Stripe Radar

Stripe Radar is Stripe's built-in fraud-detection system. It uses machine learning trained on payments across Stripe's network to evaluate each transaction and predict how likely it is to be fraudulent, then blocks, flags for review, or allows the payment based on that assessment.

How Stripe Radar works

For every payment, Radar considers hundreds of signals — card details, device and location data, past behavior, and patterns seen across millions of businesses on Stripe — and produces a risk score from 0 to 99. Each score maps to a risk level: normal, elevated, or highest.

By default, Radar's built-in rules block payments Stripe considers highest-risk. Because the model learns from network-wide fraud, it can recognize a stolen card or a bad actor even if that card has never touched your account before.

Radar for Fraud Teams

The upgraded tier, Radar for Fraud Teams, adds controls for teams that want to tune fraud handling themselves:

• Custom rules — write your own allow, block, and review logic (for example, review any payment over a threshold from a new customer).
• Manual review queues — inspect flagged payments before capture.
• Deeper insights — analytics on why payments were flagged.

Radar is priced per transaction, and pricing differs between the base machine-learning protection and Radar for Fraud Teams; the exact per-transaction cost depends on your Stripe pricing plan.

Why it matters

Fraud that slips through becomes chargebacks and Early Fraud Warnings — each carrying fees and threatening your standing with card networks. Radar catches patterns like card testing by evaluating payments together rather than in isolation, stopping many attacks before they cost you. Tuning it too tight, though, blocks good customers — so most teams balance fraud loss against declined legitimate sales.