Buyer framework

How to Choose an Identity Provider

A structured process for turning regulatory, product and risk requirements into a shortlist you can test and defend.

Start with the outcome, not the vendor list

A credible selection process begins with the business journey and the risk decision. “We need KYC” is too broad. A useful brief states who is onboarding, where they are located, which decision must be made, what evidence must be retained and what happens when automation cannot decide.

Write one decision sentence

For example: “Determine whether a new business customer and its controlling persons can open an account in our target markets, while routing uncertain cases to trained reviewers.”

Establish measurable outcomes

  • Eligible completion rate by market and document type
  • False rejection and manual-review rates
  • Time to decision, including fallback paths
  • Fraud detection performance on representative traffic
  • Operational effort per completed onboarding

Map requirements across six dimensions

Coverage: countries, documents, registries and languages
Assurance: evidence, biometrics, liveness and auditability
Compliance: screening, policies, retention and reviewer controls
Experience: accessibility, completion, fallback and support
Technology: APIs, SDKs, orchestration, latency and resilience
Commercials: minimums, overages, retries and exit costs

Separate mandatory requirements from preferences. A provider that fails a genuine legal, coverage or security constraint should not be rescued by a high score elsewhere.

Use a weighted scorecard with evidence

Assign weights before vendor demonstrations. Ask every provider the same core questions and record the evidence behind each score. Marketing claims should not receive the same weight as tested results, contractual commitments or independently verifiable documentation.

Recommended evidence hierarchy

  1. Your own proof-of-concept results on representative cases
  2. Contractual service levels and documented product behaviour
  3. Security, privacy and assurance documentation
  4. Reference architecture and implementation evidence
  5. Provider statements requiring validation

Test the difficult cases, not only the happy path

A proof of concept should include normal traffic, edge cases and known failure modes. Agree the evaluation dataset, expected outcome and decision thresholds in advance. Review performance by segment; aggregate completion can hide poor results for a country, document, device or customer group.

  • Low-light, damaged-document and transliteration cases
  • Repeated attempts, injection attacks and presentation attacks
  • Complex business ownership and incomplete registry records
  • Manual review, resubmission and customer-support journeys
  • Webhook failure, API timeout and provider outage scenarios

Contract for the operating reality

Confirm what counts as a billable check, how retries are treated, which sub-processors are used, how data is deleted and how performance is measured. Document change control for new documents, models, registries and regulatory requirements. Plan an exit path before integration becomes critical.

IA
How this guide was prepared

This framework is an editorial decision aid. It does not endorse a provider and is not legal, compliance or security advice. Adapt it to your obligations and validate claims through testing and due diligence.