What Is a Risk Score? How Fraud Risk Scoring Works

A fraud risk score is a numerical value — typically between 0 and 1 — that represents the probability a user or transaction is fraudulent. A score of 0.0 means no risk detected. A score of 1.0 means the highest possible risk. Every signal available (email validity, IP reputation, geolocation, behavioural patterns) feeds into this single number.

Risk scoring replaces the impossible task of manually evaluating dozens of fraud signals with a single, actionable metric.

How Risk Scores Are Calculated

Risk scoring uses a weighted algorithm. Each fraud signal has a weight based on how predictive it is, and a value based on the specific request being evaluated.

The Signals

Here are the most common inputs to a fraud risk score:

The Calculation

The basic approach is weighted addition with normalization:

raw_score = (signal_1 × weight_1) + (signal_2 × weight_2) + ... + (signal_n × weight_n)
risk_score = min(raw_score, 1.0)

For example, a user with a disposable email (0.4) and a VPN (0.15) scores 0.55 — moderate risk. Add a geographic mismatch (0.2) and the score jumps to 0.75 — high risk.

Fidro's API calculates this automatically and returns the composite score alongside individual signal data:

{
  "email": "user@mailinator.com",
  "disposable": true,
  "dns_valid": true,
  "risk_score": 0.72,
  "ip": {
    "vpn": true,
    "country": "RO"
  }
}

Using Risk Scores: The Threshold Framework

Risk scores become actionable through thresholds that map scores to responses:

Conservative Thresholds (Recommended Starting Point)

Implementation Example

const validation = await fidro.validate(email, ip);
const score = validation.risk_score;

if (score >= 0.8) {
  return res.status(422).json({ error: 'Please contact support to complete signup.' });
}

if (score >= 0.6) {
  return res.json({ action: 'verify_phone', reason: 'additional_verification_required' });
}

if (score >= 0.3) {
  // Allow but flag
  await flagAccount(user.id, 'elevated_risk', score);
}

// Score < 0.3: proceed normally

Tuning Your Thresholds

Start with the conservative thresholds above and adjust based on two metrics:

False Positive Rate

Legitimate users blocked or unnecessarily challenged. If this exceeds 1-2%, your thresholds are too aggressive.

Fix: Raise your blocking threshold (e.g., 0.8 → 0.85) or move some blocks to friction instead.

False Negative Rate

Fraudulent users that slip through. Measured by chargebacks, abuse reports, and manual fraud reviews.

Fix: Lower your monitoring threshold (e.g., 0.3 → 0.25) or add new signals to your scoring model.

The Two-Week Rule

Run any threshold change in monitoring mode for two weeks before enforcing it. Log what would be blocked, review the flagged accounts, and adjust before going live.

Risk Scores vs Binary Rules

Some fraud detection systems use binary rules: "if VPN, block" or "if disposable email, reject." Risk scores are superior because:

• Nuance — A VPN alone is not fraud. A VPN + disposable email + new domain is. Scores capture this nuance.
• Fewer false positives — Binary rules block entire categories of users. Scores allow graduated responses.
• Adaptability — Adjusting a threshold is one number change. Adjusting binary rules means rewriting logic.
• Transparency — A score of 0.72 tells you how risky something is. "Blocked by Rule #14" tells you nothing.

Getting Started

1. Sign up for free — Fidro returns risk scores on every validation call
2. Start in monitoring mode: log scores but don't block anyone
3. After two weeks of data, set your initial thresholds
4. Review the API documentation for the full score breakdown