How a smarter approach to identity can lead to real relationships

It’s that time of year for celebrating love and relationships, including those with your ecommerce customers. But one thing that can easily sour those relationships is falsely declining customers, which can damage both your profitability and their experience shopping with your brand.

When it comes to policy abuse, we can learn a lot from the world of online dating, where swiping right or left can mean the difference between happily ever after and a date with a dud. The real heartbreak, though, is missing out on a perfect match. For merchants, this translates to alienating valuable customers when a fraud prevention tool mistakenly flags them.

Managing increasing order approval rates and policy abuse requires understanding the true identity underneath the surface. But with naive linking, swiping wrong is all too easy—and it can be costly.

“Valentine’s Day, like Mother’s Day, is a holiday in which the bulk of sales occur week-of. But it’s also become a meaningful economic event: Valentine’s Day spending in 2024 was more than $25 billion, with jewelry sales accounting for 25 percent of that total. (The average American consumer spends just under $200 on the holiday.) It has also become a bigger online shopping holiday, with 40 percent of shoppers making their purchases on the web — up 20 percent from the year before.”

Sarah Shapiro
Retail Correspondent at Puck’s Line Sheet

Love at first sight? Not so fast. The problem with rules-based data linking

Fraud strategies that rely solely on linking oversimplify the task of capturing true customer identity and why they might match with others. Complex customers (especially savvy abusers or professional fraudsters) can easily bypass rules. All it takes is to subtly change an address (to another floor in the same building, for example) or create dozens of accounts under different names and emails to skirt a merchant rule that would otherwise link accounts.

Don’t misjudge a gem: The problem with overlinking

Conversely, rules-based linking can also lead to overlinking the connections between orders, leading merchants to interpret multiple accounts as one identity. 

Multiple customers who order from the same pickup point, for example, can be mislinked as a single identity. Or a child using a parent’s device to place an order might register the same cookie. These scenarios can cause merchants to erroneously associate good customers with abusers and thus lose out on sales.

Just like you can’t find true love with just a checklist of traits, merchants need much more than just linking rules to accurately determine good customers from bad actors.

Sounds like a task for machine learning!

Finding true love with true identity

To make the right decision in every interaction, merchants need to get to know their customers even better. This requires a broad network of data of customer orders and claims across accounts and merchants, advanced machine learning models, and a clustering strategy for identity resolution and risk evaluation.

Networked data

Tapping into a wide merchant network allows broad visibility into client behaviors across merchants to provide a full understanding of underlying identity and multiple shared traits. 

Riskified identity-based clustering and machine learning capabilities leverage our vast global merchant network to cross-reference account activities against billions of transactions across industries. This enables the detection of deeper patterns and relationships among accounts based on their behavior and activity across the network.

This extensive data pool enhances Riskified’s ability to detect unique and potentially abusive patterns that may not be visible within a single merchant’s data.

Deeper insights 

In addition, Riskified’s risk intelligence platform analyzes a comprehensive set of aggregated metrics, such as total spend, claims, and chargebacks, to create robust identity clusters that reveal real relationship signals between accounts across the network.

If an abusive customer creates multiple accounts to exploit promotional offers or submit serial item-not-received claims, for example, Riskified is able to cluster those accounts based on similar behaviors beyond the usual linking characteristics. The model can then more accurately identify those controlled by a single individual or organization.

Adjust friction to attract good customers and ghost the abusers

Using networked data and by going beyond linking, merchants can unlock a more accurate picture of the identity behind each order to ensure frictionless experiences for good customers.

Merchants can also apply tailored pressure based on the customer’s risk profile, requesting additional verification such as the credit card CVV or simply blocking them at checkout to prevent fraud or future abuse.

Happily ever after with your best customers?

With Riskified, merchants have the last word. They have visibility into the profitability of each identity in one place and the flexibility to calibrate and test their strategies according to their customers, business priorities, and appetite for risk.

Learn how activewear retailer Lorna Jane has avoided insults and shown love to their most valuable customers.