Skip to content
Case studiesPricingSecurityCompareBlog

Europe

Americas

Oceania

Back to glossary
Technology

Confidence Score

A confidence score is a numerical value, typically expressed as a percentage or on a scale of 0 to 100, that quantifies the degree of certainty an automated system has regarding the validity or authenticity of a document or an extracted data point. The higher the score, the more confident the system is in its analysis.

The confidence score is the central decision-making mechanism in any automated document verification system. It is not a simple binary indicator (true/false) but a graduated measure that reflects the probability that a document is authentic or that extracted data is correct. This granularity allows businesses to set acceptance thresholds tailored to their risk appetite: a high threshold for high-risk transactions, a lower one for routine checks.

The confidence score is calculated by aggregating the results of multiple independent checks: image quality, consistency of security features, format validity, MRZ matching, and tampering detection. Each check produces a sub-score weighted according to its reliability. A document may score highly on OCR extraction but poorly on holographic element verification, directing the final decision towards a targeted manual review.

CheckFile exposes confidence scores at three levels: per extracted field (name, date of birth, document number), per check performed (authenticity, integrity, consistency), and an overall document score. This transparency enables compliance teams to understand exactly why a document was accepted or rejected, facilitating regulatory audits and the justification of automated decisions to supervisory authorities.

Regulations

GDPREU AI Act6th Anti-Money Laundering Directive

Real-world examples

  • 1.During passport verification, the system assigns a confidence score of 98% for name and date of birth extraction, but only 72% for the MRZ zone due to a blurry image, triggering a request for document resubmission.
  • 2.An insurer configures a confidence score threshold of 85% for standard policies and 95% for high-premium contracts, automatically adapting the verification level to the financial risk.
  • 3.A bank's compliance dashboard displays the distribution of confidence scores over the past 30 days, revealing that a new type of proof of address consistently scores low, requiring a model adjustment.

Automate your compliance

Discover how CheckFile simplifies document verification for your organisation.