Kanban and Pull Systems: Managing Flow Without Overproduction

A kanban pull system manages the flow of work so that you produce only what the next step actually needs, when it needs it, rather than pushing work into the operation to keep people and machines busy. Kanban, meaning "signal" or "card", is the simple mechanism that makes pull work: a signal from a downstream step tells the upstream step to make more, and nothing is made without one. For manufacturers, a pull system is the practical cure for overproduction, the most costly of all the lean wastes because it hides every other problem.

I'm Doug Allen, and after 25 years in manufacturing — learning lean at Goodrich Aerospace and running CI at Moog — my one piece of advice on kanban is blunt: get the signal right and don't over-complicate it. Everything else is detail. A good pull signal, sized properly and honoured by everyone, does more for your flow than a wall of clever software ever will.

This article explains how pull differs from push, how kanban signals work, the benefits of managing flow this way, and how to introduce a pull system without disruption.

Push versus pull

In a push system, work is scheduled and released based on a forecast or plan, then pushed from step to step regardless of whether the next step is ready for it. Machines are kept busy, batches are made to look efficient, and work piles up wherever the flow is uneven. The result is stock everywhere, long lead times, and problems buried under inventory.

In a pull system, the logic is reversed. Downstream demand triggers upstream work: a step only makes more when the step after it signals a genuine need, and nothing is produced speculatively. This keeps inventory low, exposes problems quickly, and ties production to real demand rather than a forecast that is usually wrong in detail. Push keeps everyone busy making things that may not be needed yet; pull keeps everyone making exactly what is needed next.

How kanban signals work

Kanban is the signalling system that makes pull practical. The classic form is a card that travels with a container of parts. When the parts are used, the card is freed and sent back upstream as authorisation to make or move more. No card, no production. The number of cards in circulation sets a firm limit on how much work-in-progress can exist.

Kanban does not have to be physical cards. Common forms include:

• Cards attached to containers or batches, returned upstream when consumed.
• Two-bin systems, where an empty bin is itself the signal to replenish.
• Marked floor spaces or shelves, where an empty square signals the need to make more.
• Electronic signals, where a system flags replenishment without paper.

Whatever the form, the principle is identical: a visual signal, a firm limit on work-in-progress, and production authorised only by real downstream need.

Kanban can be as simple or as complex as the job demands. For consumables — washers, o-rings, fasteners — I think of it like a vending machine: you take what you need, the empty slot is the signal, and replenishment just happens. No paperwork, no meeting, no genius required. The trick is matching the signal to the reality of the part, not gold-plating it.

Pull is usually talked about as getting single-piece flow running, and for many operations that's the right aim. But if you're a high-volume manufacturer, you have to think about how batches flow, not just single pieces. And be clear about what you're actually chasing: the aim isn't building to stock — it's building specific amounts to meet a specific order you can ship as soon as it's finished. Pull tied to real orders is disciplined; pull that quietly drifts into topping up a warehouse is just push wearing a costume.

The benefits of managing flow with pull

Pull systems deliver several linked benefits. By capping work-in-progress, they keep inventory and the cash tied up in it low. By linking production to real demand, they shorten lead times and improve responsiveness. And, crucially, they make problems visible fast, because there is no large buffer of stock to hide them.

That last point is the deepest benefit. Overproduction is dangerous precisely because a mountain of inventory lets a business run for a long time without noticing that a machine is unreliable or a process is unbalanced. Remove the excess stock and those problems surface immediately, forcing the fixes that actually improve the operation. A pull system is, in effect, a mechanism for surfacing the truth about your flow.

There's a mirror image to this that keeps me honest about signals. I've seen assemblies worth £200,000 sitting unbuilt because the cell had run out of 10-pence o-rings. Try explaining that to the boss, or to the customer. A trivially cheap missing part can strand an enormous amount of value, and it's almost always the humble, unglamorous item nobody thought to plan for. A good pull signal on those consumables is exactly what stops it — the cheapest insurance policy in the building.

How to introduce a pull system

Introducing pull works best gradually and on a stable foundation. Trying to switch an entire, unpredictable operation to pull overnight usually fails. A measured approach works better:

1. Stabilise first. Pull works best where demand and processes are reasonably steady, so basic standardisation and reliability help before you start.
2. Start with one product or link. Choose a repeatable, higher-volume item and set up pull between two steps rather than across the whole flow.
3. Size the signals carefully. Set the number of cards or bin quantities to cover real demand and lead time, with a sensible buffer, then plan to reduce it as you improve.
4. Make the rules clear. Everyone must understand that no signal means no production, and that the limits are not to be quietly overridden.
5. Review and tighten. As reliability improves, reduce the work-in-progress limits to expose the next problem and drive the next improvement.

Start small, prove it works, and extend from there. Pull is as much a discipline as a mechanism, and the discipline is built by seeing it succeed on a small scale first.

Frequently asked questions

What is a kanban pull system?

A kanban pull system is a method of controlling production so that work is only made when a downstream step signals a genuine need, using cards or other visual signals. The signals also cap the amount of work-in-progress allowed at any time. Its purpose is to link production to real demand and prevent overproduction.

What is the difference between push and pull production?

In push production, work is released according to a forecast or schedule and pushed forward whether or not the next step is ready, which tends to create excess inventory. In pull production, downstream demand triggers upstream work, so nothing is made without a real need. Pull keeps inventory low and makes problems visible more quickly.

Does a pull system work for low-volume or custom manufacturing?

Pull works most naturally for repeatable, higher-volume items, but its principles can still help in lower-volume settings. For highly custom work you may apply pull selectively, for example to common components or sub-assemblies, while managing bespoke items differently. The key is to apply pull where demand is repeatable enough for a signal to be meaningful.

Where to go next

A pull system is one of the most effective ways to cut inventory, shorten lead times and surface the problems worth fixing. If you would like a structured way to assess your operation and get a tailored plan for where pull would pay off first, ProgressXL is coming soon. To talk through introducing kanban and pull in your own operation, get in touch.