# A vulnerability backlog is measured in analyst hours, and the count of findings is only the first input

Guide, 2 Nov 2026, read 10 Nov 2026

**The short answer.** A backlog of 20,000 findings at 25 minutes of analysis each is 8,333 analyst hours, which is 463 working days for three analysts giving 30 hours a week. Measure analysis time per finding first, then the share removed before analysis, then fix time, and keep the three apart.

## Count findings, then convert them to hours

A finding count says little about workload. Ten thousand findings that are duplicates of one base image take an afternoon. Ten thousand distinct findings across a hundred teams take a year. The conversion that makes the count useful is minutes of analyst time per finding.

The Backlog Calculator on this site uses a plain formula: findings left, times minutes per finding, divided by 60, gives analyst hours. Hours divided by (analysts times weekly queue hours divided by 5) gives working days. Its defaults are 20,000 findings, 25 minutes, 3 analysts and 30 queue hours a week, which produce 8,333 hours and 463 working days. The 25 minutes is an input to replace, not a benchmark.

| Open findings | Analyst hours at 25 minutes | Working days, 3 analysts at 30 hours a week |
| --- | --- | --- |
| 5,000 | 2,083 | 116 |
| 20,000 | 8,333 | 463 |
| 50,000 | 20,833 | 1,157 |

## Measure your own minutes per finding

Take a sample of 30 findings that were closed last month and time them from the moment an analyst opened the finding to the moment a decision was recorded: fix, mitigate, accept, or dismiss. Use the median, not the mean, because a few findings that took two days will pull the mean far above what a typical finding costs.

Split the sample by type. A dependency finding in a repository the team owns, an operating system patch on a managed host, and a cloud misconfiguration have different costs, and the split shows where an agent would remove the most time.

- Time to decide is analysis time. It ends when the finding has a recorded disposition.
- Time to fix is engineering or operations time. It belongs to another team and another queue.
- Time to verify is the re-check after the change. Count it only if your process requires one.

## Separate removal from fixing

A platform can reduce the backlog in two different ways. It can remove findings before a person reads them, by showing they cannot be used in your environment or by merging duplicates. Or it can shorten the work on findings that remain, by preparing the fix. The calculator takes the first as a percentage removed before analysis and leaves it at 0 until you set it.

Vendor figures for removal vary and carry different evidence. Maze states that 90% of the vulnerabilities it analyzes are not exploitable, and in one customer story reports 51% of findings judged not exploitable (vendor's claim, read 10 Nov 2026). Cogent states that for one disclosure it narrowed 2,266 affected assets to 16 (vendor's claim). Setting the percent removed to 25 in the calculator cuts the default case from 463 working days to 347. At 90, it falls to 46. Whichever figure you enter, test it on a sample of your own findings in a trial.

## Set the target against a public deadline

Federal agencies under CISA BOD 22-01 must remediate vulnerabilities in the Known Exploited Vulnerabilities catalog within two weeks when the CVE was assigned after 2021, and within six months for older CVEs. Private companies are not bound by it, but the directive gives a dated number to compare a queue against.

If the hours figure is far above the capacity of the team for the next two weeks, the backlog is a staffing and automation question. If it is close, the problem is routing and ownership, and the guide on ownership mapping is the better next read.

## See which lever moves the answer most

Four inputs set the result, and they do not move it equally. Cut the minutes per finding from 25 to 10 and the default case falls from 463 working days to 185. Double the analysts from 3 to 6 and it falls to 231. Remove 25% of findings before analysis and it falls to 347. Remove 90% and it falls to 46.

The comparison is the reason to run the calculator before a vendor meeting. If the lever a platform pulls is minutes per finding, test minutes per finding. If it removes findings, test the evidence for the removal. A platform that does neither is not changing the number that matters to your queue.

## Related pages

- [Guides](/guides/)
- [Automated vulnerability remediation: what an agent can fix without a human](/guides/what-an-agent-can-fix-without-a-human/)
- [AI vulnerability management in the weekly remediation meeting](/guides/ai-vulnerability-management-in-the-weekly-meeting/)
- [How approval should work when an agent changes production](/guides/how-approval-should-work/)

## Sources

- Maze Security, home page: https://mazehq.com/ (read 10 Nov 2026)
- CISA BOD 22-01: Reducing the Significant Risk of Known Exploited Vulnerabilities (archived copy): https://web.archive.org/web/2024/https://www.cisa.gov/news-events/directives/bod-22-01-reducing-significant-risk-known-exploited-vulnerabilities (read 10 Nov 2026)
- Cogent Security blog, 10 Nov 2026: https://www.cogent.com/blog/cogent-expands-its-ai-cyber-defense-factory-with-specular-acquisition (read 10 Nov 2026)

Last reviewed 10 Nov 2026. Canonical: https://automatedvulnerabilityremediation.com/guides/how-to-measure-a-backlog/
