The closed circuit a single card travels. The heartbeat of pull.
A kanban loop is the smallest functional unit of any pull system. If you can build one loop and have it run reliably, you can build a hundred of them and have a whole shop running on pull. The reverse is also true: if your loops do not run reliably, no amount of supermarket sizing or training is going to fix the system. The loop is where the discipline lives. Get the loop right and pull works. Get the loop wrong and everything else is decoration.
"Every card that moves carries a signal. Every card at rest carries a promise."
A loop has three legs. The signal leg is the card traveling from the point of use back to the supplier process. The production leg is the supplier producing a refill and attaching the card to the new bin. The delivery leg is the full bin traveling forward to the point of use and waiting in the supermarket until consumed. When the consumer takes the bin, the card detaches and starts the signal leg again. The cycle is continuous and self-contained.
Each loop is defined by four things: a single supplier process, a single consumer process, a single part variant, and a single bin size with one card attached. You can have multiple loops between the same two processes if you stock multiple variants; each variant gets its own loop and its own card. You can have multiple loops for the same variant if demand justifies multiple bins; each bin gets its own card. The card is the smallest unit of pull.
Sizing a loop is sizing the bin. The math: bin size equals consumption rate during the supplier's replenishment lead time, plus a defined cushion. If a cell consumes 50 brackets per day and the brake station takes two days to refill a bin, each bin needs at least 100 brackets, plus maybe 30 more for variation. The cushion is small. The whole point of a tight loop is that the cushion is what flow requires, not what feels safe.
The discipline that breaks most loops is cards getting lost. A card that sits at the consumer workstation instead of returning to the supplier is a refill that never happens, and eventually the supermarket stocks out. Lean shops with mature kanban have explicit card-return mechanisms: a hook near the supplier station, a small return cart that runs the route with mizusumashi, or a barcode scan that triggers a digital signal. The mechanism varies; the rule does not. Every card returns every cycle.
Picture a small fab shop running steel parts for two industrial OEMs. The shop has decided to convert one runner part, a steel bracket, to kanban pull. The bracket is consumed at the welding cell at about 40 units per day. The brake station, which forms the bracket from cut blanks, can refill a bin of 50 brackets in about half a day.
The shop builds one kanban loop. A supermarket of three bins (50 brackets each) sits between the brake and the welding cell. Each bin has a card with the part number, bin size, supplier (brake), and consumer (weld). When welding takes a bin, the card goes onto a return rack near the brake. The brake operator checks the return rack each morning and afternoon, produces a 50-bracket refill for each card, attaches the card to the new bin, and returns it to the supermarket. The loop runs without a schedule or screen.
After a month, the loop is steady. The supermarket holds two or three bins at any moment. Stockouts have not happened. The brake operator has a queue of cards to work down, in the order they returned, which is essentially sequenced pull at small scale. The shop's first loop has proved the mechanism. Six more loops get built over the next quarter, each connecting one runner variant to its supplier and consumer.
A kanban loop is the smallest functional circuit inside kanban generally. The store that loops draw from and refill is a supermarket, and the simplest physical implementation of a single loop is a two-bin system where the empty bin is the card. Multiple loops chained together form the basis of a full pull system, with one loop per part variant per supplier-consumer pair.
The questions we hear most about this term.
Long-form guides that pick up where this definition leaves off, written for manufacturers running Arda today.
Same-day setup. No distributor lock-in. Zero stockouts. Top teams double revenue in 9 months.