A pull system authorizes upstream production in response to downstream consumption rather than releasing work solely because upstream capacity is available.

A simple kanban can represent permission to produce or move one container of parts. Limiting the number of kanbans limits how much WIP can circulate in that loop.

If each container holds $q$ parts and there are $k$ authorized containers, the nominal WIP limit is

$$WIP_{max}=kq.$$

For 8 containers of 20 parts,

$$WIP_{max}=160\text{ parts}.$$

A WIP limit creates feedback. If downstream stops consuming, containers do not return and upstream eventually stops rather than continuing to build unlimited inventory.

Too little authorized WIP can starve a process when replenishment is slow or variable. Too much weakens the feedback and increases lead time and inventory.

Pull therefore does not mean 'produce exactly one item only when a customer orders it'. It means production authorization is tied to actual downstream withdrawal within a controlled replenishment structure.

The purpose is to expose flow problems and prevent unconstrained overproduction, while the chosen WIP level must still account for demand, replenishment time and variability.