Kanban in Manufacturing: Signals, Loops and Sizing

Home » Resources » Kanban in Manufacturing: Signals, Loops and Sizing

Estimated reading time: 10 minutes

Kanban in manufacturing is a pull-control method in which a signal authorizes a defined quantity of parts to be produced or moved. Downstream use creates the need; the signal tells the supplying process what to replace. A card, an empty container or an electronic message can carry that instruction.

The benefit is a clear limit on replenishment. Instead of making another batch because a machine is free, you replace what the next process has actually taken. That reduces unnecessary stock and makes missing replenishment easier to spot.

But a signal does not create machine time. If your repeat components share a press, machining centre or bending machine with customer orders, the replenishment work still needs a place in the schedule. Use kanban to authorize the quantity, and SkyPlanner APS to fit the accepted work into the factory’s available capacity and materials. You spend less time reconciling separate queues and see conflicts before they become a shortage at assembly.

Production kanban and withdrawal kanban authorize different actions

A manufacturing kanban is more than a card moving across a To Do, In Progress and Done board. It gives permission for a specific replenishment action. Two signals commonly work together:

Signal	What it authorizes	What it does not authorize
Production kanban	The supplying process makes the specified part and quantity to replace a withdrawal.	Making extra containers just to keep a machine busy.
Withdrawal kanban, also called transport or conveyance kanban	A material handler takes the specified quantity from its supplying location to its consuming location.	Manufacturing another batch merely because transport has been requested.

A supermarket is the controlled stock location between processes. Each part has a designated place and a defined replenishment quantity. The downstream process takes what it needs; the upstream process replaces that withdrawal.

This means kanban does not require zero inventory. The stock in the supermarket provides availability while replenishment takes place. The discipline is to control that stock rather than continually add to it.

Follow one container through the two-card circuit

Imagine an assembly line using a repeat bracket supplied by a shared bending machine. Brackets arrive in containers of 20. Finished containers wait in a supermarket near bending, and assembly has its own point-of-use stock.

In this example, the withdrawal signal is released when assembly opens a fresh container, not when the last bracket is gone. That gives material handling time to bring the next container while assembly uses the opened one.

1. Assembly starts using a container. The operator removes its withdrawal card and puts it in the collection point. The card requests another container of 20 brackets from the supermarket.

2. The material handler collects the request. At the supermarket, the handler takes a full container of the correct bracket. Its production card is removed and placed in the production collection point. The withdrawal card is attached to the container being delivered to assembly.

3. Bending receives the production signal. The returned production card authorizes one replacement container of 20 brackets. It does not authorize an open-ended run.

4. The replacement is made and checked. Once the correct quantity of usable brackets is ready, the production card accompanies the container back to its supermarket location.

5. The circuit repeats as assembly consumes parts. If no further production cards return, there is no authorization to make more for this loop.

The cards carry the part number, quantity, supplying and consuming locations, and an identifier that distinguishes one authorization from another. An electronic signal needs the same clarity: digitizing an ambiguous instruction does not make it a reliable replenishment rule.

Other designs use an empty bin as the trigger. That can work too, provided the remaining stock covers the full response time. Decide which event releases your signal and train everyone to use it consistently; mixing “container opened” and “container empty” changes how much time the loop has to respond.

Choose parts that can be replenished reliably

A repeat component with regular consumption, a stable specification and a reasonably predictable replenishment route is a useful starting point. A common bracket used across several assemblies may be a better candidate than a customer-specific part ordered once a year.

Before introducing cards, establish that the supplying process can keep up with average demand. If assembly consistently consumes parts faster than bending can make them, more cards only postpone the shortage. You need a capacity or demand decision, not a larger pile of authorizations.

Setup time matters as well. One card can represent one container without requiring a separate machine setup for every container. Where several returned cards are grouped into a production batch, define the batch trigger and include the waiting time in replenishment lead time. Do not quietly fill the batch with unrequested stock.

Agree on the operating rules before the pilot starts:

• Produce or move only against the appropriate valid signal, in its stated quantity.
• Keep the authorization traceable while it is waiting, being processed or accompanying finished parts.
• Do not pass defective parts downstream or count quarantined stock as usable replenishment.
• Give shortages, lost cards and urgent exceptions a named owner. Reconcile the original authorization before issuing a replacement, so one need does not become two orders.
• Change card counts deliberately rather than adding unofficial cards whenever the loop struggles.

These rules give the limit its meaning. Without them, kanban becomes another local list layered over the existing production plan.

Size the loop from demand during replenishment time

For a fixed-container production loop, a practical starting calculation is:

Number of production cards = round up [(D × L + S) ÷ C]

Here:

• D is the average demand rate at the loop’s withdrawal point, in parts per time unit.
• L is the replenishment lead time, using the same time unit: from the withdrawal that releases the production signal until a usable replacement is available at that supermarket.
• S is an additional safety allowance in parts for the variation you intend to cover.
• C is the number of usable parts represented by one card or container.

The logic is straightforward: cover expected consumption while replacement is underway, add the chosen protection, then convert that quantity into whole containers. Round up because a fraction of a card does not authorize a complete container.

A worked production-card example

For the bracket’s production loop, imagine these planning inputs:

Input	Illustrative value
Average withdrawal rate from the bending supermarket	40 brackets per operating hour
Withdrawal-to-usable-replacement lead time	3 operating hours
Additional safety allowance	30 brackets
Container quantity	20 brackets

Expected demand during replenishment is 40 × 3 = 120 brackets. Add the 30-bracket allowance to get 150. Divide by 20 and round 7.5 up to 8 production cards.

Those eight cards represent 160 brackets of authorized production-loop quantity. They are divided between finished supermarket stock and replenishment waiting or underway. They do not mean 160 brackets on the shelf plus another 160 in production. When a full container leaves, its production card returns to authorize replacement; you do not create a new permanent card.

This calculation covers the bending production loop only. The withdrawal circuit serving assembly has its own transport time and point-of-use stock requirement. Do not assume that the same eight cards also size every downstream location or the entire factory’s WIP.

The 30-bracket allowance is an illustrative policy choice, not a universal percentage or a guaranteed service level. Set your protection using actual demand and replenishment variation, the consequences of a shortage, and the service you need. Review it as the process becomes more reliable.

Measure the whole response, not just cutting or bending time

A three-hour replenishment assumption must cover more than the minutes at the machine. Include signal collection, order handling, waiting for a batch or resource, setup, processing, inspection and return to the supermarket. Include any transport or receiving activity inside the boundary you chose.

Keep the calendar consistent. If demand is expressed per operating hour, the lead time must account for the operating periods and shift gaps that affect replenishment. A supplier’s two calendar days and assembly’s hourly usage cannot be combined without that conversion.

The formula gives you a starting limit. Measure actual withdrawals, replenishment times and stockouts to find out whether that limit works. A neat calculation cannot compensate for a collection route that skips signals or a machine queue that regularly doubles the assumed response time.

Make replenishment lead time credible on shared machines

Return to the bracket example. The loop was sized around a three-hour response, but the bending machine also has customer-order work. If those jobs push replacement out to five hours, the original calculation no longer describes the process.

At the same demand and safety allowance, five hours would produce a starting requirement of 12 cards, not eight. Increasing the count may protect supply, but it also commits more inventory to a problem that could be excessive waiting. First determine whether the extra time comes from unavoidable capacity limits, avoidable queueing or unreliable execution.

This is where SkyPlanner is valuable. Put accepted replenishment orders into the same finite schedule as the other work competing for that machine. Define their required completion times, routes, quantities and priorities. SkyPlanner’s scheduling AI, Arcturus, calculates machine assignments and operation timing against the configured capacity, shifts and dependencies. The Gantt view makes that sequence visible.

With material-aware scheduling enabled, Arcturus also considers material availability and incoming supplies. A bracket refill should not look ready to start if the sheet material it needs is unavailable. SkyPlanner integrates with any ERP or MES through its open REST API, connecting existing order and production data with the schedule rather than asking you to maintain a separate planning island.

You steer the priorities; Arcturus handles the scheduling calculation. When circumstances change, recalculation helps you see how replenishment and other orders fit together. That reduces hand sequencing and lets you address a conflict without silently sacrificing another delivery.

SkyPlanner’s ASAP and JIT scheduling directions support that timing decision. ASAP seeks an early slot; JIT schedules work closer to its due date. Neither setting is the consumption signal. The kanban policy decides what is authorized, and scheduling decides when that accepted work can run.

What daily scheduling changed at KJH-Comp

KJH-Comp, a sheet-metal contract manufacturer, previously used C9000 and spreadsheets for coarse weekly loading. Day-level machine assignment remained manual, with two supervisors each spending at least half their time on scheduling.

The company kept its ERP and introduced SkyPlanner for finite scheduling and automatic machine assignment. Its account describes clearer visibility when load builds up on the four-metre bending machine, and less early work taking the place of jobs shipping sooner. Scheduling responsibility moved from two supervisors to one, with the other moving to different tasks.

That experience illustrates the scheduling problem behind a replenishment promise: a weekly load total is not enough to decide what should run next on a shared machine. SkyPlanner makes the daily sequence manageable as the range of work grows.

Start with one loop and prove the response time

Pilot one repeat part with the people who consume, move and make it. Mark the supermarket location, standardize the container quantity, issue the agreed cards and make the release event unambiguous. Confirm the starting stock and open replenishment work so the loop begins with a reconciled count.

Track actual signal-to-return time, shortages, usable stock, overdue authorizations and emergency interventions. When something fails, follow the loop: was the signal collected, was the refill scheduled, were materials ready, and did good parts reach the correct location?

If stock remains consistently higher than needed, test a controlled reduction in cards and observe the result. If shortages recur, investigate the timing and capacity assumptions before adding stock. Review the loop again when demand, container size, setup policy or shifts change.

For a factory with shared resources, choose SkyPlanner to keep authorized replenishment and customer orders in one production schedule. Bring a repeat part, its consumption rate, container quantity, route and competing machine jobs to a demo. You can then examine the question that determines whether the loop will work: can replacement finish before the stock protecting consumption runs out?

Book a SkyPlanner demo to explore that part in your production context, or try SkyPlanner with demo data to see how its scheduling works.