The valve blocks of the example plant need less than two days of actual work from saw to test. Yet an order takes 15 working days from material release to completion. Where do the other 13 days go?
They go into waiting. And the amount of waiting is set by one number most plants never manage: work in progress.
The law
Little’s Law says that in any stable system:
lead time = work in progress / output
It holds for a supermarket queue, a hospital ward and a machining shop. It needs no assumptions about variation or sequence.
Count, then divide
On one morning, count the valve blocks in progress, physically, per stage:
| Stage | Blocks | Days |
|---|---|---|
| Released, not started | 420 | 2.4 |
| Queue before CNC machining | 980 | 5.6 |
| In CNC machining | 260 | 1.5 |
| Washing and deburring | 180 | 1.0 |
| Queue before assembly | 520 | 3.0 |
| Assembly and test | 250 | 1.4 |
Total: 2,610 blocks. The plant completes 174 blocks per working day. 2,610 / 174 = 15 working days. Most of that time is spent in two queues.
The lever
The machines do not get faster when the floor is full. Output is set by the bottleneck. So the only way to cut lead time without new capacity is to cut work in progress.
With a limit of 1,050 blocks at the same output, lead time falls to 1,050 / 174 = 6 working days. Same people, same machines, same output, nine days less waiting. Capital tied up on the floor falls by about EUR 140,000 at an average value of EUR 90 per block.
Why plants resist
A full floor feels safe. Every machine has work, nobody waits, supervisors can choose. The cost is invisible: urgent orders are buried under orders that are not due for weeks, quality problems are found late, and every change of plan means searching through piles.
Limiting work in progress also feels risky: what if the bottleneck runs out of work? That is why the limit is set carefully and reduced in steps.
Setting the limit
- Count work in progress on three mornings and take the average.
- Set the first limit 20 percent below it.
- Watch the bottleneck: it must not run out of work.
- Reduce the limit by 10 percent each month while output holds.
- Stop when the bottleneck starts to wait, and go back one step.
Measure it
Lead time per order, measured from release to completion, is one of four numbers worth tracking every week, together with on-time delivery, schedule adherence at the bottleneck and work in progress itself. When work in progress falls and lead time does not, something is wrong with the measurement. Little’s Law does not take exceptions.