Push vs. pull system: producing by forecast or by consumption

In one line

Push is producing according to a plan and sending what was produced forward; pull is producing only when the next workstation consumes and signals.

What it is

Niebel and Freivalds present it when describing just-in-time: it is based on a pull system, in which the demand for parts comes from the station that is further down the flow, rather than a push system, in which parts are produced independently of the system's needs, which generates long queues and bottlenecks.

Bellon Álvarez illustrates it with the organization of work: in Taylorism the production flow (the relationship between workstations) goes from top to bottom; in kanban the process is the reverse, and the organization of work and the management of production start from the last workstation. Hence, he says, the search for the "minimal company", the lean company.

	Push	Pull
What triggers production	The plan or the forecast	Consumption by the next workstation or by the customer
Where the information goes	From front to back, in the plan	From the back (the customer) toward the start, with a signal
What happens if a workstation is faster	It keeps producing and the material piles up	It stops when the authorized space is full
Inventory between workstations	Grows until someone notices	Limited by design (cards, containers)
Typical tool	Production plan, MRP	Kanban

Why it matters

The Universidad Icesi article points out an effect that often goes unnoticed: excess inventory anywhere in the process brings storage costs, longer lead times and damaged products, and it can also hide a quality problem, because you cannot see that defective units are being produced until the conforming product has been consumed, and so the correction is delayed. A pull system, by limiting the material in circulation, makes the problem show up quickly.

Real example

Meyers and Stephens describe the case of a golf equipment manufacturer with a process-oriented layout: it had six months of work-in-process inventory. When it switched to a product-oriented layout, with the parts flowing from workstation to workstation, it reduced that inventory to less than two days. It is not kanban itself, but it is the same logic: material moves by consumption and does not pile up waiting.

An illustrative (made-up) example with numbers, to see the mechanism: if workstation A produces 60 pieces per hour and B only processes 40, in a push system 20 pieces per hour pile up in front of B, or 160 in an 8-hour shift. In a pull system, A stops when the space (or the containers) authorized for it is full, and the bottleneck problem at B is seen immediately instead of being buried under the inventory.

When each one makes sense

• Pull: repetitive flow, relatively stable demand, visible consumption between workstations, few changeovers or quick ones.
• Push: made-to-order products or those with very irregular demand, with few repetitive cycles, or long supply lead times where you cannot wait for the signal. In practice many plants combine both: pull inside the plant, push (by plan) toward distant suppliers.

Benefits of looking at the flow this way

• It changes the question from "how much can I produce" to "how much does the next one consume".
• It gives an objective criterion for sizing intermediate inventory (see the kanban calculation).

Limitations to keep in mind

• Pull without quick changeovers and without stable quality causes stoppages: this is what explains why just-in-time demands SMED and zero defects.
• Meyers and Stephens, in their glossary, contrast kanban ("that pulls") with traditional inventory systems "that push"; in practice plants mix both.

In summary

The difference between push and pull is not a vocabulary fad: it defines where material piles up and how quickly problems are discovered. Pull limits the material in circulation by design; push lets it grow until the plan turns out to be wrong.