Capacity Measurement and Performance Diagnosis

Introduction

Capacity planning begins with a deceptively simple question: how much output can an operation produce? The answer depends on the definition of capacity, the time horizon, the resource being measured, product mix, operating conditions, quality requirements, and the losses included or excluded. A machine may be able to process 100 units per hour under ideal conditions, while the plant produces only 72 saleable units per hour because of changeovers, maintenance, quality losses, absenteeism, and material shortages.

Capacity measurement therefore has two purposes. First, it establishes the supply capability available to meet demand. Second, it diagnoses why actual output differs from potential output. The second purpose is particularly important in line balancing: balancing work among stations is ineffective if the line's observed capacity is reduced by unstable processes, long breakdowns, poor yield, or excessive waiting.

This chapter develops a measurement architecture for design, effective, demonstrated, rated, and safety capacity. It then examines capacity cushions, yield, downtime, overall equipment effectiveness (OEE), flow efficiency, and capacity losses. The emphasis is not merely on computing ratios, but on interpreting them for managerial decisions.

Learning objectives

After studying this chapter, students should be able to:

1. Distinguish major capacity definitions and select an appropriate denominator.
2. Calculate design, effective, demonstrated, rated, and safety capacity.
3. Diagnose capacity differences using yield, downtime, utilization, OEE, and flow efficiency.
4. Construct a capacity-loss tree and identify the dominant constraint.
5. Interpret measures critically in the context of product mix, variability, and data quality.
6. Recommend operational and strategic responses to capacity shortfalls.

1. Capacity as a multidimensional concept

Capacity is the maximum sustainable rate of output that a resource, process, or system can deliver under specified conditions. The phrase “specified conditions” is essential. Capacity is not a single permanent property of a machine or facility. It changes with staffing, technology, product design, maintenance policy, quality standards, scheduling rules, and the time interval being considered.

Capacity may be expressed as units per hour, tonnes per day, customers per week, labor-hours per month, or revenue-producing transactions per minute. The unit must match the decision. A mixed-model assembly line should not normally be measured only in physical units if products consume very different amounts of labor or machine time. Equivalent units, standard hours, or product-family-specific rates may be more valid.

A useful measurement hierarchy is:

These measures should not be treated as interchangeable. Design capacity is not a promise of output; demonstrated capacity is not necessarily a physical limit. A low historical output may reflect poor management rather than inadequate equipment.

2. A capacity measurement architecture

2.1 Design capacity

Design capacity is the maximum output implied by the technical design under stated ideal assumptions. For a single resource operating at a constant rate:

where C_d is design capacity in units per period, R_d is the design processing rate in units per hour, and T_d is the available design time in hours per period.

For a system with multiple parallel identical resources:

where n is the number of resources. This formula assumes resources are interchangeable, supplied with adequate materials, and not restricted by downstream or upstream bottlenecks. For serial processes, system capacity is governed by the bottleneck, not by the sum of station capacities.

Design capacity is useful for initial investment analysis, but it often excludes changeovers, planned maintenance, breaks, learning effects, quality requirements, and demand variation. Presenting it as an operational forecast is a common analytical error.

2.2 Effective capacity

Effective capacity is the maximum output attainable under normal operating conditions after planned and structural restrictions are recognized. Such restrictions may include preventive maintenance, breaks, product changeovers, regulatory inspections, normal absenteeism, planned meetings, and the product mix.

A simple representation is:

where C_e is effective capacity, A_p is the fraction of design time available after planned losses, and M_p is a product-mix or process-complexity factor. Both factors range from 0 to 1. This expression is a planning model rather than a universal law; the factors must be defined consistently and should not double-count the same loss.

If planned available hours are known directly, use:

where R_n is the normal standard rate and T_p is planned productive time. Effective capacity is generally the most useful denominator for short- and medium-term operating plans.

2.3 Demonstrated capacity

Demonstrated capacity is the output actually achieved by a resource or process over a representative period, under specified conditions. It is often calculated as:

where Q_i is observed output in period i, and N is the number of comparable periods.

A representative period should include normal product mix, staffing, maintenance policy, and demand conditions. A single unusually good shift is not a reliable demonstration of capacity. Conversely, a period dominated by extraordinary shortages may understate the process's intrinsic capability. Demonstrated capacity is an empirical observation, not automatically a target or a limit.

2.4 Rated capacity

Rated capacity is the capacity obtained after applying an agreed rating to a reference capacity. In many organizations:

where C_r is rated capacity and r is a dimensionless rating, commonly reflecting historical performance or management policy.

For example, if effective capacity is 1,000 units per day and the approved operating rating is 0.92, rated capacity is 920 units per day. The rating must be governed: its source, period, product mix, confidence level, and treatment of abnormal events should be documented. Otherwise, ratings become subjective allowances that conceal losses.

2.5 Safety capacity and capacity cushion

Safety capacity is capacity deliberately held in reserve to absorb uncertainty, demand peaks, disruption, or recovery requirements. It may be defined against expected demand as:

where C_s is the absolute safety capacity, C_p is planned available capacity, and D_e is expected demand.

The capacity cushion is commonly expressed as a percentage of capacity:

Alternatively, managers may express reserve relative to demand:

These two percentages answer different questions and must not be confused. The first asks what share of capacity is unused; the second asks how much demand can increase before planned capacity is exhausted. A positive cushion improves resilience but may raise unit cost. A negative cushion indicates an expected shortfall and requires additional capacity, demand management, subcontracting, redesign, or prioritization.

Worked example 1: capacity hierarchy

A packaging cell has two identical machines. Each is designed to run at 50 units per hour for 16 hours per day. Planned maintenance and changeovers reduce available time by 12.5%, and product mix reduces the normal rate by 5%. The approved rating is 90%. Expected demand is 1,250 units per day.

Design capacity:

Effective capacity:

Rated capacity:

The rated capacity is 1,197 units per day, which is 53 units below expected demand. The shortfall is not evidence that the design is inadequate: design capacity exceeds demand. It indicates that planned restrictions and the operating rating leave insufficient reliable capacity. Management might reduce changeover time, improve reliability, add a shift, outsource a family of products, or revise the rating only if evidence supports that revision.

3. Measuring performance and capacity losses

3.1 Utilization and efficiency

Utilization compares actual output with a capacity reference. A common form is:

where Q_a is actual output. It is meaningful only when the numerator and denominator use compatible units and product mix.

Efficiency relative to effective capacity is:

A process may have low utilization but high effective-capacity efficiency if design capacity includes substantial planned reserve. Managers should report the denominator explicitly rather than saying simply that “efficiency is 85%.”

3.2 Yield and first-pass yield

Yield measures the proportion of input that becomes acceptable output. Basic yield is:

where Q_g is good output and Q_i is total input. If reworked units are included in good output, the measure is final yield; if only units passing without rework are counted, it is first-pass yield (FPY):

For a serial process with independent stage yields, rolled throughput yield is:

where Y_j is the first-pass yield of stage j and m is the number of stages. The independence assumption is a simplifying approximation; correlated defects can make the product misleading.

Worked example 2: yield and capacity

A line processes 1,000 units. At station 1, 96% pass first time; at station 2, 94% pass first time; at station 3, 98% pass first time.

Thus approximately 884 units pass all three stages without rework. Improving the weakest station from 94% to 97% would produce:

The three-percentage-point local improvement raises end-to-end first-pass yield by about 2.9 percentage points. This demonstrates why line-level diagnosis must examine cumulative losses rather than only the final station.

3.3 Downtime and availability

Downtime is time during which a resource is scheduled to operate but cannot produce at the required rate. It may be planned or unplanned, although OEE conventionally treats planned stops separately from availability.

For a defined measurement window:

In OEE terminology, planned production time is scheduled time after planned non-production is removed, and operating time is planned production time minus unplanned stops. Therefore:

The denominator must be stated because organizations frequently mix calendar time, scheduled time, and planned production time.

Common downtime categories include equipment failure, setup and adjustment, material shortage, quality hold, labor shortage, information-system failure, blocked downstream flow, and starved upstream flow. Coding downtime by cause is more valuable than merely reporting total downtime.

3.4 Overall equipment effectiveness

OEE decomposes planned production time into three dimensions:

where A is availability, P is performance, and Q is quality.

Performance is generally:

Quality is:

Thus, with consistent units:

OEE assumes a stable definition of ideal cycle time, count, planned production time, and good output. It is not a universal measure of factory productivity. It may be inappropriate for highly variable job shops, processes where ideal cycle time changes by product, or operations whose principal constraint is labor, material, or flow rather than equipment.

Worked example 3: OEE

A machine is scheduled for 480 minutes. Planned breaks total 30 minutes, so planned production time is 450 minutes. Unplanned downtime is 45 minutes, giving operating time of 405 minutes. The ideal cycle time is 0.5 minute per unit. Total output is 760 units, of which 730 are good.

The loss is not explained by downtime alone. Performance losses and quality losses also consume capacity. The equivalent good output at ideal speed during planned production time is:

Compared with 450 planned minutes, this is 81.1%, confirming the OEE result.

4. Flow efficiency and line-level diagnosis

Equipment measures do not fully describe a production system. A unit may spend only a few minutes being transformed while waiting hours for transport, inspection, batching, authorization, or downstream space. Flow efficiency exposes this distinction:

The denominator should be defined carefully. For an individual order, total elapsed lead time may include queueing and administrative time. For a process segment, touch time may include only time physically handled. The measure is therefore comparative rather than absolute unless the organization standardizes its definition.

For example, if a component receives 18 minutes of value-adding work during a 180-minute total lead time:

The implication is not that 90% of time is necessarily removable immediately. Some waiting is imposed by batch economics, safety, technical constraints, or customer requirements. However, it signals an opportunity to investigate queues, transfer batches, synchronization, and approval delays.

A line should be diagnosed at three levels:

1. Resource level: availability, speed, quality, and maintenance.
2. Station or line level: balance loss, starvation, blocking, buffers, and takt adherence.
3. Value-stream level: queues, batch sizes, release policies, transport, and customer demand.

Takt time is a demand rate, not a capacity measure:

If 420 net production minutes must satisfy 140 units of daily demand, takt time is 3 minutes per unit. A station cycle time above 3 minutes cannot meet demand without parallel capacity, overtime, inventory, or a change in demand allocation. Takt time should not be confused with the actual cycle time or the ideal machine cycle time.

5. Capacity losses and the loss tree

A capacity-loss tree organizes the difference between theoretical potential and good output. A generic sequence is:

Figure: A conceptual capacity-loss tree from ideal design capacity to saleable output.

The tree should be built using mutually exclusive categories. If setup time is deducted from planned time, it must not also be deducted as unplanned downtime. If rework is counted in output and later removed as quality loss, the definitions must be explicit. Poor category design creates false improvement.

A useful loss analysis converts time losses into output losses. If a line's standard rate is 40 units per hour, a 30-minute stoppage represents 20 units of theoretical opportunity, before considering recovery, startup scrap, and downstream effects. The opportunity cost is not necessarily actual lost sales: it depends on demand, inventory, and bottleneck position.

6. Integrated diagnosis and managerial interpretation

A disciplined diagnosis follows five steps.

First, define the object and period. Is the unit a machine, cell, line, plant, labor group, or value stream? Is the period a shift, week, month, or peak season?

Second, establish a common data model. Record scheduled time, planned stops, operating time, ideal or standard rate, total output, good output, product mix, and demand. Time stamps should reconcile: scheduled time should equal planned stops plus planned production time; planned production time should equal operating time plus unplanned downtime.

Third, segment the data. Analyze product family, shift, crew, machine, order type, and cause. Aggregate averages can hide a bottleneck that occurs only on a high-margin product.

Fourth, identify the binding constraint. A station with the lowest local OEE is not automatically the system constraint. The relevant constraint is the resource whose capacity limits throughput, considering routing, buffers, demand, and substitution possibilities.

Fifth, select a response matched to the loss. Reliability losses require maintenance and failure-mode analysis; speed losses require method, tooling, or process improvement; quality losses require root-cause control; flow losses require scheduling, batch, layout, or synchronization changes.

Capacity decisions under uncertainty

Expected demand is not the only input. Managers should examine demand scenarios, lead-time commitments, disruption probabilities, and the cost of surplus capacity. Safety capacity is a risk decision, not merely an arithmetic remainder. High utilization may reduce apparent unit cost while increasing queues and vulnerability. In variable systems, operating permanently near 100% of nominal capacity often produces unstable lead times.

A practical policy is to define a target operating range rather than a single utilization target. The range should reflect demand variability, replenishment lead time, equipment criticality, and the reversibility of capacity decisions. Permanent staff, flexible labor, overtime, subcontracting, cross-training, and modular equipment provide different forms of capacity flexibility.

7. Illustrative case: Northstar medical-device assembly

Background. Northstar assembles a regulated diagnostic cartridge on a two-shift, 16-hour schedule. The line has a designed rate of 1,200 cartridges per day. Demand is forecast at 1,000 good cartridges per day, but actual good output averages only 880.

Problem. Management proposes purchasing another assembly machine because utilization appears high. The operations manager requests a loss diagnosis before approving capital expenditure.

Analysis. Daily scheduled time is 960 machine-minutes across the line. Planned breaks and sanitation consume 96 minutes, leaving 864 planned production minutes. Unplanned downtime averages 86.4 minutes, so availability is:

The ideal cycle time is 0.72 minute per cartridge. Total output is 1,180 cartridges and good output is 1,100 before final demand allocation. Performance is:

The result above 100% reveals a data-definition problem: the stated ideal cycle time or operating-time record is inconsistent. Investigation finds that the line's “total output” includes cartridges produced during sanitation overlap, while the downtime log counts that time as unavailable. The team corrects the time boundary and obtains 1,080 total cartridges, 880 good cartridges, and performance of 99.9%. Quality is:

Corrected OEE is approximately:

Further analysis shows that 70% of defects arise from seal-temperature variation at one station, while 18% of downtime is caused by material replenishment. The station is also the line bottleneck during the high-volume product family.

Approach and solution. Northstar standardizes time definitions, installs automatic temperature feedback, changes replenishment to a two-bin system, and separates planned sanitation from production timestamps. A pilot reduces seal defects, improves first-pass yield, and lowers material-related stops. Only after these actions does the team reassess the need for capital equipment.

Outcome. The case illustrates that an apparent capacity shortage may be a quality and data-governance problem. The corrected diagnosis identifies more available good capacity than a machine purchase would have provided per dollar invested.

Learning points. OEE is only as credible as its definitions; a mathematically precise metric can still be operationally false. The system bottleneck, not the worst isolated ratio, should guide investment. In regulated operations, traceability and validated measurement definitions are part of capacity management.

8. Limitations, governance, and contemporary practice

Capacity metrics can create dysfunctional behavior. If supervisors are rewarded for utilization, they may produce excess inventory, avoid preventive maintenance, or run large batches. If OEE is treated as a universal ranking, teams may optimize equipment speed while increasing queues or defects. Measures should therefore be linked to throughput, delivery reliability, quality, safety, and total cost.

Data governance is especially important in regulated and safety-critical operations. Definitions of good output, downtime, planned stops, rework, and standard rate should be version-controlled. Changes to measurement rules should be documented and audited. Personal performance data should be handled fairly and should not be used without context to punish workers for system constraints.

Digital manufacturing systems, sensors, and machine-learning tools can improve diagnosis through automatic time stamps, anomaly detection, predictive maintenance, and loss classification. However, AI does not eliminate the need for process knowledge. Biased labels, missing downtime codes, sensor drift, and product-mix changes can produce misleading recommendations. Human review, explainable thresholds, cybersecurity, and access controls are essential. A predictive-maintenance model should be evaluated not only by forecast accuracy but also by avoided downtime, false alarms, maintenance cost, and safety consequences.

Best practice is to maintain a metric dictionary, reconcile time and quantity data, use standard loss codes, separate planned from unplanned losses, report both numerator and denominator, and review measures at resource, line, and value-stream levels. Improvement should target the economically binding constraint and be verified through before-and-after evidence.

Chapter Summary

Capacity is conditional capability, not a single permanent number. Design capacity describes ideal technical potential; effective capacity incorporates normal planned restrictions; demonstrated capacity records observed performance; rated capacity applies an agreed operating factor; and safety capacity protects against uncertainty. Capacity cushions can be expressed relative to capacity or demand, and the chosen denominator changes the interpretation.

Yield, downtime, availability, performance, quality, OEE, and flow efficiency expose different forms of loss. OEE is calculated as availability × performance × quality, but its validity depends on consistent definitions and is not universal across all operating environments. Flow efficiency reveals waiting and queueing that equipment measures may miss. A loss tree converts theoretical capacity into planned production, actual output, and good output, enabling targeted interventions.

Graduate-level capacity diagnosis requires more than calculating ratios. It requires defining system boundaries, validating data, segmenting product and process conditions, locating the true constraint, and matching interventions to the dominant loss. Ethical measurement, governance, and responsible use of digital and AI tools are necessary to ensure that capacity metrics support reliable and humane operations rather than dysfunctional local optimization.

Key Terms

Design capacity: Theoretical maximum output under ideal design conditions.

Effective capacity: Realistic maximum output after normal planned restrictions.

Demonstrated capacity: Output empirically achieved over a representative period.

Rated capacity: Reference capacity adjusted by an agreed operating rating.

Safety capacity: Deliberately protected capacity reserved for uncertainty or disruption.

Capacity cushion: Capacity held above expected demand, expressed absolutely or as a percentage.

Utilization: Actual output divided by a specified capacity reference.

Yield: Proportion of input converted into acceptable output.

First-pass yield: Proportion passing a process without rework.

Rolled throughput yield: Product of first-pass yields across sequential stages.

Downtime: Scheduled operating time during which production cannot proceed at the required rate.

Availability: Operating time divided by planned production time.

Performance: Actual speed relative to ideal speed during operating time.

Quality: Good output divided by total output.

OEE: Product of availability, performance, and quality.

Flow efficiency: Value-adding time divided by total elapsed or touch time, according to the stated definition.

Takt time: Available production time divided by required customer demand.

Capacity loss: Difference between potential capacity and output after a specified loss category.

Bottleneck: Resource or process constraint that limits system throughput.

Metric governance: Rules controlling definitions, data ownership, revision, and use of performance measures.

Review Questions

Short Answer Questions

1. Distinguish design capacity from effective capacity.
2. Why must the denominator be stated when reporting utilization or efficiency?
3. Define first-pass yield and explain how it differs from final yield.
4. What are the three components of OEE?
5. Why is a positive capacity cushion valuable in a variable operating environment?

Long Answer Questions

1. Explain the relationships among design, effective, demonstrated, rated, and safety capacity. Include appropriate managerial uses for each.
2. Discuss the assumptions and limitations of OEE as a performance measure.
3. Explain how a capacity-loss tree can support line-balancing and bottleneck decisions.
4. Compare equipment-level diagnosis with value-stream-level diagnosis.
5. Discuss how data governance and incentive design affect the validity and behavioral consequences of capacity metrics.

Analytical / Application Questions

1. A line has design capacity of 1,600 units per day. Planned losses are 12.5%, the product-mix factor is 95%, and the approved rating is 90%. Calculate effective and rated capacity, then compare rated capacity with demand of 1,250 units per day.
2. A process has availability of 0.88, performance of 0.95, and quality of 0.97. Calculate OEE and identify which component contributes the largest proportional loss from perfection.
3. A three-stage process has first-pass yields of 98%, 93%, and 97%. Calculate rolled throughput yield and recommend which information you would investigate before selecting an improvement project.

Multiple Choice Questions with Answers

Question 1

Which capacity measure represents the theoretical maximum under ideal design conditions?

A. Demonstrated capacity

B. Design capacity

C. Safety capacity

D. Rated capacity

Correct Answer: B. Design capacity

Question 2

A resource has a design capacity of 1,600 units per day. Planned losses are 12.5% and the product-mix factor is 95%. What is its effective capacity?

A. 1,197 units/day

B. 1,280 units/day

C. 1,330 units/day

D. 1,520 units/day

Correct Answer: C. 1,330 units/day

Question 3

If planned capacity is 1,000 units and expected demand is 900 units, what is the capacity cushion expressed as a percentage of planned capacity?

A. 9%

B. 10%

C. 11.1%

D. 90%

Correct Answer: B. 10%

Question 4

Which measure counts units that pass a process without rework?

A. Final yield

B. First-pass yield

C. Utilization

D. Availability

Correct Answer: B. First-pass yield

Question 5

A machine has availability 0.90, performance 0.95, and quality 0.98. What is its OEE?

A. 83.8%

B. 85.0%

C. 88.1%

D. 92.3%

Correct Answer: A. 83.8%

Question 6

Which OEE component is calculated as (Ideal cycle time × Total count) ÷ Operating time?

A. Availability

B. Quality

C. Performance

D. Yield

Correct Answer: C. Performance

Question 7

A process has 20 minutes of value-adding time and 200 minutes of total elapsed time. What is its flow efficiency?

A. 5%

B. 10%

C. 20%

D. 90%

Correct Answer: B. 10%

Question 8

For a serial system, which resource normally determines system capacity?

A. The resource with the highest nominal rate

B. The resource with the largest buffer

C. The bottleneck resource

D. The resource with the lowest labor cost

Correct Answer: C. The bottleneck resource

Question 9

A process has first-pass yields of 0.96 and 0.94 at two sequential stages. What is the rolled throughput yield, assuming the stages are modeled multiplicatively?

A. 0.9024

B. 0.9216

C. 0.9400

D. 0.9800

Correct Answer: A. 0.9024

Question 10

Why can an OEE value above 100% indicate a measurement problem?

A. Quality can never be measured in percentages

B. The ideal cycle time or time boundaries may be inconsistent

C. Availability is always excluded from OEE

D. OEE applies only to labor operations

Correct Answer: B. The ideal cycle time or time boundaries may be inconsistent