Remote Onboarding: MRZ, OCR, Liveness and Review
“Remote onboarding is not one identity check. It is a chain of evidence collection, automated validation and human judgment designed for the supported risk profile.”
Each component answers a different question
Optical character recognition extracts printed identity data so it can be compared and validated. Machine-readable zone checks use the structured area on supported documents. Liveness helps assess whether a real person is present during a capture. Document-authenticity and face-comparison services may add further evidence. None of these components, on its own, proves the entire customer relationship.
Modular Fintech offers dedicated MRZ checks and liveness capabilities that can be composed within the wider onboarding journey.
Design for evidence quality and recovery
The interface should tell users which document is supported, how to position it and why good lighting matters. Image-quality checks can prompt a recapture before a failed review. When a device, disability, document type or technical issue prevents the standard journey, provide an accessible alternative and a controlled review route.
Store the original evidence, extracted values, provider response and decision trace according to the institution’s retention and privacy requirements. Avoid relying only on a simplified pass/fail flag from an external component.
Manual review is part of the control design
A trained reviewer can assess conflicting evidence, data-quality issues and legitimate edge cases. The queue should explain why the case needs attention and present relevant evidence without overwhelming the reviewer. Sensitive vendor settings and detection logic should be restricted to authorised teams.
Test the supported document and device matrix.
Preserve raw evidence and extracted data with provenance.
Define accessible fallbacks and retry limits internally.
Train reviewers on exception categories and escalation.
Monitor false rejection, abandonment and review quality.
Test the combined journey, not each component in isolation
A document may be readable by OCR but unsupported by authenticity checks; a liveness capture may succeed while the face comparison remains uncertain; a technically valid MRZ can still conflict with application data. End-to-end tests should cover these combinations and present the reviewer with the original evidence, extracted fields and component responses.
Testing should include supported documents, devices, browsers, lighting conditions, accessibility needs and legitimate edge cases. Performance measures can track capture quality, technical failure, manual review, applicant abandonment and confirmed error outcomes without publishing sensitive control settings.
Plan for provider and device failure
A recovery design should distinguish a poor capture, an unsupported document, a temporary vendor outage and a case requiring human judgment. The customer receives an appropriate next step, while operations sees the evidence and reason for routing. Alternative providers or manual routes require their own validation and privacy controls; failover should not silently reduce the standard of evidence.
Connect the technology to a comprehensible client journey
Applicants benefit from preparation guidance that focuses on legitimate evidence rather than system mechanics. Zolvat’s article on opening a business account online [planned internal link — activate after publication] is a suitable cross-reference for the client-facing side.