Capacity, Scheduling and Flow Management

Introduction and context

Operations systems transform inputs into goods or services through people, equipment, information and facilities. Their performance depends not only on what the system can produce, but also on when work is released, how resources are assigned and how variation is absorbed. Capacity, scheduling and flow management therefore connect strategic operations design with daily execution.

For an entrepreneurial firm, these issues are especially important. Demand is uncertain, resources are constrained and premature investment can destroy cash flow. A venture that accepts more orders than its process can reliably complete may suffer late delivery, quality failures and reputational damage. Conversely, excessive capacity creates idle resources and weak returns on investment. Effective operations management seeks a viable balance among capacity, responsiveness, cost, quality and resilience.

This chapter develops the concepts of capacity, utilization, bottlenecks, queues, throughput and cycle time. It then applies loading, sequencing, workforce scheduling, line balancing, critical-path analysis and finite-capacity scheduling. The emphasis is on interpreting quantitative results for managerial decisions rather than treating formulas as substitutes for judgment.

Learning objectives

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

1. Distinguish design, effective and actual capacity and calculate utilization and efficiency.
2. Diagnose bottlenecks and explain their effects on throughput and flow.
3. Apply Little’s Law and queueing logic to operational decisions.
4. Develop workforce, production and service schedules using loading and sequencing rules.
5. Balance an assembly or service line subject to precedence and capacity constraints.
6. Construct a critical path and identify schedule flexibility.
7. Explain finite-capacity scheduling and its implementation challenges.
8. Evaluate capacity decisions using cost, service, risk and ethical criteria.

1. Capacity as a managed resource

Capacity is the maximum output rate that a process, resource or system can deliver under specified conditions. It must always be expressed with a time basis and a unit of output: for example, 240 customer calls per eight-hour shift, or 600 units per week. Capacity is not a permanent property. It changes with product mix, worker skill, maintenance, setup time, quality requirements, operating hours and technology.

Three distinctions are useful:

If demand exceeds effective capacity, management must increase resources, reduce work content, subcontract, change priorities or accept unmet demand. If demand is below capacity, it may be possible to consolidate shifts, cross-train employees or use spare time for maintenance and improvement.

Utilization and efficiency

Utilization compares actual output with design capacity:

Efficiency compares actual output with effective capacity:

Both ratios require the same output unit and time period in numerator and denominator. Utilization can be lower than efficiency when design capacity includes ideal conditions. In unusual circumstances, reported utilization may exceed 100% because of overtime, temporary outsourcing or a change in the definition of capacity; this signals that the baseline should be investigated rather than accepted uncritically.

Worked example. A packaging cell has design capacity of 1,000 cartons per shift and effective capacity of 850 cartons. It produces 765 cartons. Utilization is 765 ÷ 1,000 = 76.5%. Efficiency is 765 ÷ 850 = 90%. The gap between the two measures is largely explained by planned losses embedded in effective capacity.

Capacity planning operates at three levels. Strategic capacity concerns facilities, technology and long-term workforce commitments. Tactical capacity concerns shifts, subcontracting, inventory and workforce flexibility over months. Operational capacity concerns daily assignment, release and sequencing of work.

A capacity cushion is capacity deliberately reserved above expected demand. It protects against demand variation, absenteeism, equipment failure and urgent orders. A high cushion improves responsiveness but raises cost; a low cushion improves asset utilization but increases waiting and service risk. The appropriate cushion depends on demand volatility, consequences of lateness, replenishment lead time and the cost of capacity.

2. Flow, bottlenecks and throughput

A process consists of linked activities. The bottleneck is the resource whose available capacity is most restrictive relative to required flow. It may be a machine, employee, approval stage, supplier, information system or policy. The bottleneck determines the maximum sustainable throughput of the system when demand is sufficient.

Throughput is the rate at which completed, saleable output leaves the system. It should not be confused with activity or utilization. Producing defective units, starting too many jobs or moving work between queues can increase local activity without increasing throughput.

A practical bottleneck-diagnosis procedure is:

1. Map the process and measure demand, processing time, setup time, downtime and yield.
2. Calculate available capacity for each resource.
3. Compare required load with capacity.
4. Confirm the suspected constraint through queue growth, waiting time and output data.
5. Improve the constraint before adding capacity elsewhere.

The Theory of Constraints expresses this logic as identify the constraint, exploit it, subordinate other activities to it, elevate it when justified and repeat the cycle. “Exploit” means obtaining more useful output from existing constraint capacity—for example, reducing changeovers or preventing breaks from coinciding with peak demand. “Subordinate” means preventing upstream resources from flooding the constraint with work-in-process.

For a product requiring sequential stages, a simplified line capacity is determined by the smallest stage capacity. If stage i has processing time tᵢ minutes per unit and one worker or machine is available, its rate is 60 ÷ tᵢ units per hour. With multiple parallel resources, capacity is the sum of their rates, assuming work can be shared and quality is equivalent.

Bottlenecks can move. Improving one constrained stage may expose another. Managers should therefore monitor the entire flow rather than optimize isolated departments.

3. Queues, work-in-process and Little’s Law

A queue forms when arrivals temporarily exceed service capacity or when work is deliberately released before service is available. Waiting is caused not only by average overload but also by variability in arrivals, processing times, failures, setups and priority rules.

Little’s Law links three long-run averages:

where L is average number of items in the system (units, customers or jobs), λ is the average throughput or arrival rate (items per unit of time), and W is average time in the system (time per item). The corresponding waiting-only form is:

where L_q is average queue length and W_q is average waiting time before service. The law applies to a stable system over a sufficiently long period, with consistent definitions of arrivals, departures and system boundaries. It does not require a particular probability distribution, but throughput must equal the long-run departure rate; otherwise the system is accumulating or losing work.

Worked example. A repair centre completes 18 jobs per day. The average time from job receipt to return is 2.5 days. Therefore average work-in-process is L = 18 × 2.5 = 45 jobs. If technicians spend 0.75 day on average actually repairing each job, average queue and administrative delay are 2.5 − 0.75 = 1.75 days. The corresponding pre-service work is L_q = 18 × 1.75 = 31.5 jobs, interpreted as a long-run average rather than a fractional individual queue.

Little’s Law is a diagnostic relationship, not a scheduling rule. If throughput is fixed, reducing work-in-process reduces elapsed time. However, simply removing inventory without addressing variability can cause stockouts, starvation and unstable flow. Managers should define the system boundary carefully: a warehouse, production cell and customer order process will produce different values of L and W.

A key operational insight is that high utilization near 100% can create disproportionate waiting when arrivals and processing are variable. Consequently, the economically optimal operating point is usually below theoretical maximum utilization, particularly for emergency services, repair operations and customized production.

4. Loading and sequencing

Loading assigns work to resources; sequencing determines the order in which assigned work is processed. A load profile compares required work with available time by resource and period. If job j requires pⱼ hours on a resource and quantity qⱼ, its load is:

The summation covers all jobs assigned to that resource. If available time is A hours, the load ratio is Load ÷ A. A ratio above 1 indicates infeasibility unless overtime, subcontracting, parallel resources or a later due date is introduced.

Common sequencing rules include:

For job j, the critical ratio is:

A ratio below 1 indicates that the job is already behind its estimated processing requirement, assuming no further delay. The rule should be recalculated as conditions change.

Sequencing example. Three jobs are available at time 0:

Shortest-processing-time sequence is B–C–A. Completion times are 2, 5 and 9 hours, giving an average flow time of 16 ÷ 3 = 5.33 hours. Earliest-due-date sequence is B–C–A as well. If A had a due date of 3 hours, earliest due date would place A first, illustrating why the appropriate rule depends on the objective.

No sequencing rule optimizes every performance measure. A manager must specify whether the priority is average flow time, maximum lateness, customer criticality, fairness, setup reduction or contribution margin. Rules also require governance: priority changes should be visible, authorized and explainable.

5. Workforce and service scheduling

Workforce scheduling translates demand by time interval into shifts, skills and assignments. It must account for labor laws, contracts, breaks, skill restrictions, training, absenteeism and employee well-being. A basic schedule begins with a demand profile by hour or day, converts demand into required labor hours, and assigns qualified employees while respecting coverage constraints.

For a service operation, if forecast demand in interval t is dₜ customers and average service time is s hours, workload is dₜs labor-hours. If each employee supplies h productive hours in the interval, a minimum staffing estimate is:

The ceiling symbol means round upward to a whole employee. This estimate is only a baseline: variability, occupancy limits, shrinkage, skill mix and service-level targets may require additional staff.

A good workforce plan combines fixed staffing for predictable base demand with flexible capacity such as part-time shifts, cross-trained employees, appointment controls, overtime or temporary labor. Excessive reliance on just-in-time labor can transfer operational variability to workers through unstable hours and fatigue. Ethical scheduling includes predictable notice, lawful rest periods, safe workloads and nondiscriminatory allocation of undesirable shifts.

6. Line balancing

Line balancing assigns tasks to workstations so that precedence constraints are respected and idle time is minimized. The cycle time is the maximum available time per unit at each station. If available production time per period is T and required output is D, the required cycle time is:

where T and C use the same time unit. The theoretical minimum number of stations is:

where tᵢ is the task time. The line efficiency for N actual stations is:

These formulas assume a paced line, stable task times and a common cycle time. They do not by themselves solve precedence or ergonomic constraints.

Worked example. A line operates 420 minutes per day and must produce 84 units. Thus C = 420 ÷ 84 = 5 minutes per unit. Four tasks have times 2, 3, 4 and 5 minutes, with precedence A → B → C → D. Total work is 14 minutes, so N_min = ⌈14 ÷ 5⌉ = 3 stations. A feasible assignment is station 1: A (2); station 2: B (3); station 3: C (4) and D cannot be added because 4 + 5 > 5. Therefore this particular assignment requires four stations, for efficiency 14 ÷ (4 × 5) = 70%. The theoretical minimum is not always attainable because precedence and station-time limits constrain feasible groupings.

Balancing may involve splitting tasks, adding parallel stations, changing product design, using flexible workers or reducing task variation. A line with nominally high efficiency can still perform poorly if the bottleneck is unreliable or if material shortages interrupt stations.

7. Project scheduling and the critical path

Production launches, facility openings, software deployments and process-improvement projects are often managed as networks of activities. The critical path is the longest-duration path through the precedence network. It determines the shortest possible project completion time under the stated durations.

For activity i, the forward pass calculates earliest start and finish:

For activities with predecessors, ESᵢ is the maximum earliest finish of immediate predecessors. Beginning with the project completion time, the backward pass calculates latest finish and start:

For activities with successors, LFᵢ is the minimum latest start of immediate successors. Total float is:

Activities with zero total float form at least one critical path. If activity durations are uncertain, PERT may use optimistic a, most-likely m and pessimistic b estimates. The expected duration is:

This estimate assumes the beta-shaped approximation used by PERT and should not be treated as a guarantee. Correlated risks, resource constraints and rework can make simple precedence analysis optimistic.

Critical-path analysis identifies timing sensitivity, but a path can change when durations or resource assignments change. Resource-constrained schedules therefore require repeated review.

8. Finite-capacity scheduling

Infinite-capacity planning loads orders against a nominal calendar and assumes that resources can absorb the workload. Finite-capacity scheduling explicitly prevents a resource from being assigned more work than its available time. It considers machine calendars, labor skills, setup matrices, maintenance, material availability, alternative routings and priority rules.

A practical finite-capacity procedure is:

1. Establish the master demand and due dates.
2. Translate each order into operations, durations and precedence.
3. Create resource calendars, including breaks, maintenance and unavailable periods.
4. Load high-priority operations backward from due dates or forward from release dates.
5. Resolve overloads by changing sequence, using alternate resources, splitting lots, outsourcing, adding shifts or renegotiating dates.
6. Check material, labor and quality constraints together.
7. Freeze near-term schedules while retaining a controlled planning horizon for later periods.

Backward scheduling protects due dates by placing operations as late as possible; forward scheduling identifies the earliest completion date and is useful when materials or resources are available now. Hybrid approaches use a frozen short-term window and flexible medium-term planning.

Finite-capacity systems reduce the credibility gap between a plan and actual execution, but they do not eliminate uncertainty. Excessively detailed schedules can create false precision and frequent rescheduling. Schedule stability is itself a performance objective because constant changes create setup losses, confusion and poor supplier coordination.

9. Integrated performance management

Capacity decisions should use a balanced set of measures:

• throughput and contribution per bottleneck hour;
• utilization and efficiency;
• work-in-process and cycle time;
• on-time delivery and service level;
• queue length and abandonment;
• schedule adherence and change frequency;
• first-pass yield, rework and safety incidents.

A useful managerial principle is to maximize flow of valuable, conforming output—not local utilization. For example, increasing utilization of a non-bottleneck may increase work-in-process and delay the bottleneck. Similarly, a utilization target that encourages employees to skip maintenance can reduce long-run capacity.

Technology can support this system through enterprise resource planning, manufacturing execution systems, workforce-management platforms, digital twins and real-time dashboards. AI may forecast demand, detect emerging bottlenecks, recommend sequences or estimate processing times. Its recommendations require human oversight because training data may encode historical bias, forecasts may fail under structural change and automated priorities can disadvantage urgent customers or vulnerable workers. Data governance should define ownership, access, audit trails, model validation and override authority.

Illustrative case: scaling a meal-kit fulfilment operation

Background

FreshBox, an illustrative meal-kit venture, receives online orders for weekly delivery. Its process includes order consolidation, ingredient picking, portioning, packing and dispatch. Demand has grown from 420 to 600 boxes per day, but late deliveries and overtime are increasing.

Problem

The manager initially proposes purchasing another packing machine. However, observations show that picking completes 720 boxes per day, portioning 660, packing 540 and dispatch 700. Packing is therefore the apparent bottleneck. Actual demand of 600 exceeds packing capacity by 60 boxes per day.

Analysis

The system cannot sustainably deliver 600 boxes without changing packing capacity, subcontracting or reducing work content. Upstream picking at 720 boxes would create excessive work-in-process if released without control. Applying the bottleneck principle, FreshBox first measures changeover time, micro-stoppages, absenteeism and pack defects. It finds that product variety causes 75 minutes of daily changeover and that a poorly designed label creates rework.

Approach and solution

FreshBox groups orders into product families, reducing changeovers by 30 minutes. It redesigns labels and introduces a quality check at the start of each run. Effective packing time increases by 45 minutes per day, raising capacity from 540 to approximately 585 boxes. A cross-trained employee is scheduled for the evening peak, adding 60 minutes of packing capacity. The resulting capacity is approximately 645 boxes, providing a modest cushion above demand. A finite-capacity release rule limits picking to the packing schedule plus a controlled buffer.

Little’s Law is used to monitor flow. Before improvement, throughput was 540 boxes per day and average time in the internal system was 1.5 days, implying about 810 boxes of average work-in-process. After release control, average work-in-process falls to 600 boxes and throughput rises to 600 boxes per day. The implied average flow time becomes 600 ÷ 600 = 1 day, assuming stable daily averages and a consistent system boundary.

Outcome

The operation improves on-time delivery, reduces overtime and avoids immediate capital expenditure. The new staffing pattern is reviewed for fair notice and rest compliance. Management continues to monitor whether demand growth shifts the bottleneck to dispatch or ingredient availability.

Learning points

The case illustrates that capacity is a system property; the most visible asset is not necessarily the constraint. Exploiting existing capacity, controlling release and reducing variation can outperform indiscriminate expansion. It also demonstrates that throughput, work-in-process and cycle time must be analyzed together.

Challenges and limitations

Capacity calculations may be distorted by inaccurate standard times, hidden rework, absenteeism or demand substitution. Average values conceal peaks, and a schedule feasible on paper may fail when skills are not interchangeable. Queue formulas based on averages cannot predict every customer’s wait. Critical paths ignore resource contention unless resource constraints are added. Optimization models may produce mathematically efficient schedules that employees find unsafe or suppliers cannot support.

Managers should therefore combine quantitative analysis with observation, sensitivity analysis and stakeholder consultation. Scenarios should test demand surges, machine failure, supplier delay and labor shortages. Capacity decisions should be reversible where possible, especially for entrepreneurial firms facing uncertain growth.

Best-practice checklist

1. Define capacity in explicit units and time periods.
2. Separate design, effective and actual capacity.
3. Measure end-to-end flow, not only departmental utilization.
4. Identify and manage the current bottleneck.
5. Use Little’s Law with a stable, clearly defined system boundary.
6. Match sequencing rules to the actual performance objective.
7. Include skills, breaks, maintenance and legal constraints in schedules.
8. Use finite-capacity checks before promising dates.
9. Protect a realistic capacity cushion.
10. Review technology recommendations for bias, explainability, safety and accountability.

Chapter Summary

Capacity is the potential output of a resource or process under stated conditions. Design capacity is theoretical, effective capacity reflects realistic constraints and actual output records completed work. Utilization and efficiency compare output with different capacity baselines. The bottleneck constrains sustainable throughput, so improvement should begin by identifying, exploiting and protecting that constraint.

Queues arise from both overload and variability. Little’s Law, L = λW, connects average work-in-process, throughput and flow time when the system is stable and its boundaries are consistently defined. Loading assigns work to resources, while sequencing determines processing order. Workforce schedules must combine demand coverage with skills, legal requirements and employee well-being. Line balancing uses cycle time, precedence and station assignments to control idle capacity. Critical-path analysis identifies activities that determine project duration, while finite-capacity scheduling incorporates real calendars and resource limits. Effective operations management balances flow, cost, service, quality, resilience and ethical responsibility rather than maximizing isolated utilization.

Key Terms

Capacity: Maximum output rate under specified conditions.

Design capacity: Theoretical maximum output under ideal conditions.

Effective capacity: Realistic capacity after planned constraints.

Utilization: Actual output divided by design capacity.

Efficiency: Actual output divided by effective capacity.

Capacity cushion: Reserved capacity above expected demand.

Throughput: Rate of completed, conforming output.

Bottleneck: Resource that most restricts system flow.

Queue: Work or customers waiting for service.

Work-in-process: Items inside a process but not yet complete.

Cycle time: Time between successive completed units or required time per unit in a paced line.

Little’s Law: Relationship L = λW between inventory, throughput and flow time.

Loading: Assigning work requirements to resources.

Sequencing: Determining the order of processing jobs.

Critical ratio: Time remaining to due date divided by remaining processing time.

Line balancing: Assigning tasks to stations subject to precedence and cycle-time limits.

Critical path: Longest-duration path through a project network.

Float: Schedule flexibility available before delaying project completion.

Finite-capacity scheduling: Scheduling that does not overload available resources.

Schedule adherence: Degree to which actual execution follows the planned schedule.

Review Questions

Short Answer Questions

1. Distinguish design capacity, effective capacity and actual output.
2. Why can a non-bottleneck operate at high utilization without improving throughput?
3. State Little’s Law and define each variable.
4. What is the difference between loading and sequencing?
5. Why is a capacity cushion valuable in an uncertain market?

Long Answer Questions

1. Explain how variability and high utilization affect queues and cycle time.
2. Compare first-come, first-served, shortest-processing-time and earliest-due-date sequencing rules.
3. Describe the steps required to develop a finite-capacity production schedule.
4. Explain line balancing and discuss why the theoretical minimum number of stations may not be feasible.
5. Evaluate the managerial and ethical issues involved in AI-supported workforce scheduling.

Analytical / Application Questions

1. A service centre receives 240 customers during an eight-hour day. Average service time is 10 minutes, and each employee provides 6.5 productive hours. Estimate the minimum number of employees required and identify practical reasons why actual staffing may need to be higher.
2. A process has design capacity of 1,200 units per week, effective capacity of 1,000 units and actual output of 900 units. Calculate utilization and efficiency, then recommend two questions management should investigate.
3. A repair operation completes 30 jobs per week and has average work-in-process of 90 jobs. Use Little’s Law to calculate average flow time and explain one operational action that could reduce it without reducing throughput.

Multiple Choice Questions with Answers

Question 1

A process has design capacity of 1,000 units per shift, effective capacity of 800 units and actual output of 720 units. What is its efficiency?

A. 72%

B. 80%

C. 90%

D. 111%

Correct Answer: C. 90%

Question 2

Which resource is most likely to determine sustainable system throughput?

A. The resource with the highest utilization

B. The resource with the lowest relative capacity

C. The resource with the largest physical size

D. The resource with the most employees

Correct Answer: B. The resource with the lowest relative capacity

Question 3

A process completes 20 jobs per day and average flow time is 3 days. According to Little’s Law, average work-in-process is:

A. 6 jobs

B. 17 jobs

C. 23 jobs

D. 60 jobs

Correct Answer: D. 60 jobs

Question 4

Which sequencing rule generally prioritizes the job with the nearest due date?

A. Shortest processing time

B. Earliest due date

C. First-come, first-served

D. Random selection

Correct Answer: B. Earliest due date

Question 5

If available production time is 480 minutes and required output is 96 units, the required cycle time is:

A. 4 minutes

B. 5 minutes

C. 6 minutes

D. 10 minutes

Correct Answer: B. 5 minutes

Question 6

What does a zero total float usually indicate in critical-path analysis?

A. The activity has no processing time

B. The activity is optional

C. The activity lies on a critical path

D. The activity can be delayed indefinitely

Correct Answer: C. The activity lies on a critical path

Question 7

Which statement best describes finite-capacity scheduling?

A. It assumes unlimited resource availability

B. It ignores maintenance and breaks

C. It prevents assignments from exceeding available resource time

D. It schedules only customer demand, not operations

Correct Answer: C. It prevents assignments from exceeding available resource time

Question 8

A job has 6 hours until its due date and 8 hours of remaining processing time. Its critical ratio is:

A. 0.75

B. 1.33

C. 2

D. 14

Correct Answer: A. 0.75

Question 9

Which action is most consistent with exploiting a bottleneck?

A. Releasing unlimited work to upstream stations

B. Reducing avoidable downtime at the constrained resource

C. Adding inventory before every non-bottleneck

D. Maximizing idle time at the constraint

Correct Answer: B. Reducing avoidable downtime at the constrained resource

Question 10

Why may a line-balancing theoretical minimum not be attainable?

A. Capacity is never measurable

B. Precedence and station-time constraints restrict task combinations

C. All tasks must be assigned to separate companies

D. Cycle time is unrelated to output

Correct Answer: B. Precedence and station-time constraints restrict task combinations