A pull system is a production control method that authorizes work only when downstream demand signals that it is needed, rather than pushing material forward on a forecast or a fixed schedule. In a pull system, a workstation produces a part only after the next process (or the customer) consumes one, so inventory is replenished based on actual consumption. The result is less work in process (WIP), shorter lead times, and problems that surface quickly because there is no forecast-driven buffer hiding them.
The clearest way to understand pull is to contrast it with push. In a push system, a central schedule tells every workstation what to make and when, based on a demand forecast. Each station runs to its own plan and hands finished parts to the next, whether or not that next station is ready. When the forecast is wrong, material piles up between stations, WIP inflates, and defects can travel far before anyone notices.
A pull system inverts the trigger. Nothing moves until the downstream process pulls it. This directly attacks several of the seven wastes of lean, especially overproduction and excess inventory, which are considered the most damaging because they conceal every other problem. Pull does not eliminate planning: you still forecast at the aggregate level to size the system. It changes what authorizes the individual unit of work.
People often use "kanban" and "pull" as synonyms, but they are not the same thing. Pull is the principle. Kanban is one signaling mechanism that implements pull. A kanban (Japanese for "signboard") is a card, bin, empty square, or electronic signal that authorizes replenishment. When a downstream station consumes a container of parts, the card returns upstream as permission to make one more container, and no more.
You can build a pull system with other signals too: a two-bin setup, a marked floor location, a reorder point, or a broadcast sequence. Kanban is simply the most common and most visual option. For a deeper treatment of card mechanics, container sizing, and board design, see the guide to kanban in manufacturing.
The supermarket pull system is named after the grocery insight that inspired Toyota: a store shelf holds a fixed quantity, and staff restock only what customers take. In manufacturing, a supermarket is a controlled buffer of finished parts held between two processes. The downstream process withdraws what it needs, and a withdrawal kanban authorizes the upstream process to produce exactly the replenishment quantity.
Supermarket pull fits products with many variants and reasonably steady demand, where holding a small buffer of each part is cheaper than making everything to order. It works best when changeovers are quick and part numbers are manageable. When variety explodes, holding a supermarket for every part becomes impractical, and that is where sequential pull takes over.
In a sequential (or FIFO) pull system, you do not hold a supermarket of finished variants. Instead you make one of each order in the exact sequence the customer requested, and material flows through a first-in, first-out lane with a capped size. The FIFO lane itself acts as the signal: when it is full, the upstream process must stop, which prevents overproduction just as a kanban would.
Sequential pull suits high-variety, low-volume, or highly customized production where holding finished stock of every configuration is wasteful. Many real plants run a mixed model: supermarket pull for high-runner parts and sequential pull for the long tail. Leveling the release of orders into these lanes with heijunka (production leveling) keeps both types stable and prevents demand spikes from whipsawing the line.
The headline benefit of pull is dramatically lower work in process, which shortens lead time. Little's Law makes the link exact: average lead time equals average WIP divided by average throughput. Cut WIP without cutting throughput and lead time falls in direct proportion. Other benefits follow:
Consider an assembly cell that runs a push schedule and averages 600 units of WIP on the line. It produces (throughput) 120 units per day. By Little's Law, the average lead time is 600 divided by 120, which equals 5.0 days. Every order sits in the system for a working week before it ships.
The team converts the cell to supermarket pull. They cap replenishment with kanban so that only 15 containers of 10 units each are ever authorized between stations, capping WIP at 150 units. Throughput stays at 120 units per day because the bottleneck did not change, only the amount of material waiting did. New lead time is 150 divided by 120, which equals 1.25 days.
That is a 75 percent reduction in lead time (from 5.0 to 1.25 days) and 450 fewer units of cash tied up on the floor, achieved without buying a machine or adding a shift. If a defect appears, it now surfaces within about a day instead of a week, which shrinks the quantity of suspect parts you have to contain and rework. Pairing this with disciplined standard work keeps cycle times stable enough for the kanban count to stay accurate.
Pull is not a switch you flip. It depends on reliable equipment, stable cycle times, quick changeovers, and good quality, because a starved or flooded lane exposes any instability instantly. A pull system also runs on trustworthy consumption data: if you do not know what was actually made and used, you cannot size kanban loops or spot when a loop is drifting. This is where real-time production data matters, and it is exactly what Fabrico provides as the foundation. Fabrico gives real-time OEE and production monitoring so you can see true throughput and downtime, and it captures that data even on older machines without a PLC using computer-vision monitoring. Its CMMS keeps the assets healthy enough for small buffers to be safe. Fabrico does not run the scheduling logic of a pull system for you, but it supplies the accurate, live signal of what is actually happening on the floor, which is the data every pull design depends on.
They are closely related but not identical. Just-in-time (JIT) is the broader goal of producing and delivering only what is needed, when it is needed, in the amount needed. A pull system is the primary mechanism used to achieve JIT on the shop floor. You implement pull (often with kanban) to make JIT real, so pull is the how and JIT is the why.
No. You still forecast at the aggregate level to size the system: how many kanban to release, how big to make supermarkets, and how much capacity to staff. What changes is that the forecast no longer authorizes each individual unit of production. Consumption does. The forecast sets the guardrails; the pull signal makes the moment-to-moment decisions.
Yes, but you usually mix approaches. High-runner parts with steady demand suit supermarket pull, while low-volume or custom parts suit sequential FIFO pull made to order. Leveling incoming orders with heijunka and keeping changeovers short are what make high variety manageable. Without those, the number of supermarkets needed can become impractical.
Ready to give your pull system the accurate, real-time data it depends on? See how Fabrico turns live floor signals into lower WIP and faster flow by booking a Fabrico demo.
Zaloguj
Blog