Risk Scores

Machine Learning Engine for Fraud Detection

Paymongo Fraud Detector (FD) is powered by an advanced machine learning engine, trained on extensive historical transaction data validated by internal Paymongo fraud teams. This engine evaluates each transaction in real-time, using a dynamic risk model that incorporates multiple signals, including:

• Behavioral patterns (e.g., transaction frequency, changes in user behavior)
• Device signals (e.g., device fingerprints, browser information)
• Velocity checks (e.g., rapid transaction attempts, repeated failed payments)
• Anomaly detection (e.g., outlier transaction amounts, unusual geographic patterns)

These signals are processed to generate an accurate fraud risk score for every transaction. The risk score informs both automated rules and manual review processes, helping your team act quickly and confidently.

Risk Score Ratings and Feature Transparency

FD not only produces a risk score, but also provides clear explanations for each score. The dashboard reveals which features contributed most to a transaction’s risk—such as device changes, mismatched address details, or high transaction amounts—offering transparency and supporting informed decision-making.

Default Risk Levels

Paymongo FD assigns risk levels based on the transaction’s risk score:

• Medium Risk: Score of 500 to 800
• High Risk: Score of 800 and above

Transactions with medium or high risk are typically flagged for review or blocking, depending on your configured rules.

For details on how to configure automated actions based on risk scores, see the Rules documentation.