Radar updates sharpen fraud detection and review workflows
Ecommerce fraud is set to cost businesses $20 billion in 2021, an 18% rise over the prior year. Fraud techniques have evolved in parallel, with bad actors relying on increasingly complex strategies. Stripe has responded with a batch of updates to Radar, its machine-learning-based fraud detection product, aimed at improving detection accuracy and making fraud management less manual.
Faster model refreshes and better recall
Radar's machine learning models are now updated three times faster than before, letting the system adapt more quickly to shifting fraud patterns. As Stripe's network grows, it sees more of the key data signals across multiple payment attempts, which helps catch fraudulent usage earlier. Improved data volume has also boosted model recall—a standard accuracy metric—by more than 20%, meaning Radar catches more fraudulent attempts while keeping disruption to legitimate transactions low.
The updated machine learning models have made it even easier to manage our fraud rates—without any additional effort, we saw our fraud rates drop by half in just a few months.
A single workspace for fraud operations
Managing fraud often means juggling manual payment reviews, dispute responses, rule creation, and risk controls. The redesigned Stripe Dashboard consolidates these into one fraud workspace where teams can access reviews, disputes, rules, lists, and risk settings. A new overview chart shows how machine learning, custom rules, and manual reviews work together to block or allow each payment attempt.
Benchmarks for fraud strategy
Knowing whether your fraud strategy is effective is hard in isolation. Radar now displays relevant benchmarks in the Dashboard, letting businesses compare their fraudulent dispute rates, false positive rates, and block rates against aggregated cohorts of similar companies in the same region or industry. These comparisons can inform whether adjustments to your approach are worth making.
Expanded rules for granular control
Radar for Fraud Teams now supports dozens of additional rule attributes, enabling more precise customization of fraud setups. New rule options include the number of names associated with a card, the average amount of attempted authorizations on a card, and the time since a customer email or card was first seen by your business or anywhere on Stripe.
Per-rule performance analytics
Custom rules are only useful if you know how they're performing. Radar for Fraud Teams now provides deeper insights into each rule's impact, including the number of payments blocked and estimated false positives generated. These metrics let you identify which rules are most effective and refine your rule sets to maximize fraud blocking while minimizing effects on legitimate customers.
By leveraging Stripe’s machine learning and custom fraud rules, we’ve seen a decrease in our fraud rates and chargeback rates. We’ve even found our customer satisfaction has improved because our team settles customer disputes faster now that they spend less time managing a manual process.
Additional company testimonials echo these results, with fraud rates trending down and satisfaction improving as manual dispute management shrinks. Radar for Fraud Teams is available for a free 30-day trial, and current Stripe users can explore the new features directly from the Dashboard.
