Control Charts – X bar, R, p, np, c and u charts

Learning Objectives

By the end of this lesson, the learner will be able to:

Explain the purpose and principles of Statistical Process Control (SPC).
Distinguish between common-cause and special-cause variation.
Explain the components of a control chart.
Differentiate control limits from specification limits.
Select an appropriate control chart for variable and attribute data.
Apply Xbar-R, Xbar-S, I-MR, p, np, c, and u charts appropriately.
Recognize signals of process instability and special causes.
Interpret control-chart patterns and investigate assignable causes.
Understand rational subgrouping and its importance in SPC.
Use control charts to monitor process stability before assessing capability.
Interpret SPC results using appropriate software such as Minitab.
Explain how SPC supports process control and continuous improvement.

1. Introduction

Statistical Process Control (SPC) is a systematic approach for monitoring and controlling process performance using statistical methods.

Every process contains variation. The objective of SPC is not to eliminate all variation, because some variation is inherent in the process. Instead, SPC helps the organization:

Understand the normal level of process variation.
Detect unusual changes in process behavior.
Identify potential special causes.
Take appropriate corrective action.
Prevent recurrence of problems.
Maintain stable and predictable process performance.

A control chart is one of the principal tools used in SPC.

A control chart displays process data over time together with statistically calculated control limits. This allows the Green Belt to distinguish between routine process variation and unusual variation requiring investigation.

The fundamental SPC question is:

Is the process stable and predictable over time?

This question should generally be answered before asking whether the process is capable of meeting customer specifications.

2. Concept / Theory

2.1 Understanding Process Variation

Variation exists in virtually every process.

For example, consider a manufacturing process producing shafts with a nominal diameter of 25.00 mm.

Measurements might be:

24.98 mm
25.01 mm
25.02 mm
24.99 mm
25.00 mm

These differences may occur even when the process is operating normally.

Variation can generally be classified into two major categories:

Common-Cause Variation

Common causes are sources of variation that are inherent in the normal operation of a process.

Examples include:

Normal machine-to-machine differences
Routine environmental variation
Normal material variation
Standard operating-method differences
Ordinary measurement variation

A process affected primarily by common causes is generally considered stable or statistically controlled, provided the variation remains consistent over time.

Special-Cause Variation

Special causes are unusual sources of variation that are not part of the normal process behavior.

Examples include:

Machine breakdown
Incorrect machine setting
Tool failure
Incorrect material
New operator using an incorrect method
Sudden temperature change
Data-entry error
Equipment malfunction

Special causes can produce unusual observations or patterns on a control chart.

2.2 The Purpose of a Control Chart

A control chart helps answer:

Is the observed variation consistent with the historical behavior of the process?

A typical control chart contains:

Individual data points
A center line
Upper Control Limit (UCL)
Lower Control Limit (LCL)
A time or sequence axis

The center line generally represents the process’s estimated central tendency.

The control limits represent the expected range of process variation when the process is operating under the assumed statistical conditions.

2.3 Center Line

The center line represents the central level of the process.

Depending on the chart, it may represent:

Process mean
Average range
Average proportion
Average number of defects
Average count per unit

The center line is therefore dependent on the type of control chart being used.

2.4 Upper and Lower Control Limits

Control limits are statistically calculated boundaries used to assess process stability.

They are commonly associated with approximately three standard deviations of expected process variation from the center line, although the exact calculation depends on the chart type.

The limits are:

UCL — Upper Control Limit

LCL — Lower Control Limit

A point outside a control limit is an important signal that the process may have experienced a special cause.

However:

A point outside a control limit is a signal for investigation, not automatic proof of the root cause.

2.5 Control Limits vs Specification Limits

This distinction is extremely important.

Control Limits

Control limits are calculated from process data.

They describe the statistical behavior of the process.

Specification Limits

Specification limits are established by:

Customer requirements
Engineering requirements
Regulatory requirements
Product design
Contractual requirements

They describe what the output is required to achieve.

For example:

Specification:

LSL = 49.80 mm
USL = 50.20 mm

These limits are not calculated from the control chart.

A process can therefore be:

Stable but incapable
Stable and capable
Unstable but apparently within specifications
Unstable and incapable

This is why control and capability are different concepts.

2.6 Rational Subgrouping

Rational subgrouping is a fundamental SPC concept.

A rational subgroup should contain observations collected under conditions that are sufficiently similar to represent short-term process variation.

For example, suppose five consecutive parts are measured every hour.

Those five measurements may form a subgroup.

The objective is to compare variation within a subgroup with variation between subgroups.

Good subgrouping can help reveal changes in the process that might otherwise be hidden.

Poor subgrouping can make a control chart misleading.

Example

If five consecutive products are manufactured under essentially the same conditions, they may form a rational subgroup.

If five measurements are collected from five different machines operating under very different conditions, treating them as one subgroup may not be appropriate.

3. Key Topics

3.1 Variable Control Charts

Variable data are measured on a continuous numerical scale.

Examples include:

Length
Diameter
Weight
Temperature
Cycle time
Pressure
Thickness

Common variable control charts include:

Xbar-R Chart

The Xbar-R chart is commonly used when measurements are collected in rational subgroups, particularly with relatively small subgroup sizes.

It contains:

Xbar chart — monitors subgroup means.
R chart — monitors subgroup ranges.

The Xbar chart primarily indicates changes in process location.

The R chart monitors within-subgroup variation.

Xbar-S Chart

The Xbar-S chart uses:

Xbar — subgroup mean
S — subgroup standard deviation

It is particularly useful for larger subgroup sizes.

I-MR Chart

The Individuals-Moving Range (I-MR) chart is useful when observations are collected individually rather than in rational subgroups.

Examples:

One transaction at a time
One laboratory result at a time
One daily production measurement
One cycle-time observation per event

The chart contains:

Individuals (I) chart
Moving Range (MR) chart

3.2 Attribute Control Charts

Attribute data are generally based on classifications or counts.

Examples include:

Defective / non-defective
Pass / fail
Number of defects
Number of complaints

Common attribute charts include:

p Chart

Used to monitor the proportion or percentage of defective units.

Example:

A production line inspects 500 units per shift and records the proportion that is defective.

A p chart can be used to monitor the defective proportion.

The subgroup size may vary.

np Chart

Used to monitor the number of defective units when the sample size remains constant.

Example:

Every hour exactly 200 products are inspected and the number of defective products is recorded.

An np chart may be appropriate.

c Chart

Used to monitor the number of defects when the area of opportunity or inspection unit remains constant.

Example:

Number of scratches on each identical panel when each panel has the same inspection area.

u Chart

Used to monitor the number of defects per unit when the number of opportunities or unit size may vary.

Example:

Customer complaints per 1,000 transactions when the number of transactions varies from period to period.

3.3 Control Chart Selection

A practical selection guide is:

Data / Situation	Typical Chart
Individual continuous measurements	I-MR
Continuous data in small rational subgroups	Xbar-R
Continuous data in larger subgroups	Xbar-S
Proportion defective	p
Number defective, constant sample size	np
Number of defects, constant opportunity	c
Defects per unit, varying opportunity	u

The correct chart depends on the data structure, not simply on personal preference.

3.4 Signals of Special Cause

A process may be considered potentially unstable when the control chart shows unusual behavior.

Important signals include:

Point Beyond a Control Limit

One observation falls outside the UCL or LCL.

This should trigger investigation.

Long Run on One Side of the Center Line

Several consecutive observations fall on the same side of the center line.

This may indicate a sustained shift in process location.

Trend

A sequence of observations consistently moves upward or downward.

This may indicate:

Tool wear
Temperature drift
Material deterioration
Gradual equipment change

Cyclic Pattern

The process repeatedly rises and falls in a recognizable pattern.

Potential causes may include:

Shift changes
Daily cycles
Weekly cycles
Environmental effects
Maintenance schedules

Sudden Shift

The process suddenly moves to a different operating level.

Possible causes include:

Machine adjustment
New material
New operator
Process setting change

Formal rules such as Western Electric or Nelson rules can be used as supplementary detection criteria. Organizations should define which rules are part of their SPC standard rather than applying every possible rule indiscriminately.

3.5 Stability Before Capability

One of the most important principles in process analysis is:

Establish process stability before relying on capability analysis.

If a process is unstable, its mean and variation may change over time.

A capability index calculated from unstable data may therefore provide a misleading description of future process performance.

A typical sequence is:

Collect Data → Establish Stability → Assess Capability → Improve → Continue Monitoring

4. Tools / Methodology

4.1 General SPC Methodology

A Green Belt can follow these steps:

Step 1 — Define the CTQ or Process Measure

Determine what characteristic needs to be monitored.

Examples:

Cycle time
Defect percentage
Diameter
Customer complaints
Processing errors

Step 2 — Define the Data Collection Method

Specify:

What will be measured?
Who will measure it?
When?
Where?
How often?
What sample size?
What subgroup structure?

Step 3 — Verify the Measurement System

Before relying on the data, ensure that the measurement process is appropriate.

For continuous measurements, measurement-system analysis may be required.

For attribute classifications, Attribute Agreement Analysis or Kappa analysis may be appropriate depending on the objective.

Step 4 — Select the Control Chart

Select based on:

Variable or attribute data
Individual or subgroup observations
Constant or varying sample size
Defects versus defective units

Step 5 — Establish the Control Chart

Calculate:

Center line
UCL
LCL

Plot the observations in their correct chronological sequence.

Step 6 — Interpret the Chart

Look for:

Points beyond limits
Runs
Trends
Cycles
Shifts
Other predefined signals

Step 7 — Investigate Special Causes

When a signal occurs:

Identify when it occurred.
Review process conditions.
Check equipment.
Check materials.
Check operators and methods.
Review environmental conditions.
Examine maintenance and change records.
Identify the most plausible assignable cause.
Take appropriate action.

Step 8 — Standardize and Monitor

Once the process is stable:

Document the standard.
Continue monitoring.
Update control plans where appropriate.
React promptly to future special causes.

4.2 Minitab Workflow

Typical Minitab pathways include:

Xbar-R Chart

Stat → Control Charts → Variables Charts for Subgroups → Xbar-R

Xbar-S Chart

Stat → Control Charts → Variables Charts for Subgroups → Xbar-S

I-MR Chart

Stat → Control Charts → Variables Charts for Individuals → I-MR

p Chart

Stat → Control Charts → Attribute Charts → P

np Chart

Stat → Control Charts → Attribute Charts → NP

c Chart

Stat → Control Charts → Attribute Charts → C

u Chart

Stat → Control Charts → Attribute Charts → U

The exact menu wording can vary between Minitab versions.

The Green Belt should understand the statistical logic behind the chart rather than relying only on software output.

5. Worked Example / Case Study

Case Study: Manufacturing Shaft Diameter

A company manufactures shafts with a target diameter of 25.00 mm.

Five consecutive shafts are measured every hour.

The Green Belt wants to determine whether the process is statistically stable.

Step 1 — Identify the Data

Diameter is continuous variable data.

Five consecutive observations are collected at each sampling period.

Therefore, an Xbar-R chart is a potential choice.

Step 2 — Establish Rational Subgroups

Each set of five consecutive shafts is treated as one subgroup because the measurements are collected close together under similar operating conditions.

Step 3 — Calculate Subgroup Statistics

For each subgroup:

Calculate the subgroup mean.
Calculate the subgroup range.

The Xbar chart monitors the subgroup means.

The R chart monitors within-subgroup variation.

Step 4 — Interpret the R Chart First

The R chart should be examined for evidence of unusual within-subgroup variation.

Suppose all subgroup ranges remain within their control limits and show no unusual pattern.

This indicates that within-subgroup variation appears stable.

Step 5 — Interpret the Xbar Chart

Suppose most subgroup means remain around 25.00 mm, but one subgroup mean falls well below the lower control limit.

This is a potential special-cause signal.

Step 6 — Investigate

The Green Belt reviews the production records and finds that a cutting tool was replaced immediately before the affected subgroup.

The tool replacement becomes a potential assignable cause.

The team should verify the relationship rather than simply assuming that the replacement caused the shift.

Step 7 — Correct and Standardize

If the investigation confirms that the tool change affected the process, the organization can:

Review the tool-change procedure.
Establish appropriate setup verification.
Train operators.
Add a first-piece verification requirement.
Update the control plan if necessary.

Key Learning

The control chart did not itself identify the root cause.

It identified when the process behavior became unusual.

The Green Belt then used process knowledge and investigation to identify and address the potential cause.

6. Practical Application

Exercise 1 — Selecting a Control Chart

For each situation, select an appropriate chart.

A.

One temperature measurement is recorded every hour.

Suggested chart: I-MR

B.

Five consecutive component measurements are collected every hour.

Suggested chart: Xbar-R

C.

The number of defective units is recorded from samples of exactly 100 units.

Suggested chart: np

D.

The percentage of defective units is recorded, with sample sizes changing each day.

Suggested chart: p

E.

The number of defects on identical products with the same inspection opportunity is recorded.

Suggested chart: c

F.

The number of defects per unit is recorded when the number of units inspected changes.

Suggested chart: u

Exercise 2 — Control Limit vs Specification Limit

A component has:

LSL = 10.0 mm
USL = 20.0 mm

The control chart has:

LCL = 11.5 mm
UCL = 18.5 mm

Ask the learner:

Which limits describe customer requirements?
Which limits describe process behavior?
Can a process be statistically stable while producing output outside specifications?
Why should control and capability be assessed separately?

Expected Learning

Specification limits describe requirements.

Control limits describe statistical process behavior.

A process may be stable but still incapable of meeting specifications.

Exercise 3 — Special-Cause Investigation

A control chart shows a sudden upward shift in process cycle time.

Ask the Green Belt to investigate:

What changed?
Was there a staffing change?
Was new equipment introduced?
Was the process method changed?
Was there a material change?
Did the measurement method change?
Was maintenance performed?
Did the customer or transaction mix change?

The objective is to connect the statistical signal with actual process events.

Exercise 4 — SPC Implementation

Select one real process and develop an SPC plan containing:

CTQ or process measure.
Data type.
Measurement method.
Sampling frequency.
Rational subgroup definition.
Appropriate control chart.
Control limits.
Special-cause response procedure.
Escalation responsibility.
Documentation and monitoring method.

6.1 Practical SPC Checklist

Before implementing a control chart, confirm:

The process characteristic is clearly defined.
The operational definition is clear.
The measurement system is appropriate.
Data are collected consistently.
Sampling frequency is defined.
Rational subgrouping is appropriate.
The correct control chart has been selected.
Control limits are statistically appropriate.
Specification limits are not confused with control limits.
Special-cause response rules are defined.
Responsibilities are assigned.
The process is continuously monitored.

7. Lesson Summary

Statistical Process Control provides a structured way to monitor process behavior and distinguish normal variation from unusual variation.

The key principles are:

Every process contains variation.
Common causes are part of normal process behavior.
Special causes represent unusual sources of variation.
Control charts help detect changes in process behavior.
Control limits are calculated from process behavior.
Specification limits represent customer, engineering, regulatory, or business requirements.
Variable and attribute data require different control-chart approaches.
Rational subgrouping is essential for meaningful SPC.
Xbar-R, Xbar-S, and I-MR charts are common variable-data charts.
p, np, c, and u charts are common attribute-data charts.
A control-chart signal should trigger investigation rather than automatic assumptions.
Process stability should generally be established before capability analysis.
SPC supports prevention, early detection, and sustained process performance.

The Green Belt should remember the basic sequence:

Measure → Chart → Interpret → Investigate → Correct → Standardize → Monitor

8. Lesson Learnt / Conclusion

The central lesson of SPC is that a process should be managed through its behavior, not merely through individual results.

A single defective product may not tell the complete story. Similarly, a process that currently meets specifications may still be unstable and at risk of deterioration.

Control charts allow the Green Belt to see the process over time and identify meaningful changes.

The Green Belt should therefore develop the habit of asking:

“Is this variation part of the normal process, or is something unusual happening?”

When a special-cause signal appears, the appropriate response is not simply to adjust the process immediately. The team should investigate the process condition, identify the assignable cause where possible, take appropriate corrective action, and then standardize the improved method.

SPC is therefore not merely a statistical charting technique. It is a management system for maintaining stable, predictable, and continuously monitored processes.

Key takeaway:

Control the process before the process controls the results.