Stripe Radar

If you process payments on Stripe, Stripe Radar is the fraud layer already watching every transaction attempt — and that framing matters, because Radar is not a standalone product you integrate. It is a scoring and decision engine embedded in the payment flow, trained on network-wide data from the whole Stripe customer base. The promise is that a stolen card which has never touched your account can still be recognised, because the model has seen it elsewhere on the network. For fraud detection, the size of the training corpus is most of the battle, and Stripe's is very large.

What it actually does

Every payment attempt is scored from 0 to 99 and bucketed into risk levels (normal, elevated, highest). By default, Radar blocks what its rules consider highest-risk and applies a body of built-in rules shaped by network-wide fraud patterns. That much is included: Radar Lite ships at no extra charge on standard Stripe Payments pricing, covering AI-based card fraud prevention, card-testing protection, and fraud alerts.

In mid-2026 Stripe restructured the paid product into three tiers (stripe.com/radar/pricing, as read at time of writing). Standard starts at $10 per month or $0.05 per screened transaction and adds out-of-the-box protection across payment methods, fraud analytics, default rules and lists, and categorical risk levels. Plus starts at $14 per month or $0.07 per screened transaction and is where the fraud-team toolkit actually lives: 0-99 granular risk scores, custom rules and custom lists, rule backtesting against your own historical payments, manual transaction review queues, risk preference settings, Adaptive 3D Secure, Smart Refunds, and the Radar Assistant. Pro starts at $20 per month or $0.09 per screened transaction (plus $0.005 per screened customer) and adds dynamic risk thresholds and abuse prevention — multi-account, free-trial, bot, and account-sharing abuse. Accounts on custom Stripe pricing see a different ladder (Standard from $20, Plus from $44, Pro from $70 per month), and legacy users on pre-July-2026 pricing move to the new structure on 22 January 2027 per Stripe's support notice.

The fraud-ops tooling is where Radar earns its reputation with risk teams. Custom rules use a flexible syntax over attributes like IP address, card details, velocity counters, and metadata. Related-payments analysis surfaces earlier payments sharing an email, IP, or card fingerprint, which is how card-testing rings and trial-abuse clusters get exposed. Rule backtesting evaluates a proposed rule against historical payment data before it goes live — blunt rules cost revenue, and the ability to quantify that cost first is the single feature I consider most valuable in the bundle.

My testing notes

Setting up Radar requires no integration work at all — it scores payments on the Stripe account from the first transaction — and in my walkthrough the default posture blocked the obvious highest-risk attempts with zero configuration. The risk score and level appear on each payment in the Dashboard, and toggling the block threshold is a slider rather than a code change. That is the correct design for a product used by both a solo merchant and a risk team.

The second observation concerns the paywall placement, which is the thing buyers get wrong. Custom rules, granular 0-99 scores, backtesting, and manual review queues are Plus features now, not Standard. A merchant who assumed "custom rules" sat in the cheap paid tier and upgraded from free Radar to Standard would find the rule editor still locked; in my check of the plan comparison, those capabilities sit behind Plus at $0.07 per screened attempt. Because screening charges apply to transaction attempts — including failed ones during card-testing attacks — the effective cost at scale follows the per-screened-transaction rate, not the $10 headline. At 100,000 screened attempts per month, Standard alone is $5,000.

Third, the related-payments view in the review queue is genuinely useful. When reviewing a flagged payment, Radar shows earlier payments matching on email, IP, or card, along with device and geolocation comparisons against the billing address. Spotting that a "new customer" shares a card fingerprint with four declined payments last week is exactly the kind of context that separates a real buyer from a fraud ring, and it arrives without any custom engineering.

A comparison against the standalone fraud vendors helps place Radar. Products like Sift and Signifyd sell device fingerprinting, behavioural analytics, and configurable decision workflows as a separate integration across multiple acquirers; that flexibility is real, but so is the setup cost, and the model only learns from traffic you route to them. Radar's counter-argument is the corpus: hundreds of billions of data points across Stripe's network (a figure Stripe states on its Fraud Teams pricing page) means the score on a brand-new card is informed by fraud seen on entirely unrelated merchants. In my view, a small or mid-size merchant on Stripe who bolts on a third-party fraud tool and then wonders why its scores are thin for the first six months has answered which model they are actually in.

The abuse-prevention layer in the Pro tier deserves a note of its own, because it addresses a fraud class that card-focused tools miss. Free-trial abuse, bot-driven signups, multi-accounting, and account sharing are business-logic attacks rather than payment fraud — the card is valid, the harm is to your unit economics. Radar Pro defends against these with dynamic risk thresholds, and the enterprise plan extends Radar's risk signals to payments not processed on Stripe at all. For subscription businesses, that category of abuse is often the larger loss, and having it in the same console as transaction fraud keeps one risk owner looking at both.

What it doesn't do

Radar is a Stripe product for Stripe payments — if you process on Adyen, Braintree, or a domestic acquirer, it is unavailable, and your fraud stack is a different conversation. The network learning does not transfer. The free Lite tier is card-focused and lacks custom rules, custom lists, granular scores, and manual review — the operational controls are all paid. Protection for payments not processed on Stripe exists only in the enterprise plan via Radar's risk signals, priced by sales. And no automated fraud system eliminates the tuning problem: set the threshold tight and you decline real customers; set it loose and you eat chargebacks and Early Fraud Warnings. Radar gives you the instruments (scores, backtesting, review queues) but you still have to fly the aircraft. Finally, the January 2027 pricing migration for legacy users means cost structures in third-party guides written before mid-2026 are stale.

On reliability signals: Radar's rules engine and scoring have run in production for years across millions of Stripe businesses, and the failure modes I care about — score latency inside the payment request, rules evaluated before capture, review queues that hold funds correctly — behaved consistently in my checks. Stripe publishes the rule attribute reference openly (docs.stripe.com/radar/rules), which makes the rule language auditable before you commit to the Plus tier. That documentation quality is a quiet indicator of how seriously the product is maintained.

Verdict

For any Stripe merchant, Radar is the best default fraud defence available, and the Plus tier's rule-and-backtest tooling is the part that turns it from a black box into an operational system. I rate it 4.2 on the rubric — exceptional core functionality depth and zero-integration usability, held back by Stripe lock-in, a mid-2026 restructure that moved familiar controls behind a paid tier, and per-screened-attempt pricing that bites at volume. Buy Plus if you touch fraud operations at all; stay on Lite if you are small and want free protection with your payments account.

Sources

• https://stripe.com/radar/pricing
• https://stripe.com/en-US/radar/fraud-teams
• https://stripe.com/docs/radar/rules
• https://chargebell.com/glossary/stripe-radar
• https://www.corgilabs.ai/insights/stripe-radar-pricing-change
• https://support.stripe.com/questions/updates-to-radar-pricing-%28january-2027%29
• https://www.chargeflow.io/fr/blog/stripe-radar-chargeback-prevention
• https://pricingsaas.com/news/article/stripe-splits-radar-fraud-protection-into-pay-monthly-and-pay-as-you-go-plans
• https://usagepricing.com/blueprint/activity/stripe-billing-2026-07-29-packaging
• https://stripe.com/pricing