Risk Score

A risk score is a numerical estimate of how likely a given payment is to be fraudulent. Fraud systems like Stripe Radar assign one to every transaction so you can decide — automatically or by rule — whether to allow it, hold it for review, or block it.

How the score is calculated

Stripe Radar produces a score from 0 to 99, where 0 is the lowest risk and 99 the highest. The model weighs hundreds of factors about each payment — card and device details, location, spending patterns, and signals learned across Stripe's whole network — to reach that number.

Each score maps to a risk level rather than a hard yes/no:

• Normal — the large majority of legitimate payments.
• Elevated — some risk factors present; a common threshold for manual review.
• Highest — strong fraud signals; blocked by default.

By default on Stripe, a score of about 65 and up is treated as elevated and 75 and up as high risk, though teams using Radar for Fraud Teams can set their own thresholds.

Why it matters

A risk score turns a judgment call into a lever you can tune. Set it too permissive and fraudulent charges become chargebacks and Early Fraud Warnings; set it too strict and you decline paying customers. The score is also useful context after the fact — a disputed charge that carried a high score at checkout is a very different story than one that scored clean.

Using scores in practice

• Route elevated-risk payments to a review queue instead of blocking outright.
• Watch for clusters of high-score, low-value attempts — a classic sign of card testing.
• Combine the score with your own rules; the model is a starting point, not the whole policy.