How do we stop synthetic identities from slipping through verification when they're specifically designed to look real?

Synthetic identities are difficult to detect because they are intentionally built to appear legitimate. Fraudsters combine real and fabricated information, often using valid Social Security numbers, addresses, or other personal data to create identities that can pass basic verification checks. Traditional verification methods that only confirm whether information exists in a database may miss these sophisticated attacks

If we tighten our scoring thresholds to catch more fraud, how do we avoid punishing legitimate customers with more friction and delays?

The goal isn't to apply stricter controls to everyone, but to apply the right controls to the right users. Effective risk scoring uses a layered approach, allowing low-risk customers to move through the journey with minimal friction while reserving step-up verification and additional checks for higher-risk transactions. This helps reduce fraud without creating unnecessary delays, abandonment, or manual review volume for legitimate users.

Which risk signals actually move the needle on detection accuracy, and which ones are just adding noise to our review queues?

The most valuable risk signals are those that provide independent evidence of trust or fraud, such as document authenticity checks, biometric verification results, identity velocity indicators, device intelligence, and behavioral anomalies. Signals that are evaluated in isolation or rely on overly broad rules often generate large numbers of false positives without significantly improving detection rates. The key is not collecting more signals, but understanding which combinations consistently correlate with fraud and using them to drive more precise risk decisions.

How do we make a defensible case to leadership that our fraud losses are going down without cherry-picking the numbers?

The strongest case comes from measuring fraud outcomes alongside operational metrics, not in isolation. Fraud loss reductions should be evaluated alongside approval rates, customer pass rates, false positives, manual review volumes, and conversion metrics to show the full impact of a fraud program. When organizations demonstrate that fraud losses are declining while legitimate customers continue to move through the process efficiently, the results are both more credible and more meaningful to leadership.

Can a single platform realistically combine ID verification, biometrics, and behavioral signals into one score — or does that create new blind spots we haven't accounted for?

A unified platform can provide a more complete view of risk because it evaluates multiple signals within the same decisioning framework rather than forcing teams to correlate data across disconnected systems. The key is transparency. Organizations need visibility into how different signals contribute to risk decisions so they can understand outcomes, identify gaps, and adapt as fraud patterns evolve. When implemented effectively, combining identity, biometric, and behavioral intelligence often reduces blind spots rather than creating them.