Security teams collect hundreds of numbers from their tools, yet few can answer the three questions leadership actually asks: what are we covering, is it working, and what does it cost? A programme-level metrics framework selects a small set of measures across every security domain, defines each one precisely, sets targets, and reports them in a form each audience can act on. This guide uses NIST SP 800-55 Volumes 1 and 2 (December 2024) and the NIST Cybersecurity Framework (CSF) 2.0 to cover the types of measures, a coverage, outcome and efficiency metric set by domain, measure specifications and data quality, targets, reporting by audience, roles and a phased build. It is written for CISOs, security programme managers and GRC leads. Domain-level detail lives in companion guides, such as the Identity Security Metrics Guide and the Security Operations Metrics Guide; this page ties them into one programme.
Key takeaways
- Four kinds of measure, not one. NIST SP 800-55 Vol 1 distinguishes implementation, effectiveness, efficiency and impact measures; a programme needs all four, weighted toward effectiveness as it matures.
- Coverage comes first. Every outcome and efficiency measure is only as good as the inventory beneath it, so report the population covered next to each result.
- KPIs and KRIs are different tools. KPIs show whether controls and the programme perform; KRIs show whether exposure stays within tolerance. CSF 2.0 expects practitioners to provide both to managers and executives.
- Less is more. NIST advises starting small with quantitative metrics and excluding any metric that does not reflect a valid control or track a material risk.
- Metrics are a cycle, not a report. SP 800-55 Vol 2 runs from defining the programme through data collection and gap analysis to corrective actions tracked through a plan of action and milestones (POA&M).
What a programme metrics framework is for
NIST SP 800-55 Vol 1 defines measurement as obtaining quantitative values using quantitative methods, measures as the resulting quantifiable and objective values, and metrics as measures and assessment results used to track progress, facilitate decisions and improve performance against a target. It separates assessment (judging against criteria) and notes that qualitative ratings (high, medium, low) and semi-quantitative scales are not measurements in its sense, even though most organisations use all three.
- Manage: show control owners where performance is slipping, so they can fix it before an audit or incident does.
- Assure: give executives, boards, auditors and regulators evidence that controls operate. The UK Cyber Governance Code of Practice (April 2025) asks boards to require formal reporting at least quarterly with suitable metrics and agreed tolerances.
- Decide: inform investment, prioritisation and risk acceptance. A metric that never changes a decision is a candidate for removal.
The four types of measure
| Type (NIST SP 800-55 Vol 1) | What it shows | NIST example | Programme example |
|---|---|---|---|
| Implementation | Progress in putting specific controls in place, usually as percentages | Percentage of systems with approved system security plans | Percentage of production cloud accounts enrolled in posture management |
| Effectiveness | Whether controls work and achieve intended outcomes | Share of incidents caused by misconfigured access controls over five years | Share of red team or phishing simulation objectives detected and contained |
| Efficiency | The timeliness of controls | Mean response time to incidents compared with incident cost | Median days to remediate KEV-listed vulnerabilities on internet-facing assets |
| Impact | Effect of security on mission and goals: cost, fines, trust | Share of the budget spent on security controls | Revenue-affecting outage hours caused by security incidents per year |
Maturity changes the mix: a young programme reports mostly implementation measures because controls are still being deployed. As coverage stabilises, shift weight to effectiveness and efficiency; implementation measures that sit at 100% for a year move to an exception report.
A coverage, outcome and efficiency metric set
A practical way to apply the four types across the programme is to give each domain at least one coverage measure (implementation), one outcome measure (effectiveness) and one efficiency measure. The set below is a starting point; keep only measures you can source reliably.
| Domain | Coverage | Outcome | Efficiency |
|---|---|---|---|
| Asset management | Share of discovered assets present in the inventory with an owner | Incidents on assets not in the inventory | Days from asset discovery to owner assignment |
| Workforce identity | Share of accounts with phishing-resistant MFA, privileged first | Confirmed account takeovers | Hours from leaver event to access removal |
| Non-human identity | Share of service accounts, API keys and tokens inventoried with a named owner | Incidents involving leaked or abused machine credentials | Days to rotate an exposed secret |
| Vulnerability management | Share of assets scanned or agent-covered in the period | Exploited vulnerabilities that were known and unpatched | Median days to remediate by severity and KEV status |
| Endpoint and server | EDR coverage across servers and endpoints | Malware incidents reaching execution | Mean time to isolate a compromised host |
| Detection and response | Share of critical log sources reporting to the SIEM | Share of incidents detected internally rather than by third parties | Mean time to detect and to contain |
| Application security | Share of in-scope repositories with automated security testing | Vulnerabilities found in production that testing should have caught | Median days to fix critical findings |
| Cloud | Share of cloud accounts under posture management | Critical misconfigurations open beyond policy | Hours to remediate public data exposure |
| Third parties | Share of critical suppliers assessed in the cycle | Incidents originating at suppliers | Days from assessment to agreed remediation |
| Resilience | Share of critical systems with a tested, isolated backup | Recovery tests meeting recovery objectives | Actual recovery time in tests and incidents |
| People | Training completion for role-based modules | Phishing report rate (reports over clicks) | Minutes from first report to message quarantine |
NIST cautions that phishing simulation success rates can mislead if the test is poorly designed, which is why report rates and time to quarantine are often more useful than click rates. For domain depth, see the AppSec Programme Metrics Guide and the Security Awareness and Human Risk Management Guide.
Defining measures and protecting data quality
SP 800-55 Vol 1 lists the characteristics of a good measure: accuracy against the security objective, numeric precision, correctness (following defined collection procedures), consistency regardless of who measures, a fixed time reference, replicability and standardised units. It asks that each measure's data source, formula, responsible parties and collection frequency are documented. In practice, write a short specification for every reported metric:
| Field | Content |
|---|---|
| Name and ID | Stable identifier used in every report |
| Type | Implementation, effectiveness, efficiency or impact; KPI or KRI |
| Objective | The control, goal or risk it informs (link to the CSF 2.0 outcome or register entry) |
| Formula and unit | Numerator, denominator, unit, rounding |
| Population | In-scope assets or identities and the source of the denominator |
| Data source and method | System of record, collection method (observational logs, experiment such as a phishing test, or sampling: random, stratified or systematic) |
| Frequency | Collection and reporting cadence |
| Target and threshold | Target value, tolerance where it is a KRI, and the rationale |
| Owners | Metric owner (accountable for performance) and data steward (accountable for accuracy) |
| Known limitations | Gaps, sampling error, changes in source |
- Normalise: SP 800-55 Vol 2 asks for consistent units and formats with clear metadata, so system-level measures can be aggregated to programme level.
- Protect the data: measurement data, such as lists of unpatched systems, should be protected at the same level as the systems it describes.
- Show gaps honestly: report "not measured" rather than zero, and show the population covered next to every percentage.
Targets, baselines and trends
- Baseline first: collect two or three periods before setting targets, unless policy already sets them (for example a remediation SLA).
- Derive targets from policy and appetite: remediation days come from the vulnerability policy; tolerances for KRIs come from the risk appetite framework. See the Key Risk Indicators for Cyber Risk Guide.
- Gap analysis: SP 800-55 Vol 2 compares measurements against targets, identifies causes of poor performance and the areas needing improvement, then prioritises corrective actions by risk mitigation, cost and impact.
- Trend over snapshot: show four to eight periods; a single value invites the wrong question.
- Medians and percentiles: for time-based measures, report the median and the 90th percentile; a mean hides the long tail that attackers use.
- Profiles for maturity: CSF 2.0 Current and Target Profiles, with gap analysis and an action plan, show programme progress at outcome level, while CSF Tiers (Partial, Risk Informed, Repeatable, Adaptive) describe the rigour of governance and management. See the NIST CSF 2.0 Profiles and Tiers Guide.
Reporting by audience
| Audience | What they need | Cadence | Typical content |
|---|---|---|---|
| Board or risk committee | Exposure against appetite and tolerance, trend, decisions required | Quarterly | 8 to 12 KRIs for top risks, profile progress, major incidents, asks |
| Executive team | Programme performance, investment and risk trade-offs | Monthly or quarterly | Coverage and outcome measures by domain, efficiency against SLAs, exceptions |
| Security leadership | Where to act this month | Monthly | Full KPI set with targets and owners, corrective action status |
| Control owners and teams | Operational detail to fix problems | Weekly or continuous | Dashboards from tooling, item-level exception lists |
SP 800-55 Vol 2 names the main reporting challenge as translating technical data into terms that are relevant to the audience, and suggests asking which measures tell a precise risk story, which drive action and which the recipient will understand. CSF 2.0 describes practitioners communicating upward through risk registers and progress reports. See the Board Cyber Risk Reporting Guide and the Board Cyber Report Template.
Roles and building the programme
SP 800-55 Vol 2 assigns roles: the CEO or agency head ensures measures support planning, provides resources and approves policy; the CIO orchestrates the programme and reports on effectiveness; the CISO implements measures, standardises development, analysis and reporting and integrates measurement into remediation; programme managers and system owners confirm data feasibility and accuracy and implement corrective actions; and staff supply data and feedback. The workflow below follows the five activities NIST describes.
- Evaluate and define the existing programme (weeks 1 to 4): list assets, stakeholders, goals, policies and data already available. Map goals to CSF 2.0 outcomes.
- Identify and prioritise measures (weeks 4 to 8): draft coverage, outcome and efficiency measures per domain, then prioritise by risk strategy, mission objectives, data availability, asset value, control cost and regulatory need. Cap the first release at 20 to 30 measures.
- Pilot data collection (weeks 8 to 12): collect manually once to prove the formula, then automate. Run the first cycle without targets as a shadow period; drop measures whose data cannot be trusted.
- Collect and analyse (quarter 2 onward): aggregate system measures to programme level, run gap analysis against targets, and identify causes.
- Identify corrective actions: list options, prioritise by risk reduction and cost, and select those worth a full cost-benefit analysis.
- Apply corrective actions: assign owners and track status through the POA&M process.
- Review and retire: NIST asks for regular review and review after impactful events, including asking which measures could have predicted the event. Retire obsolete measures and add new ones as the mission and practices change. If a new metric produces disputes rather than decisions, return it to the pilot stage.
Small and large organisations: a small organisation can run 12 to 15 measures from existing tools in a spreadsheet owned by the security lead, reviewed monthly. A large enterprise needs a metric catalogue with version control, automated pipelines from tooling into a data platform or GRC tool, data stewards per domain, and separate views for each audience drawn from one data set. See the GRC Platform Buyer's Guide.
Is the metrics programme itself working?
- Share of reported metrics with a written specification and named owner.
- Share of metrics collected automatically from a system of record.
- Share of metrics below target with an open corrective action in the POA&M.
- Number of decisions (funding, prioritisation, risk acceptance) recorded as informed by the metrics in the period.
- Incidents in the year that no metric signalled beforehand.
Framework mapping
| Framework | Reference | Relevance |
|---|---|---|
| NIST SP 800-55 Vol 1 and Vol 2 (December 2024) | Measure types, characteristics, selection; programme workflow and roles | Primary method source for this guide; supersedes SP 800-55 Rev 1 (2008) |
| NIST CSF 2.0 | GV.OV-01, GV.OV-03 | Strategy outcomes and risk management performance reviewed and adjusted |
| NIST CSF 2.0 | ID.IM-01, ID.IM-02 | Improvements identified from evaluations, tests and exercises |
| NIST CSF 2.0 | ID.AM-01, ID.AM-02, PR.AA-01 | Hardware, software and service inventories, and identities and credentials for users, services and hardware: the denominators for coverage |
| NIST CSF 2.0 | DE.CM-01, PR.AT-01 | Monitoring and awareness outcomes behind detection and people metrics |
| NIST SP 800-53 Rev 5 | PM-6, PM-31, CA-7, CA-5 | Measures of performance, continuous monitoring strategy and monitoring, POA&M |
| ISO/IEC 27001:2022 | Clause 9.1 | Monitoring, measurement, analysis and evaluation; ISO/IEC 27004 gives guidance |
| NIST IR 8286C Rev 1 | Section 5 | KPIs and KRIs as the link between tolerance and controls |
Common failure patterns
- Tool exhaust as metrics: reporting whatever dashboards export, such as alerts processed, which measures volume rather than control performance.
- Coverage blind spots: 99% MFA coverage of the accounts in the directory says nothing about legacy, local or machine accounts outside it.
- Percentages without populations: a ratio improves when the inventory shrinks, so a failed discovery job looks like progress.
- Targets without owners: metrics that miss target month after month with no corrective action recorded become wallpaper.
- Too many metrics: 80 measures on a board pack dilute attention; NIST's "less is often more" applies at every level.
- Human identity only: identity metrics that ignore service accounts, API keys and tokens miss the identities attackers increasingly target. See the Non-Human Identity (NHI) Guide.
Lessons from incidents
The Colonial Pipeline Ransomware Attack of 2021 began with a legacy VPN profile that was no longer used but never disabled, protected only by a password that had appeared in leaked credentials and not by MFA. The attackers spent about eight days inside before ransomware deployment and the pipeline was shut for about six days. The metrics lesson is about denominators: an MFA coverage figure calculated over active, known remote-access accounts would not have shown this profile. Programme metrics need a coverage measure for dormant but enabled accounts, human and non-human, on every remote-access path, and an efficiency measure for how quickly such accounts are found and disabled.
Practitioner checklist
- Use the four NIST measure types and give each domain a coverage, outcome and efficiency measure.
- Write a specification for every reported metric, with formula, population, source, owner and steward.
- Show the population covered next to every percentage, and report "not measured" honestly.
- Baseline for two or three periods before setting targets not already fixed by policy.
- Report medians and 90th percentiles for time-based measures, with four to eight periods of trend.
- Build separate views for board, executives, security leadership and control owners from one data set.
- Track every below-target metric to a corrective action in the POA&M.
- Include NHI coverage and dormant account measures in the identity domain.
- Review the set after major incidents and at least annually, retiring metrics that drive no decisions.
Standards and references
- NIST: SP 800-55 Vol 1, Measurement Guide for Information Security: Identifying and Selecting Measures
- NIST: SP 800-55 Vol 2, Measurement Guide for Information Security: Developing an Information Security Measurement Program
- NIST: The NIST Cybersecurity Framework (CSF) 2.0 (CSWP 29)
- NIST: IR 8286C Rev 1, Staging Cybersecurity Risks for ERM and Governance Oversight
- NIST: SP 800-53 Rev 5, Security and Privacy Controls
- UK Government and NCSC: Cyber Governance Code of Practice
Related NHI Mgmt Group resources: Identity Security Metrics Guide · Security Operations Metrics Guide · Key Risk Indicators for Cyber Risk Guide · Board Cyber Risk Reporting Guide