The card leaves the station once the last container is emptied. It is supposed to travel back to the previous process as a signal to produce more. One day it stops making that trip, and nobody notices until material runs short on one line or piles up on another.
A manufacturing kanban is a system of visual signals, usually cards or containers, that authorizes production or the movement of material only once the next process has consumed it. It is called a pull system because nothing gets produced ahead of time. It gets produced because somebody downstream already needs it.
Why the search results point at software
Search kanban and half the results are task boards for office teams: columns labeled to do, doing and done. That tool exists and it works fine for project management. The kanban that solves a production problem on a plant floor is a different thing. It controls physical parts, not desk work.
The confusion runs deeper than the name. A task board can move with a mouse click and nothing real happens. A kanban card that does not come back stops a line or fills up a warehouse, and that happens on the floor, not on a screen.
How a real card actually moves
The simplest setup uses two containers. While one gets consumed, the other waits full. Once the first empties, its card travels to the previous process as an order to produce that exact quantity, no more and no less. The second container goes into use while the first refills.
The number of cards in circulation is not arbitrary. It gets calculated from downstream demand, replenishment time and a safety margin. Changing that number without recalculating is the most common way to break the system without anybody knowing why it started failing.
Where it breaks down in practice
It almost always breaks at the same point: the card gets lost, sits stuck on a desk, or somebody decides they will take it back later and forgets. The system depends on the signal traveling physically and on time. A card stuck for twenty four hours already distorts that day’s production.
Another common mistake is treating the kanban board as an information display, with cards that stay fixed because everybody got used to seeing them there. At that point it stopped being a signal and became decoration, the same problem I have already written about with 5S: whatever nobody audits eventually falls apart.
We covered what it is and where it tends to break. Next comes the part that decides whether it is worth installing in the first place.
When it fits and when it does not
Kanban works best with relatively stable demand and parts that repeat. On extremely high variation products, where every run is different, the card calculation gets so complex that another production control mechanism sometimes fits better. It is not the right tool for every line.
Where it does fit, it connects directly to the flow mapped in a value stream map. Kanban is the signal that makes that map work every single day, not just on the paper where it was drawn.
- Some cards have not been seen returning in weeks.
- The container count does not match the number of cards that should exist.
- Somebody produces just in case, without waiting for the card.
Start here
Count the cards you physically have today against what your original calculation says should exist. If they do not match, you already know where the system is leaking.
If you want to design or fix a kanban system on your real line, that is part of what I work on inside lean and kaizen. Tell me how many stations it connects and which part keeps failing.
Is kanban the same as a software task board?
No. A task board organizes office work into columns. A manufacturing kanban controls physical production with cards or containers that authorize making more only once the next process has consumed what came before.
How many kanban cards do I need?
It gets calculated from downstream demand, replenishment time and a safety margin. It is not a fixed or universal number, it changes with every line.
Does kanban work for any type of production?
It works best with relatively stable demand and repeating parts. On extremely high variation production, the calculation gets complicated and another mechanism is usually worth evaluating.