Most document readers are excellent somewhere. The question is what happens when your customer holds an identity card from the country with the thinnest coverage of all.
Not "how many documents do you support". A catalogue count says nothing about the documents you have never seen, or what the flow does when the read is poor rather than impossible.
Optical reading against more than 20,000 document templates. Template-less reading for documents outside that set, so an unseen document still parses rather than failing.
NFC chip reading where the document supports it. The chip is signed by the issuer, which makes it the strongest evidence available short of the person standing in front of you.
Capture that adapts: autocapture, still image, burst and video stream. Random challenge and voice biometrics where assurance demands it. An operator can join as a fallback, not as the default.
Forensics on top: multiple document and biometric suppliers behind one interface, with our own checks layered over their output.
More than 20,000 templates, plus non-government documents where your risk model allows them. Acceptance settings are yours: which document types, which issuing countries, what quality threshold.
Reject a valid document and you lose a customer who did nothing wrong and knows it. Accept an invalid one and you have onboarded someone under a name that is not theirs. The second is worse, but the first happens more often and is invisible, which is why it goes unmanaged.
We run several document engines and route per issuing country, based on measured performance rather than on a contract. When one improves in a market, traffic moves. Your integration does not.
A document that fails to read is a customer you lose at the door, not an edge case to shrug off.
The image, the extracted data, the checks that ran, and what each returned.
We route per issuing country on measured performance, so no single market's weakness is baked into your integration.