DORA metrics
Deployment frequency, lead time for changes, change failure rate, and time to restore — benchmarked against the DORA tiers.
The DORA tab on the dashboard measures delivery performance against the industry-standard DORA benchmarks. Metrics are computed only over the workflows you designate as production deployment workflows.
Setup
Open the dashboard's DORA tab and click Select deployment workflows.
Pick the workflows that deploy each service to production. Candidates come from your synced workflows and are pre-sorted by detected GitHub production environments.
Optionally set production_branches patterns per workflow (up to 20 wildcard patterns like release/*). Leave empty for auto-detection.
Save — metrics are computed immediately for the selected time range.
The four metrics
| Metric | Elite | High | Medium |
|---|---|---|---|
| Deployment frequency — successful deploys per day, median service | ≥ 1/day | ≥ 1/week | ≥ 1/month |
| Lead time for changes — commit → successful deploy | < 1 day | ≤ 1 week | ≤ 1 month |
| Change failure rate — failed deploys ÷ all deploys | ≤ 5% | ≤ 10% | ≤ 15% |
| Time to restore — failed deploy → next successful run | < 1 h | < 1 day | < 1 week |
One repository counts as one service, so a busy monorepo cannot inflate the org-wide rate — the org figure is the median of per-service medians.
What counts as a deployment
- Runs on the repository default branch, common production branches (
prod,production,release/*,releases/*,hotfix/*),releaseevents, and version tags (v1.2.3) — or yourproduction_branchespatterns. - Excluded and shown transparently:
pull_request/merge_groupruns, runs on non-production refs, and re-deploys of the already-live commit (scheduled re-runs, manual retriggers). - A successful rollback redeploy after an open incident counts as recovery for MTTR, not as a new change.
What you see on the tab
- Performance summary strip — the overall rating with limiting metrics and the tier shift vs the previous equal-length period.
- Metric cards with benchmark scales and period-over-period deltas.
- Deployment cadence chart — successful vs failed deploys per day with a change-failure-rate overlay.
- Release stability panel — outcome split, restore stats, and open-incident warnings.
- Incidents & recovery timeline — failed → restored pairs with links to the GitHub runs.
With a window shorter than 7 days the metrics are noisy — the tab warns you when the selected range is too short.
Cost analytics
Per-run attribution of GitHub Actions minutes and spend — by repository, workflow, runner type, and storage, with wasted-spend and forecasting.
PR metrics
Delivery analytics for pull requests — velocity, review turnaround, contributor load, discussions, commits, open-PR triage, and AI/bot activity.