Cycle time percentile
Also known as Cycle time p90, Cycle time p50
Definition
A cycle time percentile is a value below which a chosen percentage of completed work falls. For example, a 90th percentile cycle time of five days means 90 percent of the measured items completed in five days or less.
What a percentile answers
An average cycle time answers a mathematical question about all observations. A percentile answers an operational question about how long a specified share of work takes. The median, or p50, describes the middle item. The p90 is useful when a team wants a view that includes most routine variation.
Percentiles do not promise that the next change will finish by the displayed value. They summarize a historical sample. A small sample can produce unstable results, especially at high percentiles.
Define the measurement first
State when the clock starts and stops. One team may measure from the first commit to merge. Another may measure from pull request open to deployment. Both can be valid, but their values answer different questions.
Also define which items count. Separate abandoned requests, emergency fixes, and unusually large migrations when those differences affect interpretation. Keep the same definitions when comparing periods.
Put the number next to its distribution
The p50 and p90 become more useful when viewed alongside the full spread and stage timing. If the p90 grows while the median stays stable, a smaller group is probably encountering a new queue or source of rework.
Weave connects delivery timing with review and code signals so teams can investigate that group. It helps make the percentile a prompt for improvement rather than a score for ranking people.
How this relates to Weave
Weave helps teams examine pull request timing and compare stages of engineering work. Percentiles make it easier to see a typical change and the slower tail without hiding either in one average.
Explore Engineering intelligence