an order.

Let us look at the type of data first. Many ecommerce platforms have built-in fraud screening tools that apply simple rules to orders and flag those that do not comply. For example, a simple fraud tool might run an order through AVS (Address Verification Service) and check the card number and CVV code. If one digit in the ZIP code is off, or if the customer is ordering from a new billing address because they just moved, the program’s rules might flag the order as potential fraud. 

Of course, flagging an order or assigning it a higher-than-average risk score is not the same thing as declining the order. This is where the actions based on those results matter. If a merchant has their fraud control program set to automatically reject orders above a certain risk threshold, or to reject any order with fraud flags, they are going to generate some false declines, based on scenarios like the ones we just considered. Each decline carries the risk of customer churn. 

Using Customer Data for Better Fraud Screening & CX

To avoid turning down good orders while stopping fraud, merchants can change the data they use and the way they use it. The most important step is to access real-time and historical customer data, including the customer’s history with the merchant and the customer’s wider history of online interactions along with markers like device data, behavioral biometrics, and geolocation.

With this data, a one-digit address mismatch might still raise a small flag, but that flag might be mitigated by other data showing that the customer recently changed their address. The first-time online shopper’s historical data might show that they are using an email address they have had for 10 years and a delivery address they have lived at for 20, making it less likely that they are a fraudster. 

To minimize the risk of false positives, every order with a risk score above a certain threshold set by the merchant should be manually analyzed before decisioning. This is because human expertise still matters in “edge cases” that the fraud algorithm cannot decide. Manual review also allows the merchant to feed the results of those reviews into the machine learning program so that their fraud program gets better at identifying good and bad customer behavior over time. 

Recognizing customers at checkout, when they have committed their time and their money to your store, is the key to building and maintaining lasting relationships — and to avoiding customer churn. By working with more real-time and historical customer data, and manually reviewing suspicious orders, you can approve more orders and keep those customers coming back to your store.


Original article at: