Lean Flow, Pull and Production Control

Introduction and Context

Lean manufacturing is a socio-technical system for delivering customer value with minimum waste while maintaining quality, safety, flexibility and respect for people. Flow and pull are its central production-control ideas. Flow means that work progresses through a process with minimal interruption, waiting, transportation, rework and queue accumulation. Pull means that upstream activity is authorized by a real consumption signal from a downstream customer rather than by an isolated forecast or a large production schedule.

These ideas are not simply layout or inventory techniques. They change how capacity, quality, maintenance, scheduling and employee responsibility are managed. A plant may possess modern equipment and still lack flow if jobs are released in large batches, defects stop production, changeovers are long, or bottlenecks are overloaded. Conversely, a modest facility can achieve excellent performance through disciplined sequencing, visual control, rapid problem solving and stable processes.

Lean flow must be designed with demand characteristics, product variety, process capability, safety and supply constraints in mind. One-piece flow is an ideal and a design direction; it is not an instruction to force every product through a perfectly continuous route regardless of economics or technical realities.

Learning Objectives

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

1. Design and evaluate cellular layouts and one-piece-flow systems.
2. Distinguish push, pull, kanban and hybrid production-control systems.
3. Calculate takt time, cycle time, line efficiency, work-in-process and kanban quantities.
4. Apply heijunka, jidoka, Andon and standardised work.
5. Identify bottlenecks and use the Theory of Constraints to improve throughput.
6. Reduce batch size, queues and inventory without destabilising operations.
7. Evaluate implementation risks, digital technologies and managerial implications.

1. Flow as a Production-System Principle

A process has flow when materials, information and decisions move at the rate and sequence required by the customer. Five related conditions are important:

• Stable demand and product definition: priorities are not changed arbitrarily.
• Capable and available equipment: breakdowns and quality failures do not repeatedly interrupt the route.
• Short changeovers: products can be mixed without excessive economic batch sizes.
• Balanced work: no operation systematically accumulates a queue.
• Visible control: abnormal conditions are detected and acted upon promptly.

The basic flow measures are lead time, processing time, queue time, work-in-process (WIP), throughput and first-pass yield. Little’s Law connects three of these measures:

Here WIP is the average number of units in the system, throughput is the average completion rate in units per time period, and lead time is the average time a unit spends in the system. The units must be consistent, and the relationship assumes a stable system over the period being measured: average inflow approximately equals average outflow. Little’s Law is descriptive, not a promise that reducing WIP automatically increases throughput. If WIP is removed without addressing capacity or reliability, output may fall.

A useful diagnostic is the process-cycle-efficiency ratio:

Value-added time is time that changes the product or service in a way the customer requires and is willing to pay for. Inspection, transport, waiting and rework are generally non-value-adding, although some may be necessary under current conditions. PCE is a diagnostic measure, not a complete measure of performance; safety, quality, delivery and employee well-being must also be considered.

2. Cellular Layouts and One-Piece Flow

A functional or process layout groups similar machines together, such as all lathes in one department. It offers flexibility for diverse work but often creates long travel distances, complex dispatching and queues between departments. A product layout arranges resources according to a product route and supports high volume, but may be inflexible. A cellular layout groups dissimilar resources into a compact manufacturing cell for a family of products with similar processing requirements.

Designing a cell

Cell design normally follows these steps:

1. Collect routing, demand, equipment, quality and changeover data.
2. Group products into families using routing similarity and process requirements.
3. Map the current material and information flow.
4. Calculate demand rate and required staffing.
5. Arrange machines in route order, often in a U-shaped configuration.
6. Allocate work elements and establish standardised work.
7. Validate safety, ergonomics, maintenance access, material replenishment and abnormality response.
8. Pilot the cell and revise it using measured performance.

A U-cell can allow one operator to cover several processes, reduce walking and make WIP visible. It should not be treated as a universal solution: heavy equipment, contamination control, process isolation, automated lines and regulatory requirements may require other arrangements.

One-piece flow transfers a unit, or a very small quantity, immediately after completion at one operation. It reduces queue inventory and exposes defects earlier. However, it requires reliable equipment, compatible processing times, appropriate quality controls and manageable changeovers. Where continuous flow is technically or economically unsuitable, a controlled small-batch or supermarket arrangement may be superior.

Takt time and line balancing

Takt time expresses the production rhythm required by customer demand:

Net available production time excludes planned breaks, meetings and other unavailable periods. For example, if a shift provides 420 minutes and demand is 210 units, takt time is 2 minutes per unit. Takt is a demand signal, not the actual time a particular operation must take. Every station must normally have an effective cycle time no greater than takt, after allowing for variation and losses.

For a line with work-element times, the theoretical minimum number of stations is:

where t_i is the time of work element i, C is the selected station cycle time, and ⌈ ⌉ means round upward to the next whole station. The formula assumes work elements can be assigned feasibly without violating precedence, ergonomic or equipment constraints.

Line efficiency is:

N is the actual number of stations. The remainder represents balancing loss or theoretical idle capacity; it is not necessarily waste because some capacity is needed for variation, breaks and recovery.

Worked example. A product requires five sequential work elements of 0.8, 1.1, 0.7, 1.0 and 0.9 minutes. Demand is 180 units in a 360-minute net shift, so takt time is 360 ÷ 180 = 2.0 minutes. Total work content is 4.5 minutes, and N_min = ⌈4.5 ÷ 2.0⌉ = 3. If three stations are feasible, efficiency is 4.5 ÷ (3 × 2.0) × 100% = 75%. A manager should investigate the 25% balance loss, but should not combine tasks in a way that creates unsafe work or quality risk.

3. Push, Pull and Kanban

A push system releases work according to a central plan, forecast or material-requirements calculation. It is useful for capacity planning, long-lead procurement and situations in which demand information cannot be transmitted rapidly. Its weakness is that local plans can continue releasing work even when downstream operations are congested.

A pull system authorises replenishment when a downstream user consumes material. The upstream process produces only the quantity and mix required to replace that consumption, within agreed rules. Pull is therefore a control logic, not merely a card or software application.

Kanban is a visible or electronic signal that communicates permission to produce, move or replenish a defined quantity. Common forms include:

• Withdrawal kanban: authorises movement from a supplying point.
• Production kanban: authorises production or replenishment.
• Signal kanban: triggers a batch or setup when inventory reaches a defined level.
• Supplier kanban: communicates replenishment requirements to an external supplier.

A simple kanban quantity can be estimated as:

where n is the number of containers or kanban cards, D is average demand per time unit, L is replenishment lead time in the same time unit, S is the safety factor expressed as a decimal, and C is the standard quantity per container. Round n upward to a whole number. The formula assumes relatively stable demand, a defined replenishment loop and a safety factor that reflects lead-time and demand variation. It does not replace analysis of container space, service-level requirements, quality holds or supplier reliability.

Example. Demand is 240 units per day, replenishment lead time is 0.5 day, safety factor is 10%, and each container holds 20 units. Then n = (240 × 0.5 × 1.10) ÷ 20 = 6.6. The system requires 7 containers or signals. If demand or lead time varies substantially, the manager should reduce variation, use measured percentiles or establish a supermarket with explicit replenishment rules rather than blindly increasing cards.

Kanban limits WIP only when every signal is controlled. Extra unofficial containers, emergency releases and uncontrolled expediting convert a pull system into disguised push. A mature system includes clear ownership, FIFO lanes, card reconciliation, defect segregation and escalation rules.

4. Heijunka, Jidoka and Andon

Heijunka, or production levelling, distributes total volume and product mix more evenly over time. Instead of making a week’s demand of one model followed by another, a plant may sequence smaller repeated increments, subject to setup capability and material constraints. Heijunka reduces peaks in labour, components and capacity; it does not eliminate genuine demand variation. It depends on reliable forecasts, short changeovers, supplier responsiveness and a protected scheduling interval.

A heijunka box or electronic schedule may show the required mix by pitch interval. Pitch is commonly a convenient time interval associated with a standard production quantity, but organisations must define it consistently. Levelling can conflict with economic batch sizes, minimum supplier order quantities and campaign production; these conflicts should be made explicit and managed through total-cost analysis.

Jidoka means building quality into the process and stopping or constraining production when an abnormality occurs. It combines automatic detection, human judgment, immediate containment and structured problem solving. Stopping a line is not the final objective; restoring a capable process and preventing recurrence is.

Andon is a visual or audible abnormality signal. A team member may call for help when a defect, missing material, unsafe condition or cycle-time problem occurs. Effective Andon systems define response ownership and escalation times. A light that is ignored is decoration, not control. Managers must avoid punishing workers for revealing problems, otherwise abnormalities are hidden.

These practices are closely connected: heijunka stabilises demand on the process, jidoka protects quality, and Andon makes abnormalities visible.

5. Bottlenecks and the Theory of Constraints

A bottleneck is a resource whose effective capacity is less than or close to the demand placed upon it. The constraint may be a machine, labour skill, policy, quality approval, supplier or market condition. The Theory of Constraints (TOC) proposes five focusing steps:

1. Identify the system constraint.
2. Exploit it using existing resources; eliminate avoidable downtime and unsuitable work.
3. Subordinate other processes to the constraint’s rate and priorities.
4. Elevate the constraint by adding capacity, technology, shifts or redesign.
5. When the constraint moves, return to step one and prevent inertia from becoming the new constraint.

The drum-buffer-rope interpretation uses the constraint as the drum, a protective time or material buffer before it, and a release rope that controls upstream entry. The buffer protects throughput without allowing unlimited WIP.

A useful distinction is between utilisation and throughput. Maximising utilisation at every resource can increase queues and lead time when demand is variable. Non-constraints should often have planned idle time so that the system does not overproduce. In TOC terms, local efficiency is subordinate to system throughput.

Throughput accounting commonly distinguishes:

• Throughput contribution: sales revenue minus truly variable production costs, usually material costs in the short run.
• Investment: money tied up in inventory, equipment and other system assets.
• Operating expense: money spent to convert investment into throughput.

These categories support decisions but do not replace full financial, safety and strategic evaluation. A bottleneck improvement should be verified by increased good output, not merely by a higher local utilisation percentage.

6. Batch Size, Queues and Inventory

Large batches reduce setup frequency but increase waiting, obscure defects, consume space and make the system less responsive. Small batches reduce cycle time and expose process problems, but can increase setup effort and handling. The appropriate objective is not the smallest batch at any cost; it is an economically and operationally justified batch that supports flow.

Setup reduction uses the logic of Single-Minute Exchange of Die (SMED):

1. Separate internal activities, which require the machine to stop, from external activities, which can occur while it runs.
2. Convert internal work to external work where possible.
3. Simplify, parallelise and standardise the remaining work.
4. Use visual controls and verify first-piece quality.

Reducing setup time is valuable only when the organisation changes its planning rules and actually uses the capability for smaller lots.

Queues arise from variability, insufficient capacity, unreliability and poor release discipline. Common countermeasures include WIP caps, FIFO lanes, capacity protection at the constraint, preventive maintenance, standard work, quick changeovers and defect prevention. Inventory should be classified by purpose: cycle stock, safety stock, decoupling stock, anticipation stock or pipeline stock. Each category needs an owner, target and review rule.

7. Production Control Architecture

A robust control system connects strategic planning with shop-floor execution:

• Sales and operations planning sets aggregate volume and capacity assumptions.
• The master schedule defines near-term mix and timing.
• Heijunka translates the schedule into a levelled sequence where appropriate.
• Pull signals control replenishment within supermarkets and cells.
• Standardised work defines the current best-known method.
• Andon and daily management expose deviations.
• Problem-solving routines close the feedback loop.

Planning and pull are complementary. Pull does not mean that long-term planning is unnecessary, and a master schedule does not justify uncontrolled release. The system should specify where forecast-driven planning ends and consumption-driven replenishment begins.

Important measures include customer lead time, schedule adherence, WIP, inventory turns, throughput, first-pass yield, changeover time, downtime, response time to Andon calls, on-time delivery and employee safety. OEE can be useful for equipment analysis:

Availability, performance and quality are ratios between 0 and 1, or percentages expressed consistently. Availability usually reflects operating time divided by planned production time; performance compares actual speed with ideal speed for the produced quantity; quality is good units divided by total units. Definitions must be fixed before comparison. OEE is a local equipment measure and should not be used to encourage overproduction or to replace flow and customer-service measures.

8. Applied Case: Illustrative Appliance-Assembly Cell

Background

An illustrative appliance manufacturer produces three closely related countertop appliances. Its functional layout sends frames, wiring, testing and packaging through separate departments. Average customer lead time is 8 days, although touch time is only 42 minutes. WIP frequently exceeds 1,500 units, and urgent orders disrupt the weekly schedule.

Problem analysis

A current-state map shows long queues before final testing, inconsistent component kits and a changeover of 90 minutes between two models. Testing is the constraint: its effective capacity is 220 units per shift, while the schedule releases 260 units. Operators also discover wiring defects late, creating rework and repeated movement.

The team calculates a net shift time of 420 minutes and a demand of 210 units, giving takt time of 2.0 minutes per unit. Testing requires 2.3 minutes per unit, so it cannot meet takt without improvement. A local effort to keep assembly fully utilised has increased the testing queue rather than customer output.

Approach

The team creates a mixed-model cell for the three families, places kitting, frame assembly and wiring in route sequence, and establishes a controlled supermarket before testing. It introduces connector mistake-proofing and a first-piece verification after changeover. SMED reduces the test-equipment changeover from 90 to 30 minutes. A production kanban limits the supermarket to 120 units, while a withdrawal signal authorises replenishment.

The constraint is exploited by removing avoidable test interruptions, standardising test fixtures and scheduling maintenance outside the protected production window. Assembly is subordinated to the test rate through a drum-buffer-rope rule. A small Andon system calls the team leader when a unit fails the standard time or a defect is detected. Daily heijunka sequences the three models in repeated groups rather than large campaigns.

Outcome

After a controlled pilot, average WIP falls from 1,500 to 600 units and lead time falls from 8 days to 3.5 days. First-pass yield at wiring improves from 94% to 98%. Testing output rises to 245 good units per shift, but the team does not claim that every improvement is caused by the cell; demand, staffing and measurement definitions are monitored. The remaining constraint is reviewed monthly, and the supermarket size is changed only through an approved experiment.

Learning points

The case illustrates that layout change alone is insufficient. Flow required quality at the source, constraint protection, smaller changeovers, controlled release and employee authority to expose abnormalities. The improvement also demonstrates the danger of improving an upstream department independently of the constraint.

9. Implementation, Governance and Contemporary Technology

Implementation should begin with a clearly defined value stream and a stable pilot, not with indiscriminate conversion of every department. Leaders must provide training in standardised work, problem solving, safety and data interpretation. Operators should participate in design because ergonomic burden and practical variation are often invisible in engineering plans.

Risks include superficial use of Japanese terminology, excessive visual controls, unrealistic takt pressure, hidden overtime, unsafe work intensification, poor supplier collaboration and inaccurate data. Lean cannot justify bypassing statutory safety, quality, environmental or labour requirements. Production targets must never reward concealment of defects or unsafe shortcuts. Governance should include change control, traceable quality records, segregation of nonconforming material and protection of confidential operational data.

Digital kanban, RFID, sensors, manufacturing-execution systems and real-time dashboards can improve visibility. AI may help forecast demand, detect anomalies or optimise sequencing, but it should not replace process understanding. Algorithms can amplify biased data, create opaque priorities or trigger unstable rescheduling. Human approval, audit trails, cybersecurity, access control and clear exception rules are essential. Technology should simplify and strengthen the control logic rather than conceal excessive WIP behind screens.

Contemporary lean systems increasingly combine physical flow with environmental objectives. Smaller batches and reduced defects can reduce material and energy use, but frequent changeovers or extra transport may offset benefits. Carbon, water, energy and worker well-being should therefore be included in value-stream analysis. Resilient systems also use strategic buffers and dual sourcing where disruption risk is high; zero inventory is neither realistic nor inherently desirable.

Best-Practice Checklist

1. Define customer demand, net available time and the planning horizon.
2. Map material and information flow using measured data.
3. Identify the constraint before balancing local utilisation.
4. Design standardised work with safety and ergonomics included.
5. Use takt as a rhythm, not as permission to overload workers.
6. Introduce pull only after replenishment rules, container quantities and ownership are clear.
7. Reduce setup time and then use the capability to reduce batch size.
8. Protect the constraint with an appropriate buffer and control upstream release.
9. Make abnormalities visible and respond without blame.
10. Review performance using system-level outcomes, not isolated efficiency figures.

Chapter Summary

Lean flow synchronises work with customer demand and seeks to minimise waiting, queues, excess transport, defects and unnecessary inventory. Cellular layouts place related processes together; one-piece flow reduces interruption and exposes problems, but depends on stability and capable processes. Takt time expresses required production rhythm, while line balancing assigns work within precedence and capacity constraints.

Push systems release work from a plan; pull systems replenish actual consumption. Kanban provides a controlled signal, and its quantity must reflect demand, replenishment lead time, safety allowance and container size. Heijunka levels volume and mix, jidoka builds quality into the process, and Andon makes abnormalities visible. The Theory of Constraints directs attention to system throughput by identifying, exploiting, subordinating and elevating the constraint.

Small batches, SMED, WIP limits, FIFO discipline and quality at the source shorten lead time. Effective production control combines aggregate planning with levelled sequencing and local pull. Digital technologies can strengthen visibility but require human governance, cybersecurity and ethical control. Sustainable implementation respects safety, legal obligations, employee voice, supplier realities and the difference between local efficiency and total-system performance.

Key Terms

Andon: A visual or audible signal indicating an abnormal condition and requesting assistance.

Bottleneck: A resource whose effective capacity limits or nearly limits system output.

Cellular layout: A compact arrangement of different resources serving a product family.

Cycle time: Actual elapsed time between completed units at an operation.

Heijunka: Production levelling by volume and product mix.

Jidoka: Built-in quality supported by automatic or human detection and response to abnormalities.

Kanban: A signal authorising production, movement or replenishment.

Lead time: Total elapsed time from entry into a process or order release to completion.

Line balancing: Assigning work elements to stations to meet a required rhythm with minimal feasible imbalance.

Little’s Law: The stable-system relationship WIP = Throughput × Lead time.

One-piece flow: Moving units singly or in very small quantities between operations.

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

Push system: A system in which release is primarily based on a plan or forecast.

SMED: A method for reducing setup and changeover time.

Supermarket: A controlled inventory point replenished according to consumption.

Takt time: Net available production time divided by customer demand.

Theory of Constraints: A management approach focused on the system’s limiting constraint.

Throughput: The rate at which the system completes acceptable output.

WIP: Work-in-process inventory currently inside the production system.

Review Questions

Short Answer Questions

1. Distinguish takt time from cycle time.
2. What is the purpose of a production kanban?
3. State the five focusing steps of the Theory of Constraints.
4. Why can excessive upstream utilisation increase lead time?
5. What is the role of Andon in a jidoka system?

Long Answer Questions

1. Explain how a cellular layout can support one-piece flow, and discuss conditions under which it may be unsuitable.
2. Compare push and pull production control, including the role of forecasting, supermarkets and kanban.
3. Explain heijunka, jidoka and Andon as mutually reinforcing elements of production control.
4. Discuss how SMED, WIP limits and quality at the source reduce batch size and lead time.
5. Critically evaluate OEE as a measure in a flow-oriented production system.

Analytical / Application Questions

1. A line has 390 net minutes per shift and demand of 195 units. Its work content is 5.4 minutes, and four feasible stations are available. Calculate takt time, theoretical minimum stations and line efficiency if four stations are used. Interpret the result.
2. A supermarket serves demand of 300 units per day. Replenishment lead time is 0.4 day, the safety factor is 15%, and each container holds 24 units. Calculate the required number of containers, rounding upward, and identify two assumptions that should be validated before implementation.
3. A plant has a highly utilised upstream department but a persistent queue before a slower testing department. Using TOC and pull principles, recommend a sequence of actions and explain which performance measures should determine whether the intervention succeeded.

Multiple Choice Questions with Answers

Question 1

Which statement best describes a pull system?

A. Upstream production is authorised by downstream consumption.

B. Every department produces at maximum utilisation.

C. Production is based only on the annual sales forecast.

D. Large batches are released before customer orders arrive.

Correct Answer: A. Upstream production is authorised by downstream consumption.

Question 2

A line has 420 net minutes and demand of 210 units. What is the takt time?

A. 0.5 minute per unit

B. 2 minutes per unit

C. 210 minutes per unit

D. 630 minutes per unit

Correct Answer: B. 2 minutes per unit

Question 3

Which condition is an assumption of the simple kanban quantity formula?

A. Demand and replenishment lead time are defined in consistent time units.

B. All processes have unlimited capacity.

C. Safety stock must always equal zero.

D. The system has no container-size restriction.

Correct Answer: A. Demand and replenishment lead time are defined in consistent time units.

Question 4

What is the primary purpose of heijunka?

A. To maximise inventory at every workstation

B. To level production volume and mix over time

C. To eliminate all planned maintenance

D. To replace quality inspection with final sorting

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

Question 5

In the Theory of Constraints, what follows identification of the constraint?

A. Exploit the constraint

B. Increase every department’s batch size

C. Maximise non-constraint utilisation

D. Remove all protective buffers

Correct Answer: A. Exploit the constraint

Question 6

A process has total work content of 4.5 minutes, a cycle-time target of 2 minutes and three stations. What is line efficiency?

A. 50%

B. 66.7%

C. 75%

D. 150%

Correct Answer: C. 75%

Question 7

What is the central purpose of jidoka?

A. To hide defects until final inspection

B. To build in detection and response to abnormalities

C. To increase batch size

D. To eliminate employee involvement

Correct Answer: B. To build in detection and response to abnormalities

Question 8

Which action is consistent with SMED?

A. Separating internal and external setup activities

B. Increasing the number of unfinished units before setup

C. Delaying first-piece verification

D. Performing all preparation after the machine stops

Correct Answer: A. Separating internal and external setup activities

Question 9

According to Little’s Law, if throughput remains constant and WIP is reduced, what normally happens to average lead time?

A. It increases proportionally

B. It decreases proportionally

C. It becomes exactly zero

D. It cannot be related to WIP

Correct Answer: B. It decreases proportionally

Question 10

Why can maximising utilisation at every resource be harmful in a variable system?

A. It can create queues and obscure the system constraint.

B. It always reduces equipment availability to zero.

C. It prevents any form of customer demand.

D. It guarantees one-piece flow.

Correct Answer: A. It can create queues and obscure the system constraint.