Software factory bottleneck

Busy is not the definition

The busiest stage is not always the bottleneck. A stage can be busy because work is flowing through it easily, while another stage quietly limits completed output. The bottleneck is the capability whose constraint affects the rate of the whole system.

In software delivery, the constraint may be a required security reviewer, a slow test environment, an unavailable product decision, a deployment window, or a service that only one team understands. The evidence usually appears as a queue before the stage, starvation after it, or a sustained change in cycle time.

Test the bottleneck hypothesis

State the unit of work and observe the queue over time. Compare arrival rate, service rate, queue age, work in progress, and completed throughput. Then make a small change that protects or expands the constrained capability. Examples include adding reviewer coverage, reducing batch size, improving a test, or changing an approval rule.

After the change, check whether flow improved and whether the constraint moved elsewhere. A temporary incident or vacation can create a short-lived queue without establishing a structural bottleneck.

How Weave can help

Weave can connect queue evidence to the pull requests, review rounds, change sizes, and quality patterns behind it. That helps a team investigate the mechanism rather than infer a bottleneck from a single utilization number. Team workflow records and conversations remain necessary to choose the right intervention.