Improving reviews in Radar

Matt Duval Corporate Card

When we launched Stripe Radar to help you prevent payment fraud, we built in the ability to manually review suspicious payments. Suspicious payments are flagged for review either by Radar’s machine learning systems or when they trigger a custom rule set by your business.

Since fraud generally increases during the holiday season, we’re launching several new features today to provide more signals and contextual data to improve the review flow:

Use behavioral patterns to identify fraud

Radar’s machine learning systems use hundreds of signals and heuristics to determine whether a payment might be fraudulent. When manually reviewing suspicious payments, we’ve found that behavioral signals provide useful data, so we’re adding some of these signals to the review page.

When reviewing payments, you’ll now see the operating system and device type used to make the purchase, the number of pages viewed before a purchase, and the purchase session duration. (Fraudulent purchases will often have suspiciously quick session lengths compared to a typical session for your business.)


Additionally, to help you quickly identify discrepancies, we now also display the distance between the location of the IP address used to make the purchase and the address associated with the credit card.

See related payments when reviewing charges

A particularly useful fraud detection mechanism is to cross-reference related payments, so you’ll now see related payments inline when reviewing a charge. Stripe’s machine learning systems already look for anomalies and suspicious patterns across billions of dollars of payments data. We’re now augmenting this by directly surfacing related payments made on your account for additional human analysis.

Radar will automatically surface payments that share the same IP address, card, or customer as the payment you are currently reviewing. This should help catch common fraud patterns such as naive card testing (where many different cards would share a single IP address) or trial abuse (where many “customers” would share the same card):


These data points reveal fraud patterns that might be otherwise tricky to detect. Is the user purchasing goods on a device that is nowhere near the billing address? Did they view only one page before making a $2500 purchase?

During our beta for these features, the team at Watsi (a healthcare nonprofit) reported that it took significantly less time to review payments than before and that it was easier to accurately identify fraudulent donations.

If you’re interested in getting started or want to learn more about how to enable these signals for your business, check out our integration guide and best practices. To take full advantage of the functionality, you just need to include Stripe.js across your site (not just the checkout page) and pass relevant customer information—like name, email address, and billing and shipping addresses—to Stripe using the customer object. We’ll then automatically handle the rest.

We hope this update is useful for your business during the upcoming holiday season. If you have any questions or feedback, please let us know!

Like this post? Join our team.

Stripe builds financial tools and economic infrastructure for the internet.

Have any feedback or questions?

We’d love to hear from you.