Flow, Pull and Lean Operating Systems

Introduction

Lean operating systems are designed to deliver what the customer needs, when it is needed, in the quantity required, with minimum waste. Their central concern is not merely speed or cost reduction. It is the coordinated design of work so that materials, information and decisions move smoothly through the value stream. Three ideas are especially important: flow, in which work progresses with minimal interruption; pull, in which production is authorized by downstream demand; and Just-in-Time (JIT), in which the right item is available at the right time and in the right quantity.

Flow and pull systems connect customer demand with operational decisions. They require an understanding of takt time, process capacity, work-in-process (WIP), bottlenecks, line balancing, cellular layouts, Kanban, supermarkets and production levelling, known as heijunka. These tools are mutually reinforcing. Takt time provides the demand rhythm; flow determines how work should move; pull controls replenishment; WIP limits expose problems; and heijunka reduces instability in the production schedule.

A lean system is not simply a collection of physical arrangements. It is a management system involving standard work, visual control, problem solving, supplier coordination, maintenance, quality at source and respect for people. A flow system that hides defects or overloads employees is not lean, even if its inventory is low.

Learning Objectives

By the end of this chapter, learners should be able to:

1. Explain the relationship between flow, pull, JIT and customer value.
2. Calculate and interpret takt time, cycle time, capacity and WIP using Little’s Law.
3. Design one-piece-flow and cellular arrangements under practical constraints.
4. Explain Kanban, supermarkets, WIP limits and replenishment signals.
5. Identify bottlenecks and select appropriate improvement responses.
6. Describe how heijunka stabilizes production and supports pull systems.
7. Evaluate the operational, human and managerial limitations of lean flow systems.

1. From Functional Batches to Continuous Flow

Traditional operations are often organized by function: machining, assembly, inspection and packing may be located in separate departments. Each department produces batches and transfers them to the next department. This arrangement can achieve local efficiency, but it commonly creates queues, long lead times, transport, excess inventory and delayed detection of defects.

Flow means that work moves through a process in a timely and orderly sequence. In an ideal one-piece-flow system, one unit moves to the next operation immediately after completion. In practice, small controlled batches may be necessary because of setup times, process technology, transport, safety or economic constraints.

The principal benefits of flow include:

• shorter throughput time;
• earlier detection of defects;
• lower WIP and less capital tied up in inventory;
• clearer visibility of abnormal conditions;
• reduced transport and handling;
• faster learning and feedback.

Flow does not mean that every process must operate continuously. A process may be batch-oriented because of curing, heat treatment or minimum economic batch size. The design task is to identify where flow is technically and economically feasible and where a controlled decoupling point is required.

Push and Pull

In a push system, production is scheduled and released according to a forecast or central plan. Each process produces according to its schedule and pushes output toward the next process. In a pull system, a downstream consumption event authorizes replenishment from an upstream process. Pull therefore links production to actual use rather than merely to planned output.

Pull does not eliminate planning. Aggregate production planning, procurement, workforce planning and capacity planning remain necessary. Pull governs the short-term authorization and sequencing of replenishment within the operating system.

A useful distinction is:

2. Takt Time, Cycle Time and Capacity

Takt Time

Takt time expresses the pace of customer demand. It is calculated as available production time divided by required customer quantity:

Available production time must be stated in consistent units and normally excludes planned breaks, meetings and other non-production periods. If a shift has 27,000 seconds of net available time and customers require 450 units, then:

The result means that, on average, one completed unit must leave the process every 60 seconds to meet demand. Takt time is a demand-based reference, not a guaranteed process capability. It changes when available time or demand changes.

Cycle Time

Cycle time is the elapsed time between completion of successive units from a process or workstation. For a stable process, the effective cycle time should be no greater than takt time if the process is expected to meet demand without accumulating a queue.

However, several distinctions are essential:

• Manual task time is the time an operator spends performing work.
• Machine cycle time is the time required by equipment to complete its operation.
• Lead time is the total elapsed time from order initiation to delivery.
• Throughput time is the time a unit spends moving through a process.
• Takt time is determined by demand, not by the process itself.

A process with a 45-second cycle time can meet a 60-second takt, subject to quality, downtime and staffing. A process with a 75-second cycle time cannot meet that takt without additional capacity, work redistribution, overtime or demand adjustment.

Capacity and Line Balancing

If a resource operates for available time A and has cycle time C, its theoretical capacity is:

This assumes stable cycle time, continuous availability, acceptable quality and no significant changeovers. Real capacity is lower when downtime, defects, absenteeism or material shortages occur.

For a line with total manual work content W and takt time T, the theoretical minimum number of workstations is:

Because a fraction of a workstation is impossible, the practical minimum is rounded up. This is only a lower bound. Precedence constraints, ergonomics, equipment requirements and skill availability may require more stations.

Worked example. A product requires 210 seconds of work content. Demand creates a takt time of 60 seconds.

At least four stations are required. If the work is divided into station times of 55, 52, 53 and 50 seconds, the line can theoretically meet takt. Its balance efficiency is:

The remaining 12.5% represents balancing loss relative to the takt-based station capacity; it is not automatically waste, because some allowance may be needed for variability and human factors.

3. One-Piece Flow and Cellular Layouts

One-piece flow moves units individually or in very small quantities between linked operations. It is most appropriate where products are reasonably repetitive, work content is stable, quality is built into the process and equipment can be arranged close together.

A cellular layout groups different machines and skills needed to complete a family of products. A common arrangement is a U-shaped cell, although the correct geometry depends on safety, material presentation, equipment access and operator movement. Cells reduce transportation and make abnormalities visible, but they should not be treated as a universal replacement for functional layouts.

Important design steps include:

1. Define the product family and demand profile.
2. Measure process times, changeover times, defects and downtime.
3. Calculate takt time and compare each task with takt.
4. Identify precedence constraints and required equipment.
5. Allocate work to stations while protecting safe ergonomic limits.
6. Establish standard work, material presentation and visual controls.
7. Pilot the cell and adjust using observed data.

Cell design must address mixed-model production. If several products share a cell, the workload should be based on the mix rather than on a single product. Frequent changeovers may require SMED principles, modular fixtures or sequence planning.

One-piece flow is not automatically superior. If a high-speed machine requires a long setup or produces several units per cycle, forcing individual transfer may increase cost or risk. The proper objective is controlled flow with the smallest practical transfer batch, not an ideological commitment to one unit in every circumstance.

4. Pull Systems: Kanban, Supermarkets and WIP Limits

Kanban

Kanban is a visual or electronic signal that authorizes production, movement or replenishment. A kanban may be a card, container, barcode transaction, electronic message or other controlled signal. The signal should contain enough information to identify the item, quantity, source, destination and replenishment rule.

Two common forms are:

• Withdrawal kanban, authorizing movement of material from a supplying process to a consuming process.
• Production kanban, authorizing the supplying process to make a specified quantity.

Kanban is not simply an inventory label. It is a control mechanism with explicit rules. Typical rules include: downstream processes withdraw only what they need; upstream processes produce only what the signal authorizes; defective items are not passed forward; and the number of signals is controlled.

Supermarkets

A supermarket is a controlled inventory location from which a downstream process withdraws standard quantities. The upstream process replenishes what has been consumed. It decouples processes that cannot be directly connected by continuous flow, such as processes with different cycle times, batch requirements or geographical separation.

A supermarket should have:

• defined item locations;
• maximum and minimum quantities;
• clear replenishment signals;
• FIFO discipline where age matters;
• ownership and audit responsibility;
• controls for obsolete, damaged or quarantined stock.

A supermarket is a designed buffer, not permission to accumulate uncontrolled inventory. Excess locations and quantities conceal problems and weaken the pull signal.

WIP Limits

WIP limits restrict the quantity of work allowed between process stages. In a flow system, a WIP limit prevents upstream processes from continuing to produce when downstream capacity is unavailable. It also exposes bottlenecks and quality problems.

A practical WIP limit should consider demand, variability, replenishment frequency, container quantity and the response time required to correct abnormalities. Too much WIP increases lead time and hides problems; too little WIP may cause frequent starvation when processes are unreliable. The limit should therefore be tested and adjusted using evidence.

5. Little’s Law and the Economics of WIP

For a stable process over a suitable time interval, Little’s Law is:

where:

• L = average WIP or average number of units in the system;
• λ = average throughput rate, in units per time;
• W = average time a unit spends in the system.

The units must be consistent. If a process averages 120 units of WIP and produces 30 units per hour:

The law applies to stable averages; it does not require every unit to have the same processing time. It does, however, assume that the system is being considered over a sufficiently representative period and that the average flow into and out of the system is approximately equal.

Little’s Law is powerful because it shows that reducing WIP or increasing throughput reduces average lead time, provided the other variables do not offset the improvement. It is not a formula for predicting exact individual lead times, nor does it identify the cause of congestion.

6. Bottleneck Management

A bottleneck is a resource whose available capacity constrains total system throughput. The bottleneck may be a machine, inspection point, decision, supplier, information system or skill group. The Theory of Constraints proposes five focusing steps:

1. Identify the constraint.
2. Exploit it by eliminating avoidable losses without major investment.
3. Subordinate other activities to the constraint’s requirements.
4. Elevate the constraint through added capacity or structural change.
5. Repeat when the constraint moves.

Examples of exploitation include reducing changeover losses, ensuring material availability, assigning preventive maintenance during non-production periods and preventing the bottleneck from processing defective work. Subordination means that upstream processes should not flood the bottleneck with excess WIP.

Bottleneck measures must be system-oriented. Maximizing utilization at every resource can increase queues and lead time. A non-bottleneck may appropriately have idle time if producing more would only create inventory. This is a fundamental difference between local efficiency and total-system performance.

7. Heijunka and Just-in-Time

Heijunka, or production levelling, distributes volume and product mix more evenly over time. Without levelling, customer orders may create large peaks and valleys. These fluctuations force overtime, expedite orders, create shortages and encourage excess finished-goods inventory.

Suppose a weekly requirement is 500 units: 300 of Product A and 200 of Product B. A level sequence might repeat a pattern such as A–B–A–B–A, adjusted to the actual mix and container quantities. The exact sequence depends on demand, setup times and process capability. Heijunka does not mean producing the same quantity of every item; it means smoothing the pattern of production in accordance with the required mix.

JIT is often summarized as producing only what is needed, when it is needed, in the quantity needed. Its supporting conditions include:

• stable and visible demand;
• short and reliable changeovers;
• high equipment availability;
• quality at source;
• capable suppliers and logistics;
• standard work;
• disciplined pull signals;
• rapid response to abnormalities.

JIT should not be interpreted as zero inventory under all conditions. Strategic buffers may be justified for long replenishment times, supply risk, regulatory requirements or severe process variability. The critical question is whether the buffer is intentional, sized, visible and actively managed.

8. Integrated Design Method

A practical design sequence for a flow-and-pull system is as follows.

Step 1: Define customer demand

Determine product mix, demand rate, working calendar and service requirements. Separate genuine demand from temporary promotions or abnormal orders.

Step 2: Map the current value stream

Record material flow, information flow, processing time, waiting time, WIP, defects, changeovers and downtime. A value-stream map should be based on observation and data rather than assumptions.

Step 3: Calculate takt and compare capacity

Use net available time and demand to calculate takt. Compare each operation’s effective cycle time and available capacity with takt, including realistic allowances for downtime and quality losses.

Step 4: Identify constraints and decoupling points

Determine where direct flow is feasible and where a supermarket or controlled buffer is necessary. Locate the current bottleneck and assess whether it is stable.

Step 5: Design the future state

Select cells, workstation allocation, WIP limits, Kanban quantities, replenishment routes and production sequence. Integrate quality controls and abnormality response.

Step 6: Establish standard work and visual management

Define the standard sequence, standard WIP and standard timing. Use visual signals that make normal and abnormal conditions immediately distinguishable.

Step 7: Pilot, measure and improve

Use a controlled pilot. Measure throughput, lead time, WIP, first-pass yield, downtime, schedule adherence, changeover time, safety and employee workload. Adjust the system through structured problem solving.

9. Illustrative Case: Flow Conversion in a Medical-Device Assembly Area

This is a fictional illustrative case.

Background

A medical-device assembly area produced three related products. Components were issued weekly to functional departments. Average finished-goods lead time was 12 working days, although direct assembly time was less than one day. Quality issues were often discovered at final inspection.

Problem

Demand averaged 240 units per day over two shifts. Net available time was 25,200 seconds per day. The existing layout created long transport distances and large queues between assembly, testing and packaging.

Analysis

Takt time was:

Observation showed that assembly work content was 390 seconds per unit, requiring at least:

Testing had a practical cycle time of 140 seconds and was therefore the constraint. The team also found that defects generated in assembly were not detected until testing.

Approach

The team created a mixed-model U-cell with four assembly stations, moved in-process testing closer to the source, and established a small FIFO lane before the testing constraint. A production-and-withdrawal Kanban system controlled replenishment from component preparation. WIP between assembly and testing was limited to 12 units. Standard work included a first-piece verification and an immediate stop-and-call response for critical defects.

Because testing remained slower than takt, management added a second tester during the highest-demand period and improved the test fixture changeover. A heijunka sequence was introduced so that the three product types were interspersed rather than produced in large weekly batches.

Outcome

After stabilization, average finished-goods lead time fell from 12 days to 4 days, WIP declined substantially, and defects were detected earlier. The system did not achieve perfect one-piece flow because testing required controlled batches for calibration. The supermarket and FIFO lane therefore remained deliberate design features.

Learning Points

The improvement came from combining several mechanisms rather than applying a single tool. Takt exposed the capacity gap; the bottleneck analysis focused investment; cellular layout reduced movement; Kanban controlled replenishment; WIP limits exposed instability; and heijunka reduced product-mix disruption. The case also illustrates that lean design must preserve regulatory traceability, quality controls and safe working practices.

10. Performance Measures and Governance

A flow system should be evaluated with a balanced set of measures:

No single metric is sufficient. A reduction in WIP accompanied by increased defects or unsafe work is not a successful lean transformation. Governance should define who owns each signal, how abnormalities are escalated, and when Kanban quantities or WIP limits may be changed.

Digital Kanban, sensors, manufacturing execution systems and real-time dashboards can improve visibility. Artificial intelligence may assist demand forecasting, anomaly detection and dynamic sequencing, but it should not replace process knowledge or override safety and quality controls without validated authority. Automated signals are only as reliable as the data, master records and decision rules behind them.

11. Challenges and Limitations

Flow and pull systems face several limitations:

• highly variable or lumpy demand can make fixed takt difficult to interpret;
• unreliable equipment creates starvation and instability;
• long setup times may make small batches uneconomic;
• supplier disruptions may require strategic inventory;
• product customization may reduce the feasibility of repetitive cells;
• excessive WIP reduction can make minor disruptions immediately visible but difficult to absorb;
• poorly designed labour standards can create fatigue and resistance;
• local optimization can shift problems to suppliers or customers.

Lean implementation can also fail when management treats Kanban as an inventory-reduction programme rather than a learning system. Employees may hide problems if performance targets punish stoppages. Sustainable implementation therefore requires psychological safety, fair workload design, training, cross-functional cooperation and leadership willingness to address root causes.

Chapter Summary

Flow organizes work so that value moves with minimal interruption. Pull authorizes replenishment from downstream consumption, while JIT coordinates production with actual need. Takt time translates customer demand into an operating rhythm, whereas cycle time describes process performance. Comparing the two supports capacity analysis and line balancing.

One-piece flow and cellular layouts reduce transport, queues and delay when process conditions support them. Kanban and supermarkets provide controlled replenishment where direct flow is impractical. WIP limits prevent uncontrolled accumulation and make abnormalities visible. Little’s Law, L = λW, links average WIP, throughput and time in a stable system.

Bottleneck management requires system-level thinking: the constraint determines potential throughput, and non-bottleneck activity should be subordinated to it. Heijunka smooths volume and mix, reducing instability and supporting JIT. Effective implementation combines technical design with standard work, quality at source, maintenance, employee involvement, supplier capability and balanced performance measures. Lean systems should be designed for resilience, safety, regulatory compliance and learning—not merely for minimum inventory.

Key Terms

Takt time: The available production time divided by required customer demand.

Cycle time: The elapsed time between completed units from a process.

Lead time: Total elapsed time from request or order initiation to delivery.

Flow: The orderly movement of work through linked process steps.

One-piece flow: Movement of individual units, or the smallest practical batch, between operations.

Cellular layout: Arrangement of equipment and people to process a product family in a compact flow-oriented area.

Pull system: A system in which downstream consumption authorizes upstream replenishment.

Push system: A system in which production is released according to a forecast or central schedule.

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

Supermarket: A controlled inventory location used to decouple processes and support pull replenishment.

WIP: Work-in-process inventory currently between the start and completion of a process.

WIP limit: A defined maximum quantity of work allowed in a process or flow segment.

Bottleneck: A resource whose capacity constrains system throughput.

Heijunka: Production levelling by smoothing volume and product mix over time.

Just-in-Time: Producing and delivering what is needed, when needed, in the required quantity.

FIFO: First-in, first-out flow discipline.

Line balancing: Allocating work among stations to match demand and reduce imbalance.

Little’s Law: The relationship L = λW between average WIP, throughput and time.

Standard work: The documented best-known sequence, timing and WIP required for a process.

Review Questions

Short Answer Questions

1. Define takt time and distinguish it from cycle time.
2. What is the purpose of a supermarket in a pull system?
3. State Little’s Law and define its three variables.
4. Explain why a bottleneck should be protected from defective work.
5. What is the purpose of heijunka?

Long Answer Questions

1. Explain how flow, pull and JIT operate as an integrated lean operating system.
2. Discuss the conditions under which one-piece flow and cellular layouts are appropriate.
3. Describe the design and governance requirements of a Kanban system.
4. Explain how bottleneck management differs from maximizing utilization at every workstation.
5. Evaluate the advantages and limitations of reducing WIP in a production system.

Analytical / Application Questions

1. A plant has 28,800 seconds of net daily production time and demand of 360 units. Calculate takt time and explain how it should be used in line design.
2. A process has average WIP of 150 units and throughput of 25 units per hour. Use Little’s Law to estimate average time in the system. Identify assumptions behind your calculation.
3. A mixed-model line has unstable demand, long changeovers and a testing operation slower than takt. Propose an integrated design using heijunka, changeover reduction, WIP control and bottleneck management. Justify the sequence of actions.

Multiple Choice Questions with Answers

Question 1

A plant has 27,000 seconds of net available time and customer demand of 450 units. What is the takt time?

A. 45 seconds per unit

B. 60 seconds per unit

C. 75 seconds per unit

D. 90 seconds per unit

Correct Answer: B. 60 seconds per unit

Question 2

Which event normally authorizes replenishment in a pull system?

A. An upstream operator completing a batch

B. A forecast increase alone

C. Downstream consumption

D. Maximum machine utilization

Correct Answer: C. Downstream consumption

Question 3

If average WIP is 120 units and throughput is 30 units per hour, what is the average time in the system under Little’s Law?

A. 2 hours

B. 3 hours

C. 4 hours

D. 6 hours

Correct Answer: C. 4 hours

Question 4

What is the primary purpose of a WIP limit?

A. To increase uncontrolled production

B. To prevent accumulation and expose flow problems

C. To eliminate the need for quality control

D. To ensure every machine is always busy

Correct Answer: B. To prevent accumulation and expose flow problems

Question 5

Which statement best describes takt time?

A. The time required to repair a machine

B. The time between customer orders only

C. The demand-based pace required to meet customer demand

D. The total lead time of a product

Correct Answer: C. The demand-based pace required to meet customer demand

Question 6

What is a production kanban?

A. A signal authorizing production of a specified quantity

B. A document recording employee attendance

C. A measure of equipment depreciation

D. A forecast of annual market demand

Correct Answer: A. A signal authorizing production of a specified quantity

Question 7

Which action is consistent with the Theory of Constraints?

A. Flooding the bottleneck with excess WIP

B. Protecting the constraint from avoidable downtime and defects

C. Maximizing output at every non-bottleneck regardless of demand

D. Removing all buffers without analysis

Correct Answer: B. Protecting the constraint from avoidable downtime and defects

Question 8

What is the main purpose of heijunka?

A. To increase product variety without sequencing

B. To level production volume and mix over time

C. To replace all preventive maintenance

D. To maximize batch size

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

Question 9

A product has 210 seconds of work content and a takt time of 60 seconds. What is the theoretical minimum number of stations?

A. 2

B. 3

C. 4

D. 5

Correct Answer: C. 4

Question 10

Which condition most strongly supports one-piece flow?

A. Highly unstable processes with frequent defects

B. Long uncontrolled changeovers and no standard work

C. Stable work content, reliable quality and compatible equipment

D. Unlimited demand variability and no process data

Correct Answer: C. Stable work content, reliable quality and compatible equipment