Define the document check before onboarding
KYC document verification can refer to several different checks: reading a name, checking document consistency, establishing identity, or applying an organization's acceptance policy. Agree which operation your process needs before treating an extraction result as a decision. A correctly read document number is one input to that process, not proof of the holder's identity.
What the extraction API returns
The document extraction contract describes a tenant-configured schema. The response places field values under data.metadata, with confidence on a 0–100 scale. When a field was located on the page, the output can also include its page and bounding box. Those location keys are not guaranteed for every value.
Possible schema fields include name, document number, issuing authority, expiry date and address. Confirm the required document types, formats, languages and fields against representative samples. This page does not establish coverage of every passport, license or national ID.
Illustrative field result
This synthetic fragment demonstrates the documented field structure, not a completed identity check or a live customer record. Field names depend on the tenant schema.
{
"data": {
"metadata": {
"documentNumber": { "value": "SAMPLE-123", "confidence": 93 },
"expiryDate": { "value": "2031-07-14", "confidence": 91 }
}
}
}Keep extraction separate from identity verification
- Document fields — review missing, inconsistent or uncertain values against the original file before another system consumes them.
- Biometric or authenticity checks — face matching, liveness, hologram detection and document-authenticity services require separately confirmed capabilities and validation. They are not established by an extraction response.
- Onboarding decisions — your authorized team defines acceptance policy. A configured workflow records its decision; this page does not promise autonomous onboarding approval.
- Human review — review phases and their routing must be configured for the tenant. Confidence scores alone do not select the review path automatically.
Evaluate the documents you actually receive
Use representative native PDFs, scans and phone photos, with the languages and document variants in your intake. Agree the expected fields and reviewed ground truth. Measure field correctness, missing values, available source references and the correction work required. Keep failed inputs visible in the result and set acceptance criteria before evaluating.
The extraction benchmark guide provides an evaluation method. It does not certify KYC effectiveness. Processing time depends on the input and route; no universal sub-three-second result is claimed here.
Integration and availability
Use the document extraction API and documented REST requests to deliver structured fields to your own onboarding system. A document portal can support collection, while review and downstream stages require the agreed tenant configuration. The module overview distinguishes available modules from contract-configured capabilities.
Data handling and evidence
Identity documents contain personal information. Define access, processing location and document and audit retention for the tenant and contract before an evaluation. The security overview, data-protection terms and Trust Center describe the published control boundaries. Extraction output is not a certification or a substitute for your organization's compliance assessment.
Sources and content responsibility
Content owner: Cogneris, product documentation. Updated . Product statements above are linked to the published extraction contract, module availability and security documentation; the sample is explicitly synthetic.
Plan a scoped evaluation
Discuss your document set and required checks before integrating. For the broader processing pipeline, read the IDP buyer's guide; for transaction evidence, see bank statement extraction. Agree any biometric, authenticity or screening operation separately rather than inferring it from document extraction.