Skip to main content
This guide shows how to build risk scoring models that combine multiple factors into an overall risk assessment.

​
What you’ll build

A risk scoring decision that:
  • Evaluates multiple risk factors independently
  • Weights factors based on importance
  • Produces a composite risk score
  • Classifies into risk categories
Want to skip ahead? Download the completed decision and import it directly.

​
Decision flow

​
Example: Fraud risk scoring

​
Input data

{
  "transaction": {
    "amount": 1500,
    "currency": "USD",
    "country": "NG",
    "isFirstPurchase": true
  },
  "customer": {
    "accountAgeDays": 5,
    "verificationLevel": "email_only",
    "previousTransactions": 0
  },
  "device": {
    "isKnown": false,
    "vpnDetected": true,
    "riskScore": 75
  }
}

​
Step 1: Transaction risk factors

Score transaction-level signals:
Inputs
Outputs

​
Step 2: Customer risk factors

Evaluate customer signals:
Inputs
Outputs

​
Step 3: Device risk factors

Inputs
Outputs

​
Step 4: Calculate composite score

Combine all factors with weights:

​
Step 5: Risk classification

Map score to action:
Inputs
Outputs

​
Output structure

{
  "riskScore": 85,
  "riskLevel": "critical",
  "action": "block",
  "factors": {
    "transaction": 45,
    "customer": 65,
    "device": 50
  },
  "signals": [
    "High-risk country",
    "New account",
    "Unknown device",
    "VPN detected"
  ]
}

​
Variations

​
Credit risk scoring

Inputs
Outputs

​
Best practices

Document factor weights — Make it clear why each factor has its weight. Normalize scores — Use consistent 0-100 scales across factors. Include signal details — Output which factors contributed to the score. Tune thresholds — Adjust classification thresholds based on false positive rates.