A FIFO lane and a supermarket are the two core pull connections that link processes in a lean value stream, and choosing between them comes down to whether the downstream process consumes in the same sequence the upstream one produces. Both mechanisms cap work-in-process and signal production without a central schedule, but they solve different problems. A FIFO lane enforces first-in, first-out flow between two processes that share one sequence. A supermarket holds a controlled buffer of finished items that a downstream process draws from in any order it needs. Picking the wrong one either starves a line or floods a warehouse with inventory nobody pulled.
A FIFO lane is a physical or logical channel with a fixed maximum capacity that connects two processes directly. Parts enter at one end in a defined order and exit at the other end in exactly that order. Nothing jumps the queue. When the lane fills to its cap, the upstream process stops producing. That stop is the pull signal: an empty slot downstream is permission to make the next unit.
FIFO lanes preserve sequence, which matters when downstream operations depend on the order set upstream. Think of a paint line feeding an assembly cell where the assembly order is locked to the paint schedule, or heat-treated batches that must be processed in the order they cooled. The lane is a direct, one-to-one link. It is one of the simplest expressions of a genuine pull system because it needs no card loops or replenishment logic, just a counted capacity.
A supermarket is a managed inventory buffer between two processes. The upstream process replenishes the shelf; the downstream process (the customer) withdraws whatever mix and quantity it needs, whenever it needs it. Withdrawal triggers a signal, usually a kanban card or bin, that tells the upstream process to make exactly what was taken.
The defining feature is decoupling. The downstream process is no longer tied to the upstream sequence. It can pull part A now, part C an hour later, and part A again, in any order dictated by real demand. A supermarket is the right tool when:
Use a FIFO lane when the downstream process consumes in the same sequence the upstream process produces. Use a supermarket when it does not. That single test resolves most cases. If you can honestly connect two processes with continuous flow, do that first and skip both mechanisms. When continuous flow is not possible, ask whether one shared sequence can survive from start to end of the segment. If yes, a FIFO lane keeps it simple and cheap. If the sequence must break, because of shared demand, batching, or timing mismatch, a supermarket is the correct connection.
These are the connection decisions you draw on a value stream map, and getting them right is what turns a map into a working design rather than a picture.
Suppose a CNC machining cell (upstream) feeds a deburring and inspection station (downstream). Both run the same job sequence, so a FIFO lane fits. You size the lane to cover the largest realistic disruption without starving deburring or over-building at the machine.
Assume deburring runs at a takt of 90 seconds per part. Your data shows the machining cell can go down for up to 30 minutes during an unplanned stop before recovery. To protect deburring across that gap you need enough parts in the lane to keep it fed:
So the lane holds a maximum of 22 parts. When it hits 22, the machining cell stops. When deburring pulls one out, the machine may make one more. If your stop data comes from clean MTBF and MTTR figures, this number is defensible rather than a guess. Note the tension: a bigger lane buys more protection but adds work-in-process and lengthens lead time, a relationship Little's Law makes explicit. Size for the disruption you actually see in OEE data, not the worst case you can imagine.
Supermarket sizing follows a different logic because the buffer must absorb both replenishment time and demand variation across a mix of parts. For each part number you calculate a target level that covers three components: the average demand during the upstream replenishment lead time (cycle stock), a buffer for demand swings, and a buffer for upstream unreliability. A part with a two-hour replenishment loop and steady demand needs a small shelf; a part with a long changeover-driven loop and lumpy demand needs a larger one. Techniques from safety stock calculation and reorder point logic apply directly. High-runner parts justify a supermarket; rare parts are often better made to order to avoid dead stock, a split that ABC analysis helps you draw.
Sizing a FIFO lane or a supermarket honestly depends on trustworthy numbers about how your equipment behaves: how often it stops, how long recovery takes, and how much real output each process delivers per shift. Fabrico is the real-time data foundation for those decisions. It provides real-time OEE and production monitoring so your lane and buffer sizes rest on measured availability and throughput rather than assumptions, including computer vision on machines that have no PLC. As a field-ready CMMS, Fabrico manages work orders, assets, preventive scheduling, and spare parts, so the disruptions you are buffering against get shorter over time and your buffers can shrink. Fabrico is EU-built with EU data residency. Explore the OEE monitoring overview or the CMMS solution overview to see how the data connects to your pull design.
Yes, and most real value streams do. You choose the connection separately for each pair of adjacent processes based on whether that specific link shares a sequence. A single stream might flow continuously in one segment, use a FIFO lane in the next, and pull from a supermarket where an upstream process serves several customers.
An ordinary queue has no defined limit and no stop rule, so it grows until something forces it to stop, hiding overproduction. A FIFO lane has a fixed maximum capacity, and reaching that cap is a hard signal that stops the upstream process. The cap is what makes it a pull connection rather than a pile of inventory.
The upstream process stops producing. A full lane means downstream cannot yet consume the next part, so making more would only create excess. The stop is intentional and healthy: it exposes an imbalance or a downstream problem immediately instead of burying it under work-in-process, which is exactly the visibility a pull system is meant to create.
Ready to base your pull connections on real machine data instead of estimates? Book a Fabrico demo to see how live OEE and CMMS data help you size FIFO lanes and supermarkets that actually hold.
Login
Blog
