How to use it
- 1. Press runParts flow left to right; station colour shows working (navy), starved (amber) or blocked (red).
- 2. Pick a controlPush releases work whenever station 1 is free; Kanban caps each buffer; CONWIP caps total WIP.
- 3. Add variabilityThe CV slider makes work times erratic - the real villain of flow.
- 4. Read Little's LawWIP = throughput × cycle time. The live check at the bottom holds once the line settles (kanban, CONWIP); in push mode WIP keeps growing, so it drifts.
The line
Run
Control policy
Process
Mean work time is 10 ticks per station; station 3 runs slower by the drag factor - it sets the line's capacity.
Buffers show queued parts as dots (max 12 drawn). Under push with high CV, watch the buffer before the bottleneck fill without end.
WIP over time - push drifts upward; kanban/CONWIP hold it flat.
Throughput over time - note how little it drops when WIP is capped.
Try these experiments
- Run push at CV 1.0 for a while - WIP climbs without bound in front of the bottleneck while throughput stays flat. All that inventory buys nothing.
- Switch to kanban and find the fewest cards that still keep ~95% of the push throughput - that is your lean buffer size.
- Check Little's Law in the readout: multiply the throughput-per-tick by the cycle time and compare with WIP. It matches once the line settles (kanban, CONWIP); under push WIP keeps growing, so it drifts.
About this tool
The concept
Push releases work on schedule; pull authorises work only when downstream consumes. Pull caps WIP, and capped WIP means capped cycle time.
How it works
Each station's work time is lognormal-ish around 10 ticks. Kanban blocks a station when its output buffer is full; CONWIP blocks only the release of new work.
Winning strategy
Throughput is set by the bottleneck, not by WIP - beyond a small buffer, extra WIP only adds cycle time. Little's Law makes that arithmetic, not opinion.