Quality Control and Process Improvement

Introduction

Quality control and process improvement are central to operations management because every organisation must convert inputs into outputs that consistently satisfy customer, regulatory and organisational requirements. Quality is not achieved only by inspecting finished goods. It is designed into processes, measured through evidence, controlled using statistical and managerial methods, and improved through disciplined learning.

Quality control focuses primarily on detecting and preventing unacceptable variation. Process improvement is broader: it seeks to improve capability, efficiency, reliability, customer value and sustainability. Together they support the principle that quality is the responsibility of the process owner, not merely the inspection department.

This chapter integrates cost of quality, the seven quality tools, statistical process control (SPC), process capability, acceptance sampling, root-cause analysis, PDCA, DMAIC, benchmarking and continuous-improvement methods.

Learning Objectives

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

1. Explain the relationship between quality assurance, quality control and process improvement.
2. Classify and calculate the main categories of quality cost.
3. Select and apply the seven basic quality tools.
4. Construct and interpret control charts.
5. Evaluate process capability using capability indices.
6. Explain acceptance sampling and its limitations.
7. Conduct root-cause analysis using structured methods.
8. Compare PDCA and DMAIC and select an appropriate improvement approach.
9. Use benchmarking and continuous-improvement systems in operational settings.
10. Evaluate ethical, technological and managerial issues in quality improvement.

1. Quality, Variation and Process Thinking

A process is a linked set of activities that transforms inputs into outputs. Inputs may include materials, information, equipment, labour, energy and methods. Outputs may be products, services, decisions or records. A process is stable when its variation is predictable over time; it is capable when its predictable variation can meet specified requirements.

Variation has two broad sources:

• Common-cause variation arises from the normal design of the process, such as machine precision, standard material differences or routine environmental fluctuations.
• Special-cause variation results from identifiable abnormal events, such as a broken tool, incorrect setting, contaminated material or inexperienced operator.

This distinction is important. Common causes usually require process redesign or systemic action. Special causes require investigation and removal of the particular abnormal condition. Tampering—making adjustments in response to random common-cause variation—can increase rather than reduce variation.

Quality assurance establishes confidence that quality requirements will be met through planned systems, standards and prevention. Quality control uses operational techniques and measurements to verify and regulate performance. Process improvement changes the process so that performance becomes better, more stable or more economical.

A useful operational sequence is:

Understand the customer requirement → define the process → measure performance → analyse variation → control the process → improve the process → standardise the gain.

2. Cost of Quality

The cost of quality (COQ) is the cost associated with achieving quality and the cost arising when quality is not achieved. It is not simply the cost of inspection. The traditional classification is as follows.

The total cost of quality may be represented as:

Here, C_P is prevention cost, C_A is appraisal cost, C_IF is internal failure cost and C_EF is external failure cost. The categories must be defined consistently; otherwise comparisons across periods can be misleading.

A prevention-oriented strategy often increases prevention expenditure while reducing failure expenditure. The objective is not to maximise or minimise one category in isolation, but to reduce the total cost while protecting customer value and compliance.

Worked Example: Cost of Quality

A factory reports annual prevention costs of $40,000, appraisal costs of $60,000, internal failure costs of $90,000 and external failure costs of $110,000.

Prevention and appraisal costs total $100,000, while failure costs total $200,000. Management should investigate whether additional prevention—such as mistake-proofing, supplier development or process redesign—can reduce the much larger failure burden. A higher prevention budget is justified only when its expected benefits exceed its incremental cost and risks.

COQ data should be interpreted carefully. Some benefits, such as improved reputation or reduced risk of injury, may not appear immediately in accounting records. Managers should therefore combine financial data with customer, process and safety indicators.

3. The Seven Basic Quality Tools

The seven basic tools provide accessible methods for collecting, displaying and analysing operational data.

3.1 Check Sheet

A check sheet is a structured form for recording observations consistently at the point of occurrence. It may record defect types, locations, times, batches or causes. Clear definitions are essential; otherwise two observers may classify the same event differently.

3.2 Histogram

A histogram groups numerical observations into intervals and displays their frequency. It reveals centre, spread, skewness, multiple populations and possible outliers. The number and width of intervals influence the visual interpretation, so the analyst should examine the data using reasonable alternatives.

3.3 Pareto Chart

A Pareto chart ranks categories from largest to smallest, usually with a cumulative-percentage line. It supports prioritisation, but the familiar “80/20 rule” is an empirical tendency, not a universal law. A small number of categories may account for most impact, but this must be demonstrated with data.

3.4 Cause-and-Effect Diagram

Also called a fishbone or Ishikawa diagram, this tool organises possible causes under headings such as methods, machines, materials, people, measurement and environment. It supports hypothesis generation; it does not prove causation. Suspected causes must be tested with evidence.

3.5 Scatter Diagram

A scatter diagram plots paired observations to examine association between two variables. An apparent relationship may be nonlinear, caused by a third variable or produced by sampling effects. Correlation does not establish causation.

3.6 Flowchart or Process Map

A flowchart displays process sequence, decisions, rework loops, delays and handoffs. More advanced maps may include suppliers, inputs, outputs, customers and measures. Process mapping frequently exposes unnecessary approvals, duplicated data entry and hidden queues.

3.7 Control Chart

A control chart plots a statistic in time order with a centre line and statistically derived control limits. It helps distinguish common-cause from special-cause variation. Control limits are not the same as customer specification limits.

4. Statistical Process Control

SPC applies statistical reasoning to process data over time. It is most effective when measurements are collected in rational subgroups: observations selected so that within-subgroup variation represents short-term process variation, while differences between subgroups can reveal time-related changes.

For a variable measured on a continuous scale, an X̄ chart monitors the subgroup mean and an R chart monitors subgroup range. For subgroup size n, subgroup means X̄_i and ranges R_i, the averages are:

where k is the number of subgroups. The standard control limits are:

and:

The constants A₂, D₃ and D₄ depend on subgroup size and are obtained from standard SPC tables. These formulas assume approximately independent observations, rational subgroups and a process for which the range-based estimate is suitable. A negative lower limit for an R chart is set to zero because a range cannot be negative.

For individual measurements, an individuals–moving-range chart may be used, but its assumptions and constants differ. Attribute charts are used when data are counts or classifications. A p chart monitors the fraction defective; an np chart monitors the number defective when sample size is constant; a c chart monitors defect counts with constant opportunity; and a u chart monitors defects per unit when opportunity or sample size varies.

A point outside a control limit is evidence of a possible special cause, not automatic proof of one. Other signals include sustained runs on one side of the centre line, trends, cycles and unusual alternation. Rules should be defined before analysis to limit false alarms.

Worked SPC Example

Suppose an X̄ chart has X̄̄ = 50.2, R̄ = 1.5, and the table constant for the subgroup size is A₂ = 0.577. Then:

The limits are approximately 51.07 and 49.33 units. They describe expected process behaviour under the conditions used to estimate them; they are not customer tolerances.

5. Process Capability

Capability analysis compares the distribution of a stable process with specification requirements. It should not be performed on an unstable process without qualification, because changing special causes make estimates of future performance unreliable.

For a process with upper specification limit USL, lower specification limit LSL and within-process standard deviation σ, the potential capability index is:

This assumes a meaningful estimate of within-process variation and compares specification width with a six-standard-deviation spread. It does not account for process centring.

The centred capability index is:

where μ is the process mean. Generally, C_pk ≤ C_p because off-centre performance reduces effective capability. Industry requirements vary; values such as 1.33 are often used as practical targets, but no universal threshold applies to every risk or sector.

Worked Capability Example

A shaft specification is 20.00 ± 0.10 mm. Thus USL = 20.10 and LSL = 19.90. The stable process has μ = 20.04 mm and σ = 0.025 mm.

Although potential capability is 1.33, actual centred capability is only 0.80 because the process mean is close to the upper limit. Recentring is therefore a priority.

Capability indices do not replace engineering judgement. Non-normal data, autocorrelation, measurement error, changing conditions and asymmetric losses may require alternative methods. Measurement-system analysis should be completed before capability conclusions are trusted.

6. Acceptance Sampling

Acceptance sampling is a decision procedure in which a sample from a lot is inspected and the lot is accepted or rejected according to a sampling plan. It is useful when 100% inspection is destructive, costly, slow or less reliable than sampling. It is not a method for controlling the process that produced the lot.

A single-sampling plan is commonly written as (n,c), where n is sample size and c is the maximum number of nonconforming units permitted for acceptance. If the observed number exceeds c, the lot is rejected.

Important risks are:

• Producer’s risk (α): rejecting a lot that is actually of acceptable quality.
• Consumer’s risk (β): accepting a lot that is actually of unacceptable quality.
• AQL: a quality level regarded as acceptable for the producer’s process, within a specified sampling system.
• LTPD or RQL: a limiting or rejectable quality level associated with consumer protection.

The operating characteristic (OC) curve shows the probability of accepting lots at different quality levels. For a binomial approximation, if the defect probability is p, the probability of observing exactly x defectives in a sample of size n is:

This approximation assumes independent selections and a sufficiently large lot relative to sample size. For sampling without replacement from a finite lot, the hypergeometric distribution is more appropriate. Sampling plans must be selected using recognised standards and the organisation’s risk requirements, not improvised from a desired acceptance percentage.

7. Root-Cause Analysis

Root-cause analysis (RCA) is a structured investigation intended to identify causes whose removal prevents recurrence. A symptom is the observed problem; a proximate cause is an immediate contributor; a systemic root cause is a condition embedded in the management or process system.

A disciplined RCA process is:

1. Define the problem in measurable terms: what, where, when and extent.
2. Contain the immediate risk without confusing containment with permanent correction.
3. Collect verified evidence and stratify it by product, shift, machine, supplier, location or time.
4. Generate possible causes using a cause-and-effect diagram or process analysis.
5. Test causes against data, experiments, records and direct observation.
6. Select corrective actions that remove or control verified causes.
7. Verify effectiveness and update procedures, training, controls and risk assessments.

The 5 Whys method asks why repeatedly, but it should not be used mechanically. Complex failures may have several interacting causes. Fault-tree analysis, failure mode and effects analysis (FMEA), designed experiments and regression analysis may be needed for high-risk problems.

FMEA prioritises potential failure modes using severity, occurrence and detectability ratings. Traditional risk priority numbers multiply these ratings, but the resulting number can hide important differences and should not replace expert review or action thresholds.

8. PDCA and DMAIC

PDCA—Plan, Do, Check, Act—is a general learning cycle. In Plan, the team defines the problem, objective, measures and method. Do involves controlled implementation. Check compares results with expectations. Act standardises successful changes or begins another cycle when learning is incomplete.

DMAIC—Define, Measure, Analyse, Improve, Control—is a more structured approach commonly associated with Six Sigma:

PDCA is flexible and suitable for everyday improvement. DMAIC provides stronger structure for chronic, data-intensive and cross-functional problems. Neither method guarantees success: poor problem definition, unreliable measurement or weak leadership can undermine either approach.

Lean methods complement both frameworks. Waste categories include defects, overproduction, waiting, unused talent, transportation, inventory, motion and overprocessing. Kaizen promotes frequent incremental improvement; standard work establishes the current best-known method; 5S organises and maintains the workplace; mistake-proofing (poka-yoke) makes errors difficult or impossible; and value-stream mapping analyses end-to-end flow.

9. Benchmarking and Continuous Improvement

Benchmarking is the systematic comparison of performance, practices or processes with a reference organisation, internal unit or recognised standard. Types include internal, competitive, functional and generic benchmarking. Meaningful comparison requires common definitions, comparable operating conditions and ethical access to information.

A benchmarking programme should:

1. Select a strategically important process and metric.
2. Define the measure, denominator, scope and data period.
3. Identify a credible benchmark partner or published reference.
4. Compare performance and investigate the reasons for differences.
5. Adapt relevant practices rather than copying them blindly.
6. Implement, measure and review the change.

Continuous improvement becomes sustainable when leaders provide resources, employees can report problems without fear, process owners are accountable, and improvement is integrated into daily management. Useful measures include first-pass yield, defect rate, complaints, process lead time, on-time delivery, rework, cost of poor quality and employee suggestions. Measures must be balanced: improving speed while increasing errors is not genuine improvement.

10. Applied Case: Improving Order Fulfilment

Illustrative case. A regional medical-equipment distributor experiences late shipments and incorrect order contents. Customer complaints have increased, and management initially proposes final inspection of every order.

Background and problem

The team defines the problem as: “During the next eight weeks, at least 8% of orders are shipped late and 3% contain an item or quantity error.” A process map shows order entry, credit approval, picking, verification, packing and carrier collection.

Analysis

A check sheet stratified by shift and order type shows that most errors occur on complex multi-line orders during the afternoon shift. A Pareto chart indicates that incorrect item selection and duplicate manual data entry account for most errors. Interviews and direct observation reveal that product codes are entered manually from printed documents. A cause-and-effect analysis identifies unclear labels, similar product packaging, workload peaks and inconsistent training as contributing factors.

The team verifies that the barcode scanner frequently fails at one workstation, but scanner failure alone does not explain all errors. Data show that orders using verified barcode scanning have a substantially lower error rate. The measurement system is checked by comparing order records with physical contents and by conducting repeated observations using a common definition of “error.”

Solution

Using DMAIC, the team pilots barcode scanning, colour-coded location labels and a standard work sequence. A workload board redistributes complex orders before the afternoon peak. The scanner is repaired, and a poka-yoke prevents completion when a required item is not scanned. A control plan tracks daily error rate and late-order percentage; unusual results trigger supervisor review rather than automatic blame.

Outcome

After four weeks, the error rate falls from 3% to 0.7%, and late shipments fall from 8% to 3%. The improvement is not attributed solely to inspection; prevention, better information, equipment reliability and work redesign all contributed. The new process is audited weekly for one month, then incorporated into standard work and training.

Learning points

The case demonstrates the value of defining a measurable problem, mapping the complete process, using stratified data, testing causes, piloting changes and controlling the improved process. It also illustrates that employee involvement and reliable measurement are necessary for sustainable improvement.

11. Governance, Ethics, Technology and Contemporary Practice

Quality decisions affect safety, fairness, employment and the environment. Organisations must not conceal defects, manipulate sampling, suppress complaints or redefine measures merely to report improvement. Traceability, document control, competence records and audit evidence support accountability. Regulated sectors may require compliance with sector-specific standards, calibration rules, validation requirements and product-release controls.

Digital sensors, manufacturing execution systems, automated inspection and machine learning can improve detection and prediction. However, technology does not remove the need for process understanding. AI models may reflect biased or incomplete data, generate false alarms, drift as processes change and obscure the reasoning behind decisions. Governance should address data integrity, cybersecurity, access control, model validation, human review and retention of audit trails. Automated inspection should be validated against known reference conditions and periodically rechecked.

Contemporary quality management increasingly combines real-time data, risk-based thinking, sustainability and resilience. Organisations monitor energy, waste, carbon, ethical sourcing and supply disruption alongside traditional defect measures. The central principle remains unchanged: data should support learning and responsible action, not merely surveillance or target manipulation.

Best-Practice Checklist

• Define customer and regulatory requirements precisely.
• Stabilise the process before calculating capability.
• Validate the measurement system and operational definitions.
• Use time-ordered data and rational subgroups for SPC.
• Distinguish control limits from specification limits.
• Investigate special causes without tampering with stable processes.
• Use prevention and mistake-proofing before relying on inspection.
• Test suspected root causes rather than accepting opinions.
• Select sampling plans according to producer and consumer risks.
• Standardise successful changes and monitor for deterioration.
• Balance quality, cost, speed, safety, sustainability and employee impact.

Chapter Summary

Quality control detects and controls variation, while process improvement changes processes to produce better and more reliable outcomes. Cost of quality classifies prevention, appraisal, internal failure and external failure costs, helping managers evaluate prevention-oriented investments. The seven basic tools provide practical methods for collecting, prioritising, mapping and analysing quality information.

SPC distinguishes predictable common-cause variation from special-cause variation through time-ordered control charts. Capability indices compare stable process performance with specifications, with C_p measuring potential spread and C_pk also considering centring. Acceptance sampling makes lot decisions while managing producer and consumer risks, but it does not replace process control. Root-cause analysis requires evidence, verification and systemic corrective action.

PDCA supports iterative learning, whereas DMAIC offers a structured data-driven route for complex improvement. Benchmarking, Lean methods, Kaizen, standard work and mistake-proofing help embed continuous improvement. Ethical governance, measurement integrity, employee involvement and responsible use of technology are essential to sustainable quality performance.

Key Terms

Acceptance sampling: Decision-making about a lot using evidence from a sample.

Appraisal cost: Cost of measuring, inspecting or auditing conformity.

Benchmarking: Systematic comparison with a reference organisation, unit or standard.

Capability: The ability of a stable process to meet specifications.

Common-cause variation: Normal variation produced by the process system.

Control chart: Time-ordered chart with a centre line and statistically derived limits.

Cost of quality: Prevention, appraisal and failure costs associated with quality.

DMAIC: Define, Measure, Analyse, Improve and Control improvement framework.

External failure: Cost arising from defects discovered after delivery.

FMEA: Structured analysis of potential failure modes and their effects.

Internal failure: Cost arising from defects found before delivery.

Kaizen: Continuous, usually incremental, improvement involving people at work.

Pareto chart: Ranked display used to prioritise categories.

PDCA: Plan, Do, Check and Act learning cycle.

Process capability: Comparison of stable process variation with specification limits.

Poka-yoke: Error-proofing that prevents or immediately detects mistakes.

Root cause: Underlying condition whose removal prevents recurrence.

Special-cause variation: Unusual, identifiable variation not inherent in normal operation.

Statistical process control: Use of statistical methods to monitor and improve processes.

Specification limit: Customer, engineering or regulatory boundary for acceptable output.

Review Questions

Short Answer Questions

1. Distinguish quality assurance, quality control and process improvement.
2. What is the difference between common-cause and special-cause variation?
3. Name the four categories of cost of quality.
4. Why must capability analysis generally be based on a stable process?
5. State the five phases of DMAIC.

Long Answer Questions

1. Explain the seven basic quality tools and describe an appropriate use for each.
2. Explain how an X̄ and R chart can be constructed and interpreted.
3. Compare C_p and C_pk, including their assumptions and managerial implications.
4. Discuss the advantages and limitations of acceptance sampling compared with process control.
5. Compare PDCA, DMAIC and Lean continuous-improvement methods.

Analytical / Application Questions

1. A stable process has specification limits of 100 and 110, a mean of 108 and standard deviation of 1. Calculate C_p and C_pk, then recommend an improvement priority.
2. A service centre has rising complaints. Design a root-cause investigation using at least four of the seven quality tools and explain how you would verify causes.
3. An organisation proposes 100% final inspection instead of process redesign. Evaluate this proposal using cost of quality, SPC, risk and ethical considerations.

Multiple Choice Questions with Answers

Question 1

Which cost category includes warranty claims and product recalls?

A. Prevention cost

B. Appraisal cost

C. Internal failure cost

D. External failure cost

Correct Answer: D. External failure cost

Question 2

What is the principal purpose of a Pareto chart?

A. To establish specification limits

B. To prioritise categories by relative contribution

C. To prove that one variable causes another

D. To calculate process capability

Correct Answer: B. To prioritise categories by relative contribution

Question 3

A point outside a control limit should first be treated as evidence of what?

A. A guaranteed customer defect

B. A possible special cause

C. A specification change

D. A measurement-system failure in every case

Correct Answer: B. A possible special cause

Question 4

For a process with USL = 20.10, LSL = 19.90 and σ = 0.025, what is Cp?

A. 0.80

B. 1.00

C. 1.33

D. 2.67

Correct Answer: C. 1.33

Question 5

Which statement about Cpk is correct?

A. It ignores process centring

B. It accounts for the nearer specification limit

C. It is used only for attribute data

D. It is always greater than Cp

Correct Answer: B. It accounts for the nearer specification limit

Question 6

In acceptance sampling, consumer’s risk is the risk of:

A. Accepting an unacceptable lot

B. Rejecting every sampled lot

C. Accepting only defect-free items

D. Increasing prevention cost

Correct Answer: A. Accepting an unacceptable lot

Question 7

Which DMAIC phase establishes the current performance baseline?

A. Define

B. Measure

C. Improve

D. Control

Correct Answer: B. Measure

Question 8

What is the main weakness of using the 5 Whys mechanically?

A. It cannot be used with process maps

B. It always requires numerical data

C. Complex failures may have multiple interacting causes

D. It is limited to financial problems

Correct Answer: C. Complex failures may have multiple interacting causes

Question 9

Which method is specifically designed to make an error difficult or impossible to make?

A. Benchmarking

B. Poka-yoke

C. Acceptance sampling

D. Histogramming

Correct Answer: B. Poka-yoke

Question 10

Why should specification limits not be confused with control limits?

A. Specification limits describe customer or engineering requirements, while control limits describe process behaviour

B. Specification limits are always narrower than control limits

C. Control limits are determined only by customers

D. They are two names for the same statistical boundary

Correct Answer: A. Specification limits describe customer or engineering requirements, while control limits describe process behaviour