## Key outputs

### Address Risk Score

Probabilistic score between `0.001` and `0.999` indicating the likelihood of fraud associated with a physical address.

- Higher scores = higher risk of fraud.
- Evaluates deliverability, tenure, property type, and fraud patterns.
- Powered by internal and external data sources.
- Multiple model scores may be returned in a single response.

### Name–Address Correlation

Value between `0.01` and `0.99` showing how strongly an address is linked to the consumer’s name.

- Very high confidence for verified full-name matches.
- Supports fuzzy matching and partial matches (e.g., last name only).
- Reason codes explain the underlying correlation signals.
- Improves trust in onboarding, delivery checks, and profile updates.

## Industries it fits

- **Financial services**: Account opening, fraud prevention, progressive onboarding
- **E-commerce & retail**: Checkout fraud prevention, shipping address validation, return abuse detection
- **Logistics & delivery**: Ensure goods are sent to legitimate, deliverable addresses
- **Gaming & media**: Detect fraud in prize delivery, bonus abuse, duplicate accounts
- **Telecom**: Validate residential addresses during profile changes and service setup
- **Government & benefits**: Verify residential addresses for eligibility and fraud prevention
- **Plus** other digital onboarding and fraud risk workflows

## Business and operational outcomes

### 1. For fraud and risk teams

- Identify high-risk addresses in real time.
- Detect risky address types (e.g., correctional facilities, PO boxes, commercial drops).
- Leverage multiple model scores per call for testing and migration.
- Gain transparency with reason codes for both risk and correlation.

### 2. For growth and onboarding teams

- **Progressive onboarding**: Establish trust with minimal inputs (name + address).
- **Reduced friction**: Allow good users through while flagging suspicious addresses.
- **Adaptive signals**: Models tuned for different fraud/risk trade-offs.

### 3. For support and operations

- **Fewer escalations**: Clearer fraud signals reduce review queues.
- **Transparent explanations**: Reason codes help agents understand outcomes.
- **Lightweight integration**: Only a few required fields keep implementation fast.

## Performance considerations

- **Fast responses**: Sub-second latency on average in production.
- **Always available**: 99.99%+ uptime with global redundancy.
- **Built for scale**: Supports tens of millions of risk checks per day.
- **Smart models**: Multiple models run in parallel to optimize accuracy and fraud capture.

## Boundary conditions

- **Name + address are required**: Correlation accuracy improves when complete address details are included (e.g., city, postal code).
- **Multiple models**: Customers should test challenger vs. production models to tune thresholds.
- **Available in USA only**: This feature is limited to the U.S.A. and is not supported in other regions.