For A-Plus Writer Only

Quality Control and Improvement

Chapter outline

9.1 Design of quality control systems 9.2 Process quality control

9.3 Attribute control chart

9.4 Variables control chart

9.5 Using control charts

9.6 Process capability 9.7 Continuous improvement 9.8 Six Sigma

• 9.9 Lean and Six Sigma 9.10 Quality control and improvement in industry

9.11 Key points and terms

In the early 1900s, inspection shifted from the workers to a formal quality control department. This created tension between the workers and the inspectors that is still evident in some firms today. But those who use the modern ideas of quality control are able to avoid these tensions and create a positive environment for quality improvement.

In 1924, Walter A. Shewhart of the Bell Telephone Labs developed a statisti- cal quality control chart. Two others from the Bell Labs, H. F. Dodge and H. G. Romig, further developed the theory of statistical quality control in the 1930s. But little was done in industry until World War II in the early 1940s. The war created the demand for huge quantities of military goods from industry. The military required that industry adopt the new methods of statistical quality control to help ensure that the goods it ordered would meet government stan- dards. As a result, statistical methods for control of quality were widely ad- opted by industry during the early 1940s. In later years, however, these methods were abandoned, only to be rediscovered in the 1980s as a valid way to ensure quality products and services.

189

190 Part Three Quality

~

The service industries have been reluctant to adopt the methods of statistica quality control. While some service companies have made impressive gains in th.: use of these methods, many others have lagged behind. As a result, there is a tre- mendous opportunity to use the methods of statistical quality control an improvement in service firms.

After World War II, in 1946, the American Society for Quality (ASQ) 1 W a5 formed. While the initial emphasis was on statistical quality control methods, the focus has broadened to include customer needs, total quality management, anG. continuous improvement. The ASQ has also been focusing on spreading the ideas of quality management to service industries.

The emphasis in this chapter will be on process definition, statistical quality control, and continuous improvement. We advance the idea that quality control is the stabilization and maintenance of a transformation process (or more generically "a process") to produce consistent output. This transformation process can be a production process, a service delivery process, or an administrative process. This means the process does not vary significantly in the important quality characteris- tics that are being controlled. Continuous improvement can occur only after the process has been brought under control and stabilized.

We also advance the idea that the organization consists of many interrelated processes that need to be controlled to produce quality products and services. It follows that quality control and continuous improvement are highly cross- functional in nature and require the participation and support of the entire organization. The Operations Leader box about implementing statistical pro- cess control (SPC) at Milliken & Company explgins how these ideas are being applied in industry.

9.1 DESIGN OF QUALITY CONTROL SYSTEMS

All quality control must start with the process itself. Of course, a process is typi- cally composed of many subprocesses, each having its own intermediate product or service. A process can therefore be an individual machine, a group of machines, or any of the many clerical and administrative processes that exist in the organiza- tion. Each of these processes has its own internal customers and its own products or services that are produced. The customer is the next process (or processes) that receives the work output. For example, the customer of the design department is the machine shop that makes the parts. The customer of the machine shop is the assembly department that uses the parts. When a large production system is bro- ken down into many smaller systems or processes, quality can be defined and controlled at each point along the way.

After identifying each of the processes that need to be controlled, critical con- trol points can be chosen where inspection or measurement should take place. The types of measurement or tests required and the amount of inspection required at each of these points should be determined. Finally, management should decide who will do the inspection, the workforce itself or separate inspectors. Usually operator inspection is preferred because it places responsibility on those who make the product or service. Once these decisions are made, it is possible to design

1 This society was called the American Society for Quality Control (ASQC) until 1997, when its name was changed to reflect its scope and mission more accurately, beyond mere control of quality.

Chapter 9 Quality Control and Improvement 191

Operations Leader Implementing SPC at Milliken & Company

Milliken & Company, founded in 1865 and headquar- tered in Spartanburg, South Carolina, is a global leader in textiles, performance materials, chemicals, and floor cov- erings. An innovator at heart, Milliken & Company has

long led the way for "knowledge-based" investment, employing over 100 PhDs, and has

accumulated over 2200 U.S. patents-and more than 5000 patents worldwide. Today, Milliken & Company has 39 manufacturing facilities located in the United States, United Kingdom, Belgium, France, and China and em- ploys approximately 7000 associates globally.

Milliken & Company has a long history of deploying statistical process control (SPC) to help monitor process quality performance. Christopher Wozniak, a process im- provement specialist at Milliken & Company when SPC was initia lly deployed, writes, "When I was working on my first SPC program, a plant superintendent told me, 'Don't get too excited about this SPC stuff. I've been around for 15 years, and I give SPC a year before the next program comes along.' He was right. The program started to fizzle out. Soon after, however, a company- wide commitment was made to SPC, and the proper sys- tem developed. The SPC program has since flourished."

In one example, SPC was u,sed after an out-of- control process was found. To solve this problem,

several people from different departments spontane- ously joined together. The informal team leader ral- lied everyone around the out-of-control process by explaining the importance of the process, showing them examples of defects, and defining how team members could help. Feedback on how the process was being improved was given to everyone as the im- provement proceeded. As a result, within six months the process capability improved from a c pk of 0.15 to a c pk of 0.95. *

In 1989, Milliken & Company won the Malcolm Baldrige National Quality Award. Since then, it has won many other quality awards including the Eu- ropean Quality Award, the Canadian Quality Prize, and the British Quality Award. This dedication to quality stemmed from strong and passionate lead- ership from the late Roger Milliken, son of the founder and past chairman and CEO . The focus on quality not only helped Milliken & Company to com- pete successfully against lower-cost suppliers from the United States but also contributed to keeping jobs from being lost to overseas competitors.

*Cpk is explained later in this chapter.

Source: Christopher Wozniak, "P roactive vs Reactive SPC," Quality Progress, February 1994; www.milliken .com, 2012; Personal e-mai l correspondence dated Apr il 30, 2012.

a complete system of quality control, which allows continuous improvement of a stable system.

1. The first step in designing a quality control system is to identify the critical points in each of the processes where inspection and testing are needed. The guidelines for doing this are as follows:

• Ensure that incoming raw materials or purchased services will meet specifica- tions. Ideally, incoming inspection can be eliminated by certifying the supplier. Supplier certification is normally granted to suppliers that have demonstrated they use statistical process control and other methods to achieve consistent quality performance. In this case, the products or services of the supplier can be used directly by the customer without incoming inspection.

• Test work in process or the service while it is being delivered. As a general rule, the product or service should be inspected by operators before irrevers- ible operations take place or before a great deal of value is added to the prod- uct. In these cases, the cost of inspection is less than the cost of adding more value to the product. A precise determination of where the product should be inspected should be made from the process flowchart.

• The third critical inspection point is the finished product or service. In manufacturing, final products are frequently inspected or tested before

192 Part Th ree Quality

~

shipping or before the product is placed in inventory. At one automo~ assembly plant, for example, a random number of cars is taken directly the assembly line and thoroughly inspected for appearance and funct The defects are noted and fed back to assembly-line personnel so that ._-~ can correct the underlying causes. The defects are also used to compu-- quality score for comparison among assembly plants.

It is usually far better to prevent defects from occurring than to inspect and c rect defects after production. Nevertheless, some measurement via sampl!:- _ inspection is necessary to maintain processes in a continuous state of con:::- and to facilitate improvement. Thus, inspection (or measurement) cannor :- eliminated, but it can be reduced by a vigorous process of prevention.

2. The second step in designing a quality control system is to decide on the type measurement to be used at each inspection point. There are generally two e:- tions: measurement based either on variables or on attributes. Variables ma- surement utilizes a continuous scale for factors such as length, height, a:-.: weight. Examples of variables measurement are the dimensions of parts, ,.:... viscosity of liquids, and the time that it takes to wait on tables in a restauran:

Attribute measurement uses a discrete scale by counting the number of G.::- fective items or the number of defects per unit. When the quality specificatio:-~­ are complex, it usually is necessary to use attribute measurements. In this cas- a complicated set of criteria can be used to define a defective unit or a defec. For example, a flat screen TV may be classified as defective if any of a numl of functional tests are failed or if the appearance of the display is not satisfactor:- In inspection of cloth, a defect can be defined as a flaw in the material and number of defects per 100 yards can be counted during inspection. Determir.- ing the type of measurement to use also involves the specification of measurin~ equipment. A wide variety of devices are available for measurement. Bowen~: the selection of these devices is beyond the scope of this text.

3. The third step in defining the quality control system is to decide on the amour:- of inspection to use. Generally, statistical process control is preferred to rnin:- mize the amount of inspection needed. Exceptions to this might be when p ro- cess variables are difficult to define or when the consequences of failure are very high. For example, when human lives are at stake, both process contro~ and 100 percent testing of output may be used.

4. The final step in designing a quality control system is deciding who should d the inspection. Usually it is best to have the workers inspect their own outpu: and be responsible for the quality of their work (sometimes called quality at the source). There is much evidence to suggest that a prevention program, along with worker responsibility for quality, will be less expensive than an extensi, -e outside inspection program. In high-contact services there is no choice but to have quality at the source, since the customer immediately perceives defects.

In some cases, the customer will be involved in inspecting the product. Ser- vice customers always take this role as they receive the service. Some customers station inspectors at the vendor's plants to examine and accept or reject ship- ments before they are sent on to the customer. The government has inspectorJ in a variety of industries to ensure quality in the interest of public health and safety. Thus, many people may be involved in the inspection process.

A well-designed quality control system requires a series of management judg- ments and the participation of all functions. The control principles themselve are elementary, requiring performance standards, measurement, and feedback of

Chapter 9 Quali ty Control and Improvement 193

results to correct the process. The application of these principles in any specific sit- uation is complex. The guiding principle is to first control the system and then aim for continuous improvement of the resulting stable system.

9.2 PROCESS QUALITY CONTROL

"A Day in the Life of Quality

at Honda," Vol. VII

Statistical process control utilizes inspection (or testing) of the product or service while it is being produced. Periodic samples2 of the output of a production process are taken. When, after inspection of the sample, there is reason to believe that the process quality characteristics have changed, the process is stopped and a search is made for an assignable cause. This could be a change in the operator, the machine, or the material. When the cause has been found and corrected, the process is started again.

Process control is based on two key assumptions, one of which is that random variability is basic to any production process. No matter how perfectly a process is designed, there will be some random variability, also called common causes, in quality characteristics from one unit to the next. For example, a machine filling cereal boxes will not deposit exac tly the same weight in each box; the amount filled will vary around some average figure . The aim of process control is to find the range of natural random variation of the process and ensure that production stays within that range.

The second principle of process control is that production processes are not usu- ally found in a state of control. Due to lax procedures, untrained operators, improper machine maintenance, and so on, the variation being produced is usually much larger than necessary. The first job of process control managers is to seek out these sources of unnecessary variation, also called special causes, and bring the process under statistical control so that the remaining variation is due to random causes.

Administt ative processes in accounting, human resources, sales, marketing, and finance in most organizations are also not usually under statistical control. These processes can also be controlled. The same principles used to control production are used to control administrative processes.

A process can be brought to a state of control and maintained in that state through the use of quality control charts (also called process charts or control charts) . In the control chart shown in Figure 9.1, they axis represents the quality characteristic that is being controlled and the x axis represents time or a particular sample taken from the process. The center line of the chart is the average quality characteristic being measured. The upper control limit represents the maximum acceptable random variation, and the lower control limit indicates the minimum acceptable random variation when a state of control exists. Generally speaking, the upper and lower control limits are set at ::±: three standard deviations from the mean. If a normal probability distribution is assumed, these control limits will in- clude 99.74 percent of the random variations observed.

2 It is technically more correct to refer to the group of observat ions about some outputs from a manufacturi ng or a service delivery process as a subgroup rather than as a sample. A sample is typica lly used to denote a subset of a population for which the total number of elements is known. For example, we can say that we took a sample of the fe male students from the population of all fem ale students enrolled at the University of Minnesota . However, for a manufacturing or service delivery process, the total output produced by t he process is not fi nite. As long as th e process is operating uninterrupt ed, the total output is infinite. That is, we cannot say that this car manufacturing process makes only 3000 ca rs during t he life of the process. In this latter case, w hen we take a number of ca rs off the manufactu ri ng process to be inspected, thi s subset is more correctly referred to as a subg roup . Since students are genera lly more f am iliar w ith sample, we continue t o use sample inst ead of subgroup in ou r discussion.

194 Part Three Quality

FIGURE 9.1 Y Quality control chart. 4.

FIGURE 9.2 Quality control chart example.

Average+ 3 standard deviations

Upper control limit (UCL)

Quality I Center line (CL) measurement

average

Average- 3 standard deviations

Lower control limit (LCL)

Time ....... .X To understand this last point, take a look at the right side of Figure 9.1 showing

a normal probability distribution rotated on its side. It indicates that the distribu- tion mean (average) is located on the center line of the control chart and the tails of the distribution are outside the control limits by jus,t a small amount. Thus, 99.74 percent of the sample observations that are taken and plotted on the graph in Fig- ure 9.1 will fall inside the control limits.

In using a control chart to ensure steady-state operation, periodic samples are taken and plotted on the control chart (see Figure 9.2). When the measurement falls within the control limits, the process is continued. When measurements signal that the process is no longer in a state of control, then a search is made for assignable causes. One such signal is when measurements fall outside of the control limits. Other signals include measurements that, when plotted on the control chart, reveal points on a trend upwards or downwards; an oscillation where measurements al- ternate in an up-down manner, or a pattern with many measurements being within one standard deviation of the center line.

5 5 ~ ::s "' !U ~ .£ r;; ::s 01

UCL

CL

LCL

I I j 1. I Stop the process; look for asstgnable cause

I 1\

~ I \ \ -........,, \

Stop the process; look for assignable cause

I I I I I 1 2 3 4 5 6

Sample

Chapter 9 Q uality C ontrol and Improvem ent 195

Assignable causes have also been termed special causes, those that cause points, for example, to fall outside of the control limits . Special causes can be corrected to bring the process back under control without changing the design of the process. In contrast, common causes of variation are those that randomly occur when the process is under statistical control. Common causes cannot be removed without changing the design of the process. Through use of control charts, the process is maintained in a constant state of statistical control and there is only natural random variation (common causes) in the output from the transformation process.

Quality can be monitored using control charts for attributes or for variables. We cover each of these cases in turn below.

9. 3 ATTRIBUTE CONTROL CHART

A quality characteristic can be measured on a discrete scale (e.g., an item is either good or defective) rather than a continuous scale. An example of quality as an at- tribute is the percentage of defectives occurring in a sample. Other examples of attribute measurements are the percentage of phone calls not answered within three rings, the percentage of customers who are dissatisfied, and the percentage of parts from a supplier that are defective.

When the quality characteristic is an attribute, the appropriate type of control chart to use is an attribute control chart. One particularly useful attribute control chart is the p chart for the percentage corresponding to the quality characteristic of interest. For example, percentage defective is estimated by taking a sample of n units at random from a process at specified time intervals. For each sample, the observed percent defective (p) is computed. These observed values of pare plotted on the p control chart, one for each sample.

To get the ce~ter line and control limits of the p chart, we take a large number of samples of n units each. The p value is computed for each sample and then aver- aged over all samples to yield a value p. This value of p is used as the center line since it represents the best available estimate of the true average percent defective of the process. We also use the value of p to compute upper and lower control lim- its as follows:

UCL = p + 3~p(l : p)

LCL = p- 3~p(l: p)

In this case, the standard deviation of the process is the quantity under the square root sign. We are adding and subtracting three standard deviations from the mean to get the control limits. See the example of this computation fo r controlling com- puter data entry operations.

After the p chart is constructed with its center line and upper and lower control limits, new samples of the process are examined and plotted on the chart. If the percentage falls within the control limits, no action is taken. If the percentage falls outside the control limits, the process is stopped and a search for an assignable cause (material, operator, or machine) is made. After the assignable cause is found and corrected- or, in very rare cases, no assignable cause is found- the process is restored to operating condition and production or service is resumed.

196 Part Three Quality

9.4 VARIABLES CONTROL CHART

Attribute Control Chart: A p Chart Example

Control charts are also used for measurements of variables. In this case, a mea::. u.:-::- ment of a continuous variable is made when each item is inspected. As a result, r values are computed from the sample: a measure of central tendency (usually;:_ average) and a measure of variability (the range or standard deviation). With th values, two control charts are developed: one for the central tendency and one ~- ­ the variability of the process. When the process is found to be out of control either of these charts, it is stopped and a search for an assignable cause is made.

When variable measurement is used, two control charts are needed because !!.- normal distribution is assumed and it has two parameters (mean and variance Either the mean of the distribution can change or the variance (as measured range) can change. As a result, we monitor both the average of a process and range for control purposes.

Suppose that the average (:X) and range (R) are computed each time a samples taken. Then a control chart for average and a chart for range will be used. The control limits for the average chart are computed as follows:

CL =:X - -

UCL =:X+ A 2

R

LCL = x- A2 R where x is the grand average of several pastr:X averages and R is the average o:' several past R values. Recall that the range (R) is simply the largest value minlli the smallest value in a sample. In the above formulas, A2 is a constant that in- cludes three standard deviations in terms of the range. Table 9.1 provides values oi

Suppose 200 records are taken from a data entry operation at two-hour interva ls to monitor the data entry process. If we co llect 200 records every two hours for 11 t imes, we wou ld have 11 samples, each with 200 records. The percentage of records in error for the past 11 sam- ples is found to be .5, 1.0, 1.5, 2.0, 1.5, 1.0, 1.5, .5, 1.0, 1.5, and 2.0 percent. The average of these 11 sample percentages yields a p = 1.27 percent, which is the center line of the control chart. The upper and lower control limits are

j-0127(9873) UCL = .0127 + 3,

200 = .0364

LCL = .0127-3 .0127(.9873)

200 = - .0110

When the LCL is negative, it is rounded up to 0 because a negative percentage is impossible. Thus, we have the following p chart:

3.64 UCL

Percent defective 1.27 CL /r\ lJ

v 1\/ 0 LCL

Sample number

Since all points are found to be in control, these 11 samples can be used to estab lish the center line and control limits. New samples of 200 records can now be taken, plotted on the p chart, and interpreted to determine whether or not the process is in control.

TABLE 9.1 Control Chart Constants

Source : Factors reproduced from 1950 ASTM Manual on Quality Control of Materials by permission of the American Society for Testing and Materials, Philadelphia.

Chapter 9 Quality Control and Improvement 197

Sample Size n A2 03 04

2 1.880 3.267 3 1.023 0 2.575 4 0.729 0 2 .282 5 .577 0 2.115 6 .483 0 2 004 7 .419 .076 1.924 8 .373 .136 1.864 9 .337 .184 1.8 16

10 .308 .223 1.777 12 .266 .284 1.716 14 .235 .329 1.671 16 .212 .364 1.636 18 .194 .392 1.608 20 .180 .414 1.586 22 .167 .434 1.566 24 .157 .452 1.548

A 2

for various sample sizes for the normal distribution. The control limits for the range chart are computed as follows:

CL = R

UCL = D 4 R

LCL = D 3 R

The constants D 3 and D

4 provide three standard deviation limits for the range. Values

for these constants are also given in Table 9.1. The purpose of these constants is to help calculate the upper and lower control limits as a function of sample size. A given sample size can be used in Table 9.1 to look up the appropriate values of A2, 0 3, and D

4 for average and range charts. The variables control chart example illustrates how

to compute control limits and how to determine whether a sample is in control.

9. 5 USING CONTROL CHARTS

There are two issues of concern in using control charts. First, the problem of sample size must be faced . For an attribute control chart, samples should be fairly large, frequently in the range of 50 to 300 observations. As a general rule, the sample must be at least large enough to allow for the detection of one defective unit. For example, if the process being controlled produces 1 percent defective units, a sample size of at least 100 should be used to detect one defective unit on average. Control charts for variables require much smaller sample sizes, usually in the range of 3 to 10 items, because each variable measurement provides much more information.

Th e second issue is how frequently to sample. This issue often is decided on the basis of the rate of production and the cost of producing defects in relation to the cost of inspection. A high-volume production process should be sampled fre- quently since a large number of defective units could be produced between sam- ples. When the cost of producing defective units is high in relation to the cost of inspection, the process should also be sampled frequently. An example of a costly situation is one in which the entire production output must be screened when the process is found to be out of control. In th ese cases, samples should be taken fre- quently, provided that the cost of sampling is not too high. What is the cost of sampling from an accounting process, personnel records, or a process that receives

198 Part Three Quality

Variables Control Chart: TheX&R Charts Example

The Midwest Bolt would like to control the quality of the bolts produced by its automatic screw machines. Each machine produces 1 00 bolts per hour and is controlled by a separate control chart Every hour, a random sample of six bolts is selected from the output of each machine and the diameter of each sample bolt is mea~ured. From each six diameters, an average and range are computed . For example, one sample produced the following six measurements: .536, .507, .530, .525, .530, and .520 . The average of these measurements is x = .525, and the range is R = .029. We also know that the grand average of all past samples has been running x = .513 and the grand average range is R = .020. From these grand averages, the control chart param- eters are computed as follows (see Table 9.1 for control chart constants with n = 6).

X Chart R Chart

CL = .513 CL = .020 UCL = .513 + .483(020) = .523 UCL = 2 004(020) = .040 LCL = .513 - .483(020) = .503 LCL = 0(020) = 0

On the basis of these control limits, the sample of six bolts is found to be out of control on average measurement and in control on range. (Note: x = .525 is outside the upper control limit on the X chart.) We should therefore stop the process and look for an assignable cause that would tend to produce bolts that are too large in diameter.

sales orders? These costs can be rather high and may warrant statistical control through infrequent sampling.

Control charts are widely used in industry fo r both manufacturing and service. In manufacturing companies, control charts are frequently located on each ma- chine to control the quality output of that machine. Quality measurements are taken periodically and plotted on the chart to ensure that the machine is still pro- ducing at its required tolerances and the average and range haven't changed.

In service industries control charts are used to control the time or the percentage of defects from various processes-for example, the time it takes to answer a phone, the time it takes to serve a customer, or the time it takes to collect accounts receiv- able. Service industries also use control charts to monitor and control the percent- age of dissatisfied customers or the percentage of late payments, for example.

9.6 PROCESS CAPABILITY

Once a process has been brought under statistical control, the process capability can be assessed. Process capability is simply the ability of the process to meet or exceed the technical specifications obtained from customers. A process that is not capable of meeting specifications means that defective outputs are being produced. Moreover, it is important to remember that whether or not a process is in a state of statistical control and whether or not a process is capable are two separate issues. Just because a process is in a state of statistical control does not mean that it is ca- pable. Conversely, just because a process is not in a state of statistical control does not mean that is is not capable. When a process is in statistical control but too many defective units are being produced, then the specifications should be relaxed, or a better process used, or a 100 percent inspection instituted temporarily to screen out the bad units until the process can be improved or specifications can be changed.

Whether or not a process is capable of meeting specifications can be determined by computing the process capability index CP-the ratio of the specification (spec) width to the process width:

Spec width c = - =----- p Process width

Chapter 9 Quality C ontrol and Improvement 199

FIGURE 9.3 Process capability index examples.

LSL

100

Spec width USL- LSL CP = Process width 6cr

USL = 160 LSL = 100

cr = 10 Cp= 1

Process measure

USL

160

LSL

100

USL = 160 LSL = 100

cr = 5 Cp= 2

Specification width

Process measure

USL

160

If the process is centered within the specification range, as shown in Figure 9.3, cp ~ 1 will be a good indicator of the ability of the process to meet its specifica- tions, since the process width will be within the specification width.

In practical use, the specification width is computed as the difference between the upper specification limit (USL) and the lower specification limit (LSL). The process width is computed by using six standard deviations of the process measurement be- ing monitored (60'). The standard deviation (0') refers to the individual items being produced, not the standard deviation of the samples taken from the control chart. The logic for 60' is that most of the variation of a process measurement is included within ::±:3 standard deviations of the mean, or a total of 6 standard deviations. Thus we have

USL- LSL c = - - - - p 60'

If the process is centered in the specification range and cp = 1, the process is considered to be minimally capable of meeting the specifications. A process with cp < 1 must be improved by reducing the standard deviation or increasing the specification width, if possible, to become capable.

For the normal distribution, if CP = 1 and the process is centered within the spec- ifications and under statistical control, 99 .74 percent of the product produced will lie within the specifications, corresponding to 2600 parts per million (ppm) defective.3

The figure of 99 .74 percent can be obtained by using the normal probability distribution tables from Appendix A If CP = 1.33, then 99.994 percent of the prod- uct will lie within the specifications, corresponding to 60 ppm defective. Thus a slight increase in cp causes a 'dramatic drop in the defective rate of the process. Customers often specify CP values from 1 to 1.5 or even as high as 2.0, depending on their particular quality requirements. Table 9.2 shows why very high process capability and extremely low rates of defects may be required in some cases.

3 Note that 99.74 percen t good product co rre sponds to (100- 99.74) = .26 percent bad. The .26 percent can be con verted to 2600 ppm by multi pl yi ng .0026 by 1 ,000,000.

I -----

200 Part Three Quality

TABLE 9.2 When 99.9 Percent Quality Is Not Enough

Source: N atalie Ga bel, " Is 99. 9% Good Enough ?" Train ing Magazine, March 1991, p p . 40--41.

If 99.9 percent quality standards w ere in effect, the following would happen:

• Tw o million documents would be lost by the IRS each year. • 22,000 checks w ould be deducted from the wrong bank account each day. • 1 ,314 phone calls w ould be misrouted each day. • 12 babies would be given to the w rong parents each day. • Two plane landings daily at O' Hare w ould be unsafe.

One problem with the CP measure is that it requires the process to be centere in the specification range for an accurate measure of process capability. Because this problem, another more widely used measure (C pk) has been devised:

. (USL- f.L f.L- LSL) C =Mm ,--- ~ 3cr 3cr

where f.L = the process mean value and cr = the process standard deviation. This more complicated measure of process capability overcomes the centerir.;

problem by calculating the process capability for each half of the normal distrib::.- tion and then taking the minimum of the two calculations. The result is shown ~ Figure 9 .4a, where the value of Cpk = 0, while CP = 1. This figure illustrates that use of the CP index when the process is not centered gives the wrong answer, sine:' the process is not fully capable of meeting the specifications, whereas cpk gives tr..f- correct answer with Cpk = 0. A further example is given in Figure 9.4b, where Cpk = :. even though the distribution is not centered. In this case, the process is capable meeting the specifications but could be improved by shifting the mean closer to th.: center of the specification range. Because C k more accurately reflects the actu process capability, it is the measure commonfy used by industry.

9.7 CONTINUOUS IMPROVEMENT

If a process is unable to meet customer specifications, continuous improvemer- can be undertaken. Not all processes need to be improved. Those with strategi.: importance and low process capability should be the first ones selected fo~ improvement.

FIGURE 9.4 Computation of C pk'

>. u :::: "' g. J:

LSL

100

USL= 160 LSL = 100

a= 10 i-i.= 100

Cp= 1 Cpk = 0

130

Process measure

(a)

USL

160

>. u :::: "' :::: 0" ~

""'

100 115 130

USL= 160 LSL = 100

<J= 5 1-1.= 115

Cp= 2 cpk= 1

Process measure

(b)

USL

160

FIGURE 9.5 The seven tools of quality control. Source: Gitlow et a!., Tools and Methods for Quality Im provement, 2d ed. (Burr Ridge, IL: Irwin, 1994).

Chapter 9 Quality Control and Improvement 201

Flowchart Cause-and-Effect

D

Check Sheet Scatter Diagram

observation data

1 N • •

2 "' • • ::c • • .~ • • .. • • 3 ~ • • • 4

5 Variable 1

His togram Control Chart

"' - .. >. .--- u 1--= "' -=

= UCL "' "' "' ::E x "' 0"

I "' .. .... "' "' u 0 LCL ..

~

Measurement Time

Pare to Chart

>. r- u = ,...._ "' = 0"

11--r---, "' .. .... Type

Th e seven tools of quality control are shown in Figure 9.5. 4 These tools, first described by the Japanese, are used by small groups of workers along with managers and engineers to control and improve processes . Table 9.3 summa- rizes the purpose of each of the se tools in identifying and solving problems associated w ith process improvement. 4 There are also seven new tools of qual ity described in Mizuno (1988). These add itional seve n too ls are more suit ed to prob lem identification t han t o problem solving.

202 Part Three Quality

TABLE 9.3 Purpose of the Seven QC Tools

TABLE 9.4 Defectives in Front-End Loader Hydraulics

Tool

Flowcharts Check sheets Histograms Pareto diagrams Cause-and-effect diagrams Scatter diagrams Control charts

Purpose

Understanding the process and identifying possible problem areas Tabulating data on the problem area Illustrating the frequency of occurrence of measures Identifying the most important problems Showing possible causes of the problem Investigating causes and effects Holding the gains from process improvement

The improvement process starts with flowcharting . Flowcharts describe the re- lationships in the process and reveal any unnecessary steps and waste that can eliminated. A flowchart also identifies possible problems that need to be invec+'- gated via further data collection and analysis.

Data collection is done using check sheets, which are a tabular form used to collec: data on the process. For example, a check sheet could contain critical process mea- surements taken at periodic intervals during the day and tabulated by the time taker:

The next step in process improvement and problem solving is to display the da in terms of a histogram. A histogram is a frequency count using data from the chec.-:... sheet to show the form and shape of the distribution of the data. A histogram can in- dicate that some data points are outliers, or there may be odd shapes to the distribu- tion that indicate skewness or possibly more than one mode or peak in the distribution..

A Pareto diagram can be constructed to sl).ow the most important problems . Ir. 1906 Vilfredo Pareto observed that a few items in any population constitute a significant percentage of the entire group-the vital few. Data are tabulated tc identify the most frequently occurring modes of failure. As a result, the m os: important problems can be attacked first.

Table 9.4 provides a count of possible reasons for hydraulic leaks found in assem- bling front-end tractor loaders in a factory. As is noted in the table, the most common reason for a leak (defect) is loose connections, followed by cracked connectors, and on. These data are transferred to the Pareto diagram in Figure 9.6. Because it graphs the reasons for leaks in decreasing order of occurrence, the Pareto diagram readily shows the importance of the various types of defects that have been found. Accord- ing to Pareto's law, a few of the failure modes account for most of the defects.

The Pareto diagram shows which defects we should try to eliminate first. By reference to Figure 9.6, we see that we should investigate loose connections first because they occur most frequently. Of course, cracked connectors is a close sec- ond and should be investigated too, especially if cracked connectors are easier or less costly to correct than loose connections. Pareto analysis is very helpful when one is first studying a quality problem because it helps break the problem down

Number Inspected (N) = 2347

Defective Items Number of Defectives Percent Defective

0-rings missing 16 3.9% Improper torque 25 6.1 Loose connections 193 46.8 Fitting burrs 47 11.4 Cracked connectors 131 31.8

- --- Total 412 100.0%

FIGURE 9.6 Pareto diagram.

IGURE 9.7 ,... a use-and-effect iiagram for loose .:onnections.

400

"' 300 "' ·.6 u

"' '+< "' 0

'+< 200 0 .... "' .r:. s ::s z 100

Loose Cracked connections connectors

Fitting burrs

Chapter 9

Improper torque

0-rings missing

Quality Comrol and Improvement 203

100

75

"' OJ) .:s = 50 "' u .... "' '"'

25

into smaller pieces. Leaking hydraulics has been defined as due primarily to loose connections or cracked connectors (78.6 percent combined failures) .

The next step in the analysis is to take one of these failure modes, say, loose con- nections, and generate ideas for the causes of failure . This is done by using the cause-and-effect diagram, also called an Ishikawa diagram, after Dr. Kaoru Ishikawa (1986), who first used these diagrams in Japan.

A cause-and-effect (CE) diagram is shown in Figure 9.7 for the loose connec- tions. The problem itself, or the effect, is shown on the right side of the diagram. The various potential causes of this problem are shown along the spine of the dia- gram as materials, workers, inspection, and tools. The appearance of this diagram

204 Part Three Quality

suggests a fishbone analogy. The bones of the fish are the probable causes of qua - ity problems, but any cause can be listed. Each of the major causes is then b rok down into more detailed causes, giving rise to more bones on the fish. For exa=- ple, the worker cause is split into three possibilities: inexperience, fatigue, ar:- training. Training in turn is divided into content and method.

When a CE diagram is constructed, the potential causes of a problem becorr.- readily apparent. All these causes can be evaluated one by one to find the tr.·- causes of the problem.

CE diagrams frequently are constructed by using quality improvement teams problem-solving teams. Through the use of brainstorming, the team will identify: wide variety of possible causes for a problem. Then the team or an individual ca:- collect data to narrow down the potential causes before taking corrective action .

Additional data can be analyzed via a scatter diagram, which shows the rela- tionship between two variables. If a particular cause and effect are suspected, tt..i' relationship will be apparent as a straight line or curve on the scatter diagrarr. While statistics cannot prove cause and effect, a probable relationship is expo--- that can be corrected and then studied for results.

Using the seven tools of quality control makes it possible to reduce defects an- thus improve quality. For example, in the case of hydraulic leaks, it may be foun - that the loose connections are caused by the torque adjustment of the tools and the improper training of operators. After these causes are corrected, the number o: defects will be reduced. It is then possible to move to the second problem, whic is cracked connectors. In this way, continuous improvement is achieved.

Once improvements have been made, the new process should be stabilized rc hold the gains by using a new control chart. The limits of the original control char:: before process improvement will be narrowed by the improvements, and the pr~ cess capability will be increased. In this way, what were common causes in the original control chart have been found and removed from the process.

9.8 SIX SIGMA

The tools of quality control and improvement should be used in an organized fash- ion by firms . Tools by themselves will not lead to improvement; they need to be incorporated into an improvement approach such as Six Sigma. See the Operations Leader box "Six Sigma Quality" for background information on this approach.

Six Sigma is a systematic method for process improvement that often uses the five steps defined by the acronym DMAIC:

1. Define: The process is selected for improvement, and the project charter is specified.

2. Measure: Quality variables valued by the customer are measured, and goals are set for improvement.

3. Analyze: The root causes of the current defect levels are identified, and alterna- tives are considered for process changes.

4. Improve: The process is changed and checked for improvement.

5. Control: This step ensures that the process improvement is not lost over time.

The Six Sigma approach can be applied to processes in manufacturing, service, or administrative areas.

Before improving a specific process, however, management should make a strate- gic choice of process. Senior management should choose critical processes that are

Chapter 9 Quality Control and Improvement 205

Operations Leader Six~Sigma Quality

Motorola invented the term Six Sigma quality in the mid-80s to reflect a desire for very high levels of consis-

"MOTOROLA tent quality in all its processes. A level of Six Sigma equates

to a defect level of 3.4 parts per million (ppm), which is much better than most processes in companies can achieve.

Six Sigma quality is related to the normal probability distribution, with sigma (a) denoting the standard de- viation of the process. Motorola assumed that the pro- cess mean would experience a 1.5a shift before a ±6a change. Thus, Six Sigma corresponds to a 4.5a deviation on one side of the mean and a 7.5a deviation on the other side of the mean, resulting in 3.4 ppm defects. This can be verified by referring to the normal probabil- ity tables tails at +4.5 and -7.5 standard deviations.

The Six Sigma criterion is equivalent to a process capability of Cpk = 1.5. This can be seen by referring to the formula for Cp k with USL - fl. = 4.5a. As a re- sult, the Six Sigma criterion will ensure processes that are not just barely capable of producing the specifi-

cations but will provide somewhat better process capability.

Six Sigma quality can be applied not only to manu- facturing processes but to administrative and service processes as well. For example, Motorola has applied Six Sigma in the finance department. It measures the cycle time to close the books at the end of the month. It used to take a couple of weeks to close the books; now they are closed in about three days. The depart- ment also measures how many errors were made in closing, tracking the sigma level each month .

In another application, Texas Instruments uses the Six Sigma criterion to improve levels of customer satis- faction. A defect is defined as a level of customer ser- vice rated by the customer as less than satisfactory. The percentage of defects from a large survey of cus- tomers is then equated to the corresponding sigma level and tracked on a periodic basis for improvement.

Source: Adapted from Pandu R. Tadikamalla, "The Confusion over Six-Sigma Quality," Quality Progress, November 1994, pp. 83-85; Karen Bemowsk i, "Motorola's Fountain of Youth," Quality Progress, October 1995, pp. 29-31; www.isixsigma .com, 2012 .

needed to implement the strategy of the firm. For example, top management may determine that the sales processes, the processes for hiring new employees, or a par- ticular manufacturing process should be selected for improvement.

Once a process is selected for improvement a cross-functional team is formed since most processes cut across functional lines. A full-time trained process im- provement specialist, usually called a "black-belt," is chosen to lead the improve- ment team. The team then sets out to make improvements by using the DMAIC approach.

The team begins improvement by flowcharting the process and defining process defects, using measures that are critical to the customer. Data are collected on these measures to establish a current process baseline and goals for improvement. For example, if the process is currently producing 1 defect in 100 opportunities (1 per- cent defective), a goal might be to improve the process to 1 defect in 1000 opportu- nities, a 10 times (lOX) improvement factor. This kind of aggressive improvement approach is taken by Six Sigma teams to ensure that significant change is achieved. Of course, the improvement goal must not be set arbitrarily; rather, it is set on the basis of the economic benefit of improvement and the time available for the team to accomplish its goal, with six months being the typical project length.

Once the goal has been set, the team seeks the root causes of the current defect levels. The team must be careful to go beyond symptoms and find the real causes. This is often done by brainstorming and careful collection of data to analyze the situation. A variety of tools such as CE diagrams, scatter diagrams, and Pareto charts are used at this stage.

206 Part Three Quali ty

Operations Leader TRW Improves Legal Processes with Six Sigma

Think of automotive suppliers and TRW Automotive . comes immediately to mind . TRW Automotive, head- quartered in Livonia, Michigan, is one of the largest

and most profitable global suppliers of automotive compo- nents and subsys- tems-from braking systems to steering systems to occupant safety systems to advanced-technol- ogy systems that make driving safer for drivers, passen-

gers, and those who might be in the path of the ve- hicle. In 2011, TRW Automotive realized $16.2 billion in sales. Its 60,000+ employees working in more than 185 locations worldwide currently supply over 40 automotive assemblers and 250 vehicle brands.

The success of TRW Automotive can be described as being anchored on two pillars: pursue product and pro- cess innovations and, at the same time, continually im- prove the quality of its products and processes. For TRW Automotive, continuous improvement is not restricted to just manufacturing processes. It uses Six Sigma to im- prove and realize financial gains from all business pro- cesses cutting across the entire organization.

One example of Six Sigma deployment in a non- manufacturing process at TRW occurred in the legal department. At one point, the legal department and associated business units had to review about 2500 trademark registrations each year at a cost of $1200 per registration . TRW Automotive suspected that many trademark registrations were no longer needed or were no longer critical to its businesses and needed a way to determine which these were in a rigorous, replicable manner.

The Six Sigma method was deployed, with the start- ing point being the definition of a defect as either an

unnecessary registration that was renewed or a nec- essary registration that should have been but was not renewed. In the case of the latter, this occurrec because response for verification from the business unit was not received before the trademark expired. Based on this definition, the baseline defect ra te was computed to 25 percent; one in four registra - tions was essentially a defect.

To improve the process of renewing trademark registration, a flowchart was created to show all4 1 steps that were involved at the time. Investigatio n of these 41 steps revealed that only 11 were value- added steps. Eliminating the steps that were not adding value then streamlined the process an d saved time .

Data were also gathered to study why the defect rate was so high. Some hypothesized causes in- cluded waiting time for approval from business units, waiting time evidence to be provided about usage of a trademark, and whether a trademar k was registered in the United States or in a foreign country. Applying regression analyses to the data helped to pinpoint the root causes contributing to the high defect rate.

Root causes were then targeted for elimination with the design of a new trademark renewal pro- cess. Employees were subsequently trained on the new trademark renewal process. A new control plan was also implemented to prevent the same root causes from reappearing to create defects. As a re- sult, there was a drastic drop in the defect rate from the original 25 percent to just 0.33 percent; that is, by following the new trademark renewal process, only 1 out of 300 renewals became a defect. More- over, by reducing the defect rate and streamlining the trademark renewal process, TRW Automotive was able to save $1.8 million .

Source: R. Das, H. Collelo, and H. Davidson, "Six Sigma in Corporate Law," Six Sigma Forum 4, no. 1 (2004), pp. 30-36; www.trw.com, 2012.

After root causes are found, alternatives for improvement are considered and improvements are made. Then, further data are collected to ensure that improve- ments have occurred, savings have been generated, and a control plan is put in place to ensure that the changes are permanent. Quality control charts can be used at this point to maintain the new process in a state of statistical control.

While the use of Six Sigma is well established in manufacturing, Six Sigma can also be deployed to improve nonmanufacturing processes. The context may b e

Chapter 9 Quality Control and Improvement 207

different but the DMAIC steps are equally applicable, as are the underlying prin- ciples. For an example of how Six Sigma can improve nonmanufacturing pro- cesses, even those in the legal department, see the Operations Leader box titled "TRW Improves Legal Processes with Six Sigma."

Six Sigma has produced dramatic results in companies such as Motorola, General Electric, Citigroup, American Express, and Honeywell. These results are possible only through aggressive senior management leadership, widespread training in Six Sigma, use of full-time improvement specialists, and careful tracking of financial results. 5 Six Sigma is not merely a quality improvement approach but also a way to improve the net income of the company. GE, for example, reports adding more than $2 billion to the bottom line from Six Sigma.6

9.9 LEAN AND SIX SIGMA

"Six Sigma at Caterpillar,"

Vol. XIII

TABLE 9.5 Comparison of Lean and Six Sigma

Companies are starting to combine lean and Six Sigma programs. While these pro- grams are complementary in nature, they also have several important differences in objectives, organization, methods, and type of projects. Table 9.5 compares dif- ferences between lean and Six Sigma. These differences are based on typical im- plementations of lean and Six Sigma, but definitions and usage vary widely in practice.

Lean systems have the objective of eliminating waste defined as non-value-adding activities. Six Sigma efforts, by contrast, are aimed at reducing defects as seen by the customer. Thus, the objectives of these efforts are different, but at the same time they can be overlapping. For example, lean can attack defects as one of the seven wastes. Six Sigma can attack waste when it is causing a defect in the customers' eyes, but would not normally attack internal waste such as excessive inventory, wasted

Differences

Objectives

Organization : Team leadership Use of champions Workforce involveme nt

Training Method used :

Steps followed Data emphasis Flowcharting Use of pull systems Tracking financial impact

Type of projects : Project complexity Time for comp letion Number of projects Project selection

lean

Reduce waste (non -va lue- added activities); w aste can include, in part, defects

Part-time leaders (usua lly) Champions not used Everyone is involved

One week of training

Five -step lean thinking Less data-driven Value stream mapping Yes, pull from the customer Not typically done

Simple projects One week or less Many small projects Not necessarily strategic

Six Sigma

Reduce defects . Defects can include, in part, some non-value-added activit ies

Fu ll-time leaders (usually) Vice president champions Selected emp loyees fo r each

project Four weeks for a black belt

DMAIC or DFSS steps Statistical emphasis Any flowchart method Not part of Six Sigma By financial organization

Complex projects Typically six months Fewer la rge projects Strategic project selection

5 In a few cases compan ies have used part-time improvement spec ialists rather t han f ull-time emp loyees. 6 GE Annua l Report (1999).

208 Part Three Quality

motions, or unnecessary movement of materials. While Six Sigma tends to reduce variance in processes, lean improves process flow.

Another difference is in the way these programs are organized. Six Sigma typi- cally relies on full-time black belts as project leaders and vice president champions who oversee the projects. In contrast, lean systems rely on part-time project leaders and a more informal hierarchy. Lean improvements often involve the entire workforce, while Six Sigma projects are more selective in workforce involvement. Six Sigma black belt training is extensive and usually requires four weeks of training plus successful completion of one or more projects. Lean training is more informal and typically lasts only a week. Six Sigma programs are therefore differ- ent in their organization structure.

A third difference is in the methods employed. Lean thinking relies on a five- step process that starts with the customer need. While Six Sigma also starts with a customer need, the DMAIC sequence of steps differs from those used in lean thinking. Also, lean systems do not stress the use of data or statistical analysis to the same extent as Six Sigma. As noted in Table 9.5, value stream mapping is used in lean while Six Sigma does not use a particular type of flowcharting. Another important difference is that only lean uses the concept of pulling demand from the customer to flow the product or service. Finally, lean does not formally track proj- ect cost savings or revenue improvements, while Six Sigma insists on careful tracking by the finance organization.

The last area of difference is the type of projects undertaken. Lean projects are usually simple, but Six Sigma is used for complex" and difficult process improve- ment projects. A typical Six Sigma project takes six months and is aimed at a large impact, about $200,000 in savings or more. Lean projects can last as little as one week using kaizen events and often have much less impact from each project. Lean programs will attack many more small improvement projects than Six Sigma and may not select projects for strategic importance.

Both Six Sigma and lean are aimed at improvement, but in a different way in typical applications. Thus, an organization already using lean can benefit from Six Sigma by attacking larger complex projects with a more formalized data-driven approach, using full-time project leaders to reduce defects and variance. Con- versely, an organization using Six Sigma can benefit from the fast-hitting small- scale kaizen approach of lean that utilizes a large part of the organization, aimed at eliminating waste and improving process flow. Similarities are that both ap- proaches start with identifying a true customer need that is not being met, and both are focused on process improvement.

Some companies are developing an integrated lean and Six Sigma approach. For example, they might use the DMAIC methodology and full-time project lead- ers from Six Sigma, but then incorporate the value stream mapping, pull systems, and waste reduction focus of lean as part of the DMAIC steps. This approach would attack waste, improve flow, and reduce defects (variance). It could attack simple or complex process improvement projects with varying applications of methods to suit the particular project.

9.10 QUALITY CONTROL AND IMPROVEMENT IN INDUSTRY

Industry has made widespread use of the quality control and improvement meth- ods described in this chapter, as indicated by several surveys of industry practice. These surveys indicate that about three-fourths of all firms use process control charts. There is greater use of x and R charts than p charts due to the small samples

Chapter 9 Quality Control and Improvement 209

that are possible with variables control. Other, more sophisticated charts are not as widely used as :X, R, and p charts.

Six Sigma is rapidly gaining acceptance in manufacturing and service indus- tries as a proven approach for using the seven tools of quality control. Although there is no reliable survey of the adoption of Six Sigma, the list of companies reporting Six Sigma implementation continues to grow.

While use of quality methods has spread beyond operations, there is much room for further use in administrative and office functions in manufacturing firms and service firms. Quality control education must be concentrated on all functions in the firm and on the firm's suppliers as well. Some firms have still not internal- ized the concept of total quality management. As a result, quality control efforts are focused primarily on production operations.

Quality control in the service industries has lagged behind that in manufactur- ing for several reasons. First, services are more difficult to measure because they are intangible, whereas the characteristics of a manufactured product can be measured and specified. For example, steel can be measured by its strength, hardness, ductil- ity, and other properties. The quality of a service is related to intangibles such as atmosphere in a restaurant, the waiter's smile, and the customer's sense of well- being. Nevertheless, quality cannot be controlled unless it is measured. Therefore, it is imperative that the service industries measure what they can and develop new, innovative measurement techniques for what is now considered intangible.

One way to measure service is to quantify the transactions that take place . Transactions might include the number of tables served in a restaurant per em- ployee, the percentage of customers who are satisfied or very satisfied with the service, and the number of power outages at an electric utility company. Once the important dimensions of service delivery are determined, a way can be found to measure these dimensions with either objective or perceptual data.

Another ,way to measure service is to use the dimensions of SERVQUAL, described in the previous chapter. These perceptual measures can help management understand and improve the quality of service offered.

A characteristic of managing quality of services is the perishability of the prod- uct, which requires that quality be controlled while the service is being delivered. As a result, a great burden for service quality is placed on the workforce; when bad quality is delivered, the customer is immediately aware of it. Thus, service organizations should emphasize selection of the proper employees, workforce training, and process control. Of course, these are also good practices for manu- facturing firms to prevent errors from occurring.

Why should accounting, human resources, marketing, and finance people be interested in the ideas expressed in this chapter? First, accounting is interested in accurate costs and financial information. When all production processes are in sta- tistical control, the special causes of variation in the processes and thus in costs have been eliminated. Accounting can also benefit directly by applying these ideas to the control of quality in the input transactions received by the accounting sys- tem and the accounting outputs produced. In other words, they can bring the ac- counting system of the company under statistical control. Outside auditors can audit whether the processes producing accounting transactions are under control, rather than just auditing the transactions themselves.

From a human resources perspective, the ideas in this chapter offer many possi- bilities. Implementation of statistical quality control and improvement requires in- depth training of the workforce. The workers are no longer blamed for errors that are in fact due to the underlying system. Workers take more pride in their work

210 Part Three Quality

~~

when they are responsible for inspecting their own output and controlling their processes. A greater sense of satisfaction and productivity occurs when people contributing to reducing errors and satisfyirtg their customers (the next procesJ ..

Needless to say, marketing does not like to see dissatisfied customers who h2. received a defective product. Implementing the ideas of statistical quality con~ reduces the number of defects produced in a never-ending cycle of improvem As a result, there are fewer customer complaints and revenues can be increased:- marketing the consistent quality of the company. Marketing and sales are rna- easier by aggressive implementation of the ideas in this chapter.

Finally, finance can see the results of a progressive quality control a.:- improvement approach in the bottom line. Financial processes can also be co~ ­ trolled, and a company that implements these ideas will save money and impro' its financial results.

9.11 KEY POINTS AND TERMS

This chapter introduces methods for managing and improving output quality. Tn major points include the following:

• Quality control is defined as the stabilization and maintenance of a process : produce consistent output. Continuous improvement can occur once a stable- process is achieved.

• Operations consists of a sequence of interconnected processes, each with i~ own internal customers. Critical points must be defined for inspection an- measurement to control and improve these processes.

• Process control charts should be considered for the inputs (by the suppliers), a:: part of the process, and for the outputs. The critical control points are best de- scribed by a flowchart of the process.

• Using process quality control, periodic samples are taken from a continuoUE production or service process. As long as the sample measurements fall withir. the control limits, production is continued. When the sample measurements fa[ outside the control limits, the process is stopped and a search is made for ar. assignable cause-operator, machine, or material. With this procedure, a produc- tion or service process is maintained in a continuous state of statistical control.

• It is preferable to use statistical process control instead of inspection whenever possible, because SPC is prevention-oriented. SPC can be used as a basis for internal quality control and achieving a certified-supplier status, which requires a stable production process.

• Six Sigma is an organized and systematic approach to process improvement. It of- ten utilizes the five DMAIC steps: define, measure, analyze, improve, and controL Careful analysis using statistical tools is needed to identify the root causes of de- fects perceived by customers, analyze changes, and control the improved process.

• Companies are now combining lean and Six Sigma process improvement ap- proaches. While these approaches both start with current customer needs, they are different in their objectives, organization, methods, and types of projects . However, they are complementary in seeking process improvement and can be used in an integrated fashion.

• There are seven tools of quality control and improvement. These methods can be used to bring a process under control or to improve it.

Key Terms

STUDENT INTERNET EXERCISES

Chapter 9 Quality Control and Improvement 211

• A high percentage of manufacturing companies use the seven tools of quality control. However, the use of these methods has less acceptance in service indus- tries and administrative functions.

• Every function in the company can benefit from the application of the ideas in this chapter. All functions should bring the administrative processes they man- age under statistical quality control and improvement. Other functions outside operations, and the company, will also benefit directly when quality control and improvement principles are used.

Walter A. Shewhart 189 Process definition 190 Statistical quality

control 190 Continuous

improvement 190 Statistical process

control 190 Internal customers 190 Critical control points 190 Operator inspection 190 Supplier certification 191

Variables measurement 192

Attribute measurement 192

Assignable cause 193 State of control 193 Center line 193 Upper control limit 193 Lower control limit 193 Special cause 195 Common cause 195 Process capability 198

1. American Society for Quality (ASQ) http:/ /www.asq .org

Seven tools of quality control 201

Flowchart 202 Check sheet 202 Histogram 202 Pareto diagram 202 Cause-and-effect

diagram 203 Scatter diagram 204 Control chart 204 Six Sigma 204 DMAIC 204

Click on the Knowledge Center tab, then click on the Advanced Search link. Check the boxes for "Quality Progress/' "Six Sigma Forum Magazine/' and "Case Study." Type statistical process controt six sigma, and lean six sigma into the Keyw<Jt d field and click Find. Read one of the search results and prepare a brief summary of what you learned to share in class.

2. YouTube http://www.youtube.com

Search for videos about statistical process control, six sigma, or lean six sigma. Watch one video and come to class prepared to explain what you learned.

3. iSixSigma http://www.isixsigma.com

After finding the home page, click on "New to Six Sigma" and read more about Six Sigma history, DMAIC, certification, and the like.

SOLVED PROBLEMS.

oblem 1. P Control Chart A company that makes golf tees controls its production pro- cess by periodically taking a sample of 100 tees from the production line. Each tee is inspected for defective features. Control limits are developed using three stan- dard deviations from the mean as the limit. During the last 16 samples taken, the proportion of defective items per sample was recorded as follows:

.01

.00

.02

.01

.01

.03

.03

.02

.02

.03

.01

.02

.00

.01

.02

.00