Growth is a test for any security programme. A team that handles a few hundred assets with a small group of engineers can often manage by personal knowledge and direct conversation. As the estate grows, acquires new business units and adopts more cloud services, that approach fails. Findings multiply, ownership becomes unclear and the backlog grows in places nobody is watching.
Metrics are the natural response, but they can make the problem worse if they are chosen carelessly. A metric that rewards the wrong behaviour, or that looks precise while hiding the real position, can give leaders false confidence. This article describes a small set of measures that help a programme scale while keeping control, explains how to define each one and identifies the signals they can reveal.
Why metrics must be defined before they are reported
A metric is only as useful as its definition. Two teams can report the same figure and mean entirely different things by it. One might count a finding as closed when a ticket is resolved, while another counts it only when a rescan confirms the fix. The difference is significant, and it matters most when leaders compare teams or track trends over time.
Each metric should therefore have a written definition covering what is counted, the source of the data, the time period used and any exclusions. The definitions should be agreed by the people who produce the figures and the people who use them. They should also be reviewed when the estate or the process changes.
Coverage: is the estate being examined?
Coverage measures the proportion of the known estate that has been assessed within a defined period. It is only meaningful if the denominator, the known estate, is accurate. A coverage figure calculated against an incomplete inventory will look healthy while large areas go unexamined.
Coverage should be reported by domain, so that gaps in one area are not hidden by strong results elsewhere. Internet-facing and high-impact systems deserve separate reporting. A warning sign is coverage that improves while the inventory is shrinking, which can indicate retired systems being removed without a corresponding review of what remains.
Remediation timeliness: are important fixes moving?
Timeliness is usually measured by the time between identification and verified remediation, often grouped by priority. The average can be misleading, because a small number of very old items can be hidden by many quick closures. Reporting the distribution, or the number of items beyond their target timescale, gives a clearer view.
The timescales themselves should be set by policy, contract or regulatory expectation, with a stated basis. A metric that measures against an arbitrary target will not help leaders decide where to act.
Backlog health: is the queue growing for the right reasons?
The backlog is the set of open findings. Its size alone says little. Its composition says more. Useful breakdowns include the age of items, the proportion with a named owner, the number of items with an approved exception and the number awaiting a decision. A backlog that is growing mainly through old, unowned items is a serious signal. A backlog that grows during a planned remediation cycle may be normal.
Accepted risks are part of the backlog too. Track how many have passed their review date. An exception that has quietly expired is a common source of unnoticed exposure.
Reopened findings: are fixes holding?
The proportion of findings that reappear after being closed is a valuable indicator of process quality. A high rate may point to fixes that are not verified, to configurations that drift back to an unsafe state, or to builds that reintroduce a known weakness. The metric is most useful when reviewed alongside the causes, not only the count.
Triage effort: is analyst time being used well?
Track the time analysts spend on triage and the proportion of findings that are rejected as false positives or duplicates. A rising false positive rate can indicate a scanner that needs tuning, a change in the estate or a process that is creating noise. Triage effort per finding is often a more honest measure of efficiency than the number of findings processed.
Using the metrics together
No single metric tells the whole story. Coverage can be high while timeliness is poor. Backlog can be low because items are being accepted as risk too readily. The most useful view places the measures side by side, so that a change in one can be interpreted in light of the others.
A small dashboard with five to seven measures, each with a defined owner and a review cadence, is usually more effective than a large one that nobody reads. Include a short narrative with each report explaining what has changed and what decision is needed.
Avoiding vanity metrics
Some measures look impressive and reveal little. Total findings closed, number of scans run and count of tools deployed are activity measures. They show that work is happening, not that exposure is falling. They can be useful for operational planning, but they should not be presented as evidence of control.
The test for any metric is whether it would change a decision. If a figure would not lead anyone to act differently, it belongs in the operational detail, not the leadership report.
A practical review
Choose three metrics currently reported and write their definitions in full. Check whether each definition matches what the data actually records. Then identify one warning sign for each metric and decide in advance what action would follow if that sign appeared. This exercise often reveals that some reported figures cannot be produced reliably, which is a finding in its own right.
Common mistakes
A common error is changing definitions without recording the change, which makes trends appear and disappear. Another is comparing teams whose estates differ in risk and complexity. A third is treating a metric as a target, which encourages teams to game it. Coverage can be inflated by scanning low-value systems, and timeliness can be improved by closing items without verification.
Next steps
Agree a short list of metrics with written definitions and owners. Build the first report from data that already exists, even if it is imperfect, and note where the data is weakest. Improve the data over time rather than waiting for a perfect set before starting.
Frequently asked questions
How many metrics should a vulnerability management report include?
A small number, typically five to seven, is usually enough. Each should have a clear definition, an owner and a decision it supports.
Is mean time to remediate a useful metric on its own?
Rarely. Averages can hide long-running items. Reporting the distribution and the number of items beyond target gives a more accurate picture.
Why does a reopened finding rate matter?
It shows whether fixes are holding. A high rate often points to weak verification or configuration drift rather than to a shortage of effort.
Related reading
If you are working through this topic, these related articles cover neighbouring questions:
