Scores

Overview

Every transaction processed through your account receives a risk score between 0 and 1000 — generated in real time by PayMongo's machine learning engine. The Scores section gives you a full view of scoring results and risk assessments across your transaction volume.

Use Scores to understand how your transactions are distributed across risk levels, spot emerging fraud patterns, and calibrate your rules accordingly.

How scores are generated

The machine learning engine is trained on historical transaction data and evaluates each transaction using multiple signals:

• Behavioral patterns — transaction frequency, changes in user behavior
• Velocity checks — rapid transaction attempts, repeated failed payments
• Anomaly detection — outlier transaction amounts, unusual geographic patterns
• Device and identity signals — device changes, mismatched address details, high-risk email domains

The engine produces a score and a set of risk drivers — the specific factors that most influenced the score for that transaction. These are visible in the dashboard for every scored transaction.

Risk levels

You can override this behavior by creating custom rules based on specific score thresholds.

Using scores to calibrate rules

The Scores view helps you identify patterns that should inform your rule configuration:

• If you see a cluster of high-scoring transactions from a specific card country, consider adding a targeted block or review rule
• If your review queue is dominated by medium-risk transactions that turn out to be legitimate, consider raising your review threshold or adding an allow rule for low-risk segments
• If you see transactions with unexpectedly high scores and unusual decision reasons, investigate them individually — they may signal a new fraud pattern

Best practices

• Monitor regularly to stay current on the overall risk profile of your payment traffic
• Use score distributions to tune thresholds — don't rely solely on the default medium/high split
• Investigate anomalies — unusually high scores or unexpected risk drivers often indicate new fraud tactics worth addressing proactively