6. Pull Systems, Kanban and Lean Planning

Introduction and Context

Lean production seeks to create flow while producing only what is needed, when it is needed, and in the quantity required. This objective cannot be achieved merely by reducing waste within individual operations. It also requires a control logic that regulates the movement and replenishment of materials across the value stream. Pull systems provide that logic.

In a traditional push system, production orders are released on the basis of forecasts, planned requirements, or schedules. Each process produces according to its own plan and sends output to the next process. Such systems can create excess inventory, long queues, hidden quality problems, and poor responsiveness when demand changes. A pull system, by contrast, authorizes replenishment in response to actual consumption or a clearly defined downstream requirement.

Kanban is the most widely recognized mechanism for operating a pull system, but it is not synonymous with pull. Supermarkets, FIFO lanes, CONWIP, heijunka, electronic signals, and carefully designed replenishment policies are also part of the broader pull-system architecture. Effective lean planning therefore involves choosing the appropriate control method, sizing buffers and signals, linking operational control with MRP and ERP, and coordinating short-term execution with sales and operations planning.

This chapter develops these ideas from a managerial and analytical perspective. It emphasizes that pull is not simply a collection of cards or bins. It is a disciplined method of synchronizing capacity, inventory, information, and customer demand.

Learning Objectives

After studying this chapter, the reader should be able to:

1. Explain the logic and conditions of pull production.

2. Distinguish among Kanban, supermarket, FIFO, CONWIP, and hybrid control systems.

3. Calculate an initial number of Kanban cards or containers.

4. Design appropriate buffers and replenishment rules.

5. Explain the role of level scheduling and heijunka in stabilizing flow.

6. Integrate lean execution systems with MRP, ERP, and sales and operations planning.

7. Diagnose common causes of pull-system failure.

8. Evaluate the managerial, technological, and governance implications of lean planning.

From Forecast-Driven Release to Consumption-Driven Replenishment

A pull system begins with a demand signal from the customer or from a downstream process. When a unit is consumed, the signal authorizes replenishment. The upstream process does not produce merely because it has available capacity; it produces because a defined downstream need has been established.

The central principle can be expressed as:

Produce or move only when authorized by consumption or by a controlled downstream requirement.

This principle does not eliminate planning. Forecasts remain necessary for capacity decisions, purchasing, workforce planning, and aggregate production. Pull changes the way detailed execution is controlled. A forecast may determine that a factory should have sufficient capacity for 10,000 units next month, while actual customer consumption determines which items are replenished today.

A push system tends to separate information from material flow. Planning information moves ahead through schedules, while materials accumulate between operations. In a pull system, the movement of material and the authorization to replenish are deliberately linked. This exposes instability: if a process frequently lacks material, produces defects, or exceeds its cycle time, the shortage becomes visible rather than being hidden in a large inventory buffer.

Pull therefore requires reliable processes. Standard work, stable quality, reasonably predictable cycle times, quick changeovers, preventive maintenance, and accurate inventory records are not optional supporting activities. They are conditions that make pull control credible.

Essential Concepts and Terminology

A value stream is the complete set of activities required to deliver a product or service, including material and information flows. A customer takt is the available production time divided by customer demand during that period. Takt is a demand rhythm, not a cycle-time target for every operation. Cycle time is the time required to complete one unit or one repeated work cycle at a process. Lead time is the elapsed time from request to delivery. WIP is work-in-process inventory between operations.

A pull signal is an authorization to produce, move, or replenish. A Kanban is a visual or electronic signal that communicates such authorization. A supermarket is a controlled inventory point from which downstream users withdraw items and which upstream processes replenish. A FIFO lane is a first-in, first-out queue with a defined capacity and sequence discipline. CONWIP, or constant work-in-process, controls the total amount of work in a line by releasing a new job only when another job leaves.

Heijunka, or production leveling, distributes product volume and mix more evenly over time. A pitch is a repeated time interval used to organize and review production quantities. Safety stock protects against uncertainty in demand or supply; it is distinct from cycle stock, which exists because replenishment occurs in batches. Reorder point is the inventory position at which replenishment is triggered. Inventory position normally includes on-hand inventory plus open replenishment minus allocated or backordered demand.

Pull Architectures: Choosing the Right Control Logic

Kanban and Supermarket Pull

In a two-card Kanban arrangement, a withdrawal Kanban authorizes movement from a supermarket to the consuming process, while a production Kanban authorizes the supplying process to make a replacement quantity. In simpler systems, one signal may perform both functions. Cards, containers, barcodes, bins, and electronic messages can all serve as Kanban media, provided that the rules are unambiguous.

A supermarket works well when demand is repetitive, the product family is relatively stable, replenishment lead time is short, and inventory can be physically controlled. The supermarket should have clearly defined locations, maximum and minimum quantities, standard container sizes, and rules for abnormal conditions. It is not an uncontrolled warehouse. Its purpose is to decouple processes only to the extent necessary to preserve flow.

FIFO Lanes

A FIFO lane is useful when a direct process-to-process relationship exists but a small buffer is required. The supplying process places completed units at one end, and the consuming process withdraws from the other. Once the lane reaches its maximum capacity, the supplying process must stop or resolve the abnormality. FIFO is especially appropriate for sequential operations with limited product variety and a strong need to preserve processing sequence.

FIFO lanes are less suitable when demand is highly variable, the product mix is broad, or replenishment must be separated from production scheduling. In those situations, a supermarket or CONWIP system may provide greater flexibility.

CONWIP

CONWIP controls the total WIP in a production line rather than maintaining a separate Kanban quantity for every part number. A card travels with a job through the line. When the job is completed and exits, its card becomes available to authorize the release of another job. The number of cards therefore establishes a WIP ceiling.

CONWIP is valuable in high-mix environments where individual Kanban loops would require too many cards and excessive parameter maintenance. It can improve flow by limiting total work, but it does not by itself determine the best job sequence. Dispatching rules, due-date priorities, setup considerations, and capacity constraints remain necessary.

Hybrid Systems

Many factories use more than one pull mechanism. A supermarket may control common components, FIFO may connect closely related operations, and CONWIP may regulate a final assembly line. MRP may plan purchased materials and long-lead components, while Kanban controls repetitive internal replenishment. The appropriate question is not whether a plant is purely pull or purely push, but where each control logic creates the best combination of responsiveness, stability, and cost.

Sizing Kanban Quantities and Buffers

A commonly used starting formula for the number of Kanban containers is:

N = [D × L × (1 + S)] / C

where:

• N = number of Kanban containers or cards;
• D = average demand per time unit;
• L = replenishment lead time in the same time unit;
• S = safety factor expressed as a decimal;
• C = quantity per container.

The result is normally rounded upward. If daily demand is 480 units, replenishment lead time is 0.5 day, safety factor is 20 percent, and each container holds 40 units, then:

N = (480 × 0.5 × 1.20) / 40 = 7.2

The initial design would therefore require 8 containers. This is not a final answer. The quantity must be validated through observation of actual shortages, excess inventory, demand variation, process reliability, and container-return discipline.

The formula incorporates expected demand during replenishment lead time and a protection factor. It does not automatically account for every source of variation. If lead time, demand, quality yield, or container availability changes significantly, the loop must be recalculated. Increasing cards is often an understandable reaction to shortages, but it can conceal the real cause. A shortage may result from downtime, inaccurate records, missing cards, excessive changeover time, poor sequencing, or supplier unreliability.

Buffer design should distinguish at least three questions. First, what inventory is required for normal replenishment? Second, what protection is required for predictable variation? Third, what contingencies are required for major disruptions? The third question should not routinely be answered by enlarging every supermarket. Business continuity plans, alternate suppliers, preventive maintenance, and rapid problem escalation are usually better responses than permanently carrying excessive inventory.

A useful operational measure is inventory coverage:

Coverage = available inventory / average demand per time period.

Coverage should be evaluated together with service level, replenishment lead time, stockout frequency, inventory accuracy, and obsolescence. A low inventory figure is not inherently lean if it produces missed deliveries and emergency expediting.

Level Scheduling and the Rhythm of Replenishment

A pull system can still become unstable if production mix changes sharply. Suppose a line produces a large batch of Product A, followed by a large batch of Product B. Downstream demand may require both products every day, yet the upstream schedule creates alternating shortages and surpluses. Level scheduling addresses this problem by smoothing volume and mix.

Heijunka seeks to distribute demand across the planning horizon. If weekly demand is 500 units of A and 250 units of B over five working days, a leveled pattern might produce approximately 100 A and 50 B each day, subject to setup and capacity constraints. At a finer pitch, the sequence could repeat A-A-B, rather than producing all A followed by all B.

Leveling does not mean that every product must be made in identical quantities at every moment. It is a managerial compromise among demand responsiveness, setup time, economic batch effects, and process constraints. Quick-changeover methods make leveling more feasible by reducing the penalty associated with frequent product changeovers.

A practical leveling process includes the following steps:

1. Establish the planning horizon and demand pattern.

2. Separate stable, repetitive items from exceptional or engineered items.

3. Determine available capacity and required takt.

4. Define an achievable product sequence and pitch.

5. Protect the sequence from unnecessary schedule changes.

6. Review deviations and adjust the system through structured problem solving.

Linking Pull with MRP, ERP and Sales and Operations Planning

Pull and MRP are sometimes presented as opposing philosophies, but they operate at different levels. MRP is effective for translating a master production schedule into material requirements, coordinating dependent demand, and planning long-lead or low-volume items. Pull is effective for controlling repetitive execution and replenishment close to the point of use.

ERP systems provide the transactional backbone: purchase orders, inventory records, production orders, customer orders, financial postings, and performance data. A Kanban transaction may be represented electronically through a barcode scan, RFID signal, mobile terminal, or digital board. However, digitizing a poor process does not make it lean. The physical rules, ownership, container quantities, exception procedures, and data discipline must be established first.

Sales and operations planning, or S&OP, reconciles demand, supply, capacity, inventory, and financial objectives over a medium-term horizon. It should determine the broad volume and resource commitments within which pull systems operate. For example, S&OP may identify a sustained increase in demand that requires an additional shift or supplier capacity; Kanban cannot solve a structural capacity shortage.

A useful planning hierarchy is:

• Strategic and aggregate planning: network, capacity, suppliers, investment, and policy.
• S&OP: monthly reconciliation of demand and supply.
• Master scheduling and MRP: product-family and component planning, especially for variable or long-lead items.
• Lean execution control: Kanban, FIFO, CONWIP, dispatching, and daily management.
• Immediate response: abnormality management, escalation, and recovery.

The interfaces between these levels require explicit rules. The ERP system should not automatically generate large conflicting orders while shop-floor Kanban is regulating the same flow. Conversely, local pull signals should not ignore an approved master schedule, supplier constraint, or engineering change. Governance must specify which system is authoritative for each decision.

Designing a Pull System: A Practical Method

Design begins with demand segmentation. Products should be grouped according to demand frequency, volume, variability, value, lead time, and process similarity. High-volume stable items are candidates for supermarket Kanban. Low-volume or highly variable items may require MRP, order-based production, or CONWIP. Critical purchased items may need supplier scheduling and strategic safety stock.

The next step is value-stream mapping. The team identifies customer demand, process times, changeovers, WIP, supermarkets, information flows, quality losses, and replenishment routes. The map should reveal where a pull loop can be established and where a pacemaker process should set the rhythm for the value stream.

The team then defines the pacemaker, replenishment frequency, container size, signal type, maximum WIP, and abnormality rules. It is important to specify who owns each signal, what happens when a card is missing, how damaged or quarantined material is treated, and when management intervention is required.

Pilot implementation should focus on a manageable product family. Performance should be monitored through delivery reliability, stockouts, WIP, lead time, inventory accuracy, schedule adherence, changeover losses, defects, and emergency transactions. The number of cards should be adjusted only after the underlying causes of variation have been investigated.

Applied Illustrative Case: Pull Replenishment in an Industrial Pump Plant

Background

An illustrative industrial pump plant produces three standard pump families for distributors and maintenance customers. The final assembly area receives machined housings, impellers, seals, and fasteners from internal and external sources. Although monthly demand is reasonably stable, daily demand varies by family. The plant uses ERP-generated production orders, but supervisors frequently expedite materials because shortages occur at assembly.

Situation and Problem

The housing machining department produces in large batches to avoid setups. Assembly consequently receives excess housings for one family while waiting for another. Average WIP between machining and assembly is 1,800 units, and shortages cause approximately 12 percent of assembly orders to be rescheduled. Operators also report that physical inventory differs from ERP records.

Analysis

A value-stream review shows that machining lead time is one day when the required family is available. Assembly consumes approximately 240 housings per day. A supermarket is feasible because the housing designs are standardized and demand is repetitive. The team decides to begin with a half-day replenishment lead time and a 15 percent protection factor. Containers hold 30 housings.

The initial card calculation is:

N = (240 × 0.5 × 1.15) / 30 = 4.6

The team rounds up to five containers for the pilot family. It also establishes separate locations for available, empty, and quarantined containers. A withdrawal signal travels from assembly to the supermarket; a production signal authorizes machining to replenish the withdrawn quantity. A maximum FIFO lane is retained between machining and inspection because inspection sequence must be preserved.

Approach and Strategy

The plant introduces a daily leveling board at the machining pacemaker. Instead of producing a large batch of one housing family, the schedule follows a repeated sequence based on the assembly pitch. Setup reduction activities lower the changeover time, while barcode scanning updates ERP inventory when containers are withdrawn, completed, or quarantined.

The rules state that a missing container cannot be replaced by an informal verbal order. The shortage is recorded as an abnormality, and the supervisor investigates whether the cause is demand variation, late production, quality loss, inaccurate scanning, or capacity failure. A small, separately governed safety reserve is retained for approved disruptions rather than being mixed with normal Kanban stock.

Outcome

During the pilot, WIP falls substantially, assembly shortages become visible earlier, and the machining department receives a more stable sequence. The plant does not judge the pilot solely by lower inventory. It also evaluates delivery adherence, record accuracy, quality escapes, and the number of emergency releases. After several weeks, the team discovers that most apparent shortages were caused by containers placed in quarantine without timely status updates. Correcting the information process allows the Kanban quantity to remain at five rather than being increased.

Learning Points

The case illustrates four principles. First, card quantities must be based on demand and lead time but validated against actual causes of variation. Second, pull requires physical and informational discipline. Third, leveling and setup reduction reinforce Kanban effectiveness. Fourth, a shortage is valuable diagnostic information; simply adding inventory can hide the problem.

Common Failure Modes and Managerial Risks

A frequent mistake is implementing Kanban before stabilizing quality and process capability. Defective parts consume signals and containers while failing to satisfy demand. Another mistake is treating the number of cards as a performance target. Reducing cards without reducing variation may increase stockouts; increasing cards without analysis may create excess WIP.

Other failure modes include oversized containers, poor card visibility, unauthorized production, inaccurate inventory records, unclear ownership, excessive product variety in one loop, and failure to remove obsolete cards after engineering or demand changes. A system may also fail when managers continually override the pull rules. Every override should be visible, justified, and reviewed; otherwise the formal system becomes ceremonial.

Pull can create local optimization. A supplying process may achieve excellent utilization by producing full containers even when the downstream process does not need them. Lean control must therefore prioritize total value-stream performance over isolated machine utilization. Similarly, very low WIP can reduce resilience if supplier risk, maintenance weakness, or quality instability has not been addressed.

Ethically and professionally, planners and supervisors should not manipulate signals to conceal missed targets or shift shortages to another department. Inventory data, shortages, quality holds, and schedule deviations should be reported accurately. Governance should define approval authority for safety-stock changes, emergency releases, supplier substitutions, and engineering changes. Traceability is particularly important in regulated sectors such as medical devices, aerospace, food, and automotive components.

Digital Pull and Emerging Developments

Electronic Kanban systems can provide real-time visibility of withdrawals, replenishment status, supplier deliveries, and shortages. RFID and barcode technologies reduce manual recording, while dashboards can show WIP, aging, response time, and card circulation. Advanced analytics can identify demand variability, abnormal lead times, and recurring causes of emergency replenishment.

Artificial intelligence and machine learning may support demand classification, anomaly detection, dynamic safety-stock recommendations, and predictive maintenance. Their role should remain supportive rather than autonomous where decisions affect quality, safety, compliance, or customer commitments. A statistically optimized buffer is not a substitute for managerial understanding of process behavior.

Digital systems also introduce risks: incorrect master data, cyberattack, loss of network access, algorithmic bias, excessive dependence on automated alerts, and weak user adoption. A robust system therefore includes access controls, audit trails, backup procedures, manual contingency rules, and periodic validation of parameters.

Best-Practice Guidelines

Effective pull systems generally share the following characteristics:

• Demand and replenishment rules are clearly defined.
• The pacemaker and control points are visible.
• Container sizes and card quantities reflect actual lead time and variation.
• Quality, maintenance, setup reduction, and standard work support flow.
• Abnormalities are exposed and escalated rather than hidden.
• MRP, ERP, S&OP, and shop-floor signals have clearly separated responsibilities.
• Level scheduling protects the system from excessive mix and volume fluctuation.
• Parameters are reviewed after demand, process, supplier, or product changes.
• Digital tools improve visibility without replacing sound process design.
• Performance is judged by customer service, total lead time, quality, and total inventory—not by isolated utilization.

Chapter Summary

Pull systems regulate production and replenishment through actual consumption or controlled downstream demand. Kanban is a practical signaling method, while supermarkets, FIFO lanes, CONWIP, and hybrid designs provide alternative ways to control material and WIP. Kanban quantities can be initially estimated from demand, replenishment lead time, safety factor, and container size, but must be validated through operational evidence. Heijunka and pitch stabilize volume and mix, making pull more reliable. MRP, ERP, and S&OP remain important because they address different planning horizons and product characteristics. Successful implementation depends on stable processes, accurate information, explicit governance, disciplined abnormality management, and continuous improvement. Digital technologies can strengthen visibility and analysis, but they cannot compensate for weak process design.

Key Terms

Pull system: A production-control system in which replenishment is authorized by downstream consumption or demand.

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

Supermarket: A controlled inventory point from which users withdraw items and suppliers replenish them.

FIFO lane: A bounded first-in, first-out queue connecting processes.

CONWIP: A system that controls total work-in-process through a fixed number of circulating cards.

Heijunka: Production leveling by smoothing volume and product mix.

Takt time: Available production time divided by customer demand.

Cycle time: Time required to complete one unit or repeated work cycle.

Lead time: Elapsed time between request initiation and completion or delivery.

WIP: Work-in-process inventory located between or within production operations.

Safety stock: Inventory held to protect against uncertainty in demand or supply.

Cycle stock: Inventory created because replenishment occurs in batches.

Reorder point: Inventory position at which replenishment is triggered.

Pitch: A repeated time interval used to organize production and review.

Pacemaker process: The process whose schedule establishes the rhythm for a value stream.

Inventory position: On-hand inventory plus open replenishment minus allocated or committed demand.

MRP: Material Requirements Planning used to calculate dependent material and capacity requirements.

S&OP: Sales and operations planning that reconciles medium-term demand and supply.

Abnormality management: The structured identification, escalation, and resolution of deviations from standard conditions.

Review Questions

Short Answer Questions

1. What distinguishes a pull system from a push system?

2. Define a supermarket and explain its purpose.

3. What is the difference between a FIFO lane and a CONWIP system?

4. Why is heijunka important in a pull environment?

5. State the variables used in the basic Kanban quantity formula.

Long Answer Questions

1. Explain the operating logic of Kanban and discuss the conditions required for its successful implementation.

2. Compare supermarket Kanban, FIFO, CONWIP, and hybrid pull systems.

3. Explain how MRP, ERP, S&OP, and shop-floor pull controls can be integrated without creating conflicting instructions.

4. Discuss the role of level scheduling, pitch, and quick changeover in stabilizing production.

5. Analyze the principal failure modes of pull systems and recommend managerial responses.

Analytical / Application Questions

1. Daily demand is 360 units, replenishment lead time is 0.75 day, the safety factor is 25 percent, and container size is 30 units. Calculate the initial number of Kanban containers and identify two reasons why the result should be validated.

2. A high-mix factory experiences shortages despite holding substantial inventory. Determine whether Kanban, CONWIP, FIFO, or a hybrid system is most appropriate, and justify your choice.

3. An ERP system releases large batch orders while an internal Kanban loop is controlling replenishment. Diagnose the governance problem and propose an integrated planning arrangement.

22. Multiple Choice Questions with Answers

Question 1

Which statement best describes the central authorization rule of a pull system?

A. Each process produces whenever it has available capacity.

B. Production is authorized by actual consumption or a controlled downstream requirement.

C. Production is released only at the beginning of each month.

D. The largest possible batch is produced to minimize setups.

Correct Answer: B. Production is authorized by actual consumption or a controlled downstream requirement.

Question 2

Daily demand is 480 units, replenishment lead time is 0.5 day, the safety factor is 20 percent, and container size is 40 units. What is the appropriate initial number of containers?

A. 5

B. 6

C. 8

D. 10

Correct Answer: C. 8

Question 3

Which control method limits total work-in-process through a fixed number of circulating cards?

A. Supermarket Kanban

B. FIFO

C. CONWIP

D. MRP

Correct Answer: C. CONWIP

Question 4

What is the principal purpose of a supermarket in a lean value stream?

A. To store all finished goods for seasonal demand

B. To provide a controlled replenishment point between processes

C. To replace production scheduling entirely

D. To maximize machine utilization through large batches

Correct Answer: B. To provide a controlled replenishment point between processes

Question 5

A production line makes all of Product A for several days and then all of Product B, causing alternating shortages. Which approach most directly addresses this problem?

A. Increasing container size

B. Production leveling through heijunka

C. Removing all safety stock

D. Replacing ERP with manual records

Correct Answer: B. Production leveling through heijunka

Question 6

Which situation is most suitable for a FIFO lane?

A. Highly variable demand across many unrelated products

B. A sequential process requiring a small bounded buffer and sequence preservation

C. Long-lead imported components with uncertain supply

D. A product family with no defined processing sequence

Correct Answer: B. A sequential process requiring a small bounded buffer and sequence preservation

Question 7

What is the most appropriate relationship between S&OP and shop-floor Kanban?

A. Kanban should determine long-term capacity investment independently of S&OP.

B. S&OP sets broad demand and capacity commitments, while Kanban controls detailed replenishment.

C. S&OP should be eliminated when Kanban is introduced.

D. Both systems should release unrestricted production orders for flexibility.

Correct Answer: B. S&OP sets broad demand and capacity commitments, while Kanban controls detailed replenishment.

Question 8

A plant experiences frequent Kanban shortages because containers are placed in quality quarantine without status updates. What is the best initial response?

A. Immediately double the number of Kanban cards.

B. Remove quality inspection from the process.

C. Correct the quarantine and information-control process before changing card quantities.

D. Allow supervisors to issue unlimited verbal orders.

Correct Answer: C. Correct the quarantine and information-control process before changing card quantities.

Question 9

Which statement about digital Kanban is most accurate?

A. Digitization automatically creates a lean process.

B. Digital Kanban can improve visibility, but it still requires sound physical rules and accurate data.

C. Electronic signals eliminate the need for abnormality management.

D. Digital Kanban makes safety stock unnecessary in every situation.

Correct Answer: B. Digital Kanban can improve visibility, but it still requires sound physical rules and accurate data.

Question 10

Why should managers avoid judging a pull system solely by low inventory?

A. Low inventory always indicates poor scheduling.

B. Customer service, quality, lead time, and emergency activity must also be considered.

C. Inventory has no relationship to lean performance.

D. High utilization is always more important than delivery reliability.

Correct Answer: B. Customer service, quality, lead time, and emergency activity must also be considered.