A fast coding task can still take days to reach users. Value stream mapping makes that gap visible: draw the path work actually takes, then mark active work, waits, queues, and rework on the path. Map reality rather than the clean sequence described by the formal process. section 6.1.1 section 6.1.2
Choose the boundary that matches the question:
- Feature: request to user-visible functionality. Defect: discovery to fix in production. Deployment: merge to live behavior. Each path can have different queues and constraints. section 6.1.1
- System request: entry to user response, across services, databases, queues, and external dependencies. A fast component may sit inside a slow end-to-end path. section 11.3
For delivery work, place timestamps such as ticket creation, first commit, pull request opening, review start, merge, and deployment on the map. Compare active work with total lead time; inspect wait time and queue depth where work stalls. For a request path, separate processing from network and queue waits. The useful target is the delay affecting delivery or the user, not necessarily the step with the most visible activity. section 6.1.2 section 11.3
Small tasks, long waits
A feature takes three hours to code but spends more than eight days in the system, largely waiting in the backlog, for review, and for deployment. Its activity ratio is roughly 2%. Coding faster would leave those waits intact; review flow or deployment cadence is a more relevant question. section 6.1.2
Table 11.4 Checkout request latency breakdown by service
|
Step
|
Service
|
Processing Time
|
Network Latency
|
Wait/Queue Time
|
Total
|
|
1
|
API Gateway
|
2ms
|
—
|
0ms
|
2ms
|
|
2
|
Auth Service
|
15ms
|
3ms
|
0ms
|
18ms
|
|
3
|
Order Service
|
25ms
|
3ms
|
50ms (DB lock)
|
78ms
|
|
4
|
Inventory Service
|
10ms
|
3ms
|
120ms (DB contention)
|
133ms
|
|
5
|
Payment Service
|
5ms
|
3ms
|
800ms (external API)
|
808ms
|
|
6
|
Notification Service
|
8ms
|
3ms
|
0ms
|
11ms
|
|
Total
|
|
65ms
|
15ms
|
970ms
|
1,050ms
|
The checkout latency breakdown separates processing, network, and wait time for each service. Of 1,050ms total, 65ms is processing and 970ms is waiting, including an 800ms external API wait. section 11.3
Use the current map to name a constraint, then sketch a future path with a countermeasure and a measurable outcome. If twice-weekly breakdown meetings make a requirement wait about five days before development, rotating daily coverage could bring that wait below one day without reducing the breakdown effort. Ask which structural wait the change removes, and measure whether it did. section 6.4.2


