Kanban and Pull Replenishment Systems: How to Let Consumption Drive Supply

How kanban works: the core logic

In a push system, a central planner schedules production based on a forecast and pushes material downstream whether or not downstream processes are ready for it. This creates WIP inventory that accumulates wherever the downstream process is slower than the upstream schedule.

In a pull system, downstream processes pull material from upstream only when they need it. The kanban card or bin is the authorization mechanism: a downstream workstation can only work on material that has been authorized by a kanban signal. When it consumes a container of material, the empty container (or the card attached to it) travels back upstream and authorizes production or procurement of a refill. Nothing is produced or purchased without a kanban signal.

The pull logic controls WIP automatically. Because upstream processes can only produce when a kanban signal arrives, they cannot produce more than downstream demand authorizes. The number of kanbans in the system sets the maximum inventory level, it is a designed inventory cap, not an emergent one.

Kanban sizing: how many kanbans do you need?

The number of kanbans in a loop determines the total inventory held between two points in the value stream. The standard kanban sizing formula:

Number of kanbans = (D × LT × (1 + SF)) ÷ Container size

Where D is the average demand rate (units per day or per shift), LT is the replenishment lead time (the time from when a kanban signal is sent to when the replenished container is available), SF is a safety factor (typically 0.1 to 0.25, 10 to 25 percent buffer for variability), and Container size is the number of units per kanban container.

Example: demand is 200 units per day, replenishment lead time is 2 days, safety factor is 0.15, and container size is 100 units. Number of kanbans = (200 × 2 × 1.15) ÷ 100 = 460 ÷ 100 = 4.6, rounded up to 5 kanbans. Maximum inventory in the loop is 5 × 100 = 500 units.

The kanban formula is structurally identical to a reorder point calculation (see the Reorder Point post): it sizes a buffer against demand during replenishment lead time plus a safety allowance. The difference is implementation: a kanban system manages the buffer visually through cards and containers rather than through a system-generated replenishment order.

Two-bin kanban

The simplest physical kanban implementation is the two-bin system. Each item has two storage bins. When the first bin is empty, the empty bin is sent to the supplier or upstream process as the replenishment signal, and the second bin feeds consumption in the meantime. When the replenishment arrives, the bins are refilled and swapped. The bin itself is the kanban.

Two-bin systems work well for C-class items, consumables, MRO supplies, and any item where the administrative cost of a formal purchase order or MRP transaction exceeds the value of tight inventory control. They are common in maintenance and repair environments, medical supply rooms, and high-mix manufacturing cells for fasteners and standard hardware.

The bin size is set to cover demand during the replenishment lead time plus safety allowance, the same logic as the formula above, but applied to a physical bin rather than a card count.

Supermarket kanban versus signal kanban

Supermarket kanban (also called replenishment kanban or withdrawal kanban) manages a finished goods or WIP buffer between two processes. The downstream process withdraws from the supermarket inventory. When the withdrawal drops the supermarket below a trigger level (the kanban quantity), a replenishment signal is sent upstream. The supermarket holds buffer stock that decouples the upstream and downstream processes, allowing them to operate independently within the lead time window.

Signal kanban (also called production kanban or make kanban) is used when the upstream process produces in batches rather than single units. A signal kanban is not triggered by each individual withdrawal but by the supermarket inventory level dropping below a defined reorder point. When it does, the signal kanban authorizes the upstream process to produce a batch. Signal kanbans are appropriate where setup costs make single-unit production impractical, the upstream process needs to produce a batch to justify the setup, but it only produces when the downstream supermarket calls for it.

Electronic kanban (e-kanban)

Physical kanban cards and bins are effective but require manual handling and can be lost or misplaced. Electronic kanban (e-kanban) replaces the physical card with a system-generated replenishment signal, triggered by a barcode scan when a container is consumed, by a WMS transaction, or by a direct ERP integration. The logic is identical: consumption triggers replenishment authorization. E-kanban scales better than physical cards for large SKU counts and for replenishment signals that must travel to remote suppliers.

Many ERP systems support e-kanban as a replenishment method that can be configured alongside MRP. For standard, repetitive items on short lead times, e-kanban produces cleaner, lower-overhead replenishment signals than MRP-generated planned orders, which must be reviewed and released by a planner.

Kanban versus MRP versus reorder point

These three replenishment methods are not competitors, they address different parts of the demand and lead time spectrum.

Kanban works best for high-frequency, stable demand items with short, reliable lead times. It is self-regulating: consumption generates the signal, and the number of kanbans caps inventory automatically. Planning overhead is minimal once the system is designed and the kanban quantities are set.

Reorder point (ROP) works for similar demand profiles but uses a calculated trigger level rather than a physical container. It is better suited to items where container sizes are impractical, bulk materials, items purchased in variable quantities, and where the demand variability is better captured by a statistical safety stock calculation than by a fixed safety factor. See the Reorder Point post and Safety Stock Calculator.

MRP works for complex, multi-level products where component requirements are derived from a finished goods production schedule (MPS), not from direct consumption signals. MRP is necessary when lead times are long, BOMs are deep, and demand is lumpy or project-driven. The Production Planning and Scheduling post covers MRP in detail.

The practical decision: use kanban for any item where you can size a physical or electronic buffer that fits between the consumption rate and the replenishment lead time. Use ROP for items where the demand variability is high enough to require a statistically sized safety stock rather than a fixed safety factor. Use MRP for dependent demand, components whose requirements are driven by a parent item production schedule rather than independent consumption.

Kanban and the push-pull boundary

In most manufacturing and distribution operations, kanban and MRP coexist. The push-pull boundary is the point in the value stream where the supply strategy switches from push (MRP-driven, schedule-based) to pull (kanban or ROP-driven, consumption-based). Long-lead items upstream of the boundary are planned by MRP. Standard components downstream of the boundary, or held in a supermarket buffer at the boundary, are replenished by kanban.

Placing the push-pull boundary correctly is a strategic decision. Moving the boundary further upstream (toward raw materials) increases the portion of the supply chain managed by pull and reduces WIP inventory, but requires shorter and more reliable replenishment lead times throughout the upstream chain. Moving it downstream (toward finished goods) extends the MRP planning horizon and may increase forecast-driven inventory, but reduces the lead time risk of relying on pull for long-lead items.

Common mistakes

• Setting kanban quantities once and never updating them. Demand rates and lead times change. A kanban sized for 200 units per day of demand will either create shortages (if demand has grown) or excess inventory (if demand has fallen) if not reviewed periodically. Review kanban quantities whenever demand changes by more than 10 to 20 percent or when replenishment lead times change.
• Using kanban for items with highly variable or lumpy demand. Kanban buffers are sized for average demand with a modest safety factor. A SKU with highly variable demand, spikes followed by weeks of no consumption, will either run out (buffer too small for spikes) or accumulate excess inventory (buffer too large for quiet periods). For high-variability items, a statistically sized safety stock and reorder point is more appropriate than a fixed kanban buffer.
• Losing physical kanban cards. A missing card means the replenishment signal is never sent, and the downstream process runs out of material without warning. Physical kanban systems require discipline in card handling. E-kanban eliminates this failure mode.
• Applying kanban to items with long or unreliable lead times. A kanban buffer sized for a 2-day lead time is completely different from one sized for a 30-day lead time. If the lead time is long or highly variable, the kanban buffer becomes large and the visual simplicity that makes kanban attractive is lost. For long-lead items, MRP with statistical safety stock is more appropriate.

Frequently asked questions

What is the difference between a push and a pull system?

In a push system, production and replenishment are driven by a schedule or forecast, material is pushed downstream whether or not downstream processes are ready for it. This creates WIP inventory wherever downstream processes are slower than upstream schedules. In a pull system, production and replenishment are triggered by actual consumption, a downstream process pulls material from upstream only when it needs it, and the signal (kanban card, empty bin, or system transaction) authorizes the upstream process to replenish. Pull systems limit WIP to the number of authorized kanbans and reduce the inventory that accumulates from forecast errors and push scheduling.

How do you calculate the number of kanbans needed?

The standard formula is: Number of kanbans = (D × LT × (1 + SF)) ÷ Container size. D is the average daily demand rate, LT is the replenishment lead time in days, SF is the safety factor (typically 0.10 to 0.25), and Container size is the number of units per kanban container. The result gives the number of containers (kanbans) that must circulate in the loop to ensure the downstream process is never starved. Multiply by the container size to get total system inventory. Round up to a whole number of kanbans.

What is a two-bin kanban system?

A two-bin kanban is the simplest physical kanban implementation. Each item has two bins. The first bin is consumed first. When it is empty, the empty bin is returned upstream as the replenishment signal, and the second bin feeds consumption in the meantime. When the replenishment arrives, the bins are restocked and swapped. The bin itself is the kanban, no card is needed. Two-bin systems work well for low-value C-class items, consumables, fasteners, and MRO supplies where the administrative cost of formal purchase orders exceeds the benefit of tight inventory control.

When should you use kanban versus MRP?

Use kanban for items with stable, high-frequency demand and short, reliable replenishment lead times, standard components, consumables, C-class items. Kanban is self-regulating and requires minimal planning overhead once designed. Use MRP for items with dependent demand (components driven by a finished goods production schedule), long lead times, deep bill-of-materials structures, or lumpy/project-driven demand patterns where material requirements must be calculated from a time-phased production plan rather than triggered by direct consumption. Many operations use both: MRP for long-lead and engineered items, kanban for standard replenishment items downstream of the push-pull boundary.

What is the safety factor in the kanban formula and how do you set it?

The safety factor (SF) in the kanban formula is a percentage buffer added to the base demand × lead time calculation to account for variability in demand and replenishment lead time. A safety factor of 0.10 (10 percent) adds 10 percent more inventory to the loop as a buffer against variability. Start with 0.15 to 0.25 for a new kanban system where process variability is not yet well understood. As the replenishment process becomes more reliable, shorter lead time variability, fewer quality rejects causing partial containers, more consistent supplier delivery, reduce the safety factor toward 0.05 to 0.10. A high safety factor is a signal that the underlying process needs improvement, not that the kanban quantity should permanently remain high.

What is electronic kanban (e-kanban) and how does it differ from physical kanban?

Electronic kanban (e-kanban) replaces physical kanban cards or bins with a system-generated replenishment signal triggered by a transaction, a barcode scan when a container is consumed, a WMS pick confirmation, or a direct ERP integration. The pull logic is identical to physical kanban: consumption triggers a replenishment authorization. E-kanban eliminates the failure modes of physical systems (lost cards, cards not returned upstream, illegible cards) and scales better for large SKU counts and for replenishment signals that must reach remote suppliers. Most modern ERP and WMS systems support e-kanban as a configurable replenishment method alongside MRP.