Radar catches network fraud; Prevent catches chargeback fraud. Radar's model is trained on Stripe's cross-merchant view of card-testing, stolen-card use, and true fraud rings. Prevent's model is trained on Aurai's book of chargeback outcomes — friendly fraud, dispute-prone customer patterns, and reason-code-specific risk. Merchants should run both. Radar blocks obvious fraud upstream; Prevent flags high-chargeback-risk orders for review.
What Stripe Radar catches
Stripe Radar is Stripe's built-in fraud detection layer. It scores every incoming Stripe transaction 1–100 based on:
- Cross-merchant patterns visible to Stripe (device use across sites, velocity across merchants)
- Card BIN risk (issuer, country, product type)
- AVS/CVV mismatch signals
- Rules the merchant configures
Radar excels at stolen-card fraud and card-testing — patterns where a bad actor uses the same card across many merchants. Because Stripe sees the entire ecosystem, its signal on this category is unmatched.
Radar is less effective at [friendly fraud](/blog/friendly-fraud-65-percent-problem) — where a legitimate cardholder disputes a legitimate charge. There's no signal at authorization that a specific customer will file a dispute; the transaction looks fine.
What Aurai Prevent catches
Aurai Prevent is Aurai's transaction-level chargeback risk model. It scores incoming orders based on:
- Aurai's book of chargeback outcomes
- Reason-code-specific risk signals
- Customer-history signals (prior chargebacks, dispute patterns, cancellation history)
- Merchant-vertical risk (SaaS trial-to-paid has different signals than physical goods)
Prevent excels at friendly fraud — the category Radar misses. Prevent's signal isn't "is this fraud" but "is this likely to produce a chargeback." Different question, different model.
Where they overlap
Meaningful but not total.
Both catch: Very-high-risk transactions with obvious fraud signals (mismatched geography, high-risk BINs, extreme velocity). Radar for card-testing patterns; Prevent for chargeback risk.
Only Radar catches: Cross-merchant card-testing. Aurai has no visibility into other merchants' books.
Only Prevent catches: Chargeback-history patterns. Radar optimizes for fraud broadly, not chargeback outcomes specifically.
At Aurai's book of Stripe merchants, running both catches ~30% more chargebacks than only Radar and ~15% more than only Prevent.
Running both
The right stack for a Stripe merchant:
- Radar as the gate. Any transaction Radar scores above block threshold gets blocked. Handles card-testing and true fraud at payment level.
- Prevent as the review layer. Transactions clearing Radar but Prevent-flagged get reviewed. Depending on ops, that's holding fulfillment, requiring 3DS on high-risk orders, or auto-refunding above a threshold.
- Alerts as safety net. For orders clearing both models but disputed anyway, Ethoca/Verifi alerts let you refund pre-chargeback.
- Representment as last resort. Chargebacks that get filed anyway → Aurai's representment engine handles them.
Each layer catches what the previous missed.
Frequently asked questions
Is Stripe Radar enough by itself?
For preventing true fraud (stolen cards, card-testing), Radar is excellent. For preventing chargebacks — especially friendly fraud — Radar misses categories that chargeback-specific models catch. Most merchants get real value from running both.
How is Aurai Prevent different from Radar?
Radar's training signal is cross-merchant fraud. Prevent's is chargeback outcomes across Aurai's book. Different metric, different model.
Does running both models slow down checkout?
No. Both score in milliseconds and run in parallel. Neither adds meaningful latency.
Post reflects public documentation, industry surveys, and Aurai's own book of disputes as of the publish date. Visa, Mastercard, American Express, Discover, Stripe, PayPal, and Shopify are trademarks of their respective owners; Aurai is independent and not endorsed by any of them. Network rules change — always verify current official rules before acting on any specific tactic.