Flow distribution

Why the spread matters

Imagine ten changes that finish in one day and two changes that wait for two weeks. The average suggests a middle that no developer actually experiences. A distribution preserves the difference between routine work and work caught in a queue.

The long tail often deserves attention. It may contain changes that need security review, depend on another team, or require repeated fixes. The tail does not make every change slow, but it can reveal a constraint that averages conceal.

Read percentiles with the workflow

Teams commonly look at a median and a high percentile. The median describes the midpoint. A 90th percentile answers a different question: how long do almost all items finish within? The exact percentile is a choice, so document the calculation and the population being measured.

Pair the chart with stages such as coding, review, checks, and deployment. A broad distribution in review points to a different intervention than a broad distribution during implementation.

Use it for learning

Flow distribution is useful for forecasting and improvement when the underlying work is defined consistently. It becomes misleading when teams mix tiny fixes, large migrations, and abandoned work without labeling the difference.

Weave gives engineering teams a way to explore these patterns across repositories and pull requests. The distribution starts the investigation. The team supplies the explanation and chooses the experiment.