Mapping Value Streams
This domain tests whether you can build a current-state map that is honest, calculate lead time and flow efficiency from it, and design a future state that addresses the real constraint.
The vocabulary of the map
A map is built from process boxes with data boxes underneath, inventory or queue triangles between them, push arrows, FIFO lanes, supermarkets and kanban posts, information-flow arrows, and kaizen bursts marking problems. The timeline ladder underneath is the point of the exercise: lead time (including waiting) runs along the top, process time (actual work) along the bottom, so the ratio of the two is visible at a glance. In software, the inventory is work waiting in a backlog, in a review queue, or blocked; process time is hands-on work.
Building it honestly
Scope the stream first with a start and end trigger. Walk the flow where the work happens rather than in a meeting room. For each step capture process time, wait time, WIP, batch size, number of people, and percent complete and accurate (%C&A), which exposes rework loops. Add the numbers to get total lead time and value-added time, then compute flow efficiency. Mark problems as kaizen bursts and only then draw the future state, with a target condition and an improvement backlog.
Worked example
Four steps: refinement waits 2 days and works 4 hours; development waits 8 days and works 16 hours; test waits 6 days and works 8 hours; deploy waits 4 days and works 2 hours. Lead time = 2 + 8 + 6 + 4 = 20 working days. Active work = 4 + 16 + 8 + 2 = 30 hours. At 8 hours per working day, 20 days is 160 hours, so flow efficiency is 30 ÷ 160 = 18.75%. The reasoning that earns marks: 130 of 160 hours are wait, so the future state should attack queues and batching before it attacks coding speed.
Traps that cost marks
- Mapping the ideal process or the documented process rather than the actual one.
- Omitting wait time, rework loops, and %C&A.
- Mapping alone at a desk instead of with the people who do the work.
- Trying to map an entire enterprise at once instead of one stream at a time.