Process maturity and composite scores: what a single number cannot carry
What a maturity model measures, and why its single composite score is its weakest part.

A maturity model answers a question an inventory cannot: not whether a practice exists, but how completely it is performed and how deeply it is embedded in the way the organisation works. Its architecture is a scale of levels; its output is a rating per domain, usually consolidated into one synthetic figure. That figure reports which practices are performed and says nothing about how well the organisation’s risks are managed, because a level is fixed by a set of described attributes — which is why the composite score is the weakest part of the model.
A level is defined by its attributes
what the final figure is built from
attributes ─▶ practice ─▶ level (MIL) ─▶ domain rating ─▶ composite score
defined for performed cumulative: reported one number
the level or not the lowest per domain produced by a
level caps weighting the
the domain score hides
C2M2 states the general property first: “To measure progression, maturity models typically have a scale defining levels of maturity” (line 547). Its own scale is maturity indicator levels 0–3 (line 548), and the model “defines four maturity indicator levels (MILs), MIL0 through MIL3, which apply” independently to each domain (line 843) — one organisation may sit at different levels in different domains. The design decision that matters is what a level is: “A set of attributes defines each level. If an organization demonstrates these attributes, it has achieved both that level and the capabilities that the level represents” (lines 549–550). A level is a set of demonstrated attributes, not a quantity of risk reduced.
Two progressions, one scale
Within a level, two things move. The MILs “define a dual progression of maturity: an approach progression and a management progression” (lines 844–845): the completeness of the security activity, and “the extent to which a practice or activity is ingrained in an organization’s operations (or institutionalized)” (lines 976–978). The management characteristics are largely documentary — at MIL2, “Practices are documented” (line 889) — which is where a score first flatters, since a described process and an operating process earn the same rating. The model is applied through a self-evaluation that “provides a measurement of the implementation of cybersecurity activities” (lines 1261–1262) and counts a practice as performed “if a response of Fully Implemented or Largely Implemented is selected” (lines 858–859). Its output is a report, not a verdict: on completion “a scoring report is generated that provides summary-level depictions of performance relative to the model” (lines 1301–1303), and that report “will identify gaps in the performance of model practices” (lines 1311–1312).
The framework-side analogue: Tiers are a characterisation
CSF 2.0 carries the same idea without the number. “An organization can choose to use the Tiers to inform its Current and Target Profiles” (line 426), because “Tiers characterize the rigor of an organization’s cybersecurity risk governance and management practices” (lines 426–428). There are four — “Partial (Tier 1), Risk Informed (Tier 2), Repeatable (Tier 3), and Adaptive (Tier 4)” (line 1275) — and the glossary calls a Tier “a characterization of the rigor” of those practices (line 1274), not a score. They “describe a progression from informal, ad hoc responses to approaches that are agile, risk-informed, and continuously improving” (lines 431–432, 438). The Core refuses the same reduction: its outcomes “are not a checklist of actions to perform” (lines 243–244).
What aggregation assumes
Aggregation is a decision, not an arithmetic given. SP 800-55 Vol. 1 warns that “A single measure alone may not provide sufficient data to make risk-based decisions” (lines 387–388), so consolidation is wanted. Vol. 2 says what it is for — data “can be aggregated to help executives understand information security” (line 628) — and names the hazard in the same breath: “A common challenge is determining what data to include and how to aggregate large amounts of data to tell a meaningful story” (lines 635–636). SP 800-137 adds the duty behind any such figure: “Metrics are designed and frequencies determined to ensure that information needed to manage risk to within organizational risk tolerances is available” (lines 1425–1427), and the strategy is reviewed for “accuracy in reflecting organizational risk tolerances, correctness of measurements, and applicability of metrics” (lines 1091–1092).
What a single score cannot carry
A level is not a risk statement. Because a level is fixed by its attributes (lines 549–550), it reports which practices are performed, not how well risks are managed; two organisations can hold the same level with very different exposure.
The lowest level caps the domain. The MILs “are cumulative within each domain. To earn a MIL in a given domain, an organization must perform all practices in that level and in the preceding level” (lines 853–855). A domain’s rating is decided by its weakest practice, and averaging across domains spreads that floor effect into a figure that no longer shows where it came from.
The weighting is hidden. A composite blends domains that do not carry equal risk, and the weight each receives is a policy choice the score presents as a result, so comparing two figures can be comparing two weightings.
Documentation can be the improvement. Documentation is itself an attribute of the level (line 889), so a programme can raise its number by writing procedures it does not yet operate, and the rating cannot tell the two apart.
Comparability is narrow. A score is comparable only between organisations that applied the same model, version and sampling: C2M2 counts a practice as performed on a self-evaluation response of “Fully Implemented or Largely Implemented” (line 859), and its report is a “point-in-time view” of the function in scope (line 1303), so a published benchmark invites a comparison the model does not license.
The model warns against the goal a single number suggests: “Striving to achieve the highest MIL in all domains may not be optimal” (lines 866–867). The score belongs in a governance conversation with its domain detail attached, not in a dashboard alone.
Level and prerequisites
L4 — assurance: this sheet takes the criteria, the evidence chain and the records as given and asks what the aggregate built on them can support. It assumes the L1 vocabulary of controls and effectiveness, the L2 records (control objectives and verification criteria; the risk register) and the L3 chain from risk to control to owner to evidence. The purpose and limits of indicators in general, coverage and timeliness, residual-risk indicators and the CSF Profile mechanics belong to companion sheets.
Where to go next
References
- U.S. Department of Energy — Cybersecurity Capability Maturity Model (C2M2), Version 2.1 — https://www.energy.gov/C2M2 — the MIL scale, the attributes that define a level, the dual progression, the self-evaluation and the scoring report.
- NIST — The NIST Cybersecurity Framework (CSF) 2.0, NIST CSWP 29 — https://doi.org/10.6028/NIST.CSWP.29 — the Tiers as a characterisation and the Core’s outcomes as not a checklist.
- NIST — Measurement Guide for Information Security, Volume 1: Identifying and Selecting Measures (SP 800-55v1) — https://doi.org/10.6028/NIST.SP.800-55v1 — the single-measure warning.
- NIST — Measurement Guide for Information Security, Volume 2: Developing an Information Security Measurement Program (SP 800-55v2) — https://doi.org/10.6028/NIST.SP.800-55v2 — aggregation for executives and the aggregation challenge.
- NIST — SP 800-137, Information Security Continuous Monitoring (ISCM) for Federal Information Systems and Organizations — https://csrc.nist.gov/publications — metrics designed and frequencies determined, and the review of accuracy, correctness and applicability.