A small code change can take minutes to write and days to reach a user. A value stream makes that gap visible. In software delivery, it is the end-to-end path from an idea or request to working behavior and user benefit—not just the coding step. section 6.1 section 6.1.3
Map every transformation, queue, handoff, check, and rework loop. Typical stops include a backlog, branch, pull-request review, build pipeline, deployment window, and production operation. Unmerged or undeployed work is inventory: effort already invested but not yet validated by users. A map helps locate delays that a measure of coding speed would miss. section 6.0 section 6.1
A quick fix with a long wait
A session-timeout change took twenty minutes to make, yet its path through delivery took nine days. The mapped stream recorded about 40 minutes of process time and seven days of waiting. Review availability, deployment scheduling, and queue depth—not the edit itself—were the constraints. section 6.1
Classify steps before trying to remove them:
- Value-adding work changes the product in a way the customer values. section 6.1
- Necessary but non-value-adding work includes safety validation or required checks; make it effective rather than simply deleting it. section 6.1
- Pure waste includes avoidable waiting, defects, and rework. Ask where work stalls or loops, not which person looks busy. section 6.1
Do not stop the map at a successful deploy if the user still cannot benefit. A feature may wait for legal approval, billing changes, support readiness, or a feature flag. Likewise, passing tests does not settle whether a service can be operated, scaled, and changed sustainably. The slowest necessary step across that wider path governs delivery. section 6.1.3 section 1.1.5
Use the stream as a diagnostic lens in improvement discussions: compare lead time, review queues, and rework instead of only refining sprint ceremonies. Production signals can point upstream too; a rollback may reveal a missing test or an unrepresentative staging environment. Improve the constraint affecting end-to-end flow, then check whether the change helped. section 10.3 section 12.5.3




