1. What Square Actually Is
Square, by Block, is one account for in-person POS, online checkout, invoices, and payment links — with inventory and reporting synced across all of them. That matters because the risk engine sees the whole picture, not just one checkout. It is built for small and omnichannel sellers, not high-volume card-not-present operations.
2. The Flow — Card to Decision
Card presented → Square checkout → Machine-learning risk evaluation → Risk Manager rules → Approve / 3D Secure / Decline. Every online payment taken through Square, or through its Connect v2 Transactions API, gets evaluated. No evaluation shown usually means the models saw few fraud signals — not that no check happened.
3. What the Machine Learning Actually Sees
Square's own docs say it plainly: risk evaluations come from proprietary machine-learning models, informed by thousands of signals from its payments ecosystem, continuously evolving.
4. Risk Manager — The Rules You Can Actually Set
Risk Manager is free, and this is where a seller controls their own tolerance.
5. Why Payments Die on Square
6. What Actually Matters — Checklist
7. Square vs Stripe, In One Line
Stripe Radar is built for e-commerce scale with deep custom rules. Square is built for a shop that sells in person and online under one roof, with fraud screening bundled into the rate. Pick the tool that matches the volume and the channel — the risk engine follows the business model.
Bottom Line
Square does not kill payments randomly. It scores thousands of signals, flags Moderate or High, and hands the decision to the seller. The sellers who lose money are the ones who never set a rule, never read an evaluation, and accept everything that clears.
Square, by Block, is one account for in-person POS, online checkout, invoices, and payment links — with inventory and reporting synced across all of them. That matters because the risk engine sees the whole picture, not just one checkout. It is built for small and omnichannel sellers, not high-volume card-not-present operations.
2. The Flow — Card to Decision
Card presented → Square checkout → Machine-learning risk evaluation → Risk Manager rules → Approve / 3D Secure / Decline. Every online payment taken through Square, or through its Connect v2 Transactions API, gets evaluated. No evaluation shown usually means the models saw few fraud signals — not that no check happened.
3. What the Machine Learning Actually Sees
Square's own docs say it plainly: risk evaluations come from proprietary machine-learning models, informed by thousands of signals from its payments ecosystem, continuously evolving.
- Device and behaviour signals — new device, unusual location, rushed checkout
- Card signals — card origin, AVS mismatch, CVV results, card history across the network
- Velocity signals — too many attempts, too fast, too many cards on one account
- Order signals — amount, item type, shipping vs billing mismatch
4. Risk Manager — The Rules You Can Actually Set
Risk Manager is free, and this is where a seller controls their own tolerance.
- Alert rules — get emailed when a payment is Moderate or High risk
- Block rules — decline online payments that match conditions you set, like card origin, AVS mismatch, or payment velocity
- 3D Secure rules — trigger 3DS on risky payments. 3DS verifies the buyer and shifts fraud-dispute liability back to the card issuer, free of charge
5. Why Payments Die on Square
- AVS mismatch — billing address does not match the card
- Velocity — several cards, several attempts, short window
- New card, new device, new location — no history, high uncertainty
- Card-not-present at scale — Square's model is trained on small-business volume, so large CNP volume outgrows it
- A Moderate/High evaluation the seller ignored — then a chargeback lands anyway
6. What Actually Matters — Checklist
- Turn on Risk Manager alerts for Moderate and High risk
- Use 3DS on high-risk orders instead of blindly declining them
- Keep billing, shipping, and device history consistent
- Treat a first payment from a new card as a test, not a green light
- Remember the pricing underneath it all: in-person 2.6% + $0.15, online 2.9% + $0.30 on paid plans, 3.3% + $0.30 on the free plan
7. Square vs Stripe, In One Line
Stripe Radar is built for e-commerce scale with deep custom rules. Square is built for a shop that sells in person and online under one roof, with fraud screening bundled into the rate. Pick the tool that matches the volume and the channel — the risk engine follows the business model.
Bottom Line
Square does not kill payments randomly. It scores thousands of signals, flags Moderate or High, and hands the decision to the seller. The sellers who lose money are the ones who never set a rule, never read an evaluation, and accept everything that clears.




