CHAPTER 18

Statistical Applications in Quality

and Productivity Management
USING STATISTICS: Service Quality at the Beachcomber Hotel

18.1 TOTAL QUALITY MANAGEMENT 18.7 CONTROL CHARTS FOR THE RANGE
AND THE MEAN
18.2 SIX SIGMA MANAGEMENT
The R Chart
18.3 THE THEORY OF CONTROL CHARTS The X Chart
18.4 CONTROL CHART FOR THE 18.8 PROCESS CAPABILITY
PROPORTION OF NONCONFORMING Customer Satisfaction and Specification Limits
ITEMS—THE p CHART Capability Indexes
18.5 THE RED BEAD EXPERIMENT: CPL, CPU, and Cpk
UNDERSTANDING PROCESS VARIABILITY A.18 USING SOFTWARE FOR CONTROL
18.6 CONTROL CHART FOR AN AREA CHARTS
OF OPPORTUNITY— THE c CHART A18.1 Microsoft Excel
A18.2 Minitab
A18.3 (CD-ROM Topic) SPSS

LEARNING OBJECTIVES
In this chapter, you learn:
• The basic themes of quality management and Deming’s
14 points
• The basic aspects of Six Sigma management
• How to construct various control charts
• Which control chart to use for a particular type of data
• How to measure the capability of a process
752 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

U S I N G S TAT I S T I C S
Service Quality at the Beachcomber Hotel
In the Chapter 12 “Using Statistics” scenario, you were the manager of
T. C. Resort Properties. For this scenario, consider that you manage only
the Beachcomber Hotel. As the hotel manager, you would want to contin-
ually improve the quality of service that your guests receive so that over-
all guest satisfaction increases. To help you achieve this improvement,
T. C. Resort Properties has provided its managers with training in Six
Sigma management. In order to meet the business objective of increasing
the return rate of guests at your hotel, you have decided to focus on the
critical first impressions of the service that your hotel provides. Is the
assigned hotel room ready when a guest checks in? Are all expected
amenities such as extra towels and a complimentary guest basket in the
room when the guest first walks in? Are the video-entertainment center
and high-speed Internet access working properly? And, do guests receive their luggage in a rea-
sonable amount of time?
To study these guest satisfaction issues, you have embarked on an improvement project
that measures two critical-to-quality (CTQ) measurements, the readiness of the room and the
time it takes to deliver luggage. You would like to learn the following:
■ Are the proportion of rooms ready and the time required to deliver luggage to the rooms
acceptable?
■ Are the proportion of rooms ready and the luggage delivery time consistent from day to
day, or are they increasing or decreasing?
■ On the days when the proportion of rooms that are not ready or the time to deliver lug-
gage is greater than normal, is this due to a chance occurrence or is there a fundamental
flaw in the process used to make rooms ready and deliver luggage?

I n this chapter the focus is on quality and productivity management. Companies manufactur-
ing products, as well as those providing services, such as the Beachcomber Hotel in the
“Using Statistics” scenario, realize that quality and productivity are essential for survival in the
global economy. Among the areas in which quality has an impact on our everyday work and
personal lives are:
• The design, production, and subsequent reliability of our automobiles
• The services provided by hotels, banks, schools, retailing operations, and mail-order
companies
• The continuous improvement in computer chips that makes for faster and more powerful
computers
• The ever-expanding capability of communication devices such as data transmission lines,
paging devices, facsimile machines, and cellular telephones
• The availability of new technology and equipment that has led to improved diagnosis of ill-
nesses and the improved delivery of health care services

18.1 TOTAL QUALITY MANAGEMENT

During the past twenty-five years, the renewed interest in quality and productivity in the United
States followed as a reaction to improvements of Japanese industry that had began as early as
1950. Individuals such as W. Edwards Deming, Joseph Juran, and Kaoru Ishikawa developed
an approach that focuses on continuous improvement of products and services through an
 18.1: Total Quality Management 753

increased emphasis on statistics, process improvement, and optimization of the total system.
This approach, widely known as total quality management (TQM), is characterized by these
themes:
• The primary focus is on process improvement.
• Most of the variation in a process is due to the system and not the individual.
• Teamwork is an integral part of a quality management organization.
• Customer satisfaction is a primary organizational goal.
• Organizational transformation must occur in order to implement quality management.
• Fear must be removed from organizations.
• Higher quality costs less not more, but requires an investment in training.
In the 1980s the federal government of the United States increased its efforts to improve qual-
ity in American business. Congress passed the Malcolm Baldrige National Improvement Act
of 1987 and began awarding the Malcolm Baldrige Award to companies making the greatest
strides in improving quality and customer satisfaction. W. Edwards Deming became a promi-
nent consultant to many Fortune 500 companies including Ford, General Motors, and Proctor
and Gamble. Through four-day seminars, Deming widely promoted his “14 points for man-
agement” listed below, and many companies adopted some or all of them.
1. Create constancy of purpose for improvement of product and service.
2. Adopt the new philosophy.
3. Cease dependence on inspection to achieve quality.
4. End the practice of awarding business on the basis of price tag alone. Instead, minimize
total cost by working with a single supplier.
5. Improve constantly and forever every process for planning, production, and service.
6. Institute training on the job.
7. Adopt and institute leadership.
8. Drive out fear.
9. Break down barriers between staff areas.
10. Eliminate slogans, exhortations, and targets for the workforce.
11. Eliminate numerical quotas for the workforce and numerical goals for management.
12. Remove barriers that rob people of pride of workmanship. Eliminate the annual rating or
merit system.
13. Institute a vigorous program of education and self-improvement for everyone.
14. Put everyone in the company to work to accomplish the transformation.
Point 1, create constancy of purpose, refers to how an organization deals with problems
that arise both at present and in the future. The focus is on the constant improvement of a prod-
uct or service. This improvement process is illustrated by the Shewhart-Deming cycle shown
in Figure 18.1. The Shewhart-Deming cycle represents a continuous cycle of “plan, do, study,
and act.” The first step, planning, represents the initial design phase for planning a change in a
manufacturing or service process. This step involves teamwork among individuals from differ-
ent areas within an organization. The second step, doing, involves implementing the change,
preferably on a small scale. The third step, studying, involves an analysis of the results using
statistical tools to determine what was learned. The fourth step, acting, involves the acceptance
of the change, its abandonment, or further study of the change under different conditions.

FIGURE 18.1
Shewhart-Deming Cycle
Act Plan

Study Do
754 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Point 2, adopt the new philosophy, refers to the urgency with which companies need to
realize that there is a new economic age of global competition. It is better to be proactive and
change before a crisis occurs than to react to some negative experiences that may have
occurred. Rather than take the approach “if it’s not broke, don’t fix it,” it is better to continually
work on improvement and avoid expensive fixes.
Point 3, cease dependence on inspection to achieve quality, implies that any inspection
whose purpose is to improve quality is too late because the quality is already built into the
product. It is better to focus on making it right the first time. Among the difficulties involved in
inspection (besides high costs) are the failure of inspectors to agree on the operational defini-
tions for nonconforming items, and the problem of separating good and bad items. The follow-
ing example illustrates the difficulties inspectors face.
Suppose your job involves proofreading the sentence in Figure 18.2 with the objective of
counting the number of occurrences of the letter “F.” Perform this task and record the number
of occurrences of the letter F that you discover.

FIGURE 18.2
An Example of a
Proofreading Process
FINISHED FILES ARE THE RESULT
Source: W. W. Scherkenbach, OF YEARS OF SCIENTIFIC
The Deming Route to
Quality and Productivity: STUDY COMBINED WITH THE
Road Maps and Roadblocks
(Washington, DC: CEEP
EXPERIENCE OF MANY YEARS
Press, 1986).

People usually see either three Fs or six Fs. The correct number is six Fs. The number you
see depends on the method you use to examine the sentence. You are likely to find three Fs if
you read the sentence phonetically and six Fs if you count the number of Fs carefully. The point
of the exercise is to show that if such a simple process as counting Fs leads to inconsistency of
inspectors’ results, what will happen when a much more complicated process fails to contain a
clear operational definition of nonconforming? Certainly, in such situations, a large amount of
variability occurs from inspector to inspector.
Point 4, ending the practice of awarding business on the basis of price tag alone, represents
the antithesis of lowest-bidder awards. There is no real long-term meaning to price without
knowledge of the quality of the product.
Point 5—improve constantly and forever every process for planning, production and ser-
vice—reinforces the importance of the continuous focus of the Shewhart-Deming cycle and the
belief that quality needs to be built in at the design stage. Attaining quality is a never-ending
process in which reduction in variation translates into a reduction in the financial losses result-
ing from products and services experiencing large fluctuations in quality.
Point 6, institute training, reflects the needs of all employees, including production work-
ers, engineers, and managers. It is critically important for management to understand the dif-
ferences between special causes and common causes of variation (see section 18.3) so that
proper action is taken in each circumstance.
Point 7, adopt and institute leadership, relates to the distinction between leadership and
supervision. The aim of leadership should be to improve the system and achieve greater con-
sistency of performance.
Points 8 through 12—drive out fear, break down barriers between staff areas, eliminate
slogans, eliminate numerical quotas for the workforce and numerical goals for management,
and remove barriers to pride of workmanship (including the annual rating and merit system)—
are all related to the evaluation of employee performance. An emphasis on targets and exhorta-
tions may place an improper burden on the workforce. Workers cannot produce beyond what
 18.3: The Theory of Control Charts 755

the system will allow (this is clearly illustrated in section 18.5). It is management’s job to
improve the system, not to raise the expectations on workers beyond the system’s capability.
Point 13, encourage education and self-improvement for everyone, reflects the notion that
the most important resource of any organization is its people. Efforts that improve the knowl-
edge of people in the organization also serve to increase the assets of the organization.
Education and self-improvement can lead to reduced turnover within an organization.
Point 14, take action to accomplish the transformation, again reflects the approach of man-
agement as a process in which one continuously strives toward improvement in a never-ending
cycle.
Although Deming’s points were thought-provoking, some criticized his approach for lack-
ing a formal, objective accountability. Many managers of large organizations, used to seeing
financial analyses of policy changes, needed a more prescriptive approach.

18.2 SIX SIGMA MANAGEMENT

Six Sigma management is a quality improvement system originally developed by Motorola in
the mid-1980s. Six Sigma offers a more prescriptive and systematic approach to process
improvement than TQM, and places a higher emphasis on accountability and bottom line
results. Many companies all over the world are using Six Sigma management to improve effi-
ciency, cut costs, eliminate defects and reduce product variation.
The name Six Sigma comes from the fact that it is a managerial approach designed to cre-
1The Six Sigma approach ate processes that result in no more than 3.4 defects per million.1 One of the aspects that dis-
assumes that the process tinguishes Six Sigma from other approaches is a clear focus on achieving bottom-line results in
may shift as much as 1.5 a relatively short three- to six-month period of time. After seeing the huge financial successes
standard deviations over at Motorola, GE, and other early adopters of Six Sigma, many companies worldwide have now
the long term. Six standard instituted Six Sigma programs (see references 1, 8, 9, and 15).
deviations minus a 1.5 To guide managers in their task of improving short- and long-term results, Six Sigma uses
standard deviation shift a five-step process known as the DMAIC model—named for the five steps in the process:
produces a 4.5 standard Define, Measure, Analyze, Improve, and Control.
deviation goal. The area
under the normal curve • Define The problem is defined along with the costs, benefits, and the impact on the customer.
outside 4.5 standard • Measure Operational definitions for each critical-to-quality (CTQ) characteristic are
deviations is approximately developed. In addition, the measurement procedure is verified so that it is consistent over
3.4 out of a million repeated measurements.
(0.0000034). • Analyze The root causes of why defects occur are determined, and variables in the process
causing the defects are identified. Data are collected to determine benchmark values for
each process variable. This analysis often uses control charts (to be discussed in sections
18.3 through 18.7).
• Improve The importance of each process variable on the CTQ characteristic is studied
using designed experiments (see Chapter 11). The objective is to determine the best level
for each variable.
• Control The objective is to maintain the benefits for the long term by avoiding potential
problems that can occur when a process is changed.

Implementation of Six Sigma management requires a data-oriented approach that is heavily

based on using statistical tools such as control charts and designed experiments. It also involves
training everyone in the organization in the DMAIC model.

18.3 THE THEORY OF CONTROL CHARTS

Both total quality management and Six Sigma management make use of a wide array of statis-
tical tools. One tool widely used in each approach to analyze process data collected sequen-
tially over time is the control chart.
756 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

The control chart monitors variation in a characteristic of a product or service over time.
You can use a control chart to study past performance, to evaluate present conditions, or to pre-
dict future outcomes (see reference 7). Information gained from analyzing a control chart
forms the basis for process improvement. Different types of control charts allow you to analyze
different types of critical-to-quality (CTQ) variables—for categorical variables such as the pro-
portion of hotel rooms that are nonconforming in terms of the availability of amenities and the
working order of all appliances in the room, discrete variables such as the number of hotel
guests registering a complaint during a week, and continuous variables such as the length of
time required for delivering luggage to the room. In addition to providing a visual display of
data representing a process, a principal focus of the control chart is the attempt to separate spe-
cial causes of variation from common causes of variation.

Special causes of variation represent large fluctuations or patterns in the data that are not
inherent to a process. These fluctuations are often caused by changes in the process that
represent either problems to correct or opportunities to exploit. Some organizations refer to
special causes of variation as assignable causes of variation.
Common causes of variation represent the inherent variability that exists in a process.
These fluctuations consist of the numerous small causes of variability that operate
randomly or by chance. Some organizations refer to common causes of variation as chance
causes of variation.

The distinction between the two causes of variation is crucial because special causes of
variation are not part of a process and are correctable or exploitable without changing the sys-
tem. Common causes of variation, however, can be reduced only by changing the system. Such
systemic changes are the responsibility of management.
Control charts allow you to monitor a process and identify the presence or absence of spe-
cial causes. By doing so, control charts help prevent two types of errors. The first type of error
involves the belief that an observed value represents special-cause variation when it is due to
the common cause variation of the system. Treating common-cause variation as special-cause
variation often results in overadjusting a process. This overadjustment, known as tampering,
increases the variation in the process. The second type of error involves treating special-cause
variation as common-cause variation. This error results in not taking immediate corrective
action when necessary. Although both of these types of errors can occur even when using a
control chart, they are far less likely.
To construct a control chart, you collect samples from the output of a process over time.
The samples used for constructing control charts are known as subgroups. For each subgroup
(i.e., sample), you calculate the value of a statistic associated with a CTQ variable. Commonly used
2Recall from section 6.2 that statistics include the fraction nonconforming (see section 18.4), the number of nonconformities
in the normal distribution, (see section 18.6), and the mean and range of a numerical variable (see section 18.7). You then plot
µ ± 3σ includes almost all the values versus time and add control limits to the chart. The most typical form of a control chart
(99.73%) of the observations sets control limits that are within ±3 standard deviations2 of the statistical measure of interest.
in the population. Equation (18.1) defines, in general, the upper and lower control limits for control charts.

CONSTRUCTING CONTROL LIMITS

Process mean ±3 standard deviations (18.1)

so that
Upper control limit (UCL) = process mean +3 standard deviations
Lower control limit (LCL) = process mean −3 standard deviations

When these control limits are set, you evaluate the control chart by trying to find any pattern
that might exist in the values over time and by determining whether any points fall outside the
control limits. Figure 18.3 illustrates three different situations.
 18.3: The Theory of Control Charts 757

FIGURE 18.3
Three Control Chart Common cause variation
only: No points outside Special cause
Patterns X 3σ Limits; no pattern X variation X Pattern over time:
over time Special cause variation
UCL

Center
Line

LCL
Special cause variation

Time Time Time

Panel A Panel B Panel C

In panel A of Figure 18.3, there is no apparent pattern in the values over time and there are
no points that fall outside the 3 standard deviation control limits. The process appears stable
and contains only common-cause variation. Panel B, on the contrary, contains two points that
fall outside the 3 standard deviation control limits. You should investigate these points to try to
determine the special causes that led to their occurrence. Although panel C does not have any
points outside the control limits, it has a series of consecutive points above the mean value (the
centerline) as well as a series of consecutive points below the mean value. In addition, a long-
term overall downward trend is clearly visible. You should investigate the situation to try to
determine what may have caused this pattern.
3Minitabuses different The detection of a trend is not always so obvious. Two other simple rules3 (see references 7
rules—see reference 13. and 8) that allow you to detect a shift in the mean level of a process are:

• Eight or more consecutive points that lie above the center line or eight or more consecutive
points that lie below the center line.
• Eight or more consecutive points move upward in value or eight or more consecutive
points move downward in value.

A process whose control chart indicates an out-of-control condition (a point outside the control
limits or exhibiting a trend) is said to be out of control. An out-of-control process contains
both common causes of variation and special causes of variation. Because special causes of
variation are not part of the process design, an out-of-control process is unpredictable. Once
you determine a process is out of control, you must identify the special causes of variation that
are producing the out-of-control conditions. If the special causes are detrimental to the quality
of the product or service, you need to implement plans to eliminate this source of variation.
When a special cause increases quality, you should change the process so that the special cause
is incorporated into the process design. Thus, this beneficial special cause now becomes a com-
mon-cause source of variation and the process is improved.
A process whose control chart is not indicating any out-of-control conditions is said to be
in control. An in-control process contains only common causes of variation. Because these
sources of variation are inherent to the process itself, an in-control process is predictable. In-
control processes are sometimes said to be in a state of statistical control. When a process is
in control, you must determine whether the amount of common-cause variation in the process
is small enough to satisfy the customers of the products or services. (In section 18.8, you will
learn statistical methods that allow you to compare common-cause variation to customer
expectations.) If the common-cause variation is small enough to consistently satisfy the cus-
tomer, you then use control charts to monitor the process on a continuing basis to make sure
that the process remains in control. If the common-cause variation is too large, you need to alter
the process itself.
758 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

18.4 CONTROL CHART FOR THE PROPORTION

OF NONCONFORMING ITEMS—THE p CHART
Various types of control charts are used to monitor processes and determine whether special-
cause variation is present in a process. Attribute charts are used for categorical or discrete
variables. This section introduces the p chart, used when sampled items are classified accord-
ing to whether they conform or do not conform to operationally defined requirements. Thus,
the p chart helps you monitor and analyze the proportion of nonconforming items there are in
repeated samples (i.e., subgroups) selected from a process. Section 18.6 introduces the c chart,
an attribute chart for the number of nonconformances (or occurrences) in a given area of
opportunity.
To begin the discussion of p charts, recall that you studied proportions and the binomial
distribution in section 5.3. Then, in section 6.4, the sample proportion is defined as p = X/n, and
the standard deviation of the sample proportion is defined as
4In this chapter the term π(1 − π )
σp =
nonconforming items n
is used, while in earlier
chapters the term Using Equation (18.1) on page 756, control limits for the proportion of nonconforming4 items
success was used. from the sample data are established in Equation (18.2).

CONTROL LIMITS FOR THE p CHART

p(1 − p )
p±3 (18.2)
n
p(1 − p )
UCL = p + 3
n
p(1 − p )
LCL = p − 3
n
For equal ni,
k
∑ pi
i =1
n = ni and p =
k
or in general,
k k
∑ ni ∑ Xi
i =1 i =1
n = and p = k
k
∑ ni
i =1

where Xi = number of nonconforming items in subgroup i

ni = sample (or subgroup) size for subgroup i
pi = Xi /ni = proportion of nonconforming items in subgroup i
k = number of subgroups selected
n = mean subgroup size
p = estimated proportion of nonconforming items
 18.4: Control Chart for the Proportion of Nonconforming Items—The p Chart 759

Any negative value for the lower control limit means that the lower control limit does not exist.
To show the application of the p chart, return to the “Using Statistics” scenario concerning the
Beachcomber Hotel.
During the Measure phase of the Six Sigma DMAIC model, a nonconformance was
operationally defined as the absence of an amenity in the room or a room appliance not in
working order upon check-in. During the Analyze phase of the Six Sigma DMAIC model,
data on the nonconformances were collected daily from a sample of 200 rooms. Table 18.1
lists the number and proportion of nonconforming rooms for each day in the 4-week period.
HOTEL1.

TABLE 18.1 Rooms Rooms Rooms Rooms

Nonconforming Hotel Day Studied Not Ready Proportion Day Studied Not Ready Proportion
Rooms at Check-In over
1 200 16 0.080 15 200 18 0.090
a 28 Day Period
2 200 7 0.035 16 200 13 0.065
3 200 21 0.105 17 200 15 0.075
4 200 17 0.085 18 200 10 0.050
5 200 25 0.125 19 200 14 0.070
6 200 19 0.095 20 200 25 0.125
7 200 16 0.080 21 200 19 0.095
8 200 15 0.075 22 200 12 0.060
9 200 11 0.055 23 200 6 0.030
10 200 12 0.060 24 200 12 0.060
11 200 22 0.110 25 200 18 0.090
12 200 20 0.100 26 200 15 0.075
13 200 17 0.085 27 200 20 0.100
14 200 26 0.130 28 200 22 0.110

k
For these data, k = 28, ∑ pi = 2.315 and because the ni are equal ni = n = 200. Thus,
i =1
k
∑ pi 2.315
i =1
p = = = 0.0827
k 28

Using Equation (18.2),

( 0.0827 )(0.9173)
0.0827 ± 3
200

so that

UCL = 0.0827 + 0.0584 = 0.1411

and

LCL = 0.0827 − 0.0584 = 0.0243

Figure 18.4 displays the Microsoft Excel control chart for the data of Table 18.1. Figure
18.5 provides Minitab output. Figure 18.4 and 18.5 indicate a process in a state of statistical
760 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

control, with the individual points distributed around p without any pattern and all the points
within the control limits. Thus, any improvement in the process of making rooms ready for
guests in the Improve phase of the DMAIC model must come from the reduction of common-
cause variation. Such reductions require a change in the process. These changes are the respon-
sibility of management. Remember that improvements cannot occur until improvements to the
process itself are successfully implemented.

FIGURE 18.4
Microsoft Excel p Chart
for the Nonconforming
Hotel Room Data

FIGURE 18.5
Minitab p Chart for the
Nonconforming Hotel
Room Data

The first example illustrated a situation in which the subgroup size did not vary. As a gen-
eral rule, as long as none of the subgroup sizes ni differ from the mean subgroup size n by
more than ±25% of n (see reference 7), you can use Equation (18.2) on page 758 to compute
the control limits for the p chart. If any subgroup size differ by more than ±25% of n , you use
alternative formulas for calculating the control limits (see reference 7). To illustrate the use of
the p chart when the subgroup sizes are unequal, Example 18.1 studies the production of gauze
sponges.

EXAMPLE 18.1 USING THE p CHART FOR UNEQUAL SUBGROUP SIZES

Table 18.2 indicates the number of sponges produced daily and the number of nonconforming
sponges for a period of 32 days. SPONGE Construct a control chart for these data.
 18.4: Control Chart for the Proportion of Nonconforming Items—The p Chart 761

TABLE 18.2 Non- Non-

Nonconforming Sponges Conforming Sponges Conforming
Sponges over a Day Produced Sponges Proportion Day Produced Sponges Proportion
32-Day Period
1 690 21 0.030 17 575 20 0.035
2 580 22 0.038 18 610 16 0.026
3 685 20 0.029 19 596 15 0.025
4 595 21 0.035 20 630 24 0.038
5 665 23 0.035 21 625 25 0.040
6 596 19 0.032 22 615 21 0.034
7 600 18 0.030 23 575 23 0.040
8 620 24 0.039 24 572 20 0.035
9 610 20 0.033 25 645 24 0.037
10 595 22 0.037 26 651 39 0.060
11 645 19 0.029 27 660 21 0.032
12 675 23 0.034 28 685 19 0.028
13 670 22 0.033 29 671 17 0.025
14 590 26 0.044 30 660 22 0.033
15 585 17 0.029 31 595 24 0.040
16 560 16 0.029 32 600 16 0.027

SOLUTION For these data,

k k
k = 32, ∑ ni = 19, 926, and ∑ X i = 679
i =1 i =1

Thus, using Equation (18.2) on page 758,

19, 926 679

n = = 622.69 and p = = 0.034
32 19, 926

so that

( 0.034 )(1 − 0.034 )

0.034 ± 3
622.69
= 0.034 ± 0.022

Thus,

UCL = 0.034 + 0.022 = 0.056

and

LCL = 0.034 − 0.022 = 0.012

Figure 18.6 displays the Microsoft Excel control chart for the sponge data. Figure 18.7
shows the Minitab control chart. An examination of either of these figures indicates that day
26, in which there were 39 nonconforming sponges produced out of 651 sampled, is above the
upper control limit. Management needs to determine the reason (i.e., root cause) for this special
cause variation and take corrective action. Once actions are taken, you can remove the data
from day 26 and then construct and analyze a new control chart.
762 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Notice that the UCL and LCL on Figure 18.7 are represented by a jagged line. Minitab is
calculating separate control limits for each day depending on the subgroup size for that day.
FIGURE 18.6
Microsoft Excel p Chart
for the Proportion
of Nonconforming
Sponges

FIGURE 18.7
Minitab p Chart
for the Proportion
of Nonconforming
Sponges

PROBLEMS FOR SECTION 18.4

Learning the Basics b. What are the LCL and UCL?
PH Grade 18.1 The following data were collected on non- c. Are there any special causes of variation?
ASSIST conformances for a period of 10 days.
PH Grade 18.2 The following data were collected on non-
Day Sample Size Nonconformances ASSIST conformances for a period of 10 days.
1 100 12
Day Sample Size Nonconformances
2 100 14
3 100 10 1 111 12
4 100 18 2 93 14
5 100 22 3 105 10
6 100 14 4 92 18
7 100 15 5 117 22
8 100 13 6 88 14
9 100 14 7 117 15
10 100 16 8 87 13
9 119 14
a. On what day is the proportion of nonconformances 10 107 16
largest? smallest?
 18.4: Control Chart for the Proportion of Nonconforming Items—The p Chart 763

a. On what day is the proportion of nonconformances Number Number

largest? smallest? Day Nonconforming Day Nonconforming
b. What are the LCL and UCL?
c. Are there any special causes of variation? 11 22 22 24
12 26 23 26
13 25 24 23
Applying the Concepts 14 29 25 27
15 20 26 28
You can solve problems 18.3–18.8 manually or by 16 19 27 24
using Microsoft Excel, Minitab, or SPSS. 17 23 28 22
18.3 A medical transcription service enters medical data 18 19 29 20
on patient files for hospitals. The service studied ways to 19 18 30 25
improve the turnaround time (defined as the time between 20 27 31 27
receiving data and the time the client receives completed 21 28 32 19
files). After studying the process, it was determined that
turnaround time was increased by transmission errors. A a. Construct a p chart.
transmission error was defined as data transmitted that did b. Is the process in a state of statistical control? Why?
not go through as planned, and needed to be retransmitted.
Each day a sample of 125 record transmissions were ran- PH Grade 18.5 A hospital administrator is concerned with
ASSIST the time to process patients’ medical records after
domly selected and evaluated for errors. The table below
presents the number and proportion of transmissions with discharge. She determined that all records should
errors. TRANSMIT be processed within 5 days of discharge. Thus, any record
not processed within 5 days of a patient’s discharge is non-
Number Proportion Number Proportion conforming. The administrator recorded the number of
Day of Errors of Errors Day of Errors of Errors patients discharged and the number of records not
1 6 0.048 17 4 0.032 processed within the 5-day standard for a 30-day period in
2 3 0.024 18 6 0.048 the file MEDREC.
3 4 0.032 19 3 0.024 a. Construct a p chart for these data.
4 4 0.032 20 5 0.040 b. Does the process give an out-of-control signal? Explain.
5 9 0.072 21 1 0.008 c. If the process is out of control, assume that special
6 0 0.000 22 3 0.024 causes were subsequently identified and corrective
7 0 0.000 23 14 0.112
8 8 0.064 24 6 0.048 action taken to keep them from happening again. Then
9 4 0.032 25 7 0.056 eliminate the data causing the out-of-control signals,
10 3 0.024 26 3 0.024 and recalculate the control limits.
11 4 0.032 27 10 0.080
12 1 0.008 28 7 0.056 SELF 18.6 The bottling division of Sweet Suzy’s
13 10 0.080 29 5 0.040 Test Sugarless Cola maintains daily records of the
14 9 0.072 30 0 0.000 occurrences of unacceptable cans flowing from
15 3 0.024 31 3 0.024 the filling and sealing machine. The following table lists
16 1 0.008 the number of cans filled and the number of nonconform-
ing cans for one month (based on a 5-day workweek).
a. Construct a p chart. COLASPC
b. Is the process in a state of statistical control? Why?
18.4 The following data represent the findings from a
Cans Unacceptable Cans Unacceptable
study conducted at a factory that manufactures film canis-
Day Filled Cans Day Filled Cans
ters. For 32 days, 500 film canisters were sampled and
inspected. The following table lists the number of defective 1 5,043 47 12 5,314 70
film canisters (i.e., nonconforming items) for each day 2 4,852 51 13 5,097 64
(i.e., subgroup). CANISTER 3 4,908 43 14 4,932 59
4 4,756 37 15 5,023 75
Number Number 5 4,901 78 16 5,117 71
Day Nonconforming Day Nonconforming 6 4,892 66 17 5,099 68
1 26 6 20 7 5,354 51 18 5,345 78
2 25 7 21 8 5,321 66 19 5,456 88
3 23 8 27 9 5,045 61 20 5,554 83
4 24 9 23 10 5,113 72 21 5,421 82
5 26 10 25 11 5,247 63 22 5,555 87
764 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

a. Construct a p chart for the proportion of unacceptable b. Based on your answer to (a), if you were the manager of
cans for the month. Does the process give an out-of-con- the accounting office, what would you do to improve the
trol signal? process of account number entry?
b. If you want to develop a process for reducing the pro-
portion of unacceptable cans, how should you proceed? 18.8 A regional manager of a telephone company is
responsible for processing requests concerning additions,
18.7 The manager of the accounting office of a large hos- changes, or deletions of telephone service. She forms a
pital is studying the problem of entering incorrect account service improvement team to look at the corrections in
numbers into the computer system. A subgroup of 200 terms of central office equipment and facilities required
account numbers is selected from each day’s output, and to process the orders that are issued to service requests.
each account number is inspected to determine whether it Data collected over a period of 30 days are in the file
is a nonconforming item. The results for a period of 39 TELESPC.
days are in the file ERRORSPC. a. Construct a p chart for the proportion of corrections.
a. Construct a p chart for the proportion of nonconform- Does the process give an out-of-control signal?
ing items. Does the process give an out-of-control b. What should the regional manager do to improve the
signal? processing of requests for changes in telephone service?

18.5 THE RED BEAD EXPERIMENT: UNDERSTANDING

PROCESS VARIABILITY
This chapter began with a discussion of total quality management, Deming’s 14 points, Six
Sigma management, and definitions of common-cause variation and special-cause variation.
Now that you have studied the p chart, this section presents a famous parable, the red bead
experiment, to enhance your understanding of common cause and special cause variation.
The red bead experiment involves the selection of beads from a box that contains 4,000
beads. Unknown to the participants in the experiment, 3,200 (80%) of the beads are white and
800 (20%) are red. You can use several different scenarios for conducting the experiment. The
one used here begins with a facilitator (who will play the role of company supervisor) asking
members of the audience to volunteer for the jobs of workers (at least four are needed), inspec-
tors (two are needed), chief inspector (one is needed), and recorder (one is needed). A worker’s
job consists of using a paddle that has five rows of 10 bead-size holes to select 50 beads from
the box of beads.
When the participants have been selected, the supervisor explains the jobs to them. The
job of the workers is to produce white beads, because red beads are unacceptable to the cus-
tomers. Strict procedures are to be followed. Work standards call for the daily production of
exactly 50 beads by each worker (a strict quota system). Management has established a stan-
dard that no more than 2 red beads (4%) per worker are to be produced on any given day. Each
worker dips the paddle into the box of beads so that when it is removed, each of the 50 holes
contains a bead. The worker carries the paddle to the two inspectors, who independently
record the count of red beads. The chief inspector compares their counts and announces the
results to the audience. The recorder writes down the number and percentage of red beads next
to the name of the worker.
When all the people know their jobs, “production” can begin. Suppose that on the first
“day,” the number of red beads “produced” by the four workers (call them Alyson, David, Peter,
and Sharyn) was 9, 12, 13, and 7, respectively. How should management react to the day’s pro-
duction when the standard says that no more than 2 red beads per worker should be produced?
Should all the workers be reprimanded, or should only David and Peter be given a stern warn-
ing that they will be fired if they don’t improve?
Suppose that production continues for an additional two days. Table 18.3 summarizes the
results for all three days.
 18.5: The Red Bead Experiment: Understanding Process Variability 765

TABLE 18.3 Day

Red Bead Experiment Name 1 2 3 All Three Days
Results for 4 Workers
over 3 Days Alyson 9 (18%) 11 (22%) 6 (12%) 26 (17.33%)
David 12 (24%) 12 (24%) 8 (16%) 32 (21.33%)
Peter 13 (26%) 6 (12%) 12 (24%) 31 (20.67%)
Sharyn 7 (14%) 9 (18%) 8 (16%) 24 (16.0%)
All 4 workers 41 38 34 113
Mean 10.25 9.5 8.5 9.42
Percentage 20.5% 19% 17% 18.83%

From Table 18.3, on each day, some of the workers were above the mean and some below the
mean. On day 1, Sharyn did best, but on day 2, Peter (who had the worst record on day 1) was
best, and on day 3, Alyson was best. How can you explain all this variation? Using Equation
(18.2) on page 758 to develop a p chart for these data,

k
k = 4 workers × 3 days = 12, n = 50, and ∑ X i = 113
i =1

Thus,

113
p= = 0.1883
(50)(12)

so that

p(1 − p )
p±3
n
0.1883(1 − 0.1883)
= 0.1883 ± 3
50
= 0.1883 ± 0.1659

Thus,

UCL = 0.1883 + 0.1659 = 0.3542

and

LCL = 0.1883 − 0.1659 = 0.0224

Figure 18.8 represents the p chart for the data of Table 18.3. In Figure 18.8, all of the points are
within the control limits and there are no patterns in the results. The differences between the
workers merely represent common-cause variation inherent in a stable system.
766 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Four morals to the parable of the red beads are:

• Variation is an inherent part of any process.
• Workers work within a system over which they have little control. It is the system that pri-
marily determines their performance.
• Only management can change the system.
• There will always be some workers above the mean and some workers below the mean.

FIGURE 18.8
p Chart for the Red UCL
Bead Experiment
.30
Proportion of Red Beads

.20
P = 0.1883

.10

LCL
0
Alyson Peter Alyson Peter Alyson Peter
David Sharyn David Sharyn David Sharyn

PROBLEMS FOR SECTION 18.5

Applying the Concepts a. Conduct the experiment in the same way as described in
18.9 In the Red Bead experiment, how do you think many this section.
managers would have reacted after day 1? day 2? day 3? b. Remove 400 red beads from the bead box before begin-
ning the experiment. How do your results differ from
18.10 (Class Project) Obtain a version of the red bead those in (a)? What does this tell you about the effect of
experiment for your class. the “system” on the workers?

18.6 CONTROL CHART FOR AN AREA OF OPPORTUNITY—

THE c CHART
In this text, you are introduced to two types of attribute charts, the p chart in section 18.4
and the c chart in this section. Recall that you use the p chart for monitoring and analyzing
the proportion of nonconforming items. You use the c chart for monitoring and analyzing
the number of nonconformities in an area of opportunity. Areas of opportunity can be indi-
vidual units of products or services, or units of time, space, or area. Nonconformities are
sometimes referred to as defects or flaws. Examples of the number of nonconformities in
areas of opportunity are the number of flaws in a square foot of carpet, the number of typo-
graphical errors on a printed page, the number of breakdowns per day in an academic com-
puter center, and the number of hotel customers filing a complaint in a given week. This sit-
uation differs from the one in which you used the p chart; instead of classifying each unit as
conforming or nonconforming, you count the number of nonconformities in each area of
opportunity.
 18.6: Control Chart for an Area of Opportunity—The c Chart 767

The above situation fits the assumptions of a Poisson distribution (see section 5.4). For the
Poisson distribution, the standard deviation of the number of nonconformities is the square root
of the mean number of nonconformities (λ). Assuming that the size of each area of opportunity
5Ifthe size of the unit varies, remains constant,5 you can compute the control limits for the number of nonconformities per
you should use the u chart area of opportunity using the observed mean number of nonconformities as an estimate of λ.
instead of the c chart (see Equation (18.3) defines the control limits for the c chart, which you use to monitor and analyze
references 7, 8, and 11). the number of nonconformities per area of opportunity.

CONTROL LIMITS FOR THE c CHART

c±3 c (18.3)

UCL = c + 3 c

LCL = c − 3 c

k
∑ ci
i =1
where c =
k
k = number of units sampled
ci = number of nonconformities in unit i

To help study the hotel service quality in the “Using Statistics” scenario, you can use a c chart
to monitor the number of customer complaints filed with the hotel. If guests of the hotel are
dissatisfied with any part of their stay, they are asked to file a customer complaint form. At the
end of each week, the number of complaints filed is recorded. In this example, a complaint is a
nonconformity and the area of opportunity is one week. Table 18.4 lists the number of com-
plaints from the last 50 weeks. COMPLAINTS

TABLE 18.4 Number of Number of Number of

Number of Complaints Weeks Complaints Weeks Complaints Weeks Complaints
in the Last 50 Weeks
1 8 18 7 35 3
2 10 19 10 36 5
3 6 20 11 37 2
4 7 21 8 38 4
5 5 22 7 39 3
6 7 23 8 40 3
7 9 24 6 41 4
8 8 25 7 42 2
9 7 26 7 43 4
10 9 27 5 44 5
11 10 28 8 45 5
12 7 29 6 46 3
13 8 30 7 47 2
14 11 31 5 48 5
15 10 32 5 49 4
16 9 33 4 50 4
17 8 34 4
768 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

For these data,

k
k = 50 and ∑ ci = 312
i =1

Thus,

312
c = = 6.24
50

so that using Equation (18.3) on page 767,

c ±3 c
= 6.24 ± 3 6.24
= 6.24 ± 7.494

Thus,

UCL = 6.24 + 7.494 = 13.734

LCL = 6.24 − 7.494 < 0

Therefore, the LCL does not exist.

Figure 18.9 displays the Minitab control chart for the complaint data of Table 18.4. Figure
18.9 does not indicate any points outside the control limits. However, there is a clear pattern to
the number of customer complaints over time. During the first half of the sequence almost all
the weeks had more than the mean number of complaints, and almost all the weeks in the sec-
ond half had less than the mean. There are more than 8 points in a row above the centerline
(weeks 6–23) and below the centerline (weeks 31–50), thus signaling a trend. This change,
which is an improvement, is due to a special cause of variation. The next step is to investigate
the process and determine the special cause that produced this pattern. When identified, you
then need to ensure that this becomes a permanent improvement, not a temporary phenomenon.
In other words, the source of the special-cause variation must become part of the permanent
ongoing process in order for the number of customer complaints not to slip back to the high
levels experienced in the first half of the data.

FIGURE 18.9
Minitab c Chart for
Hotel Complaints
 18.6: Control Chart for an Area of Opportunity—The c Chart 769

PROBLEMS FOR SECTION 18.6

Learning the Basics Items Items
PH Grade 18.11 The following data were collected on the Returned Returned
ASSIST number of nonconformities per unit for 10 time Day For Rework Day For Rework
periods:
7 6 16 4
Nonconformities Nonconformities 8 4 17 10
Time Per Unit Time Per Unit 9 8 18 9
10 6 19 6
1 7 6 5 11 5 20 5
2 3 7 3 12 12 21 8
3 6 8 5 13 5 22 6
4 3 9 2 14 8 23 7
5 4 10 0 15 3 24 9

a. Construct the appropriate control chart and determine a. Construct a c chart for the number of items per day that
the LCL and UCL. are returned for rework. Do you think that the process is
b. Are there any special causes of variation? in a state of statistical control?
b. Should the owner of the dry cleaning store take action to
PH Grade 18.12 The following data were collected on the investigate why 12 items were returned for rework on
ASSIST number of nonconformities per unit for 10 time day 12? Explain. Would your answer change if 20 items
periods: were returned for rework on day 12?
c. On the basis of the results in (a), what should the owner
Nonconformities Nonconformities
of the dry cleaning store do to reduce the number of
Time Per Unit Time Per Unit
items per day that are returned for rework?
1 25 6 15
2 11 7 12 PH Grade SELF 18.14 The branch manager of a sav-
3 10 8 10 ASSIST Test ings bank has recorded the number of
4 11 9 9 errors that each of 12 tellers has made during the
5 6 10 6 past year. The results were as follows: TELLER

a. Construct the appropriate control chart and determine Number Number

the LCL and UCL. Teller of Errors Teller of Errors
b. Are there any special causes of variation? Alice 4 Mitchell 6
Carl 7 Nora 3
Applying the Concepts Gina 12 Paul 5
You can solve problems 18.13 to 18.17 manually or Jane 6 Susan 4
by using Microsoft Excel, Minitab, or SPSS. Linda 2 Thomas 7
PH Grade 18.13 To improve service quality, the owner of a Marla 5 Vera 5
ASSIST dry cleaning business wants to study the number
of dry-cleaned items that are returned for rework a. Do you think the bank manager will single out Gina for
per day. Records were kept for a 4-week period (the store is any disciplinary action regarding her performance in the
open Monday through Saturday) with the results given last year?
below. DRYCLEAN b. Construct a c chart for the number of errors committed
by the 12 tellers. Is the number of errors in a state of sta-
Items Items tistical control?
Returned Returned c. Based on the c chart developed in (b), do you now think
Day For Rework Day For Rework that Gina should be singled out for disciplinary action
regarding her performance? Does your conclusion now
1 4 4 7
agree with what you expected the manager to do?
2 6 5 6
d. On the basis of the results in (b), what should the branch
3 3 6 8
manager do to reduce the number of errors?
770 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

PH Grade 18.15 Falls are one source of preventable hospi- does it appear that special causes of variation occurred
ASSIST tal injury. Although most patients who fall are not in these weeks?
hurt, a risk of serious injury is involved. The data e. Explain how the fire department can use these data to
in the file PTFALLS represent the number of patient falls per chart and monitor future weeks in real-time (i.e., on a
month over a 28-month period in a 19-bed AIDS unit at a week-to-week basis)?
major metropolitan hospital.
a. Construct a c chart for the number of patient falls per 18.17 Rochester-Electro-Medical Inc. is a manufactur-
month. Is the process of patient falls per month in a state ing company based in Tampa, Florida, that produces
of statistical control? medical products. Recently, management felt the need to
b. What effect would it have on your conclusions if you improve the safety of the workplace and began a safety
knew that the AIDS unit was started only one month sampling study. The following data represent the number
prior to the beginning of data collection? of unsafe acts observed by the company safety director
c. What other factors might contribute to special cause over an initial time period in which he made 20 tours of
variation in this problem? the plant. SAFETY

18.16 A member of the volunteer fire department for Number of Number of

Trenton, Ohio, decided to apply the control chart method- Tour UnsafeActs Tour Unsafe Acts
ology he learned in his business statistics class to data
collected by the fire department. He was interested in 1 10 11 2
determining whether weeks containing more than the 2 6 12 8
mean number of fire runs were due to inherent, chance- 3 6 13 7
cause variation, or if there were special causes of varia- 4 10 14 6
tion such as increased arson, severe drought, or holiday- 5 8 15 6
related activities. The data in the file FIRERUNS contain the 6 12 16 11
number of fire runs made per week (Sunday through 7 2 17 13
Saturday) during 2001. Week one begins on Sunday, 8 1 18 9
December 31, 2000, and week 53 ends on Saturday, 9 23 19 6
January 5, 2002. 10 3 20 9
Source: Extracted from The City of Trenton 2001 Annual Report, Source: H. Gitlow, A. R. Berkins, and M. He, “Safety Sampling:
Trenton, Ohio, February 21, 2002. A Case Study,” Quality Engineering, 14, 2002, 405–419.
a. What is the mean number of fire runs made per week?
b. Construct a c chart for the number of fire runs per week. a. Construct a c chart for the number of unsafe acts.
c. Is the process in a state of statistical control? b. Based on the results of (a), is the process in a state of
d. Weeks 15 and 41 experienced seven fire runs each. Are statistical control?
these large values explainable by common causes, or c. What should management do next to improve the process?

18.7 CONTROL CHARTS FOR THE RANGE AND THE MEAN

You use variables control charts to monitor and analyze a process when you have numerical
data. Common numerical variables include time, money, and weight. Because numeric vari-
ables provide more information than attribute data such as the proportion of nonconforming
items or the number of nonconformities, variables control charts are more sensitive in detect-
ing special-cause variation than the p chart or the c chart. Variables charts are typically used
in pairs. One chart monitors the dispersion (or variability) in a process, and the other moni-
tors the process mean. You must examine the chart that monitors dispersion first because if it
indicates the presence of out-of-control conditions, the interpretation of the chart for the
mean will be misleading. Although businesses currently use several alternative pairs of
charts (see references 7, 8, and 11), this text considers only the control charts for the range
and the mean.
 18.7: Control Charts for the Range and the Mean 771

The R Chart
You can use several different types of control charts to monitor the dispersion (i.e., variability)
in a numerically measured characteristic of interest. The simplest and most common is the
control chart for the range, the R chart. You use the range chart only when the sample size is
10 or less. If the sample size is greater than 10, a standard deviation chart is preferable.
Because sample sizes of five or less are typically used in many applications, the standard devi-
ation chart is not illustrated in this text. (For a discussion of standard deviation charts, see ref-
erences 7, 8, or 11.) The R chart enables you to determine whether the variability in a process
is in control or whether changes in the amount of variability are occurring over time. If the
process range is in control, then the amount of variation in the process is consistent over time,
and you can use the results of the R chart to develop the control limits for the mean.
To develop control limits for the range, you need an estimate of the mean range and the
standard deviation of the range. As shown in Equation (18.4), these control limits depend on
two constants, the d2 factor, which represents the relationship between the standard deviation
and the range for varying sample sizes, and the d3 factor, which represents the relationship
between the standard deviation and the standard error of the range for varying sample sizes.
Table E.13 contains values for these factors. Equation (18.4) defines the control limits for the R
chart.

CONTROL LIMITS FOR THE RANGE

d3
R ± 3R (18.4)
d2
d3
UCL = R + 3R
d2
d3
LCL = R − 3R
d2

k
∑ Ri
i =1
where R =
k

You can simplify the calculations in Equation (18.4) by using the D 3 factor, equal to
1 − 3(d3/d2), and the D4 factor, equal to 1 + 3(d3/d2), to express the control limits as shown
in Equations (18.5a) and (18.5b).

CALCULATING CONTROL LIMITS FOR THE RANGE

UCL = D4 R (18.5a)

LCL = D3R (18.5b)

To illustrate the R chart, return to the “Using Statistics” scenario concerning hotel service qual-
ity. During the Measure phase of the Six Sigma DMAIC model, the amount of time to deliver
luggage was operationally defined as the time from when the guest completes check-in proce-
dures to the time the luggage arrives in the guest’s room. During the Analyze phase of the Six
Sigma DMAIC model, data were recorded over a 4-week period. Subgroups of five deliveries
were selected from the evening shift on each day. Table 18.5 summarizes the results for all 28
days. HOTEL2
772 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

TABLE 18.5 Day Luggage Delivery Times in Minutes Mean Range

Luggage Delivery Times 1 6.7 11.7 9.7 7.5 7.8 8.68 5.0
and Subgroup Mean
2 7.6 11.4 9.0 8.4 9.2 9.12 3.8
and Range for 28 Days
3 9.5 8.9 9.9 8.7 10.7 9.54 2.0
4 9.8 13.2 6.9 9.3 9.4 9.72 6.3
5 11.0 9.9 11.3 11.6 8.5 10.46 3.1
6 8.3 8.4 9.7 9.8 7.1 8.66 2.7
7 9.4 9.3 8.2 7.1 6.1 8.02 3.3
8 11.2 9.8 10.5 9.0 9.7 10.04 2.2
9 10.0 10.7 9.0 8.2 11.0 9.78 2.8
10 8.6 5.8 8.7 9.5 11.4 8.80 5.6
11 10.7 8.6 9.1 10.9 8.6 9.58 2.3
12 10.8 8.3 10.6 10.3 10.0 10.00 2.5
13 9.5 10.5 7.0 8.6 10.1 9.14 3.5
14 12.9 8.9 8.1 9.0 7.6 9.30 5.3
15 7.8 9.0 12.2 9.1 11.7 9.96 4.4
16 11.1 9.9 8.8 5.5 9.5 8.96 5.6
17 9.2 9.7 12.3 8.1 8.5 9.56 4.2
18 9.0 8.1 10.2 9.7 8.4 9.08 2.1
19 9.9 10.1 8.9 9.6 7.1 9.12 3.0
20 10.7 9.8 10.2 8.0 10.2 9.78 2.7
21 9.0 10.0 9.6 10.6 9.0 9.64 1.6
22 10.7 9.8 9.4 7.0 8.9 9.16 3.7
23 10.2 10.5 9.5 12.2 9.1 10.30 3.1
24 10.0 11.1 9.5 8.8 9.9 9.86 2.3
25 9.6 8.8 11.4 12.2 9.3 10.26 3.4
26 8.2 7.9 8.4 9.5 9.2 8.64 1.6
27 7.1 11.1 10.8 11.0 10.2 10.04 4.0
28 11.1 6.6 12.0 11.5 9.7 10.18 5.4
Sums: 265.38 97.5

For the data in Table 18.5,

k

k ∑ Ri
∑ Ri = 97.5, and R = i =1 97.5
k = 28, = = 3.482
i =1
k 28

Using Equation (18.4) on page 771 and from Table E.13 for n = 5, d2 = 2.326, and d3 = 0.864,
 0.864 
3.482 ± 3( 3.482) 
 2.326 
= 3.482 ± 3.880
Thus,
UCL = 3.482 + 3.880 = 7.362
LCL = 3.482 − 3.880 < 0
Therefore, the LCL does not exist because it is impossible to get a negative range.
Alternatively, using Equation (18.5) on page 771, and D3 = 0 and D4 = 2.114, from Table E.13,
UCL = D4 R = ( 2.114 )(3.482 ) = 7.36

and the LCL does not exist.

 18.7: Control Charts for the Range and the Mean 773

Figure 18.10 displays the Microsoft Excel R chart for the luggage delivery times. Figure
18.10 does not indicate any individual ranges outside the control limits or any trends.

FIGURE 18.10
Microsoft Excel R Chart
for the Luggage
Delivery Times

The X Chart
Now that you have determined that the control chart shows that the range is in control, you con-
tinue by examining the control chart for the process mean, the X chart.
The control chart for X uses subgroups each of size n for k consecutive periods of time.
6R / d 2 is used to estimate To compute control limits for the mean, you need to compute the mean of the subgroup
the standard deviation of means (called X ) and the standard deviation of the mean (which is called the standard error
the population, and R / d 2 n of the mean σ X in Chapter 7). The estimate of the standard deviation of the mean is a func-
is used to estimate the tion of the d2 factor, which represents the relationship between the standard deviation and the
standard deviation of the range for varying sample sizes.6 Equations (18.6) and (18.7) define the control limits for the
mean. X chart.

CONTROL LIMITS FOR THE X CHART

R
X ±3 (18.6)
d2 n
R
UCL = X + 3
d2 n
R
LCL = X − 3
d2 n

k k
∑ Xi ∑ Ri
i =1 i =1
where X = R =
k k
X i = sample mean of n observations at time i
Ri = range of n observations at time i
k = number of subgroups

You can simplify the calculations in Equation (18.6) by utilizing the A2 factor, equal to
( )
3/ d2 n . Equations (18.7a) and (18.7b) show the simplified control limits.
774 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

CALCULATING CONTROL LIMITS FOR THE MEAN USING THE A2 FACTOR

UCL = X + A2 R (18.7a)

LCL = X − A2 R (18.7b)

From Table 18.5 on page 772,

k k
k = 28, ∑ X i = 265.38, and ∑ Ri = 97.5
i =1 i =1

so that

k k
∑ Xi 265.38
∑ Ri 97.5
i =1 i =1
X = = = 9.478 and R = = = 3.482
k 28 k 28

Using Equation (18.6) on page 773 and Table E.13 for n = 5, d2 = 2.326,

R
X ±3
d2 n
3.482
= 9.478 ± 3
( 2.326) 5
= 9.478 ± 2.008

Thus,

UCL = 9.478 + 2.008 = 11.486

LCL = 9.478 − 2.008 = 7.470

Alternatively, using Equations (18.7a) and (18.7b), and A2 = 0.577 from Table E.13,

UCL = 9.478 + ( 0.577 )(3.482 ) = 9.478 + 2.009 = 11.487

LCL = 9.478 − ( 0.577 )(3.482 ) = 9.478 − 2.009 = 7.469

These results are the same as those using Equation (18.6), except for rounding error.
Figure 18.11 displays the Microsoft Excel X chart for the luggage delivery time data.
Figure 18.12 presents Minitab R and X charts. Figures 18.11 and 18.12 do not reveal any
points outside the control limits or any trend. Although there is a considerable amount of vari-
ability among the 28 subgroup means, since both the R chart and the X chart are in control, the
luggage delivery process is in a state of statistical control. If you want to reduce the variation or
lower the mean delivery time, then you need to change the process.
 18.7: Control Charts for the Range and the Mean 775

FIGURE 18.11
Microsoft Excel X Chart
for the Luggage
Delivery Times

FIGURE 18.12
Minitab X and R Charts
for the Luggage
Delivery Times

PROBLEMS FOR SECTION 18.7

Learning the Basics a. Compute control limits for the range.
PH Grade 18.18 For subgroups of n = 4, what is the value b. Is there evidence of special cause variation in (a)?
ASSIST of: c. Compute control limits for the mean.
a. the d2 factor? d. the D4 factor? d. Is there evidence of special-cause variation in (c)?
b. the d3 factor? e. the A2 factor?
c. the D3 factor? Applying the Concepts
PH Grade 18.19 The following summary of data is for You should use Microsoft Excel, Minitab, or SPSS to
ASSIST subgroups of n = 4 for a 10-day period. solve problems 18.20–18.26.
SELF 18.20 The manager of a branch of a local bank
Day Mean Range Day Mean Range Test wants to study waiting times of customers for
teller service during the 12:00 noon to 1:00 P.M.
1 13.6 3.5 6 12.9 4.8
lunch hour. A subgroup of four customers is selected (one
2 14.3 4.1 7 17.3 4.5
at each 15-minute interval during the hour), and the time in
3 15.3 5.0 8 13.9 2.9
minutes is measured from the point each customer enters
4 12.6 2.8 9 12.6 3.8
the line to when he or she reaches the teller window. The
5 11.8 3.7 10 15.2 4.6
results over a 4-week period are in the data file: BANKTIME
776 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Day Time in Minutes Hour 1 2 3 4

1 7.2 8.4 7.9 4.9 9 20.30 20.42 20.68 19.60
2 5.6 8.7 3.3 4.2 10 20.19 20.00 20.23 20.59
3 5.5 7.3 3.2 6.0 11 19.66 21.24 20.35 20.34
4 4.4 8.0 5.4 7.4 12 20.30 20.11 19.64 20.29
5 9.7 4.6 4.8 5.8 13 19.83 19.75 20.62 20.60
6 8.3 8.9 9.1 6.2 14 20.27 20.88 20.62 20.40
7 4.7 6.6 5.3 5.8 15 19.98 19.02 20.34 20.34
8 8.8 5.5 8.4 6.9 16 20.46 19.97 20.32 20.83
9 5.7 4.7 4.1 4.6 17 19.74 21.02 19.62 19.90
10 1.7 4.0 3.0 5.2 18 19.85 19.26 19.88 20.20
11 2.6 3.9 5.2 4.8 19 20.77 20.58 19.73 19.48
12 4.6 2.7 6.3 3.4 20 20.21 20.82 20.01 19.93
13 4.9 6.2 7.8 8.7 21 20.30 20.09 20.03 20.13
14 7.1 6.3 8.2 5.5 22 20.48 21.06 20.13 20.42
15 7.1 5.8 6.9 7.0 23 20.60 19.74 20.52 19.42
16 6.7 6.9 7.0 9.4 24 20.20 20.08 20.32 19.51
17 5.5 6.3 3.2 4.9 25 19.66 19.67 20.26 20.41
18 4.9 5.1 3.2 7.6 26 20.72 20.58 20.71 19.99
19 7.2 8.0 4.1 5.9 27 19.77 19.40 20.49 19.83
20 6.1 3.4 7.2 5.9 28 19.99 19.65 19.41 19.58
29 19.44 20.15 20.14 20.76
a. Construct control charts for the mean and the range. 30 20.03 19.96 19.86 19.91
b. Is the process in control?
Source: Extracted from Susan K. Humphrey, and Timothy C. Krehbiel,
18.21 The manager of a warehouse for a local telephone “Managing Process Capability,” The Mid-American Journal of
company is involved in a process that receives expensive cir- Business, 14, Fall 1999, 7–12.
cuit boards and returns them to central stock so that they can
be reused at a later date. Speedy processing of these circuit a. Construct a control chart for the range.
boards is critical in providing good service to customers and b. Construct a control chart for the mean.
reducing capital expenditures. The data in the file WAREHSE c. Is the process in control?
represent the number of circuit boards processed per day by 18.23 The data in the following table are the tensile
each of a subgroup of five employees over a 30-day period. strengths of bolts of cloth. The data were collected in sub-
a. Construct control charts for the mean and the range. groups of 3 bolts of cloth over a 25-hour period. TENSILE
b. Is the process in control?
Hour 1 2 3 Hour 1 2 3
18.22 An article in the Mid-American Journal of Business
presents an analysis for a springwater bottling operation. 1 15.06 14.62 15.10 14 16.29 14.61 15.67
One of the characteristics of interest is the amount of mag- 2 17.58 15.75 16.72 15 15.84 12.16 15.40
nesium, measured in parts per million (ppm), in the water. 3 13.83 14.83 15.61 16 15.12 15.60 13.83
The data in the following table represent the magnesium 4 17.19 15.75 15.42 17 18.48 16.07 16.31
levels from 30 subgroups of 4 bottles collected over a 5 14.56 15.37 15.67 18 17.55 14.73 16.95
30-hour period. SPWATER 6 14.82 17.25 15.73 19 13.57 17.55 15.81
7 17.92 14.76 14.40 20 16.23 16.92 16.45
Hour 1 2 3 4 8 16.53 14.52 17.31 21 14.60 16.83 15.34
9 13.83 14.53 15.32 22 16.73 18.60 16.76
1 19.91 19.62 19.15 19.85
10 16.45 13.85 16.32 23 18.03 14.55 13.87
2 20.46 20.44 20.34 19.61
11 15.20 14.61 18.45 24 16.61 16.45 16.95
3 20.25 19.73 19.98 20.32
12 14.49 16.15 17.80 25 15.86 17.00 18.28
4 20.39 19.43 20.36 19.85
13 15.89 15.04 16.67
5 20.02 20.02 20.13 20.34
6 19.89 19.77 20.92 20.09
a. Construct a control chart for the range.
7 19.89 20.45 19.44 19.95
b. Construct a control chart for the mean.
8 20.08 20.13 20.11 19.32
c. Is the process in control?
 18.8: Process Capability 777

PH Grade 18.24 The director of radiology at a large met-

Sample Weight (in grams)
ASSIST ropolitan hospital is concerned about scheduling
in the radiology facilities. On a typical day 250 9 5.53 5.25 5.67 5.53
patients are transported to the radiology department for 10 5.41 5.55 5.51 5.53
treatment or diagnostic procedures. If patients do not reach 11 5.55 5.58 5.58 5.56
the radiology unit at their scheduled time, backups occur 12 5.58 5.36 5.45 5.53
and other patients experience delays. The time it takes to 13 5.63 5.75 5.46 5.54
transport patients to the radiology unit is operationally 14 5.48 5.44 5.45 5.60
defined as the time between when the transporter is 15 5.49 5.57 5.43 5.36
assigned to the patient and the time the patient arrives at 16 5.54 5.62 5.66 5.59
the radiology unit. A sample of n = 4 patients was selected 17 5.46 5.46 5.38 5.49
each day for 20 days and the time to transport each patient 18 5.72 5.36 5.59 5.25
(in minutes) was determined, with the results in the data 19 5.58 5.50 5.36 5.40
file TRANSPORT. 20 5.43 5.51 5.37 5.32
a. Construct control charts for the mean and the range. 21 5.59 5.58 5.60 5.46
b. Is the process in control? 22 5.42 5.41 5.40 5.69
23 5.64 5.59 5.42 5.56
18.25 A filling machine for a tea bag manufacturer pro-
24 5.62 5.38 5.75 5.47
duces approximately 170 tea bags per minute. The process
25 5.51 5.54 5.73 5.77
manager monitors the weight of the tea placed in individ-
ual bags. A subgroup of n = 4 teabags is taken every 15 a. What are some of the sources of common-cause varia-
minutes for 25 consecutive time periods. The results are tion that might be present in this process?
given below: TEA3 b. What problems might occur that would result in special
causes of variation?
Sample Weight (in grams)
c. Construct control charts for the range and the mean.
1 5.32 5.77 5.50 5.61 d. Is the process in control?
2 5.63 5.44 5.54 5.40
18.26 A manufacturing company makes brackets for book-
3 5.56 5.40 5.67 5.57
shelves. The brackets provide critical structural support and
4 5.32 5.45 5.50 5.42
must have a 90-degree bend ±1 degree. Measurements of the
5 5.45 5.53 5.46 5.47
bend of the brackets were taken at 18 different times. Five
6 5.29 5.42 5.50 5.44
brackets were sampled at each time ANGLE.
7 5.57 5.40 5.52 5.54
a. Construct control charts for the range and the mean.
8 5.44 5.61 5.49 5.58
b. Is the process in control?

18.8 PROCESS CAPABILITY

Often it is necessary to analyze the amount of common-cause variation present in an in-control
process. Is the common-cause variation small enough to satisfy customers with the product or
service? Or is the common-cause variation so large that there are too many dissatisfied cus-
tomers and a process change is needed?
Analyzing the capability of a process is a way to answer these questions. There are many
methods available to analyze and report process capability (see reference 2). This section
begins with a relatively simple method to estimate the percentage of products or services that
will satisfy the customer. Later in the section, the use of capability indexes is introduced.

Customer Satisfaction and Specification Limits

Quality is defined by the customer. A customer who believes that a product or service has
met or exceeded his or her expectations will be satisfied. The management of a company
must listen to the customer and translate the customer’s needs and expectations into easily
measured critical-to-quality (CTQs) variables. Management then sets specification limits for
these CTQs.
778 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Specification limits are technical requirements set by management in response to

customers’ needs and expectations. The upper specification limit (USL) is the largest
value a CTQ can have and still conform to customer expectations. Likewise, the lower
specification limit (LSL) is the smallest value a CTQ can have and still conform to
customer expectations.

For example, a soap manufacturer understands that customers expect their soap to produce
a certain amount of lather. The customer can become dissatisfied if the soap produces too much
or too little lather. Product engineers know that the level of free fatty acids in the soap controls
the amount of lather. Thus, the process manager, with input from the product engineers, sets
both a USL and an LSL for the amount of free fatty acids in the soap.
As an example of a case in which only a single specification limit is involved, consider the
“Using Statistics” scenario concerning hotel service quality on page 752. Since customers want
their bags delivered as quickly as possible, hotel management sets a USL for the time required
for delivery. In this case, there is no LSL. As you can see in both the luggage delivery time and
soap examples, specification limits are customer-driven requirements placed on a product or a
service. If a process consistently meets these requirements, the process is capable of satisfying
the customer.

Process capability is the ability of a process to consistently meet specified customer-

driven requirements.

One way to analyze the capability of a process is to estimate the percentage of products or
services that are within specifications. To do this, you must have an in-control process
because an out-of-control process does not allow you to predict its capability. If you are deal-
ing with an out-of-control process, you must first identify and eliminate the special causes of
variation before performing a capability analysis. Out-of-control processes are unpredictable,
and therefore, you cannot conclude that such processes are capable of meeting specifications
or satisfying customer expectations. In order to estimate the percentage of product or service
within specifications, first, you must estimate the mean and standard deviation of the popula-
tion of all X values, the CTQ variable of interest for the product or service. The estimate for
the mean of the population is X , the mean of all the sample means [see Equation (18.6) on
page 773]. The estimate of the standard deviation of the population is R divided by d2. You
can use the X and R from in-control X and R charts, respectively. You need to find the
correct d2 value in Table E.13.
In this text, the population of X values is assumed to be approximately normally distrib-
uted. (If your data are not approximately normally distributed, see reference 2 for an alter-
native approach.) Assuming that the process is in control and X is approximately normal,
you can use Equation (18.8) to estimate the probability that a process outcome is within
specifications.

ESTIMATING THE CAPABILITY OF A PROCESS

For a CTQ variable with a lower specification limit and an upper specification limit:

P ( an outcome will be within specifications ) = P ( LSL < X < USL) (18.8a)

 LSL − X USL − X 
= P < Z < 
R R
 
 d2 d2 
 18.8: Process Capability 779

For a CTQ variable with only an upper specification limit:

P ( an outcome will be within specifications ) = P ( X < USL) (18.8b)

 USL − X 
= P Z < 
R
 
 d2 

For a CTQ variable with only a lower specification limit:

P ( an outcome will be within specifications ) = P ( LSL < X ) (18.8c)

 LSL − X 
= P < Z
R
 
 d2 

where Z is a standardized normal random variable

In section 18.7, you determined that the luggage delivery process was in control. Suppose that
the hotel management has instituted a policy that 99% of all luggage deliveries must be com-
pleted in 14 minutes or less. From the summary computations on page 774

n=5 X = 9.478 R = 3.482 and from Table E.13, d2 = 2.326

Using Equation (18.8b),

P ( delivery is made within specifications ) = P ( X < 14 )

 14 – 9.478 
= P Z <
 3.482 
 
 2.326 
= P ( Z < 3.02 )
Using Table E.2, P ( Z < 3.02 ) = 0.99874

Thus, you estimate that 99.874% of the luggage deliveries will be made within the specified
time. The process is capable of meeting the 99% goal set forth by the hotel management.

Capability Indexes
A common approach in business is to use capability indexes to report the capability of a
process. A capability index is an aggregate measure of a process’s ability to meet specification
limits. The larger the value of a capability index, the more capable a process is of meeting cus-
tomer requirements. Equation (18.9) defines Cp , the most commonly used index.

THE CP INDEX
USL − LSL
Cp = (18.9)
6( R / d 2 )
specification spread
=
process spread
780 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

The numerator in Equation (18.9) represents the distance between the upper and lower specifi-
cation limits, referred to as the specification spread. The denominator, 6( R / d2 ) , represents a 6
standard deviation spread in the data (the mean ±3 standard deviations), referred to as the
process spread. (Recall from Chapter 6 that approximately 99.73% of the values from a normal
distribution fall in the interval from the mean ±3 standard deviations.) You want the process
spread to be small in comparison to the specification spread in order for the vast majority of the
process output to fall within the specification limits. Therefore, the larger the value of Cp, the
better is the capability of the process.
Cp is a measure of process potential, not of actual performance, because it does not con-
sider the current process mean. A value of 1 indicates that if the process mean could be cen-
tered (i.e., equal to the halfway point between USL and LSL), then approximately 99.73% of
the values would be inside the specification limits. A Cp value greater than 1 indicates that a
process has the potential of having more than 99.73% of its outcomes within specifications.
A Cp value less than 1 indicates that the process is not very capable of meeting customer
requirements, for even if the process is perfectly centered, less than 99.73% of the process
outcomes will be within specifications. Historically, many companies required a Cp greater
than or equal to 1. Now that the global economy has become more quality conscious, many
companies are requiring a Cp as large as 1.33, 1.5, and for companies adopting Six Sigma
management, 2.0.
To illustrate the calculation and interpretation of the Cp index, suppose a soft-drink
producer bottles its beverage into 12-ounce bottles. Each hour four bottles are selected and
control charts for the range and the mean are constructed. At the end of 24 hours the capa-
bility of the process is studied. The lower specification limit is 11.82 ounces and the upper
specification limit is 12.18 ounces. Suppose that the control charts indicate that the
process is in control and the following summary calculations were recorded on the control
charts:

n= 4 X = 12.02 R = 0.10

To calculate the Cp index assuming that the data are normally distributed, from Table E.13,
d2 = 2.059 for n = 4. Using Equation (18.9) on page 779,

USL − LSL
Cp =
6( R / d2 )
12.18 − 11.82
= = 1.24
6( 0.10 / 2.059 )

Because the Cp index is greater than 1, the bottling process has the potential to fill more than
99.73% of the bottles within the specification limits.
In summary, the Cp index is an aggregate measure of process potential. The larger the value
of Cp, the more potential the process has of satisfying the customer. In other words, a large Cp
indicates that the current amount of common-cause variation is small enough to consistently
produce items within specifications. For a process to reach its full potential, the process mean
needs to be at or near the center of the specification limits. The following introduces capability
indexes that measure actual process performance.

CPL, CPU, and Cpk

To measure the capability of a process in terms of actual process performance, the most
common indexes are CPL, CPU, and Cpk. Equation (18.10) defines the capability indexes CPL
and CPU.
 18.8: Process Capability 781

CPL AND CPU

X − LSL
CPL = (18.10a)
3( R / d2 )

USL − X
CPU = (18.10b)
3( R / d2 )

Since the process mean is used in the calculation of the CPL and CPU indices, the value of the
index gives a measure of process performance—unlike Cp, which measures only potential. A
value of CPL (or CPU) equal to 1.0 indicates that the process mean is 3 standard deviations
away from the lower specification limit (or upper specification limit). For CTQ variables with
only an LSL, the CPL measures the process performance. For CTQ variables with only a USL,
the CPU measures the process performance. In either case, the larger the value of the index, the
better the capability of the process.
In the “Using Statistics” scenario, the Beachcomber Hotel has a policy that luggage deliver-
ies are to be made in 14 minutes or less. Thus, the CTQ variable delivery time has an upper spec-
ification limit of 14, and there is no lower specification limit. Because you previously deter-
mined that the luggage delivery process was in control, you can now calculate the CPU. From
the summary computations on page 774,
X = 9.478 and R = 3.482

And, from Table E.13, d2 = 2.326.

Then, using Equation (18.10b),
USL − X 14 − 9.478
CPU = = = 1.01
3( R / d2 ) 3( 3.482 / 2.326 )

The capability index for the luggage delivery CTQ variable is 1.01. Because this value is
slightly more than 1, the upper specification limit is slightly more than 3 standard deviations
above the mean. To increase CPU even farther above 1.00 and therefore increase customer sat-
isfaction, you need to investigate changes in the luggage delivery process. To study a process
that has a CPL and a CPU, see the bottling process in Example 18.2.

EXAMPLE 18.2 CALCULATING CPL AND CPU FOR THE BOTTLING PROCESS
In the soft-drink bottle filling process described on page 780, the following information was
presented:
n= 4 X = 12.02 R = 0.10 LSL = 11.82 USL = 12.18 d2 = 2.059

Calculate the CPL and CPU for these data.

SOLUTION You compute the capability indexes using Equations (18.10a) and (18.10b):

X − LSL
CPL =
3( R / d2 )
12.02 − 11.82
= = 1.37
3( 0.10 / 2.059 )
USL − X
CPU =
3( R / d2 )
12.18 − 12.02
= = 1.10
3( 0.10 / 2.059 )
782 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Both the CPL and CPU are greater than 1, indicating that the process mean is more than 3 stan-
dard deviations away from both the LSL and USL. Since the CPU is less than the CPL, you
know that the mean is closer to the USL than the LSL.

The capability index Cpk measures actual process performance for quality characteristics with
two-sided specification limits. Cpk is equal to the value of either the CPL or CPU, whichever is
smallest:

Cpk
Cpk = MIN[CPL, CPU] (18.11)

A value of 1 for Cpk indicates that the process mean is 3 standard deviations away from the
closest specification limit. If the characteristic is normally distributed, then a value of 1 indi-
cates that at least 99.73% of the current output is within specifications. Like all capability
indices, the larger the value of Cpk the better. Example 18.3 illustrates the use of the Cpk index.

EXAMPLE 18.3 CALCULATING Cpk FOR THE BOTTLING PROCESS

The soft-drink producer in Example 18.2 requires the bottle filling process to have a Cpk
greater than or equal to 1. Calculate the Cpk index.
SOLUTION In Example 18.2, CPL = 1.37 and CPU = 1.10. Using Equation (18.11):
C pk = MIN [ CPL, CPU ]
= MIN [1.37, 1.10 ] = 1.10

The Cpk index is greater than 1, indicating that the actual process performance exceeds the
company’s requirement. More than 99.73% of the bottles contain between 11.82 and 12.18 ounces.

PROBLEMS FOR SECTION 18.8

Learning the Basics Applying the Concepts
PH Grade 18.27 For an in-control process with subgroup SELF 18.30 Referring to the data of problem 18.22
ASSIST data n = 4, X = 20 and R = 2 find the esti- Test SPWATER on page 776, the researchers stated,
mate of: “Some of the benefits of a capable process are
a. the population mean of all X values. increased customer satisfaction, increased operating effi-
b. the population standard deviation of all X values. ciencies, and reduced costs.” To illustrate this point, the
authors presented a capability analysis for a springwater bot-
PH Grade 18.28 For an in-control process with subgroup tling operation. One of the critical-to-quality (CTQ) vari-
ASSIST data n = 3, X = 100 and R = 3.386 calculate the ables is the amount of magnesium, measured in parts per
percentage of outcomes within specifications if: million (ppm), in the water. The LSL and USL for the level of
a. LSL = 98 and USL = 102. magnesium in a bottle are 18 ppm and 22 ppm, respectively.
b. LSL = 93 and USL = 107.5. a. Estimate the percentage of bottles within specifications.
c. LSL = 93.8 and there is no USL. b. Calculate Cp, CPL, CPU, and Cpk.
d. USL = 110 and there is no LSL.
18.31 Refer to the data in problem 18.23 on page 776
18.29 For an in-control process with subgroup data n = 3, concerning the tensile strengths of bolts of cloth. There is
X = 100, and R = 3.386, calculate the Cp, CPL, CPU, no upper specification limit for tensile strength, and the
and Cpk if: lower specification limit is 13. TENSILE
a. LSL = 98 and USL = 102. a. Estimate the percentage of bolts within specifications.
b. LSL = 93 and USL = 107.5. b. Calculate Cp and CPL.
 Key Formulas 783

PH Grade 18.32 Refer to problem 18.25 on page 777 con- 18.33 Refer to problem 18.20 on page 775 concerning
ASSIST cerning a filling machine for a tea bag manufac- waiting time for customers at a bank. (Note: Ignore the fact
turer. In this problem you should have concluded that in this problem, X is out of control). Suppose manage-
that the process is in control. The label weight for this prod- ment has set an upper specification limit of 5 minutes on
uct is 5.5 grams, the lower specification limit is 5.2 grams, waiting time and that at least 99% of the waiting times
and the upper specification limit is 5.8 grams. Company must be less than 5 minutes in order for the process to be
policy states that at least 99% of the tea bags produced considered capable. BANKTIME
must be inside the specifications in order for the process to a. Estimate the percentage of the waiting times that are
be considered capable. TEA3 inside the specification limits. Is the process capable of
a. Estimate the percentage of the tea bags that are inside meeting the company policy?
the specification limits. Is the process capable of meet- b. If management implemented a new policy stating that
ing the company policy? 99.7% of all waiting times are required to be within
b. If management implemented a new policy stating that specifications, is this process capable of reaching that
99.7% of all tea bags are required to be within specifica- goal? Explain.
tions, is this process capable of reaching that goal? Explain.

SUMMARY
This chapter has introduced you to quality and productiv- ity of a process by estimating the percentage of items
ity including total quality management, Deming’s 14 within specifications and by calculating capability
points, and Six Sigma management. You have learned indices. By applying these concepts to the services pro-
how to use several different types of control charts to dis- vided by the Beachcomber Hotel, you learned how a
tinguish between common causes and special causes of manager can identify problems and continually improve
variation. You have learned how to measure the capabil- service quality.

KEY FORMULAS
Constructing Control Limits Control Limits for the Range
Process mean ±3 standard deviations (18.1) d3
R ± 3R (18.4)
Upper control limit (UCL) = d2
process mean +3 standard deviations d3
UCL = R + 3R
Lower control limit (LCL) = d2
process mean −3 standard deviations
d3
LCL = R − 3R
Control Limits for the p Chart d2

p(1 − p ) Calculating Control Limits for the Range

p±3 (18.2)
n UCL = D4 R (18.5a)
p(1 − p ) LCL = D3R (18.5b)
UCL = p + 3
n
Control Limits for the X Chart
p(1 − p )
LCL = p − 3 X ±3
R
(18.6)
n
d2 n
Control Limits for the c Chart R
(18.3) UCL = X + 3
c ±3 c d2 n
UCL = c + 3 c R
LCL = X − 3
LCL = c − 3 c d2 n
784 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

Calculating Control Limits for the Mean Using For a CTQ variable with only a lower specification limit:
the A2 Factor P ( an outcome will be within specifications ) (18.8c)
UCL = X + A2 R (18.7a) = P ( LSL < X )
 LSL − X 
LCL = X − A2 R (18.7b) = P < Z
R
 
 d2 
Estimating the Capability of a Process
The Cp Index
For a CTQ variable with a lower specification limit and an
upper specification limit: USL − LSL
Cp = (18.9)
P ( an outcome will be within specifications ) (18.8a) 6( R / d 2 )
= P ( LSL < X < USL)
specification spread
 LSL − X =
USL − X  process spread
= P < Z < 
R R
  CPL and CPU
 d2 d2 
X − LSL
For a CTQ variable with only an upper specification limit: CPL = (18.10a)
3( R / d2 )
P ( an outcome will be within specifications ) (18.8b)
= P ( X < USL) USL − X
CPU = (18.10b)
 USL − X  3( R / d2 )
= P Z < 
R Cpk
 
 d2  Cpk = MIN[CPL, CPU] (18.11)

KEY TERMS
A2 factor 773 D4 factor 771 Six Sigma Management 755
area of opportunity 766 Deming’s 14 points special causes of variation 756
assignable causes of variation 756 for management 753 specification limits 778
attribute charts 758 DMAIC model 755 state of statistical control 757
c chart 766 in-control process 757 subgroup 756
capability index 779 lower control limit (LCL) 756 tampering 756
chance causes of variation 756 lower specification limit (LSL) 778 total quality management (TQM) 753
common causes of variation 756 out-of-control process 757 upper control limit (UCL) 756
control chart 756 p chart 758 upper specification limit (USL) 778
critical-to-quality (CTQ) 755 process capability 778 variables control charts 770
d2 factor 771 R chart 771 X chart 773
d3 factor 771 Red Bead experiment 764
D3 factor 771 Shewhart-Deming cycle 753

CHAPTER REVIEW PROBLEMS

Checking Your Understanding 18.36 What should you do to improve a process when
18.34 What is the difference between common-cause only common causes of variation are present?
variation and special-cause variation? 18.37 Under what circumstances do you use a p chart?
18.35 What should you do to improve a process when 18.38 What is the difference between attribute control
special causes of variation are present? charts and variables control charts?
 Chapter Review Problems 785

18.39 Why are the X and R charts used together? Cases Sold Before the Promotion
18.40 What principles did you learn from the red bead Day Total Rudybird Day Total Rudybird
experiment? 1 154 35 16 177 56
18.41 How do you decide if you should use a p chart or a 2 153 43 17 143 43
c chart? 3 200 44 18 200 69
4 197 56 19 134 38
18.42 What is the difference between process potential 5 194 54 20 192 47
and process performance? 6 172 38 21 155 45
7 190 43 22 135 36
18.43 A company requires a Cpk value of 1 or larger. If a 8 209 62 23 189 55
process has a Cp = 1.5 and a Cpk = 0.8, what changes 9 173 53 24 184 44
should you make to the process? 10 171 39 25 170 47
18.44 Why is a capability analysis not performed on out- 11 173 44 26 178 48
of-control processes? 12 168 37 27 167 42
13 184 45 28 204 71
14 211 58 29 183 64
Applying the Concepts 15 179 35 30 169 43
You should use Microsoft Excel, Minitab, or SPSS to
solve problems 18.45–18.53. Cases Sold After the Promotion
18.45 A producer of cat food constructed control charts Day Total Rudybird
and analyzed several quality characteristics. One character-
istic of interest is the weight of the filled cans. The lower 31 201 92
specification limit for weight is 2.95 pounds. The data file 32 177 76
CATFOOD contains the weights of five cans tested every fif- 33 205 85
teen minutes during a day’s production. 34 199 90
a. Construct a control chart for the range. 35 187 77
b. Construct a control chart for the mean. 36 168 79
c. Is the process in control? 37 198 97
d. If the process is in control, estimate the percentage of Source: Extracted from Charles T. Crespy, Timothy C. Krehbiel, and
the cans whose weight is inside the specification limits. James M. Stearns, “Integrating Analytic Methods into Marketing
e. If the process is in control, calculate CPL. Research Education: Statistical Control Charts as an Example,”
f. If the manufacturer requires that 99.7% of all cans be Marketing Education Review, 5, Spring 1995, 11–23.
within the specification limits, comment on the capability a. Construct a p chart using data from the first 30 days
of the process based on your calculations in (d) and (e). (prior to the promotion) to monitor the market share for
18.46 Researchers at Miami University in Oxford, Ohio, RudyBird Diskettes.
investigated the use of p charts to monitor the market b. Is the market share for RudyBird in control before the
share of a product and to document the effectiveness of start of the in-store promotion?
marketing promotions. Market share is defined as the c. On your control chart, extend the control limits gener-
company’s proportion of the total number of products sold ated in (b) and plot the proportions for days 31 through
in a category. If a p chart based on a company’s market 37. What effect, if any, did the in-store promotion have
share indicates an in-control process, then their share in on RudyBird’s market share?
the marketplace is deemed to be stable and consistent over 18.47 The manufacturer of “Boston” and “Vermont”
time. In the example given in the article, the RudyBird asphalt shingles constructed control charts and analyzed
Diskette Company collected daily sales data from a several quality characteristics. One characteristic of inter-
nationwide retail audit service. The first 30 days of data in est is the strength of the sealant on the shingle. During
the accompanying table indicate the total number of cases each day of production, three shingles are tested for their
of computer diskettes sold and the number of RudyBird sealant strength. (Thus, a subgroup is operationally
diskettes sold. The final 7 days of data were taken after defined as one day of production, and the sample size for
RudyBird launched a major in-store promotion. A control each subgroup is 3.) Separate pieces are cut from the upper
chart was used to see if the in-store promotion would and lower portions of a shingle, and then reassembled to
result in special-cause variation in the marketplace. simulate shingles on a roof. A timed heating process is
RUDYBIRD used to simulate the sealing process. The sealed shingle
786 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

pieces are pulled apart, and the amount of force (in pounds) PH Grade 18.49 The funds-transfer department of a bank
required to break the sealant bond is measured and recorded. ASSIST is concerned with turnaround time for investiga-
This variable is called the sealant strength. The lower and tions of funds-transfer payments. A payment may
upper specification limits for sealant strength are 1.0 and 1.5 involve the bank as a remitter of funds, a beneficiary of
pounds, respectively. The data file SEALANT contains sealant funds, or an intermediary in the payment. An investigation
strength measurements on 25 days of production for is initiated by a payment inquiry or query by a party
“Boston” shingles and 19 days for “Vermont” shingles. involved in the payment or any department affected by the
For the “Boston” shingles: flow of funds. When a query is received, an investigator
a. Construct a control chart for the range. reconstructs the transaction trail of the payment and veri-
b. Construct a control chart for the mean. fies that the information is correct and the proper payment
c. Is the process in control? is transmitted. The investigator then reports the results of
d. If the process is in control, estimate the percentage of the investigation and the transaction is considered closed.
the shingles whose sealant strength is inside the specifi- It is important that investigations are closed rapidly, prefer-
cation limits. ably within the same day. The number of new investiga-
e. If the process is in control, calculate Cp, CPL, CPU, tions and the number and proportion closed on the same
and Cpk. day that the inquiry was made are in the file FUNDTRAN.
f. If the manufacturer requires that 99.7% of all shingles be a. Construct a control chart for these data.
within the specification limits, comment on the capability b. Is the process in a state of statistical control? Explain.
of the process based on your calculations in (d) and (e). c. Based on the results of (a) and (b), what should manage-
g. Repeat (a) through (f) using the 19 production days for ment do next to improve the process?
“Vermont” shingles.
PH Grade 18.50 A branch manager of a brokerage com-
PH Grade 18.48 A professional basketball player has ASSIST pany is concerned with the number of undesirable
ASSIST embarked on a program to study his ability to trades made by her sales staff. A trade is consid-
shoot foul shots. On each day in which a game is ered undesirable if there is an error on the trade ticket.
not scheduled, he intends to shoot 100 foul shots. He main- Trades with errors are canceled and resubmitted. The cost
tains records over a period of 40 days of practice, with the of correcting errors is billed to the brokerage company. The
following results: FOULSPC branch manager wants to know whether the proportion of
Foul Foul Foul undesirable trades is in a state of statistical control so she
Shots Shots Shots can plan the next step in a quality improvement process.
Day Made Day Made Day Made Data were collected for a 30-day period with the following
results: TRADE
1 73 15 73 29 76
2 75 16 76 30 80 Undesirable Total Undesirable Total
3 69 17 69 31 78 Day Trades Trades Day Trades Trades
4 72 18 68 32 83
5 77 19 72 33 84 1 2 74 16 3 54
6 71 20 70 34 81 2 12 85 17 12 74
7 68 21 64 35 86 3 13 114 18 11 103
8 70 22 67 36 85 4 33 136 19 11 100
9 67 23 72 37 86 5 5 97 20 14 88
10 74 24 70 38 87 6 20 115 21 4 58
11 75 25 74 39 85 7 17 108 22 10 69
12 72 26 76 40 85 8 10 76 23 19 135
13 70 27 75 9 8 69 24 1 67
14 74 28 78 10 18 98 25 11 77
11 3 104 26 12 88
a. Construct a p chart for the proportion of successful foul 12 12 98 27 4 66
shots. Do you think that the player’s foul-shooting 13 15 105 28 11 72
process is in statistical control? If not, why not? 14 6 98 29 13 118
b. What if you were told that the player used a different 15 21 204 30 15 138
method of shooting foul shots for the last twenty days?
How might this information change your conclusions a. Construct a control chart for these data.
in (a)? b. Is the process in control? Explain.
c. If you knew the information in (b) prior to doing (a), c. Based on the results of (a) and (b), what should the man-
how might you do the analysis differently? ager do next to improve the process?
 Case Study 787

18.51 As chief operating officer of a local community 18.53 (Class Project) Use the table of random numbers
hospital, you have just returned from a 3-day seminar on (Table E.1) to simulate the selection of different colored
quality and productivity. It is your intention to implement balls from an urn as follows:
many of the ideas that you learned at the seminar. You have 1. Start in the row corresponding to the day of the month
decided to maintain control charts for the upcoming month you were born plus the year in which you were born. For
for the following variables: number of daily admissions, example, if you were born October 15, 1976, you would
proportion of rework in the laboratory (based on 1,000 daily start in row 15 + 76 = 91. If your total exceeds 100, sub-
samples), and time (in hours) between receipt of a specimen tract 100 from the total.
at the laboratory and completion of the work (based on a 2. Select two-digit random numbers.
subgroup of 10 specimens per day). The data collected are 3. If you select a random number from 00 to 94, consider
summarized in the file HOSPADM. You are to make a presen- the ball to be white; if the random number is from 95 to
tation to the chief executive officer of the hospital and the 99, consider the ball to be red.
board of directors. Prepare a report that summarizes the Each student is to select 100 such two-digit random num-
conclusions drawn from analyzing control charts for these bers and report the number of “red balls” in the sample.
variables. In addition, recommend additional variables to Construct a control chart for the proportion of red balls.
measure and monitor using control charts. What conclusions can you draw about the system of select-
ing red balls? Are all the students part of the system? Is any-
18.52 On each morning for a period of 4 weeks, record one outside the system? If so, what explanation can you
your pulse rate (in beats per minute) just after you get out give for someone who has too many red balls? If a bonus
of bed and also before you go to sleep at night. Set up X were paid to the top 10% of the students (the 10% with the
and R charts and determine whether your pulse rate is in a fewest red balls), what effect would that have on the rest of
state of statistical control. Explain. the students? Discuss.

CASE STUDY
THE HARNSWELL SEWING MACHINE COMPANY CASE
Phase 1 starting with a trial project in the machine parts area. As
For almost 50 years, the Harnswell Sewing Machine she is walking to Mr. Harnswell’s office for the meeting,
Company has manufactured industrial sewing machines. she has second thoughts about whether this is such a good
The company specializes in automated machines called idea. After all, just last month Mr. Harnswell told her,
pattern tackers that sew repetitive patterns on such mass- “Why do you need to go to graduate school for your mas-
produced products as shoes, garments, and seat belts. Aside ter’s degree in business? That is a waste of your time and
from the sales of machines, the company sells machine will not be of any value to the Harnswell Company. All
parts. Because the company’s products have a reputation those professors are just up in their ivory towers and don’t
for being superior, Harnswell is able to command a price know a thing about running a business like I do.”
premium for its product line. As she enters his office, Mr. Harnswell, ever courteous
Recently, the production manager, Natalie York, pur- to her, invites Natalie to sit down across from him. “Well,
chased several books relating to quality at a local book- what do you have on your mind this morning?” Mr.
store. After reading them, she considered the feasibility of Harnswell asks her in an inquisitive tone. She begins by
beginning some type of quality program at the company. At starting to talk about the books that she has just completed
the current time, the company has no formal quality pro- reading and about how she has some interesting ideas for
gram. Parts are 100% inspected at the time of shipping to a making production even better than it is now and improving
customer or installation in a machine, yet Natalie has profits. Before she can finish, Mr. Harnswell has started to
always wondered why inventory of certain parts (in partic- answer. “Look, my dear young lady,” he says, “everything
ular the half-inch cam roller) invariably falls short before a has been fine since I started this company in 1955. I have
full year lapses, even though 7,000 pieces have been pro- built this company up from nothing to one that employs
duced for a demand of 5,000 pieces per year. more than 100 people. Why do you want to make waves?
After a great deal of reflection and with some appre- Remember, if it ain’t broke, don’t fix it.” With that he ush-
hension, Natalie has decided that she will approach John ers her from his office with the admonishment of, “What
Harnswell, the owner of the company, about the possibility am I going to do with you if you keep coming up with these
of beginning a program to improve quality in the company, ridiculous ideas?”
788 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

EXERCISES Cam Roller

HS.1 Based upon what you have read, which of Deming’s
14 points of management are most lacking in the Batch 1 2 3 4 5
Harnswell Sewing Machine Company? Explain. 9 .5074 .5076 .5075 .5075 .5076
HS.2 What changes if any, do you think that Natalie York 10 .5076 .5077 .5075 .5075 .5075
might be able to institute in the company? Explain. 11 .5075 .5075 .5075 .5076 .5075
DO NOT CONTINUE UNTIL YOU HAVE COMPLETED 12 .5075 .5076 .5075 .5077 .5075
THE PHASE 1 EXERCISES 13 .5076 .5076 .5073 .5076 .5074
Phase 2 14 .5075 .5076 .5074 .5076 .5075
15 .5075 .5075 .5076 .5074 .5073
Natalie slowly walks down the hall after leaving Mr.
16 .5075 .5074 .5076 .5075 .5075
Harnswell’s office, feeling rather downcast. He just won’t
17 .5075 .5074 .5075 .5074 .5072
listen to anyone, she thinks. As she walks, Jim Murante, the
18 .5075 .5075 .5076 .5075 .5076
shop foreman, comes up beside her. “So,” he says, “did you
19 .5076 .5076 .5075 .5075 .5076
really think that the old man would just listen to you? I’ve
20 .5075 .5074 .5077 .5076 .5074
been here more than 25 years. The only way he listens is if
21 .5075 .5074 .5075 .5075 .5075
he is shown something that worked after it has already been
22 .5076 .5076 .5075 .5076 .5074
done. Let’s see what we can plan out together.”
23 .5076 .5076 .5075 .5075 .5076
Natalie and Jim decide to begin by investigating the pro-
24 .5075 .5076 .5075 .5076 .5075
duction of the cam rollers that are a precision ground part.
25 .5075 .5075 .5075 .5075 .5074
The last part of the production process involves the grinding
26 .5077 .5076 .5076 .5074 .5075
of the outer diameter. After grinding, the part mates with the
27 .5075 .5075 .5074 .5076 .5075
cam groove of the particular sewing pattern. The half-inch
28 .5077 .5076 .5075 .5075 .5076
rollers technically have an engineering specification for the
29 .5075 .5075 .5074 .5075 .5075
outer diameter of the roller of 0.5075 inch (the specifica-
30 .5076 .5075 .5075 .5076 .5075
tions are actually metric, but in factory floor jargon they are
referred to as half-inch), plus a tolerable error of 0.0003 inch
on the lower side. Thus, the outer diameter is allowed to be EXERCISE
between 0.5072 and 0.5075 inch. Anything larger is reclassi- HS.3 a. Is the process in control? Why?
fied into a different and less costly category, and anything b. What recommendations do you have for improv-
smaller is unusable for anything other than scrap. ing the process?
The grinding of the cam roller is done on a single DO NOT CONTINUE UNTIL YOU HAVE COMPLETED
machine with a single tool setup and no change in the THE PHASE 2 EXERCISE
grinding wheel after initial setup. The operation is done by
Dave Martin, the head machinist, who has 30 years of expe- Phase 3
rience in the trade and specific experience producing the Natalie examined the X and R charts developed from the
cam roller part. Since production occurs in batches, Natalie data presented in Table HS.1. The R chart indicated that the
and Jim sample five parts produced from each batch. Table process is in control, but the X chart revealed that the mean
HS.1 presents data collected over 30 batches. HARNSWELL for batch 17 was outside the lower control limit. This imme-
diately gave her cause for concern because low values for the
TABLE HS.1 roller diameter could mean that parts had to be scrapped.
Diameter of Cam Rollers (in inches) Natalie went down to see Jim Murante, the shop foreman, to
try to find out what had happened on batch 17. Jim looked
Cam Roller
up the production records to determine when this batch was
Batch 1 2 3 4 5 produced. “Aha,” He exclaims, “I think I’ve got the answer!
This batch was produced on that really cold morning we had
1 .5076 .5076 .5075 .5077 .5075
last month. I’ve been after Mr. Harnswell for a long time to
2 .5075 .5077 .5076 .5076 .5075
let us install an automatic thermostat here in the shop so that
3 .5075 .5075 .5075 .5075 .5076
the place doesn’t feel so cold when we get here in the morn-
4 .5075 .5076 .5074 .5076 .5073
ing. All he ever tells me is that people aren’t as tough as they
5 .5075 .5074 .5076 .5073 .5076
used to be and if I want to see real cold, I should have been
6 .5076 .5075 .5076 .5075 .5075
back in that foxhole during the Korean winter of 1952.”
7 .5076 .5076 .5076 .5075 .5075
Natalie stood there almost in shock. What she realized
8 .5075 .5076 .5076 .5075 .5074
had happened is that, rather than standing idle until the
 Running Case 789

environment and the equipment warmed to acceptable tem- c. Write a report expressing your conclusions con-
peratures, the machinist opted to manufacture parts that cerning the cam roller diameters. Be sure to discuss
might have to be scrapped. In fact, Natalie recalled that a the diameters as they relate to the specifications.
major problem had occurred on that same day when several DO NOT CONTINUE UNTIL YOU HAVE COMPLETED
other expensive parts had to be scrapped. Natalie said to Jim, THE PHASE 4 EXERCISE
“We just have to do something. We can’t let this go on now
that we know what problems it is potentially causing.” Phase 5
Natalie and Jim decided to take enough money out of petty Natalie noticed immediately that the overall mean diameter
cash to get the thermostat without having to fill out a requisi- with batch 17 eliminated is 0.507527, which is higher than
tion requiring Mr. Harnswell’s signature. They installed the the specification value. Thus, the mean diameter of the
thermostat and set the heating control so that the heat would rollers produced is so high that they would be downgraded
turn on a half hour before the shop opened each morning. in value. In fact, 55 of the 150 rollers sampled (36.67%)
were above the specification value. If this percentage is
EXERCISES
extrapolated to the full year’s production, 36.67% of the
HS.4 What should Natalie now do concerning the cam 7,000 pieces manufactured, or 2,567, could not be sold as
roller data? Explain. half-inch rollers, leaving only 4,433 available for sale. “No
HS.5 Explain how the actions of Natalie and Jim to wonder we often have shortages that require costly emer-
avoid this particular problem in the future has gency runs,” she thought. She also notes that not one diam-
resulted in quality improvement. eter is below the lower specification of 0.5072, so not one
DO NOT CONTINUE UNTIL YOU HAVE COMPLETED of the rollers had to be scrapped.
THE PHASE 3 EXERCISES Natalie realized that there had to be a reason for all
Phase 4 this. Along with Jim Murante, she decided to show the
Because corrective action was taken to eliminate the spe- results to Dave Martin, the head machinist. Dave said that
cial cause of variation, the data for batch 17 were removed the results didn’t surprise him that much. “You know,” he
from the analysis. The control charts for the remaining days says “there is only 0.0003 inch in diameter that I’m
indicate a stable system with only common causes of varia- allowed in variation. If I aim for exactly halfway between
tion operating on the system. Thus, Natalie and Jim sat 0.5072 and 0.5075, I’m afraid that I’ll make a lot of short
down with Dave Martin and several other machinists to try pieces that will have to be scrapped. I know from way
to determine all the possible causes for the existence of back when I first started here that Mr. Harnswell and
oversized and scrapped rollers. Natalie was still troubled by everybody else will come down on my head if they start
the data. After all, she wanted to find out whether the seeing too many of those scraps. I figure that if I aim at
process is giving oversizes (which are downgraded) and 0.5075, the worst thing that will happen will be a bunch of
undersizes (which are scrapped). She thought about which downgrades, but I won’t make any pieces that have to be
tables and charts would be most helpful. scrapped.”

EXERCISE EXERCISES
HS.6 a. Construct a frequency distribution and a stem- HS.7 What approach do you think the machinist should
and-leaf display of the cam roller diameters. take in terms of the diameter he should aim for?
Which one do you prefer? Explain.
b. Based on your results in (a), construct all HS.8 What do you think that Natalie should do next?
appropriate graphs of the cam roller diameters. Explain.

RUNNING CASE
MANAGING THE SPRINGVILLE HERALD
Phase 1
An advertising production team is charged with reducing Monday to Saturday basis. Table SH18.1 includes the total
the number and dollar amount of the advertising errors, number of ads and the number containing errors for a
with initial focus on the ran-in-error category. The team period of one month. (Sundays are excluded because a spe-
collected data including the number of ads with errors on a cial type of production is used for that day.) SH18-1
790 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

TABLE SH18.1 SH18.5 In addition to the number of ads with errors,

Number of Ads with Errors and Daily Number what other information concerning errors on a
of Display Ads daily basis should the team collect?
DO NOT CONTINUE UNTIL YOU HAVE COMPLETED
Number Number THE PHASE 2 EXERCISES
of Ads of Ads
with Number with Number Phase 3
Day Errors of Ads Day Errors of Ads A print production team also is charged with improving
the quality of the Herald. The team has chosen as its first
1 4 228 14 5 245
project the blackness of the print of the newspaper. Each
2 6 273 15 7 266
day the print production team determines how “black” the
3 5 239 16 2 197
newspaper is printed. Blackness is measured on a densi-
4 3 197 17 4 228
5 6 259 18 5 236 mometer that records the results on a standard scale. Five
6 7 203 19 4 208 spots on the first newspaper printed each day are randomly
7 8 289 20 3 214 selected and the blackness of each spot is measured. Table
8 14 241 21 8 258 SH18.2 presents the results for 25 days. SH18-2
9 9 263 22 10 267
10 5 199 23 4 217 TABLE SH18.2
11 6 275 24 9 277 Newsprint Blackness for 25 Consecutive Days
12 4 212 25 7 258
13 3 207 Spot
Day 1 2 3 4 5
EXERCISES 1 0.96 1.01 1.12 1.07 0.97
SH18.1 What is the first thing that the team from the 2 1.06 1.00 1.02 1.16 0.96
advertising production department should do to 3 1.00 0.90 0.98 1.18 0.96
reduce the number of errors? Explain. 4 0.92 0.89 1.01 1.16 0.90
SH18.2 a. Construct the appropriate control chart for 5 1.02 1.16 1.03 0.89 1.00
these data. 6 0.88 0.92 1.03 1.16 0.91
b. Is the process in a state of statistical control? 7 1.05 1.13 1.01 0.93 1.03
Explain? 8 0.95 0.86 1.14 0.90 0.95
c. What should the team recommend as the next 9 0.99 0.89 1.00 1.15 0.92
step to improve the process? 10 0.89 1.18 1.03 0.96 1.04
DO NOT CONTINUE UNTIL YOU HAVE COMPLETED 11 0.97 1.13 0.95 0.86 1.06
THE PHASE 1 EXERCISES 12 1.00 0.87 1.02 0.98 1.13
13 0.96 0.79 1.17 0.97 0.95
Phase 2 14 1.03 0.89 1.03 1.12 1.03
The advertising production team examined the p chart 15 0.96 1.12 0.95 0.88 0.99
developed from the data of Table SH18.1. Using the rules 16 1.01 0.87 0.99 1.04 1.16
for determining out-of-control points, they observed that 17 0.98 0.85 0.99 1.04 1.16
day 8 is above the upper control limit. Upon investigation, 18 1.03 0.82 1.21 0.98 1.08
it was determined that on that day there was an employee 19 1.02 0.84 1.15 0.94 1.08
from another work area assigned to the processing of the 20 0.90 1.02 1.10 1.04 1.08
ads because several employees were out ill. The group 21 0.96 1.05 1.01 0.93 1.01
brainstormed ways of avoiding the problem in the future 22 0.89 1.04 0.97 0.99 0.95
and recommended that a team of people from other work 23 0.96 1.00 0.97 1.04 0.95
areas receive training on the work done by this area. 24 1.01 0.98 1.04 1.01 0.92
Members of this team could then cover the processing of 25 1.01 1.00 0.92 0.90 1.11
the ads by rotating in one- or two-hour shifts.
EXERCISES EXERCISE
SH18.3 What should the advertising production team SH18.6 a. Construct the appropriate control charts for
now do concerning the data of Table SH18.1? these data.
Explain. b. Is the process in a state of statistical control?
SH18.4 Explain how the actions of the team to avoid this Explain.
particular problem in the future has resulted in c. What should the team recommend as the next
quality improvement. step to improve the process?
 Appendix 791

REFERENCES
1. Arndt, M., “Quality Isn’t Just for Widgets,” Business 9. Hahn, G. J., N. Doganaksoy, and R. Hoerl, “The
Week, July 22, 2002, 72–73. Evolution of Six Sigma,” Quality Engineering, 12 (2000):
2. Bothe, D. R., Measuring Process Capability (New York: 317–326.
McGraw-Hill, 1997). 10. Halberstam, D., The Reckoning (New York: Morrow,
3. Deming, W. E., Out of the Crisis (Cambridge, MA: 1986).
MIT Center for Advanced Engineering Study, 1986). 11. Levine, D. M., P. P. Ramsey, and R. K. Smidt, Applied
4. Deming, W. E., The New Economics for Business, Statistics for Engineers and Scientists Using Microsoft
Industry, and Government (Cambridge, MA: MIT Excel and Minitab (Upper Saddle River, NJ: Prentice
Center for Advanced Engineering Study, 1993). Hall, 2001).
5. Friedman, T. L., The Lexus and the Olive Tree: 12. Microsoft Excel 2003 (Redmond, WA: Microsoft Corp.,
Understanding Globalization (New York: Farrar, Straus 2002).
and Giroux, 1999). 13. Minitab for Windows Version 14 (State College, PA:
6. Gabor, A., The Man Who Discovered Quality (New Minitab, Inc., 2004).
York: Time Books, 1990). 14. Scherkenbach, W. W., The Deming Route to Quality and
7. Gitlow, H., A. Oppenheim, R. Oppenheim, and D. Productivity: Road Maps and Roadblocks (Washington,
Levine, Tools and Methods for the Improvement of DC: CEEP Press, 1987).
Quality, 3rd ed. (Homewood, IL: Irwin, 2005). 15. Snee, R. D., “Impact of Six Sigma on Quality,” Quality
8. Gitlow, H. and D. Levine, Six Sigma for Green Belts Engineering, 12 (2000): ix–xiv.
and Champions (Upper Saddle River, NJ: Financial 16. Walton, M., The Deming Management Method (New
Times-Prentice-Hall, 2005). York: Perigee Books, 1986).

Appendix 18 Using Software

for Control Charts
A18.1 MICROSOFT EXCEL 1. Select Stat  Control Charts  Attribute Charts 
P. In the P Chart dialog box (see Figure A18.1) enter
For p Chart C3 or ‘Not ready’ in the Variables: edit box. Since the
See section G.41 (p Chart) if you want PHStat2 to produce subgroup sizes are equal, enter 200 in the Subgroup
a p chart as a Microsoft Excel chart. (There are no Microsoft sizes: edit box. Click the P Chart Options button.
Excel commands that directly produce this chart.)

For R and X Chart

See section G.42 (R & XBar Chart) if you want PHStat2
to produce an R or X chart as a Microsoft Excel chart.
(There are no Microsoft Excel commands that directly pro-
duce these charts.)

A18.2 MINITAB
Using Minitab for the p Chart
To illustrate how to construct a p chart, refer to the data of
Table 18.1 on page 759 concerning the number of rooms
not ready. Open the HOTEL1.MTW worksheet. FIGURE A18.1 Minitab P Chart Dialog Box
792 CHAPTER EIGHTEEN Statistical Applications in Quality and Productivity Management

2. In the P Chart - Options dialog box, click the Tests tab

(see Figure A18.2). In the drop-down list box, select
Perform all tests for special causes. Click the OK
button to return to the P Chart dialog box. (These val-
ues will stay intact until Minitab is restarted.) Click the
OK button to get the p chart.

FIGURE A18.3 Minitab c Chart Dialog Box

3. If there are points you want to omit when estimating

the center line and control limits, click the Estimate
tab in the C Chart - Options dialog box. Enter the
points to omit in the edit box shown. Click the OK but-
ton to return to the C Chart dialog box. Click the OK
button to get the c Chart.

FIGURE A18.2 Minitab p Chart - Options Dialog Box,

Tests Tab Using Minitab for the R and X̄¯¯ Charts
To construct R and X charts using Minitab, select Stat 
Control Charts  Variable Charts for Subgroups 
3. If there are points you want to omit when estimating Xbar-R from the menu bar. The format for entering the
the center line and control limits, click the Estimate variable name is different, depending on whether the data
tab in the P Chart - Options dialog box. Enter the points are stacked down a single column or unstacked across a set
to omit in the edit box shown. Click the OK button to of columns with the data for each time period located in a
return to the P Chart dialog box. In the P Chart dialog single row. If the data for the variable of interest are stacked
box, click the OK button to get the p chart. down a single column, choose All observations for a
chart are in one column in the drop-down list box and
enter the variable name in the edit box below. If the sub-
Using Minitab for the c Chart groups are unstacked with each row representing the data
To illustrate how to construct a c chart, refer to the data of for a single time period, choose Observations for a sub-
Table 18.4 on page 767 concerning the number of com- group are in one row of columns in the drop-down list box
plaints at the hotel. Open the COMPLAINTS.MTW and enter the variable names for the data in the edit box
worksheet. below.
To illustrate how to construct R and X charts, refer to
1. Select Stat  Control Charts  Attribute Charts
the data of Table 18.5 on page 772 concerning the luggage
 C. In the C Chart dialog box (see Figure A18.3),
delivery times. Open the HOTEL2.MTW worksheet.
enter C2 or COMPLAINTS in the Variables: edit
box. 1. Select Stat  Control Charts  Variable Charts for
2. Click the C Chart Options button. In the C Chart Subgroups  Xbar-R. Since the data are unstacked,
Options dialog box, click the Tests tab. In the drop- select Observations for a subgroup are in one row of
down list box, select Perform all tests for special columns in the drop-down list box. In the Xbar-R
causes. Click the OK button to return to the C Chart Chart dialog box (see Figure A18.4) enter C2 or ‘Time
dialog box. (These values will stay intact until Minitab 1,’ C3 or ‘Time 2,’ C4 or ‘Time 3,’ C5 or ‘Time 4,’ and
is restarted.) Click the OK button to produce the c C6 or ‘Time 5’ in the edit box. Click the Xbar-R
chart. Options button.
 Appendix 793

more than one set of R and X charts in the same ses-

sion, reset the points to omit before creating new charts.
These values will stay intact until Minitab is restarted.)
4. In the Xbar-R Chart dialog box, click the OK button to
produce the R and X charts.

FIGURE A18.4 Minitab Xbar-R Chart Dialog Box

2. In the Xbar-R Chart - Options dialog box, click the

Tests tab. In the drop-down list box, select Perform all
tests for special causes. (These values will stay intact
until Minitab is restarted.)
3. Click the Estimate tab in the Xbar-R Chart Options dia-
log box (see Figure A18.5). Click the Rbar option but-
ton. If there are points you want to omit when estimating
the center line and control limits, enter the points to omit
in the edit box shown. Click the OK button to return to FIGURE A18.5 Minitab Xbar-R Chart Options Dialog
the Xbar-R Chart dialog box. (Note: When creating Box, Estimate Tab