Bottlenecks, Constraints and Flow Management

Introduction

Capacity requirement planning is not only the task of comparing aggregate demand with available machine hours. In real production systems, output is governed by the interaction of resources, product mix, variability, information, quality, material availability and operating policies. A plant may possess sufficient total capacity and still fail to meet demand because one resource, policy or process condition restricts the flow of work. That restricting condition is a constraint.

A constraint may be physical, such as a heat-treatment furnace, skilled technician or inspection station. It may also be managerial, such as a batch-size rule, release policy, approval procedure, inaccurate planning parameter or incentive system that rewards local utilisation rather than system throughput. Constraints may persist for years or shift from one resource to another as demand, product mix and operating policies change.

This chapter develops an integrated approach to bottlenecks and flow management. It combines the Theory of Constraints (TOC), drum-buffer-rope (DBR), throughput accounting, line balancing and protective capacity. Particular attention is given to the effects of constraints on throughput, work-in-process (WIP), lead time and delivery reliability.

Learning Objectives

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

1. Distinguish bottlenecks, capacity constraints, non-bottlenecks and temporary constraints.
2. Diagnose shifting and persistent bottlenecks using capacity, flow and variability data.
3. Apply the five focusing steps of the Theory of Constraints.
4. Design a drum-buffer-rope control system and select appropriate buffers.
5. Use throughput accounting to evaluate operational and investment decisions.
6. Balance production lines and assess the effects of protective capacity.
7. Evaluate improvement proposals by their effects on throughput, WIP and delivery performance.
8. Recognise the managerial, technological and ethical risks of local optimisation.

1. Bottlenecks and Constraints

A constraint is anything that limits a system from achieving a higher level of its goal. In a manufacturing context, the goal is commonly expressed as generating sustainable profit by delivering customer value, although safety, compliance and quality are non-negotiable conditions.

A bottleneck is a resource whose effective capacity is less than, or close to, the demand placed upon it. Effective capacity is the output realistically available after allowing for planned downtime, changeovers, quality losses, absenteeism and other restrictions. A resource can be highly utilised without being a bottleneck; conversely, a resource can be a bottleneck even if its average utilisation appears moderate when variability and due-date requirements are considered.

Important distinctions are shown below.

The relevant capacity calculation is generally resource- and period-specific. For resource r in period t, the capacity load ratio is:

where D_r,t is required workload in hours, C_r,t is effective available capacity in hours, and L_r,t is dimensionless. A ratio above 1 indicates an overload under the stated assumptions. Workload should include processing time, setup time and other resource-specific time requirements, not merely run time.

For multiple products, workload is calculated as:

Here Q_i,t is quantity of product i, p_i,r is processing time per unit on resource r, s_i,r is setup time per setup, and B_i,r is batch size. The formula assumes setup frequency is represented by quantity divided by batch size and that all times use the same unit, such as hours.

Capacity ratios alone are insufficient. A product may require the same machine at different times, queues may be created by release policies, and a machine may be unavailable when its output is needed. Bottleneck identification therefore combines capacity calculations with queue lengths, starvation, blocking, due-date performance and material availability.

2. Flow, WIP and Delivery Performance

Flow management concerns the movement of material, information and decisions through the system. The central relationship is Little’s Law:

where WIP is average work-in-process, TH is average throughput rate, and CT is average cycle time or flow time. Units must be consistent—for example, units, units per day and days. The relationship applies to a stable system over the measurement period, where average input and output rates are approximately equal and the same system boundary is used for all three measures.

Thus:

If a plant produces 240 units per day and holds an average of 1,200 units within a defined production system, average flow time is 5 days. Reducing WIP to 960 units while maintaining throughput reduces average flow time to 4 days. However, reducing WIP without protecting the constraint may starve the bottleneck and reduce throughput. The objective is not minimum WIP at any cost; it is controlled WIP consistent with reliable throughput and delivery.

A bottleneck creates two characteristic effects. Upstream, material accumulates because release exceeds the bottleneck’s output. Downstream, resources may be starved unless they have work or finished-goods protection. Variability amplifies both effects. Breakdowns, quality rework, arrivals of urgent orders and uneven processing times reduce the practical capacity available to the system.

3. Theory of Constraints

The Theory of Constraints views the organisation as a system whose performance is limited by a small number of constraints. Its five focusing steps are:

1. Identify the constraint. Locate the resource, policy or market condition that limits system performance.
2. Exploit the constraint. Obtain more output from existing capacity without major investment. Examples include reducing avoidable downtime, ensuring materials and tools are ready, improving first-pass yield, reducing unnecessary setups and assigning skilled operators.
3. Subordinate everything else. Align release, scheduling, maintenance, quality and purchasing decisions with the constraint’s requirements. Non-constraints should not produce merely to remain busy.
4. Elevate the constraint. Add capacity or remove the restriction through overtime, outsourcing, equipment, redesign, cross-training, parallel processing or policy change.
5. Repeat the process. Once one constraint is broken, another normally governs system output. The process must return to identification, while avoiding inertia and local optimisation.

The third step is frequently the most difficult. It challenges conventional utilisation measures. Producing inventory at a non-bottleneck can increase WIP and hide quality problems without increasing sellable output. A non-bottleneck should have enough work to support the constraint and customer demand, but not unlimited work.

Physical and policy constraints

Physical constraints are visible and often easier to measure. Policy constraints may be more consequential. Examples include large minimum batch sizes, a rule that every machine must operate at high utilisation, excessive approval levels, a forecast-driven release system, or accounting practices that treat all inventory as an asset regardless of obsolescence risk.

A useful diagnostic question is: “If this rule were removed, would system throughput or delivery performance improve?” If yes, the system may be constrained by policy rather than equipment.

4. Drum-Buffer-Rope Control

Drum-buffer-rope is a TOC scheduling and release mechanism.

• The drum is the schedule set by the constraint. It establishes the pace and sequence of critical production.
• The buffer is strategically positioned protective time or inventory that protects the drum from disruption.
• The rope is the release signal that authorises upstream work at a rate and sequence compatible with the drum.

The drum is not necessarily the first operation or the fastest possible schedule. It is a feasible schedule that recognises constraint capacity, due dates, setups, material availability and maintenance requirements.

Buffers can be located before the constraint, before shipping, and sometimes at critical assembly or merging points. A constraint buffer protects the constraint from starvation. A shipping buffer protects customer delivery from internal disruption. Buffers should be measured by the time remaining before a potential interruption, not only by the number of pieces present.

A practical buffer-management method divides buffer status into zones:

DBR is especially useful where detailed finite scheduling of every resource would create excessive nervousness. It does not eliminate planning. Rather, it concentrates planning precision where it has the greatest system effect.

5. Throughput Accounting

Throughput accounting evaluates decisions by their effect on system performance rather than by allocating fixed overhead to individual products. Three core measures are:

• Throughput (T): the rate at which the system generates money through sales. A common formulation is sales revenue minus totally variable cost, usually direct materials and other costs that vary directly with each unit sold.
• Investment (I): money tied up in inventory, equipment and other assets required to operate the system.
• Operating expense (OE): money spent to convert investment into throughput, including labour and operating costs that do not vary directly with each additional unit in the short run.

A commonly used relationship is:

and, when investment is considered:

The formulation assumes consistent accounting periods and definitions. Throughput accounting is a managerial decision framework, not a replacement for statutory financial reporting.

For product j, unit throughput may be represented as:

where P_j is selling price per unit and DM_j is direct material cost per unit. If the bottleneck time per unit is b_j, throughput per bottleneck hour is:

This measure is useful when the constraint is fully loaded and products compete for the same scarce time. It should not be used mechanically: market demand, quality, strategic importance, contractual commitments, setup effects and future capacity must also be considered.

Worked example: product mix at a constrained machine

Product A sells for 90 monetary units and has direct material cost of 30; it requires 6 minutes at the constraint. Product B sells for 84, has direct material cost of 24, and requires 4 minutes.

For A, unit throughput is 60 and throughput per constraint hour is:

For B, unit throughput is 60 and throughput per constraint hour is:

If demand exists for both products and all other conditions are equal, B should receive priority because it generates more throughput per constrained hour. The conclusion changes if B’s demand is limited, its material supply is unreliable, or its production creates disproportionate quality or delivery risk.

6. Line Balancing and Protective Capacity

A production line contains tasks with precedence relationships. Line balancing assigns tasks to workstations so that required output is achieved with acceptable idle time and workload variation.

For a required output of Q units during available production time A, the cycle time is:

where A and C use the same time unit. If daily available time is 420 minutes and required output is 60 units, cycle time is 7 minutes per unit.

If total task time is ∑t_i, the theoretical minimum number of stations is:

The ceiling function rounds upward because a fraction of a station is not feasible. This is a lower bound, not necessarily an achievable solution; precedence restrictions and task indivisibility may require more stations.

Line efficiency is:

where N is the actual number of stations. Balance delay is 1 − E. These measures assume a common cycle time and comparable station availability. They do not capture all effects of changeovers, learning, absenteeism, rework or ergonomic constraints.

Protective capacity is intentionally unused or flexible capacity maintained to absorb variability and protect flow. It may take the form of cross-trained workers, spare machine time, alternate suppliers, parallel equipment, flexible shifts or capacity at downstream operations. A system designed with every resource at 100% average utilisation has little ability to absorb disruptions. The effective capacity of the system may therefore be lower than its nominal capacity.

Protective capacity should be positioned strategically. Capacity immediately before a bottleneck may be valuable if it prevents starvation, while excess capacity far from the constraint may merely increase WIP. The appropriate level depends on demand variability, process variability, recovery time, customer penalty, safety requirements and the cost of flexibility.

7. Diagnosing Shifting Bottlenecks

A bottleneck can shift because of product mix, seasonal demand, engineering changes, maintenance, absenteeism, learning, outsourcing or improvement at the former constraint. A decision based on last month’s bottleneck may therefore be wrong for next month.

A robust diagnostic process includes:

1. Establish the system boundary and planning horizon.
2. Convert each product plan into resource-specific workload.
3. Include setups, yield losses, planned downtime and realistic staffing.
4. Compare workload with effective capacity by period.
5. Validate calculations against queues, starvation, blocking and schedule adherence.
6. Test alternative mixes and disruption scenarios.
7. Identify whether the constraint is physical, policy-based, supply-based or market-based.
8. Reassess after every major improvement or product change.

Useful indicators include constraint utilisation, queue time before the constraint, starvation minutes, blocked minutes, first-pass yield, schedule adherence, buffer penetration, WIP by operation and throughput per constraint hour. No single indicator is sufficient. High utilisation may reflect excessive WIP rather than healthy flow.

8. Integrated Case: Illustrative Precision-Pump Plant

Background

A precision-pump plant produces two families, P1 and P2. The machining centre is shared by both products and supplies an assembly line. The plant works 20 days per month, with 7 productive hours per day, giving 140 available hours per resource before additional losses. A realistic effective capacity of the machining centre is 126 hours after planned maintenance and normal losses.

Problem

Monthly demand is 900 units of P1 and 600 units of P2. Machining times are 5 minutes for P1 and 8 minutes for P2. Assembly has sufficient nominal capacity, but customer delivery is poor and WIP before assembly is rising.

Analysis

Machining workload is:

The machining load ratio is:

Machining is therefore almost fully loaded before unplanned disruption. A small breakdown or quality loss will create a queue. Assembly may appear underutilised because it is starved whenever machining misses the sequence or material is unavailable.

Suppose unit throughput contributions are 52 monetary units for P1 and 68 for P2. Throughput per machining hour is:

• P1: 52 ÷ (5/60) = 624 monetary units per hour.
• P2: 68 ÷ (8/60) = 510 monetary units per hour.

If demand and contractual conditions permit, P1 has priority for scarce machining time. However, simply producing more P1 is not a complete solution because customer demand is finite and P2 commitments must be respected.

Approach

The plant applies the five focusing steps. First, machining is identified as the constraint. Second, it is exploited by pre-staging tools and materials, separating preventive maintenance from peak production, reducing avoidable changeovers and improving first-pass yield. Third, release of work is subordinated to the machining schedule; upstream processes stop producing components that cannot be accepted by machining. A time buffer is placed before machining and a smaller shipping buffer protects confirmed customer orders. The rope releases work according to the machining sequence.

Fourth, the plant elevates the constraint by cross-training operators and outsourcing selected low-risk machining operations. It also evaluates a second machining centre. The investment decision is assessed using throughput per constrained hour, incremental operating expense, quality risk and expected demand rather than machine utilisation alone.

Outcome

After improvement, effective machining capacity rises to 138 hours. Demand workload remains 125 hours, leaving 13 hours of protective capacity. WIP before machining falls because release is controlled, while assembly receives a more reliable sequence. The plant should continue monitoring the system because the next constraint may become assembly, an external supplier or the market.

Learning points

The case demonstrates that total nominal capacity can be misleading. The decisive issue is effective capacity at the resource that governs flow. It also shows why local utilisation, uncontrolled release and product-margin analysis can produce conflicting decisions. A constraint-centred policy improves throughput only when it is connected to demand, quality and delivery requirements.

9. Technology, Governance and Contemporary Practice

Manufacturing execution systems, finite-capacity planning, industrial IoT sensors and digital twins can improve constraint visibility. Real-time data may reveal queue growth, micro-stoppages, starvation and quality losses that periodic reports conceal. AI-based forecasting and anomaly detection can support maintenance and demand planning, but they do not automatically identify the system constraint. Poor master data, biased historical schedules and inappropriate optimisation objectives can cause an algorithm to maximise local utilisation while worsening flow.

Governance is therefore essential. Organisations should define data ownership, audit scheduling rules, protect personal and operational data, and maintain human accountability for safety, quality and customer commitments. AI recommendations should be explainable enough for planners and supervisors to challenge them, particularly when they reallocate scarce capacity among customers.

Ethically, constraint decisions must not encourage unsafe overtime, concealment of quality defects, unfair prioritisation of customers, or excessive work pressure on constrained employees. Regulatory requirements for safety, traceability, environmental compliance and labour conditions override throughput objectives. A faster flow of defective or unsafe products is not operational excellence.

10. Limitations and Best Practices

TOC and throughput accounting are powerful but not universal. A constraint may be difficult to define in project, service or highly customised environments. Throughput measures can oversimplify long-term product economics if research, selling, environmental and lifecycle costs are ignored. Line-balancing efficiency can encourage excessive standardisation and overlook ergonomics. DBR requires reliable data and disciplined execution; poorly sized buffers can either starve the constraint or create excessive inventory.

Best practice is to:

• Define the system goal and boundary explicitly.
• Use effective, not nominal, capacity.
• Analyse workload by period, product and resource.
• Combine quantitative ratios with direct observation of flow.
• Protect the constraint from starvation and quality problems.
• Release work according to system need, not local utilisation targets.
• Use throughput per constrained hour alongside strategic and contractual considerations.
• Maintain protective capacity where variability and recovery risk justify it.
• Review the constraint after every significant improvement.
• Treat safety, quality, legality and ethics as boundary conditions.

Chapter Summary

Bottlenecks and constraints govern the performance of production systems. A constraint may be physical, policy-based, supply-related or market-based, and it may persist or shift as product mix and operating conditions change. Capacity analysis must include processing, setup, downtime, yield and staffing effects.

The Theory of Constraints provides five focusing steps: identify, exploit, subordinate, elevate and repeat. Drum-buffer-rope translates this logic into operational control through a constraint schedule, protective buffers and controlled release. Throughput accounting evaluates decisions by their effects on throughput, investment and operating expense, while throughput per constrained hour supports product-mix decisions.

Line balancing aligns task workloads with cycle time, but its efficiency measures are only lower-level indicators. Protective capacity is strategically unused or flexible capacity that absorbs variability and protects flow. Across all approaches, Little’s Law demonstrates the relationship among WIP, throughput and cycle time. Sustainable improvement requires system-level optimisation, reliable data, ethical governance and continuous reassessment of the active constraint.

Key Terms

Constraint: Anything that limits a system from achieving a higher level of its goal.

Bottleneck: A resource whose effective capacity is less than or close to required demand.

Effective capacity: Realistically available capacity after normal losses and restrictions.

Persistent bottleneck: A constraint that remains restrictive across planning periods.

Shifting bottleneck: A constraint that changes with mix, time or operating conditions.

Theory of Constraints: A system-improvement approach centred on managing the limiting constraint.

Exploit: Improve output from an existing constraint without major investment.

Subordinate: Align non-constraints and policies with the constraint’s requirements.

Elevate: Add capacity or remove the constraint through investment or redesign.

Drum: The schedule and pace established by the constraint.

Buffer: Protective time or inventory positioned to protect flow.

Rope: The release signal that controls upstream work.

Throughput: The rate at which the system generates money through sales.

Throughput accounting: A decision framework based on throughput, investment and operating expense.

Work-in-process: Inventory currently inside the production system.

Little’s Law: The stable-system relationship WIP = throughput × cycle time.

Line balancing: Assigning tasks to stations to meet output with acceptable workload balance.

Protective capacity: Flexible or unused capacity retained to absorb variability.

First-pass yield: The proportion of output that meets requirements without rework.

Local optimisation: Improving one resource or measure while harming overall system performance.

Review Questions

Short Answer Questions

1. Distinguish a bottleneck from a constraint.
2. What are the five focusing steps of the Theory of Constraints?
3. Explain the roles of the drum, buffer and rope in DBR.
4. State Little’s Law and define its three variables.
5. Why can a resource with high utilisation fail to improve system throughput?

Long Answer Questions

1. Explain how product mix and variability can cause a bottleneck to shift between resources.
2. Discuss the relationship between throughput accounting and conventional cost accounting.
3. Explain how protective capacity can improve delivery reliability while increasing apparent idle time.
4. Describe a complete diagnostic method for identifying a policy constraint.
5. Critically evaluate the limitations of applying TOC in a highly customised production environment.

Analytical / Application Questions

1. A plant has 480 available minutes per day and must produce 80 units. Calculate cycle time and explain how it would be used in line balancing.
2. Product X generates 45 monetary units of throughput and uses 6 minutes of bottleneck time; Product Y generates 50 monetary units and uses 8 minutes. Which product has the higher throughput per bottleneck hour, and what additional information should a manager examine before changing the mix?
3. A system produces 150 units per day with average WIP of 900 units. Calculate average flow time. If WIP is reduced to 750 units while throughput remains constant, calculate the new flow time and discuss one risk of reducing WIP too aggressively.

Multiple Choice Questions with Answers

Question 1

Which condition most directly indicates a bottleneck?

A. A resource has low utilisation and no queue

B. Required workload exceeds effective available capacity

C. A resource has the highest purchase price

D. A resource has the largest floor area

Correct Answer: B. Required workload exceeds effective available capacity

Question 2

In the Theory of Constraints, what does “exploit the constraint” mean?

A. Purchase a completely new production system immediately

B. Increase output from existing constrained capacity without major investment

C. Keep every non-bottleneck operating continuously

D. Increase finished-goods inventory regardless of demand

Correct Answer: B. Increase output from existing constrained capacity without major investment

Question 3

In drum-buffer-rope, the drum primarily represents:

A. The schedule and pace established by the constraint

B. The total quantity of raw material purchased

C. The accounting depreciation schedule

D. The inspection procedure after shipping

Correct Answer: A. The schedule and pace established by the constraint

Question 4

A system has average throughput of 200 units per day and average WIP of 1,000 units. According to Little’s Law, average flow time is:

A. 2 days

B. 5 days

C. 200 days

D. 1,200 days

Correct Answer: B. 5 days

Question 5

A product sells for 100 monetary units, has direct material cost of 40, and uses 6 minutes of bottleneck time. Its throughput per bottleneck hour is:

A. 600 monetary units per hour

B. 400 monetary units per hour

C. 1,000 monetary units per hour

D. 60 monetary units per hour

Correct Answer: A. 600 monetary units per hour

Question 6

Which action is an example of subordinating a non-bottleneck?

A. Releasing work whenever the non-bottleneck is idle

B. Scheduling upstream output to match the constraint’s requirements

C. Maximising local utilisation regardless of WIP

D. Removing all protective buffers

Correct Answer: B. Scheduling upstream output to match the constraint’s requirements

Question 7

If available production time is 420 minutes per day and required output is 60 units per day, the line cycle time is:

A. 5 minutes per unit

B. 6 minutes per unit

C. 7 minutes per unit

D. 8 minutes per unit

Correct Answer: C. 7 minutes per unit

Question 8

Protective capacity is best described as:

A. Capacity deliberately retained to absorb variability and protect flow

B. Capacity permanently removed from the system

C. Inventory that has already become obsolete

D. Theoretical capacity calculated without downtime

Correct Answer: A. Capacity deliberately retained to absorb variability and protect flow

Question 9

Which event can cause a bottleneck to shift?

A. A change in product mix or improvement at the former constraint

B. Using the same planning horizon every month

C. Defining all resources in the same unit of measure

D. Maintaining accurate master data

Correct Answer: A. A change in product mix or improvement at the former constraint

Question 10

Throughput accounting commonly defines net profit as:

A. Investment minus throughput

B. Throughput minus operating expense

C. Operating expense minus investment

D. Sales revenue plus inventory

Correct Answer: B. Throughput minus operating expense