Measure driver completion separately from verification outcome

General lesson: In identity-proofing journeys, user completion, process failure, abandonment, provider acceptance, and final verification verdict are separate outcomes. Optimize the user's ability to submit all required inputs without redefining success around downstream provider behavior. Use a stable transaction start/end, match save-and-return sessions to one transaction, report mature cohorts, and stratify by proofing route and step. Track capture failure-to-acquire and false reject separately from abandonment, and preserve fraud/false-accept guardrails.

Applicability and limits: This applies to mobile document/photo workflows and similar multi-step proofing funnels. It does not justify removing required evidence, lowering assurance, or treating support/bypass as ordinary self-serve completion. External completion percentages should not be adopted as targets unless their start point, end point, eligibility, support inclusion, time window, and fraud treatment match.

Evidence/example (2026-09-17): NIST SP 800-63-4 separately defines pass, fail, abandonment, per-step failure, completion time, help-desk, and redress metrics: https://pages.nist.gov/800-63-4/sp800-63/dirm/ . GOV.UK defines completion as completed transactions divided by all started transactions, says save-and-return must be correlated, and explicitly notes that an application may be complete even if unsuccessful: https://www.gov.uk/service-manual/measuring-success/measuring-completion-rate . FIDO defines document failure-to-acquire, false reject, and false accept separately: https://fidoalliance.org/specs/idv/docauth/document-authenticity-verification-requirements-v1.0-fd-20220815.html . A historical UK Verify report illustrates denominator risk: its 48% verification success excluded people who dropped before finishing: https://www.nao.org.uk/press-releases/investigation-into-verify/ . For Shield, the qualified hypothesis is to define the product north star at driver-owned evidence completion, retain invitation/open/start diagnostics, and treat verification integrity as a guardrail.