A strong KYC program should produce more than an approval or rejection: it should leave a clear, protected record of what was checked, which rules applied, who acted, and why the decision was reasonable. This checklist helps compliance, risk, and engineering teams build an audit-ready evidence trail and review it on a practical monthly or quarterly cadence.
Overview
KYC audit evidence is the collection of records that allows an authorized reviewer to reconstruct a customer onboarding decision without relying on memory, informal messages, or an opaque vendor dashboard. The objective is not to retain every possible data point indefinitely. It is to retain the evidence needed to demonstrate that your process was designed, followed, monitored, and governed appropriately for the customer, product, and applicable obligations.
A useful evidence trail connects five questions:
- Who was assessed? Identify the customer, applicant, business, beneficial owner, or account involved, using an internal case ID rather than exposing unnecessary personal data in reports.
- What was checked? Record the identity information, documents, screenings, signals, and verification steps used.
- Which rules applied? Preserve the relevant policy, workflow, risk model, and configuration versions.
- What decision was made? Capture the outcome, reason codes, confidence or risk indicators, reviewer actions, and any conditions or restrictions.
- Can the record be trusted? Protect evidence with access controls, timestamps, integrity controls, and a documented retention schedule.
The exact records and retention period depend on the jurisdictions, products, customer types, and obligations applicable to your organization. Treat this article as an operational checklist, not a substitute for legal or compliance advice. For broader process design, compare it with your KYC audit trail requirements and review workflow.
What to track
Organize evidence by the point in the customer journey where it is created. A consistent structure makes sampling faster and exposes missing records before an examination or internal audit.
1. Notice, consent, and collection context
Retain the version of the privacy notice, consent record where consent is the applicable basis, disclosures, and collection timestamp presented to the applicant. Record the channel, jurisdiction or market, purpose of collection, and whether the customer was informed about automated checks or manual review when relevant. Avoid placing full identity documents or sensitive attributes in general-purpose audit exports.
2. Applicant and account linkage
Keep a stable case or application identifier that links the onboarding event to the account, customer profile, business record, and related parties. The record should show when the case was created, whether it was a new application or a refresh, and which product or risk tier triggered the workflow. For business onboarding, include the ownership and control relationships used in the assessment; the KYB verification workflow provides a useful reference for structuring those relationships.
3. Identity and document verification
Record the type of identity evidence requested, document country or issuer where appropriate, document status, verification method, and result. If an image or document is retained, document its location, protection, and retention class rather than copying it into multiple systems. Capture important technical outcomes such as document authenticity checks, expiry checks, image-quality failures, name or date-of-birth mismatches, and manual overrides. A result from an identity verification API should be linked to the provider event ID and timestamp.
4. Screening and risk signals
Preserve the screening type, search date, data source or provider, match status, disposition, and reviewer rationale. Track sanctions, politically exposed person, adverse media, fraud, device, location, velocity, and other signals only when they are relevant to the documented risk model. A risk score without its model version, input category, and threshold is difficult to interpret later, so store those references together.
5. Decision logs and human review
Every final outcome should have a structured decision record: approved, declined, pending, restricted, or another defined state. Include the decision timestamp, actor or service identity, reason code, free-text explanation where needed, escalation path, and any follow-up requirement. For manual reviews, record what the reviewer examined and whether the reviewer confirmed, changed, or overrode an automated result. Do not allow an override to be saved without a reason.
6. Policy, vendor, and system context
Link the case to the policy version, ruleset, risk matrix, form configuration, and vendor configuration active at the time. Maintain vendor records covering the service used, processing purpose, integration version, service status, and relevant assurance or review materials. Engineering teams should also retain webhook delivery status, signature-validation results, replay protection outcomes, and processing errors where those events affect the decision record. See the identity verification webhook security guide for implementation considerations.
7. Access, retention, and deletion evidence
Track who viewed, exported, changed, or deleted a case and when. Access logs should distinguish ordinary case handling from administrative access. Every evidence category should have an owner, retention class, review date, and deletion or anonymization rule. If a legal hold, investigation, complaint, or other approved exception pauses deletion, record the reason and authorization.
Practical evidence matrix
| Evidence area | Minimum record to locate | Control to test |
|---|---|---|
| Consent and notice | Notice version, timestamp, channel, and applicable collection context | Correct version was presented and retained |
| Verification | Check type, result, provider event, timestamp, and exception details | Result is linked to the correct applicant |
| Risk and screening | Signal, source, model or ruleset version, and disposition | Thresholds and escalations were applied consistently |
| Decision | Outcome, reason, actor, timestamp, and override explanation | No unexplained or orphaned decisions |
| Governance | Policy version, access log, retention class, and vendor reference | Evidence is protected, reviewable, and disposed of correctly |
Cadence and checkpoints
Audit readiness is easier when evidence is checked continuously rather than assembled under deadline. Use a tiered cadence that matches the speed and risk of your onboarding program.
- Per case: Validate that required fields, timestamps, decision reasons, policy versions, and provider references are present before the case is closed.
- Monthly: Sample completed cases across approval, decline, escalation, manual review, and exception paths. Check for missing evidence, inconsistent reason codes, broken links, and unauthorized access.
- Quarterly: Review trends by product, market, risk tier, reviewer, vendor, and workflow version. Reconcile retention schedules, access roles, exception registers, and vendor records.
- After a change: Run targeted sampling after a policy update, new market launch, identity verification integration change, model threshold change, or material vendor change.
Sampling should be risk-based but not limited to successful customers. Declines, overrides, abandoned applications, repeat attempts, and cases that entered escalation often reveal control weaknesses that approval-only samples miss. A metrics review can complement the evidence review; use the identity verification metrics and QA template to connect operational trends with case-level records.
How to interpret changes
A change in a metric is a prompt for investigation, not proof that the program improved or deteriorated. Start by separating process changes from customer or fraud-pattern changes. For example, a higher manual-review rate may follow a new threshold, a provider configuration change, a market expansion, or a genuine shift in applicant risk.
For each material change, ask:
- When did the change begin, and which policy, code, vendor, or operational release preceded it?
- Is the change concentrated in one market, product, document type, reviewer group, or risk tier?
- Do case records contain the expected reason codes, evidence links, and policy versions?
- Did false-positive indicators, completion rates, escalation volume, or review time move at the same time?
- Does a targeted sample confirm that decisions remain explainable and consistently applied?
Record the conclusion, supporting sample, owner, and corrective action. If the issue concerns privacy, access, retention, or a potentially unlawful decision process, escalate it through the organization’s established governance route rather than resolving it only in an operational dashboard.
When to revisit
Revisit this KYC audit checklist at least monthly for operational completeness and quarterly for governance review. Update it sooner when applicable AML or KYC requirements change, when the business enters a new market, or when products, customer types, risk appetite, or verification methods change.
Use this action list at the next review:
- Select a documented sample that includes ordinary cases and exceptions.
- Trace each case from notice and identity evidence through screening, decision, and retention status.
- Compare stored policy and vendor versions with the versions approved for that period.
- Test access logs, export controls, deletion handling, and exception approvals.
- Assign owners and due dates for every missing, inconsistent, or excessive record.
- Update the checklist when a control, workflow, provider, or applicable requirement changes.
A review-ready KYC program is not defined by the volume of data it stores. It is defined by whether an authorized reviewer can follow a proportionate, privacy-conscious, and reproducible path from customer application to final decision. Keep that path structured, protected, and current, and your evidence will support both compliance review and better operational control.