Mean time to remediate is honest when it reports a median, the open items and the clock it used
Closed items alone flatter the number
The usual calculation averages the time to close items that were closed in a period. That leaves out every item still open. A team that closes ten easy findings in an hour and leaves one hard finding open for a year reports an average of one hour.
Add the open items at their current age, or report them separately as aged items. A good report has two lines: median time to remediate for items closed this quarter, and the count and age of items open beyond the SLA.
Use a median and a percentile
Time to remediate has a long tail. Because of it, the mean is pulled up by a few very old items and can hide a typical week. The median shows the typical item. A 90th percentile shows the tail.
Report both by tier. A 4-day median for emergency items and a 40-day median for standard items says more than a 12-day mean across all of them.
| Measure | What it hides if used alone |
|---|---|
| Mean of closed items | Open items; the long tail |
| Median of closed items | The tail; the open items |
| 90th percentile of closed items | The typical case |
| Count and age of open items past SLA | How quickly easy items close |
State the clock and the population
Say when the clock starts (disclosure, detection, ticket) and stops (merge, deploy, verified closure). Say which findings are in the population: all, critical and high only, or exploitable only. Moving from all findings to exploitable only can make mean time to remediate look worse, because the easy noise is gone, and that is a fair reading.
Vendor figures rarely say either. Averlon's example reports 45 days to 6 for an unnamed customer. Backline states more than two weeks to one hour for a Fortune 500 manufacturer it does not name, and 5 weeks to 2 hours for a data-analytics company. Vicarius states months to hours. Qualys cites a customer patching the most severe threats automatically in under 24 hours (Cintas). Only the last names its customer, and the register records a clock definition for none of them.
A starting definition
Write the definition at the start of the report. When a platform is added, keep the definition the same so the before and after compare.
- Start: public disclosure date, or the date your scanner first saw the asset if later.
- Stop: verified closure, by rescan or version check, or recorded mitigation with an expiry date.
- Population: findings assigned to a tier above planned.
- Report: median, 90th percentile, and open items past SLA with their ages, per tier, per month.
Read a before-and-after with care
A reduction from 45 days to 6 compares two periods, and many things differ between periods. Patch cycles, staffing, the mix of findings and the number of new disclosures all move the figure. A single quarter's change can come from a quiet quarter.
To trust a before-and-after, ask for at least two quarters on each side, the same population in both, and the same clock. Then measure your own baseline for the same length of time before the trial, so that the comparison is yours.
Split the metric by who owns the delay
Mean time to remediate combines waiting and working. Waiting for an owner, waiting for approval, and waiting for a change window are different delays with different fixes. Working time is the engineering or operations effort.
Break the total into three segments: time to decide, time to assign and accept, and time to change and verify. Each has a different cure: evidence for the first, ownership mapping for the second, approval design and rollback for the third. A platform that shortens only one segment will move the total less than its demo suggests, and the segment view shows which one it moved.
Review the definition once a year. When tiers, scanners or the ticket system change, the clock can shift without anyone deciding it should, and a trend that looks like improvement may be a change in what is counted. A short note beside each chart, saying what was counted and when the definition last changed, prevents that.
Related pages
- Guides
- Vulnerability ownership at scale: grouping, routing and expiry
- The remediation SLA that holds
- What to ask in a vulnerability remediation platform trial
Sources
5 sources, each with its read date
- Averlon blog on automating remediation, read 10 Nov 2026
- Backline case study, read 10 Nov 2026
- Vicarius, patchless protection, read 10 Nov 2026
- Qualys, Cintas case study, read 10 Nov 2026
- CISA BOD 22-01: Reducing the Significant Risk of Known Exploited Vulnerabilities (archived copy), read 10 Nov 2026
Last reviewed 10 Nov 2026.