Vendor-only checks carry two costly failure modes:
fraudulent attempts that go undetected, and
legitimate customers who are incorrectly rejected.
Improving one typically comes at the cost of the other. This update introduces a dedicated face match model, purpose-built to catch cases vendor-only checks miss, alongside a transition to in-house models for PAN and Aadhaar verification.
What changed
Digital Contact Point Verification now runs on a dedicated face match model trained specifically to catch cases vendor-only checks miss. Matching accuracy stands at 93.6%, and incorrect rejections of legitimate customers drop by 18% compared to vendor-only checks. PAN and Aadhaar verification also transition to in-house models, replacing third-party classifiers.
What does this mean for your organization?
Stronger fraud detection alongside fewer incorrect rejections. The dedicated face match model catches more of the cases vendor-only checks miss, at 93.6% matching accuracy, while reducing incorrect rejections of legitimate customers by 18%. Both failure modes improve together, rather than trading one off against the other.
No disruption to existing workflows. The model integrates directly into your current digital contact point verification process. Fallback mechanisms preserve reliability when a check requires a second pass, so no configuration changes are required.
Faster turnaround and reduced vendor dependency. PAN and Aadhaar verification now run on in-house models rather than third-party classifiers, improving processing speed. This also reduces dependency on vendor timelines: extending verification to a new document type in future draws on existing infrastructure rather than requiring a new vendor integration, contract, or procurement cycle.
Who this update is built for
This update is designed for lending, BFSI, and insurance teams managing high-volume onboarding, where both unchecked fraud and incorrect rejections of legitimate customers carry measurable operational and financial costs.
