A supermarket pull system is a lean method where each process keeps a small, controlled store of finished parts, a supermarket, and only produces to replace exactly what a downstream process withdraws. The withdrawal sends a signal, usually a kanban, that authorizes the upstream process to replenish. It is the alternative to scheduling every step from a forecast.
The supermarket exists for the awkward reality that you cannot always connect two processes in true one-piece flow. When a machine has to run in batches, sits far from the next step, or serves several downstream lines, you need somewhere to hold parts without losing control of inventory. A supermarket is that somewhere, run by a strict rule: the upstream process makes nothing until the downstream process takes something. That single rule turns a warehouse of guesswork into a self-regulating buffer, and it is one of the foundational mechanisms of lean manufacturing.
What Is a Supermarket in Lean Manufacturing?
A supermarket is a controlled store that holds a set amount of each part a process produces, positioned so the downstream customer can take what it needs, when it needs it, off the shelf. Each process has its own supermarket, and each process produces only to refill what has been withdrawn from that store. The total inventory is capped at the number of parts the supermarket is sized to hold, so it can never grow without bound the way a forecast-driven pile can. The Lean Enterprise Institute defines it plainly: each process produces to replenish what is withdrawn from its supermarket.
The reason it works is that it inverts who decides. In a push system, a central schedule decides what each process makes and when, based on a forecast that is always at least a little wrong. In a supermarket pull system, the downstream process decides, by consuming, and the upstream process reacts by replacing. Nobody upstream has to predict demand; they only have to see what was taken. That is why a supermarket is inseparable from kanban: the card returning from an emptied slot is the instruction to make one more, and the fixed number of cards is what caps the inventory.
Where Did the Supermarket Idea Come From?
From an actual grocery store. Toyota's Taiichi Ohno took the concept from photographs of American supermarkets, where goods sat on shelves in fixed locations and customers took only what they wanted, when they wanted it, and the store restocked each shelf based on what had been removed rather than pushing product onto it by guess. Ohno saw a replenishment system that regulated itself, and Toyota installed its first in-plant supermarket in 1953, in the machine shop of its main plant in Toyota City. The name stuck because the analogy is exact: the shelf, the withdrawal, the restock. What a shopper does to a cereal aisle is what a downstream process does to a lean supermarket.
The insight was that a store is not the enemy of flow if it is bounded and pulled. A pile of parts made to a forecast is waste; a small, capped store that only refills what leaves is a control system. That distinction, between inventory that grows on prediction and inventory that is replaced on consumption, is the whole difference between push and pull.
How Does a Supermarket Pull System Work?
In operation, the loop is simple and unforgiving. The downstream process withdraws a container of parts from the supermarket to use it. That withdrawal frees a kanban card, which travels back to the supplying process. The card is the only authorization the supplying process gets to make more, it produces exactly the withdrawn quantity, in the sequence the returning cards dictate, and places the finished parts back on the supermarket shelf. No card, no production. Because the number of cards is fixed, the shelf can never hold more than its designed maximum, and overproduction becomes physically impossible.
The discipline is the hard part. It is always tempting for the supplying process to make a bigger batch "while the machine is set" or to build ahead because a shift is slow. Every time that happens, the cap is broken and the supermarket stops being a control system. A supermarket run honestly holds inventory flat and exposes real demand; a supermarket run loosely is just a warehouse with a nicer name. This is the same rigor a full pull system demands, which is why supermarkets appear on nearly every future-state value stream map as the connection between processes that cannot flow directly.
How Do You Size a Supermarket?
Sizing a supermarket means deciding how many parts each shelf holds, enough to cover demand during replenishment, plus protection against variation, and no more. Oversize it and you are back to hiding problems in inventory; undersize it and the downstream process starves. The parts break into three layers.
- Calculate cycle stock. This is the average demand during the time it takes the supplying process to replenish. If downstream uses a container an hour and replenishment takes three hours, you need at least three containers just to cover normal consumption.
- Add buffer stock for demand variation. Real downstream demand is not perfectly level. Buffer stock covers the swings above average so a busy stretch does not empty the shelf before it can be refilled.
- Add safety stock for supply variation. This layer covers the supplying side failing, a breakdown, a bad lot, a missed changeover. It is safety stock in the classic sense, held against internal unreliability, and it is the layer you attack first when you improve.
- Set the container size and kanban quantity. Divide the total by a sensible container quantity to get the number of kanban cards. The card count is your inventory cap, so it is a deliberate decision, not a default.
- Level the demand where you can. A supermarket sized for wild demand swings is huge; smoothing the schedule with heijunka shrinks the buffer you need. Sizing and leveling are two ends of the same lever.
- Shrink it over time. Treat the initial size as a starting point. As you improve replenishment speed and reliability, remove cards, re-observe, and let the smaller supermarket expose the next constraint, the same reduce-and-expose discipline kanban uses.
Supermarket Pull vs FIFO vs Sequential Pull
A supermarket is not the only way to connect two processes without full flow. It fits when the downstream process needs many different parts at unpredictable times and you can afford to hold a little of each. When you cannot, because parts are large, expensive, or too varied to stock, you use other pull methods instead.
| Method | How it connects processes | Best when |
|---|---|---|
| Supermarket pull | Hold a store of each part; replenish what is withdrawn | Many part types, unpredictable withdrawal, parts cheap enough to stock |
| FIFO lane | A capped first-in-first-out lane between two steps, no store | Processes that must stay in sequence and cannot hold every part variant |
| Sequential (make-to-order) pull | Upstream makes to a signal in the exact order downstream will use | Too many variants or parts too costly to keep any in a supermarket |
Most plants use all three in different places. A future-state map often runs continuous flow inside a cell, a FIFO lane to the next cell, and a supermarket where a shared upstream process feeds several lines. Choosing among them is a matter of part variety, cost, and how directly two steps can be linked, and the cycle time and demand data that drives that choice comes straight off the floor.
By the Numbers
The supermarket concept is documented by the Lean Enterprise Institute, which credits Ohno with adapting it from American grocery stores and dates Toyota's first in-plant supermarket to 1953 in the machine shop of its main plant (Lean Enterprise Institute, Supermarket; Lean Enterprise Institute, Pull Production). It remains a pillar of the Toyota Production System's just-in-time approach, in which the supermarket implements pull at exactly the points where continuous one-piece flow is not possible (Toyota Motor Corporation, Toyota Production System). Its power is structural: by capping each shelf at a fixed card count, it converts inventory control from a daily argument into a property of the system. For the card mechanics underneath it, see kanban in manufacturing.
When Should You Use a Supermarket, and When Not?
Use a supermarket when two processes cannot be joined in direct flow and the downstream one pulls many part types at times you cannot predict: a shared molding or stamping process feeding several assembly lines is the textbook case. Do not use one when you can achieve continuous flow instead, a supermarket is a controlled compromise, and flow is always better where it is possible. Also avoid it for parts so numerous or costly that stocking even one of each is wasteful; those call for sequential make-to-order pull. The mistake to avoid is treating a supermarket as permission to hold inventory. It is permission to hold the minimum, capped, pulled inventory that a process which cannot flow requires, and every improvement should be shrinking it.
Where this connects to running a plant: a supermarket only regulates itself if the withdrawal and replenishment signals are accurate and timely, and on most floors those signals live on paper cards that get lost and counts that are a shift stale. Harmony captures real consumption and production at the line and connects it across your systems, so pull signals reflect what has actually been withdrawn rather than a count from this morning. That keeps the shelf honest and the inventory capped, the way the system was designed to work, and it is the same reason standard work only holds when reality is visible. See it on a running line in the CLS case study or the platform overview.