Delivery, Velocity and Flow

Open Analytics → Flow (/analytics/delivery) for delivery detail. This guide keeps its existing velocity URL so earlier links remain valid.

Velocity Metrics

Cycle Time

The cycle-time calculator measures from the earliest available commit to merge for merged PRs with usable commit timestamps. The backend provides median and percentile statistics; check the displayed aggregation rather than assuming every card is an arithmetic mean.

Review Time

Review-time measures depend on actual review timestamps. Missing review events are not interchangeable with an immediate review. Inspect the underlying PR and selected timeframe when a value is absent.

Worked Timeline

Consider a merged PR with this available history, with all times in UTC:

Event Time
Earliest recorded commit Monday 09:00
PR opened Monday 12:00
First timestamped review Tuesday 12:00
PR merged Wednesday 09:00

The first-commit-to-merge cycle time is 48 hours. The PR-open-to-first-review wait is 24 hours. The remaining intervals describe different stages: three hours before PR creation and 21 hours between first review and merge. Those intervals help locate where time accumulated; they do not prove that work was active throughout each interval.

Open PRs have no completed merge timestamp. Missing commit or review timestamps also limit which calculations are meaningful. Inspect the aggregation named by a card: a median for a group of PRs is different from one PR's duration or the group's arithmetic mean.

Rework Rate

The calculator uses commits after the first timestamped review divided by total commits, expressed as a percentage in the interface. With no commits or timestamped reviews it cannot establish meaningful post-review iteration. A high value signals iteration, not its cause by itself.

rework percentage = commits after first review / total commits × 100

For a PR with ten commits, three of which occurred after the first timestamped review, the result is 30%. Those commits could implement review feedback, add planned work or address a changed requirement. The percentage alone cannot distinguish these causes or establish initial code quality.

When review history or commits are absent, the calculator can return zero as a fallback. Check source history before treating a displayed zero as evidence that no iteration happened.

Bottleneck Detection

Flow includes stale PRs, bottlenecks, risky PRs and review flags. Time thresholds are configurable in Analytics Settings; read the current setting rather than relying on the old fixed threshold table.

Visibility Metrics

Review repository and contributor activity in the selected account/team scope. Use the current repository and timeframe controls consistently across comparisons.

Charts

Flow includes velocity trends/overview, reviewer bottleneck information, PR-type trends and review-quality signals when data is available. Open the detail lists to inspect the PRs behind a summary.

Action Items

Check the underlying PRs, reviewer availability and actual blockers before changing team process. These metrics help locate work to investigate; they do not identify its cause automatically.

Observation Investigation Possible response after verification
Long wait for first review Check assignments, reviewer availability and missing provider review events. Adjust review rotation or clarify who owns the review.
Large first-commit-to-merge duration Inspect time before PR creation, review wait and later iteration separately. Split future changes or address the stage that is actually delayed.
High post-review commit share Read the feedback and subsequent commits. Clarify requirements or discuss recurring review issues when the evidence supports it.
A newly stale PR Check whether it is blocked, waiting for review or intentionally paused. Unblock, reassign or close it according to the team's decision.

Compare the same repositories and timeframe when checking whether a process change helped. Use Analytics Settings to inspect current time thresholds rather than applying a universal good/warning/critical table.