Lead time for changes: the trend line that tells you if shipping is getting harder
Lead time for changes is the median hours from a PR or MR being opened to the production deploy that shipped it, plotted weekly. Where a repository records no production deploys, the clock stops at merge instead — and the caption above the chart tells you which one you are reading.
Lead time for changes, p50 by week, with benchmark bands.
What it measures, and where the data comes from
The p50 (median) hours from a pull or merge request being opened to the first successful production deploy of the same repository at or after its merge — the release that shipped it. Failed deploys and deploys to staging or other non-production environments ship nothing, so they are skipped. Median, not average, so one three-week monster PR doesn't drag the whole line around.
- GitHub — pull requests, paired with GitHub deployments flagged as production.
- Azure DevOps — pull requests, paired with release deployments that reached a production stage. The repository is identified from the commit the release was built from.
- GitLab — merge requests, measured to merge: GitLab deployments are not ingested yet.
Three rules decide what lands in the median, and you should know all of them before quoting the number:
- Repository records production deploys → measured to the deploy. A change merged but not yet deployed is left out until it ships; it has no lead time yet.
- Repository records no production deploys → measured to merge. That leaves out any wait in a release queue, so it reads shorter than DORA's definition. The widget is captioned "→ merge", and the DORA scorecard marks the tile (proxy) — or (partly proxied) when a board mixes both kinds of repository.
- The clock starts at PR open, not first commit. Deckgauge does not link a PR to the commits that preceded it, so work done before the PR was opened is not counted. If your team opens PRs late, expect this number to read low.
Before relying on it for a decision, check it against a handful of your own changes: pick ten recent PRs and compare opened, merged and deployed times with what the widget reports.
How to read it
The shaded bands are the DORA tiers: Elite under 24 hours, High under a week, Medium under a month, Low beyond that. Look at which band the recent weeks sit in before you look at the exact hour count — the tier is the story, the number is the detail.
A flat line in a good band is boring, and boring is the goal. Watch the direction, not any single week. One bad week is noise — a bad three-week run is a trend, and trends are what you act on.
What it tells you over time
This line is a leading indicator for review health, not coding speed. Engineers rarely get slower at writing code; what changes is how long a finished change sits waiting on someone else. If lead time climbs while commit volume holds steady, the bottleneck is downstream of the keyboard — reviews, batch size, or a flaky pipeline gate.
Measured over the same definition week to week, it's also the fastest way to prove or disprove a process change. Ship a review SLA or a smaller-PR norm, then watch this line for two to three weeks. If it doesn't move, the SLA isn't the fix — go looking elsewhere.
Example situations
1. The review process is quietly slowing down
Three straight weeks climbing, commit volume flat.
What you're seeing: the line has climbed for three consecutive weeks — Elite to High to Medium — while the team's commit count didn't budge. Nobody wrote less code. Something between "done" and "merged" got slower.
How to react: don't tell people to code faster; that's not where the time is going. Pull up Review Pickup Time first — if PRs are sitting hours before anyone even looks, that's your answer. If pickup is fine but merge still lags, check PR Size Distribution for creeping batch size. Either way, this is a process signal, not an effort signal.
2. A process change actually worked
A step down that holds, after smaller PRs became the norm.
What you're seeing: the line drops sharply around the week you introduced smaller PRs or a review SLA, and — this is the part that matters — it stays down. Not a one-week blip; a new baseline, holding in the Elite band.
How to react: this is rare enough to be worth calling out explicitly. Confirm it's real by checking that PR volume and merge count didn't just drop with it (a quiet quarter can fake a good lead-time line). If the work kept flowing and the line still fell, you have proof a specific change worked.
Frequently asked
- What is lead time for changes?
- The median hours from a pull or merge request being opened to the production deploy that shipped it, plotted weekly. Repositories with no production deployment records are measured to merge instead, and the widget says which. It is one of the four DORA keys.
- Is lead time measured or estimated in Deckgauge?
- Measured from real events — PR opened, merged, and deployed — with two caveats. DORA starts the clock at the first commit; Deckgauge starts it at PR open, because that is the earliest event it observes. DORA stops the clock in production; Deckgauge does too wherever your repositories record production deployments, and otherwise stops at merge and labels the line accordingly.
- What should I do when lead time drifts upward?
- Check PR size and review pickup time first. A rising lead time is far more often caused by larger batches or slower first review than by people writing code more slowly.
Related widgets
- DORA Metrics — this is the measured lead-time tile inside the full four-key scorecard.
- Review Pickup Time — isolates the reviewer-wait slice of lead time.
- PR Size Distribution — checks whether batch size, not review speed, is driving the trend.
- Back to the widget reference.
Last updated