How to Measure What Your GitHub Actions Actually Cost

By Kaveh Alemi·Latchkey

Most teams optimise the job they assume is expensive. Pull the numbers first: the answer is usually dependency install, a wide matrix, or re-runs, and almost never the step people were about to rewrite.

GitHub shows you a total, which tells you that CI costs money and nothing about where it goes. Before changing a single workflow, get a per-workflow and per-job breakdown, because optimisation aimed at the wrong step is the most common way teams spend a sprint and save nothing.

There are three numbers worth having: minutes by workflow, minutes by job within the expensive workflows, and the share of runs that are re-runs of a previous failure. The third is the one almost nobody measures and frequently the largest.

Minutes by workflow

# per-workflow usage for the repository
gh api /repos/{owner}/{repo}/actions/workflows --paginate \
  --jq '.workflows[] | "\(.id)\t\(.name)"' |
while IFS=$'\t' read -r id name; do
  ms=$(gh api "/repos/{owner}/{repo}/actions/workflows/$id/timing" \
        --jq '[.billable[]?.total_ms] | add // 0')
  printf "%8.1f min  %s\n" "$(echo "$ms/60000" | bc -l)" "$name"
done | sort -rn

Where the minutes go inside a workflow

# per-job duration across recent runs of one workflow
gh run list --workflow=ci.yml --limit 50 --json databaseId --jq '.[].databaseId' |
while read -r id; do
  gh api "/repos/{owner}/{repo}/actions/runs/$id/jobs" \
    --jq '.jobs[] | [.name, (((.completed_at|fromdate) - (.started_at|fromdate))/60)] | @tsv'
done | awk -F'\t' '{s[$1]+=$2; n[$1]++} END {for (k in s) printf "%8.1f min avg  %4d runs  %s\n", s[k]/n[k], n[k], k}' | sort -rn

The number nobody measures: re-run share

A re-run buys the same work twice. If a meaningful share of your runs are retries of a previously failed run, that is pure waste and no per-minute discount touches it.

# what fraction of recent runs are attempt 2 or later?
gh run list --limit 200 --json databaseId,conclusion --jq '.[].databaseId' |
while read -r id; do
  gh api "/repos/{owner}/{repo}/actions/runs/$id" --jq '.run_attempt'
done | awk '{t++; if ($1>1) r++} END {printf "re-runs: %d of %d (%.1f%%)\n", r, t, 100*r/t}'

Convert minutes to money correctly

Now decide what to change

• Install dominates: cache the dependency directory and key it to the lockfile hash.
• A wide matrix of short jobs dominates: the per-minute rounding is the cost. Consolidate jobs or move to a per-second biller.
• One long job dominates: shard it, or move that job alone to a larger or faster runner.
• Re-runs dominate: nothing about runner choice fixes this. Find the flaky failures and address them at source.
• macOS dominates: it is 10x Linux per minute. Move anything that does not genuinely need macOS.

Frequently asked questions

Related guides

References

• GitHub Actions workflow timing API (verified 2026-08-20)
• GitHub Actions billing and per-minute rates (verified 2026-08-20)
• GitHub-hosted runner specifications (verified 2026-08-20)
• GitHub Actions documentation