Lead Time for Changes
The elapsed time from code committed to code running in production.
One of the four DORA metrics, and the one that most reliably exposes where a delivery system is actually stuck, because it measures elapsed time rather than effort. Waiting counts. Waiting is usually the majority.
Measure it as a distribution, never a mean. The median tells you about the healthy path; the p90 tells you about the changes that hit an approval queue, a shared environment booking or a weekly release train, and that tail is where the improvement lives.
The common misuse is treating it as a productivity metric for individuals, which corrupts it immediately — it is a property of the system, and the fix is nearly always a queue or a handoff, not effort. Its second misuse is optimising it alone: lead time improved by removing tests is not an improvement, which is exactly why the four metrics are read together with change failure rate and time to restore.
The decomposition that makes it actionable: commit to merge, merge to artifact, artifact to staging, staging to production approval, approval to live. Time each segment. The answer is usually embarrassing and sits in one of them.