Pull Systems, Kanban and Production Leveling

Introduction

Lean manufacturing coordinates material, information, people, and equipment so that value is created with minimum waste. A central design choice is how production is authorized. In a push system, work is released according to a plan or forecast and moved forward whether or not the next process is ready. In a pull system, a downstream user authorizes replenishment only when it has consumed material or capacity.

Pull control is not simply a collection of cards. It is a system of rules for limiting work-in-process (WIP), making shortages visible, synchronizing processes, and responding to actual customer demand. Kanban, supermarkets, FIFO lanes, CONWIP, heijunka, pitch, pacemaker processes, and small-lot replenishment are complementary methods for implementing this logic.

Learning Objectives

After studying this chapter, students should be able to:

1. Compare push and pull production control.
2. Explain the operation and design requirements of kanban and supermarket systems.
3. Calculate kanban quantities using demand, lead time, safety allowance, and container size.
4. Distinguish FIFO, CONWIP, kanban, and hybrid control systems.
5. Apply heijunka, takt time, pitch, pacemaker processes, and small-lot replenishment.
6. Evaluate implementation risks, performance measures, and improvement opportunities.

1. Push and Pull Control

Push control

Push control begins with a forecast, master production schedule, or material requirements plan. Orders are released to work centres in advance, and each department attempts to meet its scheduled quantity. Push systems are useful when demand is highly predictable, production is strongly dependent on long procurement lead times, or regulatory and capacity constraints require advance planning.

However, push control can create excess WIP. A work centre may produce according to its local schedule even when the next operation is congested or customer demand has changed. Forecast error is then amplified through the supply chain. Large batches can conceal quality problems and lengthen lead time.

Pull control

Pull control authorizes production or movement after a downstream process has used material. Its basic logic is:

Do not make or move an item unless a defined downstream signal authorizes it.

Pull does not eliminate planning. Forecasts are still needed for capacity, purchasing, workforce planning, and aggregate scheduling. The distinction is that detailed replenishment is triggered by consumption rather than by an isolated forecast.

Pull is most effective when processes are stable, quality is built in, changeovers are controlled, and replenishment rules are respected. A poorly maintained pull system can merely create frequent shortages.

2. Core Pull-System Concepts

Supermarket

A supermarket is a controlled inventory location containing standard quantities of completed units, components, or materials. The supplying process replenishes what the consuming process withdraws. It is analogous to a retail shelf: consumption creates a replenishment requirement.

A supermarket should have:

• clearly defined items and locations;
• standard container quantities;
• maximum and minimum quantities;
• an agreed replenishment route and frequency;
• visual status identification;
• rules for abnormal conditions and urgent demand.

Supermarkets are appropriate where two processes cannot be directly connected because of distance, different operating speeds, product variety, or batch requirements. They are not a justification for uncontrolled inventory.

Kanban

Kanban means a signal that authorizes production, movement, or replenishment. The signal may be a card, electronic message, empty container, barcode transaction, or visual position. A kanban loop normally includes a consuming process, a withdrawal signal, a supermarket, a supplying process, and a return of the signal after replenishment.

Common kanban types include:

• Withdrawal kanban: authorizes movement from a supermarket.
• Production kanban: authorizes the supplying process to make a quantity.
• Supplier kanban: communicates replenishment requirements to an external supplier.
• Signal kanban: releases a batch when a threshold is reached.
• Electronic kanban: uses digital identification and transaction records.

A kanban is not a general work order. It should identify the item, quantity, source, destination, container, and relevant handling information. The number of kanbans limits the quantity in the loop.

The supermarket loop

A typical loop operates as follows:

1. The customer process withdraws a full container.
2. The withdrawal signal travels to the supplying process.
3. The supplying process makes the specified item and quantity.
4. The replenished container is delivered to the supermarket.
5. The signal is returned to its standard location.

Only authorized containers should circulate. Lost cards, mixed items, unrecorded emergency production, and bypassed inspection destroy the integrity of the loop.

3. Designing Kanban Quantity

A practical design equation for the required number of kanban containers is:

where:

• N = number of kanban containers or cards;
• D = average demand per time period, such as units/day;
• L = replenishment lead time in the same time unit;
• S = safety factor expressed as a decimal;
• C = units per container;
• ⌈ ⌉ = round upward to the next whole container.

The equation assumes relatively stable average demand, a defined replenishment lead time, consistent container quantity, and that the safety factor is intended to cover variability and minor disruption. It does not replace analysis of demand distributions, supplier reliability, transport frequency, or capacity constraints.

If lead time is separated into processing, waiting, transport, and information time, then:

The units must be consistent. For example, demand in units/day requires lead time in days. The safety allowance should not be added twice if it is already included in measured lead time.

Worked example: kanban sizing

A machining cell consumes 480 components per day. Replenishment lead time is 0.75 day, the desired safety factor is 10%, and each container holds 30 components.

Required circulating quantity:

Number of containers:

Thus, the initial design requires 14 kanban containers. Their nominal capacity is 420 units, which exceeds the calculated requirement by 24 units because containers cannot normally be divided. The result should be validated in a pilot. If actual shortages occur, management should first investigate quality losses, delays, inaccurate demand, and noncompliance before simply adding cards.

Refining the design

For highly variable systems, designers may use a service-level or statistical approach. If demand and lead-time variation are measured, safety stock can be estimated from a chosen protection period and variability. Such models require valid data and assumptions about demand distribution. A simple percentage allowance is often preferable during initial implementation because it is transparent, but it should be reviewed as evidence accumulates.

Kanban quantities should be reduced gradually after improvement in setup time, uptime, quality, transport reliability, and demand stability. Lowering the number of cards exposes problems; it does not solve them automatically.

4. FIFO, CONWIP, and Hybrid Systems

FIFO lanes

A first-in, first-out (FIFO) lane is a bounded queue in which work enters at one end and must be processed in sequence. FIFO is useful when a direct connection between processes is desirable but continuous flow is not yet possible. It prevents older orders from being indefinitely bypassed and makes excess WIP visible.

FIFO is governed by a maximum queue size. When the lane is full, the upstream process must stop or switch to another authorized activity. A FIFO lane is therefore both a sequencing rule and a WIP-control mechanism.

CONWIP

CONWIP means constant work-in-process. A fixed number of authorization cards circulates through an entire production line or route. When a finished order leaves the system, its card is released to authorize the entry of another order.

CONWIP is particularly useful when there are many product routes and item-specific kanban would create excessive administrative complexity. It controls total WIP but does not, by itself, specify the exact product mix at each operation. Dispatching rules and a reliable release policy are therefore necessary.

A simplified Little’s Law relationship supports WIP analysis:

where WIP is average work-in-process, TH is average throughput in units per time, and CT is average cycle time in the same time unit. The relationship assumes a stable system over the measurement period, with average inflow approximately equal to average outflow. It is not a scheduling rule, but it shows why excessive WIP generally increases cycle time when throughput is constrained.

Comparison

A hybrid may use CONWIP to limit total line WIP, FIFO between critical operations, and supermarkets for purchased components.

5. Production Leveling and Heijunka

Pull systems perform poorly when customer orders arrive in large, irregular waves. Heijunka, or production leveling, distributes product volume and mix over time. It does not mean producing the same quantity every hour regardless of demand. Rather, it converts a variable demand pattern into a more manageable sequence while preserving the required total output.

Leveling has two dimensions:

• Volume leveling: smoothing the quantity produced per period.
• Mix leveling: distributing product types instead of making long campaigns of one type followed by another.

Suppose daily demand is 240 units over an eight-hour net production day. Takt time is:

The calculation assumes that the eight hours represent net available production time. Planned breaks, meetings, and known downtime should be removed before calculating takt. Takt is a demand pace, not the actual cycle time of every operation.

If daily demand is 120 units of product A and 120 units of product B, producing all A before all B may create long customer waiting times and large finished-goods inventories. A repeating sequence such as A–B–A–B may provide better mix leveling if changeovers are short enough. If changeover time is substantial, a sequence such as A–A–B–B may be a practical compromise.

Pitch

Pitch is the quantity or time interval used to replenish and review production in manageable increments. For a container quantity C and takt time T, pitch time is:

For a container of 15 units and takt of 2 minutes per unit:

Every 30 minutes, the pacemaker should complete or replenish approximately one container, subject to the production sequence. Pitch connects customer demand, container size, scheduling visibility, and material delivery routes.

6. Pacemaker Processes and Small-Lot Replenishment

The pacemaker process is the process selected to receive the leveled production schedule. It is normally the furthest downstream process that can control the product family and is close enough to the customer to represent demand. Upstream processes should be connected through pull signals rather than independently scheduled wherever practical.

A pacemaker should have:

• a stable and visible schedule;
• standardized work and reliable quality;
• a manageable product family;
• controlled changeovers;
• a defined pitch and replenishment route;
• performance measures tied to customer demand.

Small-lot replenishment means delivering materials in quantities and frequencies that support the leveled schedule rather than large, infrequent batches. It reduces line-side inventory and exposes defects sooner, but it increases the importance of route discipline, container standards, supplier reliability, and material-handling capacity.

The economic argument for large batches often relies on setup cost. Lean improvement attacks the setup time itself through methods such as separating internal and external setup, preparing tools in advance, using quick-release fixtures, and standardizing adjustment. The objective is not to ignore cost, but to reduce the cost and risk of frequent changeovers.

7. Implementation Method

A practical implementation sequence is:

1. Select a product family and map its material and information flow.
2. Measure demand, process times, changeovers, quality losses, transport time, and WIP.
3. Identify the pacemaker and calculate takt time.
4. Establish supermarkets or FIFO connections only where direct flow is impractical.
5. Define container quantities, kanban types, routes, and escalation rules.
6. Calculate an initial number of kanbans using measured lead time and a transparent allowance.
7. Create a leveled schedule and pitch board at the pacemaker.
8. Train operators and supervisors in the standard response to each signal.
9. Pilot the system, record abnormalities, and correct root causes.
10. Reduce WIP and kanban quantities as process capability improves.

The system should make abnormalities visible. A red card, empty location, overdue replenishment, or full FIFO lane should trigger a defined response—not informal bypassing.

8. Performance Measurement and Control

Useful measures include:

• customer service level and on-time delivery;
• replenishment lead time;
• WIP quantity and age;
• stockout frequency;
• kanban accuracy;
• schedule adherence at the pacemaker;
• changeover duration;
• first-pass yield;
• equipment availability;
• transport and material-handling frequency.

Measures must be interpreted together. For example, reducing WIP while delivery performance collapses may indicate an undersized system or unstable processes. High utilization at one machine may be undesirable if it creates a bottleneck and excessive queues. Local efficiency should not override end-to-end flow.

Digital kanban can improve traceability, transaction speed, and exception reporting. Barcode, RFID, manufacturing-execution systems, and electronic dashboards are useful when they support clear physical standards. Technology cannot compensate for inaccurate master data, unclear ownership, or poor process discipline. Artificial intelligence may assist demand sensing, anomaly detection, or route optimization, but its recommendations require human governance, data validation, and protection against automated overproduction.

9. Illustrative Case: A Small Motor Assembly Line

Background

A fictional motor manufacturer produces two variants, M1 and M2, for a stable industrial-equipment market. Customer demand is 160 motors per day: 96 M1 and 64 M2. The net production day is 480 minutes. Assembly is downstream of machining and component supermarkets.

Problem

The company uses a weekly push schedule. Machining produces large batches, assembly frequently waits for the required variant, and finished-goods inventory fluctuates between 300 and 600 motors. Average demand is met, but delivery reliability is only 86%.

Analysis

Takt time is:

The team selects final assembly as the pacemaker. With a container quantity of 8 motors, pitch is:

The mix ratio is 3 M1 for every 2 M2. A practical repeating sequence is M1–M1–M1–M2–M2, repeated through the day, subject to actual order requirements and changeover capability.

A machined housing supermarket supplies 320 housings per day. Replenishment lead time is 0.5 day, safety allowance is 15%, and the container holds 20 housings.

The ten containers provide 200 housings of nominal capacity. A withdrawal kanban is attached to each container, and machining receives the associated production signal.

Approach and outcome

The company introduces a pacemaker pitch board, a two-hourly small-lot delivery route, a ten-container housing loop, and a FIFO lane between machining and surface treatment. Changeover work is moved outside the equipment downtime window, reducing average changeover from 42 to 24 minutes. After a four-week pilot, finished-goods inventory falls to approximately 220 motors, delivery reliability rises to 97%, and stockouts are reduced. The results are illustrative rather than a guarantee; the improvement depends on stable quality, disciplined replenishment, and continued problem solving.

Learning points

The principal gain does not come from cards alone. It comes from combining a pacemaker, leveled sequence, bounded WIP, improved changeovers, and rapid reaction to abnormalities. The team also discovers that two kanban containers are frequently delayed by inspection. Rather than increasing the loop, it improves inspection response and standardizes the release process.

10. Challenges, Limitations, and Governance

Pull systems are vulnerable to unstable demand, unreliable equipment, long changeovers, poor quality, inaccurate inventory records, supplier disruption, and excessive product variety. A kanban loop can transmit defects quickly if quality is not controlled at the source. Emergency signals should be rare, authorized, and reviewed; otherwise they become a hidden push system.

Managers should avoid using pull to pressure employees into unsafe work rates. Takt is a planning reference, not permission to bypass ergonomic safeguards, maintenance, inspection, or legally required breaks. Material containers should be designed for safe lifting and transport. Digital records should protect traceability and worker privacy, and automated decisions should remain auditable.

When demand is highly erratic or products are engineered to order, a pure kanban system may be inappropriate. CONWIP, make-to-order release, capacity reservations, or a hybrid system may provide better control. The correct design is contingent on demand pattern, variety, process capability, lead time, and strategic service requirements.

Chapter Summary

Push systems release work from schedules or forecasts; pull systems authorize replenishment from downstream consumption. Supermarkets decouple processes while controlling inventory, and kanban signals authorize movement or production. Kanban quantities can be estimated from demand, replenishment lead time, safety allowance, and container capacity, but the result must be validated against actual variability and process performance.

FIFO lanes control sequence and queue size. CONWIP limits total WIP and is useful for multi-route production. Heijunka levels volume and mix, while takt time expresses the customer demand pace. Pitch converts takt and container size into a practical replenishment interval. A pacemaker receives the leveled schedule and coordinates upstream pull. Small-lot replenishment reduces inventory and response time but requires reliable quality, transport, changeovers, and data.

Effective implementation is evolutionary: establish standards, pilot the loop, expose abnormalities, solve root causes, and progressively reduce WIP. Pull is therefore both a control method and a management system for continuous improvement.

Key Terms

Pull system: Production control in which downstream consumption authorizes replenishment.

Push system: Production control based primarily on schedules or forecasts.

Kanban: A signal authorizing production, movement, or replenishment.

Supermarket: Controlled inventory location replenished according to withdrawal.

Withdrawal kanban: Signal authorizing movement from a supplying location.

Production kanban: Signal authorizing manufacture of a specified quantity.

FIFO lane: Bounded queue operated in first-in, first-out order.

CONWIP: Constant work-in-process control using circulating authorization cards.

Heijunka: Production leveling by volume and product mix.

Takt time: Available net production time divided by customer demand.

Pitch: Replenishment interval based on container quantity and takt.

Pacemaker process: Process that receives the leveled schedule and sets the production rhythm.

Work-in-process: Items currently being processed or waiting between operations.

Replenishment lead time: Time from consumption signal to restored availability.

Safety factor: Allowance for measured uncertainty or disruption.

Small-lot replenishment: Frequent delivery of limited quantities.

Cycle time: Elapsed time required to complete a unit or process activity.

Throughput: Average rate at which completed units leave a system.

Review Questions

Short Answer Questions

1. What is the main difference between push and pull production control?
2. What information should a kanban signal contain?
3. Why is a supermarket used between two processes?
4. Define takt time and distinguish it from cycle time.
5. What is the purpose of a pacemaker process?

Long Answer Questions

1. Explain how kanban, supermarkets, FIFO, and CONWIP control WIP, and compare their appropriate applications.
2. Describe the assumptions and limitations of the basic kanban-quantity equation.
3. Explain how heijunka levels both production volume and product mix.
4. Discuss why small-lot replenishment requires improvement in changeover, quality, and material handling.
5. Propose an implementation plan for converting a push-controlled assembly line to a pull system.

Analytical / Application Questions

1. A line consumes 600 units per day. Replenishment lead time is 0.4 day, safety allowance is 20%, and containers hold 24 units. Calculate the required number of kanban containers and state one reason the result should be validated in a pilot.
2. A plant has 420 net minutes per day and customer demand of 140 units. Calculate takt time. If the container quantity is 7 units, calculate pitch and explain how it could be used on a pacemaker board.
3. A high-variety factory has unstable product routes and excessive item-specific kanban cards. Recommend either kanban, CONWIP, FIFO, or a hybrid design, and justify the choice using product variety, WIP control, and sequencing requirements.

Multiple Choice Questions with Answers

Question 1

Which event normally authorizes replenishment in a pull system?

A. A forecast increase alone

B. Downstream consumption

C. Maximum machine utilization

D. A monthly financial review

Correct Answer: B. Downstream consumption

Question 2

A process consumes 480 units per day, has a 0.75-day replenishment lead time, a 10% safety allowance, and containers of 30 units. How many containers are required?

A. 10

B. 12

C. 14

D. 16

Correct Answer: C. 14

Question 3

What is the primary purpose of a supermarket in a pull system?

A. To hide excess inventory

B. To decouple processes while controlling replenishment

C. To eliminate all material handling

D. To maximize batch size

Correct Answer: B. To decouple processes while controlling replenishment

Question 4

Which system controls total work-in-process with circulating authorization cards across a production route?

A. CONWIP

B. Heijunka

C. Takt production

D. Economic order quantity

Correct Answer: A. CONWIP

Question 5

A line has 480 net minutes available per day and demand is 160 units per day. What is its takt time?

A. 2 minutes per unit

B. 3 minutes per unit

C. 4 minutes per unit

D. 8 minutes per unit

Correct Answer: B. 3 minutes per unit

Question 6

If container quantity is 8 units and takt time is 3 minutes per unit, what is the pitch?

A. 11 minutes

B. 24 minutes

C. 32 minutes

D. 64 minutes

Correct Answer: B. 24 minutes

Question 7

What is the principal role of a pacemaker process?

A. To receive the leveled schedule and set the production rhythm

B. To store all finished goods

C. To perform every upstream operation

D. To eliminate the need for demand data

Correct Answer: A. To receive the leveled schedule and set the production rhythm

Question 8

Which statement best describes a FIFO lane?

A. An unlimited queue processed in any order

B. A bounded queue operated in first-in, first-out order

C. A finished-goods supermarket only

D. A forecast-based release system

Correct Answer: B. A bounded queue operated in first-in, first-out order

Question 9

Using WIP = TH × CT, if throughput is 20 units/hour and cycle time is 2 hours, what is average WIP?

A. 10 units

B. 22 units

C. 40 units

D. 400 units

Correct Answer: C. 40 units

Question 10

What is the principal objective of heijunka?

A. To increase batch size regardless of demand

B. To level production volume and mix over time

C. To remove all safety allowances

D. To schedule every process independently

Correct Answer: B. To level production volume and mix over time