> Index of every page on this site: [llms.txt](https://docs.a-line.cloud/llms.txt)

# Actions insights

> Queue and run time, flaky jobs, duration regressions and your spend forecast.

The **Usage** tab of the Actions page shows how your jobs behave. Pick 24 hours, 7 days or 30 days at the top of the Insights section. The forecast always covers the current month.

## Queue time and run time

**Queue time is not billed; only run time counts.** Queue time is how long a job waits for a runner. Run time is how long the job is on a runner, and it is the time you pay for.

You pay per **vCPU-minute**: a job's run time, rounded up to whole minutes, times its runner's vCPUs. The usage chart and the workflow costs table show billed vCPU-minutes, so an 8 vCPU job that runs 2 minutes counts 16 vCPU-minutes.

Both show as a median (p50) and a 95th percentile, per workflow and per job, with a daily trend.

## Flaky and failing jobs

Each job that failed in the range is listed with its failure rate and the number of runs that failed. **Passed on retry** counts the workflow runs where a failed attempt was followed by a pass, which is the usual sign of a flaky job. Cancelled and skipped runs do not count.

## Duration regressions

A job is flagged when its recent median duration jumps above its own baseline. The table shows both medians and the change.

## Spend forecast

The forecast takes this month's Actions and build spend so far and projects it to the end of the month at the same daily pace. If your organization has a monthly cap, it shows whether the projection stays under the cap and, if not, the day the cap would be reached. Once the cap is reached, new runners wait.
