Fraud Intelligence
Transaction fraud patterns, behavioral signal methodology, chargeback benchmarks, and risk ops guides — written by the team that builds the scoring model, for the risk ops and engineering teams that run it.
All articles
Velocity Checks in Fraud Detection: Beyond the Basics
Simple transaction velocity is table stakes. We walk through composite velocity features — cross-device, cross-merchant, cross-BIN — that catch what basic rate limits miss.
Rules Engine vs. ML for Fraud Detection: The Honest Comparison
Rules are auditable and fast to deploy. ML catches what rules cannot. We make the case for when you need both — and when you can retire the rules.
BNPL Fraud: Why Buy-Now-Pay-Later Has a Unique Risk Profile
BNPL products let fraud slip through at the application stage. We look at the signal patterns that distinguish synthetic BNPL applicants from real ones.
Chargeback Ratio Benchmarks by Industry: What Is Normal?
Chargeback rates above 1% trigger card network monitoring programs. We publish current benchmarks across e-commerce, travel, and BNPL verticals.
Building a Fraud Feedback Loop That Actually Improves Your Model
Outcome labeling is only valuable if it reaches the model. We describe the feedback API design pattern that closes the loop between chargebacks and signal weights.
Fraud Liability Shift for PSPs: What the 2025 Rule Changes Mean
New Visa and Mastercard liability frameworks put more fraud cost on PSPs who lack real-time scoring. We break down the thresholds and what risk ops should prepare.
Card Testing Attacks: Detection Patterns That Work
Card testing attacks use hundreds of small-value probes before the real hit. Velocity clustering and merchant pattern analysis are the two levers that work.
What It Actually Takes to Score Fraud in Under 50ms
Latency budgets for fraud scoring are brutal. Here is how we keep median decision time at 47ms without compromising signal depth.
Why Your Fraud Model Has a False Positive Problem (and How to Fix It)
A 2% false positive rate sounds low until you run 5 million transactions a month. We walk through threshold calibration and feedback loop design.
Account Takeover Patterns in 2025: What Changed
Credential stuffing kits now rotate device fingerprints. We document the behavioral drift signatures that still surface ATO even when hardware looks clean.
The 12 Behavioral Signals That Predict Transaction Fraud
Not all signals are equal. We break down the twelve behavioral features with the highest fraud predictive value in our 140+ signal set.
How Synthetic Identity Fraud Gets Past Identity Verification
Synthetic identities combine real SSN fragments with fabricated names and addresses. Here is why rule-based checks miss them and what behavioral signal clusters catch instead.