Read Article 21(2)(f) alongside the wider Article 21 duty
Article 21(1) requires appropriate and proportionate technical, operational and organisational cybersecurity risk-management measures. Paragraph 2 identifies minimum areas to cover, including policies and procedures to assess effectiveness under point (f). The assessment obligation therefore concerns the measures chosen to manage actual risks, not merely whether a policy document exists.
An organisation can have a formally approved security policy and still lack assurance that its controls operate. The relevant question is whether implementation and day-to-day performance align with the risk-management decision.
What a useful policy should establish
A written effectiveness policy should identify the scope of the programme, accountable roles, how testing methods are selected and how results are reported. It should explain when assessments are triggered, how significant failures are escalated and who verifies completed remediation.
Procedures translate those decisions into repeatable work. For example, a procedure for access-control assurance can specify which systems are sampled, which access records are compared against approved roles and how exceptions are recorded. Avoid treating a generic checklist as a substitute for system-specific testing.
How to judge whether a control works
Define an expected outcome before collecting evidence. For privileged access, the expected outcome might be that administrative rights are approved, periodically reviewed and promptly revoked when no longer needed. Examine actual permissions and supporting approvals, then investigate discrepancies rather than counting completed review forms alone.
Consider whether a control protects the relevant service under realistic conditions. A disaster-recovery process can appear adequate on paper but fail in a restoration exercise because the required credentials or dependent services are unavailable. Testing helps expose that gap.
Use incidents and material changes to refine assessments
An incident can reveal that a detection rule missed important activity, escalation was delayed or a third-party dependency was undocumented. Feed those lessons into the assurance plan. Major migrations, acquisitions and changes to critical suppliers can also invalidate previous conclusions.
A risk-based cadence is better supported than an unexplained promise to test everything once per year. The exact cadence must still be checked against binding national and sector-specific obligations.
Document a decision that can be reviewed
Keep a traceable record of the measure, the risk it addresses, assessment scope, procedure, evidence, exceptions, conclusion, assessor and date. Record material limitations, such as environments excluded from a penetration test or log sources unavailable during a review.
Findings should produce named corrective actions and repeat checks. Reporting both failures and unresolved uncertainty gives management a more accurate picture than reporting only the percentage of tests passed.
Points of clarification
Is a cybersecurity policy an effectiveness assessment?
No. The policy explains how assessments are governed. An effectiveness assessment tests or evaluates whether the measures perform as intended.
Must every control be assessed with the same method?
No. Evidence appropriate for backup recovery differs from evidence for supplier due diligence, access review or incident detection.
Do national laws matter?
Yes. NIS2 is implemented through Member State law, and applicable national and sector-specific rules may add detail.