The Day the Board Stopped Lying: Kanban as a Pull Signal at a Stretched Supplier
A mid-sized fabrication supplier came out of 2020 with a backlog it was proud of and a shop floor it could not see clearly. Orders were piling in as customers rebuilt their own supply chains, and the natural response had been to push more work into the plant as fast as it arrived. By the time we looked at it, every station had a queue, every queue had a story, and nobody could honestly say when a given order would ship. The planning board on the wall said one thing; the plant said another. This is a Lean problem with a Lean answer, and the answer is kanban used the way Taiichi Ohno meant it — as a signal, not a sticky note.
Push felt like control. It was the opposite.
Pushing work means releasing jobs to the floor according to a forecast or a schedule, regardless of whether the next operation is ready for them. It feels productive because every machine is busy. But busy is not the same as flowing. The supplier had heaps of work-in-process — half-finished assemblies waiting between stations — which tied up cash, hid quality defects until much later, and made the true lead time impossible to measure. The board lied because it tracked what had been released, not what was actually moving.
Kanban inverts the logic. In a pull system, a downstream operation only signals for more work when it has capacity to take it. Nothing is produced until there is a real demand signal from the next step in the line. The signal — historically a physical card, here a card on a board and a bin convention on the floor — is permission to make exactly what is needed, when it is needed.
What we actually did
Mapped the flow and counted WIP. We walked the line and recorded how many jobs sat at each station. Two operations held more than half of all work-in-process. That was the bottleneck, hiding in plain sight.
Set WIP limits per station. Each column on the board got a cap. When a column is full, the upstream station may not push more in — it must stop and help, or wait. Limiting WIP is what turns a board into a pull system.
Replaced the schedule with cards. A job moved forward only when the next station pulled a card, freeing a slot. Release into the plant was governed by what the constraint could absorb, not by the order book.
Made the bottleneck everyone's problem. Because work could no longer pile up in front of the constraint, the whole team's attention turned to keeping it fed and running — the heart of Lean flow.
None of this added machinery or people. It changed who decided when work moved, and on what authority.
What changed, and what to watch
Within several weeks the visible chaos drained out of the plant. Work-in-process fell sharply, which freed up both floor space and cash. Because jobs were no longer hidden in queues, the genuine lead time became measurable for the first time, and it was shorter than anyone expected once the queues were gone. Quality issues surfaced faster, because a defect at the constraint stopped the line instead of being buried under three more half-built orders.
A pull signal exposes the bottleneck; a push schedule hides it under inventory.
WIP limits are the active ingredient — a board without limits is just a to-do list on the wall.
Lower work-in-process means faster feedback on defects and a lead time you can actually quote to a customer.
In a recovering, disruption-wary market, a short and reliable lead time beats a long and impressive backlog.
Kanban is not a scheduling app and it is not magic. It is a discipline that forces the system to tell the truth: produce only what the next step has asked for, limit what is in progress, and let the constraint set the pace. Done honestly, it turns a wall of optimistic stickies into a working signal — and a plant you can finally see.
If your operation is busy but slow, and you suspect work-in-process is hiding the real problem, XNM's strategic advisory can help you design a pull system that fits how your shop actually runs.