Sigma Identity Fraud Troubleshooting & FAQs

Known issues

Sigma Identity Fraud is United States-only. It is not optimized for global consumers. Using the product for consumers outside the U.S. will have unpredictable performance.

Why do some evaluations receive a high Sigma Identity Fraud score without accompanying reason codes?

Reason codes provide specific explanatory conditions applicable to a consumer.

Some applications may not match any predefined conditions but are still considered high risk by the model.

These are edge cases, but they can occur.

How is the model's accuracy affected if certain PII values are missing?

The impact on model accuracy depends on model version and your specific population.

Reference benchmark metrics are available in the Model Governance documentation, which can be requested from your Account Manager.

Key considerations:

Third-Party Fraud is defined as the use of stolen identities, often combined with a fraudster-controlled contact channel.
Performance is most affected when phone number, email, address, or IP address are missing. These are expected inputs.
If IP address is unavailable, you can be configured with an alternative model flavor optimized for performance without IP.

Other important fields:

Missing a nine-digit national ID or date of birth impacts resolution accuracy.
Alternative model flavors are available if either of these fields cannot be provided. Contact your Account Manager for configuration support.

General impact:

Missing PII typically increases overall fraud scores, leading to a higher false positive rate.