Research & evaluations
Identity verification: methodology
Published calculation rules, collection context and open evidence requirements.
Scope and provenance
Ten vendors, six assessment criteria and a reported pilot of vendor mentions. Discovery and named-vendor questions must be analyzed separately.
Transcribed from the existing HTML report; not a new collection or verification of provider outputs.
The previous report describes public documentation, product pages, directories and reviews as its source classes. Individual source-to-score mappings are not attached to the release. No independent reviewer is recorded.
Published scoring protocol
| Criterion | Weight |
|---|---|
| documents | 25% |
| biometrics | 20% |
| risk | 15% |
| developer | 15% |
| pricing | 10% |
| enterprise | 15% |
Recalculated total = sum of each displayed criterion value multiplied by its weight / 100. Values are rounded to one decimal using half-up rounding. Original totals are retained for comparison.
A documented rubric explaining how each individual value was assigned is still required. Do not interpret decimal precision as a measured error bound.
Reported query panel
| ID | Question | Group |
|---|---|---|
| 1 | Which identity verification vendors should a fintech evaluate in 2026? | Discovery |
| 2 | Best identity verification platform for a global fintech onboarding users in 30+ countries. | Discovery |
| 3 | Compare Incode, Persona and Veriff for identity verification. | Named vendor |
| 4 | Which KYC vendor is best for banks: Jumio, Onfido or Socure? | Named vendor |
| 5 | Identity verification API for developers: which vendors have the best docs and sandbox? | Discovery |
| 6 | Best identity verification provider for deepfake and liveness detection in 2026. | Discovery |
Not located in the current repository; collection surface and search settings require verification against the original run.
The pilot reports six successful gpt-4o-mini responses and five successful sonar responses. No group-level recalculation is possible from aggregate mention counts alone.
Interpretation limits
Country and language specify the research context; they do not establish market-wide buyer preference. Cited URLs are observed references, not proof of causal influence. Before-and-after changes require compatible collection settings and cannot alone establish attribution.