STATISTICS

Use SPC for

Everyday Work
Processes
by Greg Gruska and Chad Kymal

S
tatistical quality control (SQC), also evolved into statistical process control (SPC) to
known as control charting, started with reflect the move away from product control to a
Walter Shewhart’s work at the Western systems focus.
Electric plant outside Chicago in the 1920s. But why must SPC be periodically revitalized?
Since then, SQC has been reintroduced into If it is all people say it was and is, shouldn’t it be
industry every couple of decades or so and has self-sustaining? Partly the problem is that if times
are good, management focuses not on economic
control but on volume control. So we see many
organizations embracing SPC only during times of
trouble. When times are good, the attitude is “We
In 50 Words don’t have time for such luxuries.”
Or Less Even organizations that implement SPC as part of
their continual improvement efforts fail to sustain
• Theoretically, statistical process control (SPC) its use, sometimes because the results of applying
is viewed as useful for economically producing SPC to processes have a variation model different
from the one shown in most books. It is a case of
consistently acceptable products and services.
using the right toolbox but the wrong tool.
To help organizations use SPC tools the right
• In practice, however, SPC isn’t being used for way, the Automotive Industry Action Group’s
continually improving unique industrial supplier requirements task force, representing
General Motors Corp., Ford Motor Co. and
processes.
DaimlerChrysler, recently released a second edi-
tion of its SPC Manual.1 The entire first chapter
• The right tools, such as advanced charts, can explains the philosophy and use of SPC, main-
make SPC effective in these situations. taining it should not be applied to processes but
integrated into an organization’s continual
improvement activities.

QUALITY PROGRESS I JUNE 2006 I 25

STATISTICS

The edition discusses a useful three-stage improvement, but it must be within a plan-do-
improvement cycle for integration (see Figure 1): study-act cycle, not haphazardly applied without
1. Analyze the process. an understanding of its impact.
2. Maintain (control) it. Right idea/wrong tool, or not understanding
3. Improve it. the physics of the product and process. Applying
SPC without understanding the physics of the
Deployment Shortcomings product or process and the dominant sources of
Despite the advantages of SPC, why have many variation will lead to frustration among both oper-
organizational implementation efforts not been ators and management. Much of this happens
successful or self-sustaining? because most people have been exposed to only
Many of the contributing causes have nothing to basic SPC control charts.
do with the underlying methodology but with the Although the four basic variable charts and four
organization and deployment. Some examples fol- basic attribute charts are applicable to a wide vari-
low. ety of processes, advanced charts are better suited
Constant change. SPC assumes the process con- to many processes. The term “advanced” does not
trols maintain the common cause variation system. necessarily imply the use of more sophisticated sta-
All too often this is not possible because there are tistics. Often, these charts are a modification of
ongoing changes to the process resulting from: basic charts for specific conditions of the process to
• Special causes of variation. optimize the detection of special causes.2
• Physical changes to the process—with the If they use the wrong charts, the operators will
intent of improving it. not see any benefit from the extra work necessary
• Administrative changes to the control activi- for the SPC implementation, and management will
ties for logistical—or whimsical—reasons. still see inconsistencies in the process output.
• Changes in management direction regarding Limited understanding. SPC’s application is
what is desired or needed. often limited to processes similar to the examples
Change is a necessary element of continual provided in an SPC class. But SPC can be useful in
a wide variety of sectors
outside manufacturing.
Within healthcare, many
FIGURE 1 SPC Improvement Cycle organizations such as the
Joint Commission on Ac-
1. Analyze the process creditation of Healthcare
• Determine what the process
Organizations and the
should be doing.
2. Maintain the process Institute for Healthcare
• Determine what can go wrong.
• Monitor process performance. Improvement have recog-
• Determine what the process is doing.
• Detect special cause nized the need to under-
• Achieve a state of statistical control.
variation and act on it.
• Determine capability. stand common and special
causes of variation and the
Plan Do Plan Do use of SPC in process analy-
1 2 sis. This goes beyond ad-
ministrative processes to
Act Study Act Study
also include clinical pro-
cesses and improvement
actions.3
Lack of patience. Even
Plan Do when the SPC deployment
3. Improve the process
3 is the right idea using the
• Change the process to better
understand common cause variation. right tool, management and
Act Study
• Reduce common cause variation. workers seem to expect

26 I JUNE 2006 I www.asq.org

instant gratification. When they don’t immediately is not enough. It sometimes can be difficult to dis-
observe consistency and improvement, manage- tinguish between a white noise and a merely sta-
ment may withdraw support or workers may not tionary process—in fact, the white noise process is
follow through. often called a weakly stationary process.
Several charts can monitor and control a station-
Assumptions ary process:
W. Edwards Deming taught us the role of man- Autoregressive charts. These include the autore-
agement is to make predictions.4 The purpose of gressive, and autoregressive and moving average
SPC activities is to enable management to predict models. This approach seeks to model the underly-
the future state of a process by identifying and ing relationships among the process output values
ameliorating special causes of variation. For SPC to and use this knowledge to better identify other
be implemented effectively, some assumptions special causes of variation.
hold: Cumulative sum (CUSUM) and exponentially
• Variation and interdependencies exist in all weighted moving average charts (EWMA).
things. Although the CUSUM and EWMA were devel-
• Few systems and processes are constantly sta- oped to detect small shifts in the mean in random
ble. processes, they are robust enough to handle
• When applying the basic control charts, real- processes with minor autocorrelation.
ize the process being analyzed, monitored or Individuals (I) and moving range (MR) charts.
controlled must be a purely random (or white If the within subgroup variation is less than or
noise stochastic) process. equal to the discrimination of the measurement
A random process satisfies: system appropriate for the process, an I and MR
chart may be a suitable way to control the process
E [ε t ] = µt = µε for all t. variation. However, very strong autocorrelation
may still display itself in a nonrandom pattern.
V ar (t) = σ t2 = σ ε2 for all t.
Structured samples. If the source of the autocor-
ε t is uncorrelated with ε t – k for all k. relation is a consistent and predictable special
cause, the selection of the sampling quantity and
A common deviation from the standard assump- frequency should reflect this dominant source of
tions lies in processes with outputs correlated with variation. For example, if the process is material
each other. Some include stamping, machining that
is tool wear dominant, chemical processing, the
stock market and an individual’s medical readings
(for example, temperature, blood sugar level and
blood pressure). FIGURE 2 Rapidly Drifting Process Center
These processes are called stationary processes
and satisfy:5 Upper specification limit

E [xt ] = µt = µx for all t.

Var ( xt ) = σ t2 = σ ε2 for all t.
with the correlation between xt
and xt – 1 equals ρk .

This is also called an autoregressive process or a

process with autocorrelated data.
The Shewhart chart control limits and the stan-
dard calculations for capability indexes depend on Lower specification limit
the assumption of a white noise process. Stationary Time

QUALITY PROGRESS I JUNE 2006 I 27

STATISTICS

dominant, the sampling should occur whenever How Common Are Common Causes?
the material changes—for example, with the Shewhart charts require the center of the process
change of coils. to remain constant over time and the variation due
Structured charting. If the source (special cause) to only common causes for process behavior to be
of the autocorrelation is predictable, it is possible predictable.
However, one type of special cause is not part of
the common cause variation but, within bounds, is
predictable. This type of cause is often called an
Allow the operator to stay as economically allowable special cause (EC), because
only minor cost benefit results from its elimination.
close as possible to traditional Another name for this type is environmental cause.
If the EC has a consistent and predictable behav-
Shewhart procedures. ior, the behavior of the total process variation can
also be predicted within bounds. If the EC is incon-
sistent or exhibits chaotic behavior, the control
methods in the example that follows will not be
effective.
A classic example of a process with an EC is a
to control the process by segregating the within screw machine. In this process, the tool wear is
subgroup variation from the between subgroup rapid. Figure 2 (p. 27) is an example of a rapidly
variation on separate charts. The between-within drifting process center, often evident within the
chart uses an I and MR chart approach and the typ- span of a single shift. Because of this EC, the process
ical range chart: can have additional variation caused by setup varia-
• The I chart plots the subgroup averages as tion.
individuals against the control limits based on Organizations often use conventional control
the moving ranges. limits, with the process center forced to be at the
• The MR chart plots the between subgroup midpoint of the specification limits. This can result
variation using the moving ranges based on in overcontrol and decreased productivity.
the subgroup averages. A screw machine is used to manufacture small
• The range—or standard deviation—chart shafts. The shafts are produced continuously on
plots the within subgroup variation (common the same machine by two shifts per day, six days
cause variation). per week. The measurement under study is the flat
width of the shaft spindle.
The process center m increases rapidly as the
machine tools wear. The process may run for an
FIGURE 3 Sample Averages entire eight-hour shift before tool maintenance is
required. The process standard deviation, σX,
remains fairly constant during the course of a tool
68
wear cycle.
Once the effect of tool wear is removed, there is
Average for sample

64
strong evidence the product measurements are
being generated by a white noise process.
60 The objectives of the control plan are:
• Keep tool change time to a minimum.
56 • Minimize operator overcontrol.
• Maximize the length of the tool wear cycle.
52 • Ensure the process remains in control once the
0 10 20 30 40 50 60 70 80 90 100 tool wear variation has been taken into
Sample number account.

28 I JUNE 2006 I www.asq.org

 FIGURE 4 Sample Data How To Do It Right
The first step in dealing with tool
wear trends is to collect the right kind
68
of data. For this process the approach
selected is:
subgroup reading
Average value of

64
• Draw samples of size three once
every 675 pieces, which keeps the
60 sampling frequency close to its
original value of a sample once
56 every hour.
• Use a pan holding approximately
52 675 pieces to determine when a
0 10 20 30 40 50 60 70 80 90 100 sample needs to be drawn.
Sample number The values from the first 101 samples are
shown in Figure 3. Each dot represents the average
for a sample of size three.
• Minimize the probability of producing a non- It is difficult to see tool wear cycles in these data.
conforming part. Examination suggests tool wear trends in the first
The first step in trying to meet these objectives is part of the data. These trends can be highlighted by
to realize once again the fundamental equation of removing the lines connecting the last point in one
process control must be revised in this situation to cycle with the first point in the next cycle.
read: Extensively annotated control charts provided
by the operators helped identify individual cycles.
Total variation = common cause variation + tool It was determined most of the later data were not
wear variation + other special cause variation. collected under the proper conditions and could
not be used for estimating tool wear.
Isolate and measure tool wear variation sepa- Tool wear patterns lurking in the data emerge
rately from other sources of variation. Understand once extraneous lines and dots are removed (see
the trade-off between long tool wear cycles and the Figure 4).
need to deal with fussy tool change problems. The first step is to determine whether the tool
Allow the operator to stay as close as possible to (the special cause) is consistent over time (tool-to-
traditional Shewhart procedures. tool variability is predictable). To find a common

FIGURE 5 Superimposed Patterns With FIGURE 6 Random Sample

Regression Line and X-bar Chart
62 Slope = 0.7962
Subgroup average
subgroup reading
Average value of

60

58

56

54

52
0 1 2 3 4 5 6 7 Subgroup or time
Sample number (675 pieces per pan)

QUALITY PROGRESS I JUNE 2006 I 29

STATISTICS

trend in the various tool wear patterns, all cycles exhibited the properties of white noise.
were superimposed and a single simple linear The tool wear slope of 0.7962 was estimated
regression model was fit to the entire collection of using a simple linear regression model. This regres-
tool wear trends (see Figure 5, p. 29). sion model also estimated a standard error of
The regression model explained 77% of the 0.0571 for the slope coefficient. Three standard
data’s variation. The unexplained variation error tool wear growth limits can be calculated by
± (3 x .0571) applied to the expected aver-
age line.
When someone is manually controlling
FIGURE 7 Control Chart processes with a tool, he or she can lay a
transparency of Figure 6 (p. 29) on top of
the control chart.
The control limits in Figure 5 (p. 29) are
Subgroup average

the same as those in a conventional X-bar

chart except they follow the average tool
wear line rather than a horizontal center-
line. This approach requires the special
cause—tool wear—to exhibit a predictable
and consistent behavior, which requires
control by the supplier and purchasing.
Subgroup or time The second step is to verify and quanti-
fy the tool wear behavior over time by
studying a random sample of tools over
FIGURE 8 Controlling Features their life (see Figure 6). This may require
100% sampling if the tool life is short.
This analysis must determine whether
the expected tool life pattern and common
cause variation are consistent—and thus
Change control predictable. Then this information should
be used to establish process control charts
Wear control to control the process variation and tool
wear variation (see Figure 7).
Setup control As confidence is gained in the process,
the diagonal control charts can be replaced
Features by I charts monitoring specific tool life fea-
tures (see Figure 8):
FIGURE 9 Controlling Features — • Setup control. Ensures the setup—the
Continual Improvement starting point of the process trend—of
each tool is consistent. This, with the
change control, also determines the
capability indexes of the process.
• Wear control. Ensures there is no
Change control change in the tool life model.
• Change control. Ensures the useful
life of the tool has been reached and
the end life model of the tool has not
Setup control
changed. This, with the setup control,
also determines the capability indexes
Expected number of parts of the process.

30 I JUNE 2006 I www.asq.org

 As confidence in the process and FIGURE 10 Controlling Features —
tooling increases, monitoring can use Further Continual Improvement
the tool life or number of parts before a
required change (see Figure 9).
In this stage of the continual improve-
ment process, there is confidence the
tooling wear trend will be consistent and
acceptable due to the actions of the sup-
pliers of the tooling and materials. The
need for redundant inspection can be Setup control
eliminated.
As confidence in the process and
tooling further increases, monitoring Expected number of parts
can use the tool life and periodic sys-
tems audits (see Figure 10).
Based on process knowledge gained, FIGURE 11 Controlling Features— Continual
there is confidence the tooling wear Improvement Final Stage
trend and life will be consistent and
acceptable due to the actions of the
suppliers of the tooling and materials.
The need for redundant inspection can
be eliminated.
In the final stages, setup control
moves offline, and control uses the
tool life or block tool change (see
Figure 11).
In this stage of the continual improve-
Expected number of parts
ment process, there is confidence:
Offline
• The tooling wear trend and life setup control
will be consistent and acceptable
due to the actions of the suppliers
of the tooling and materials.
• The variation caused by the machine will be need to understand the underlying model of vari-
consistent and acceptable due to preventive ation and physics of the process. To enable the
maintenance. reader to select the appropriate tool for a specific
• Setup consistency will be controlled by offline process model, the manual includes sections on
setup activities. the following charts:6
The need for any online inspection can be elimi- • Probability based charts.
nated. This does not eliminate the need for period- • Short run control charts.
ic system and product audits to verify the process • Charts for detecting small changes.
controls are still valid or ensure an unknown spe- • Non-normal charts.
cial cause hasn’t crept into the process. • Multivariate charts.
At any stage, the controlling cycle must restart if: • Regression control charts.
• The tooling exhibits erratic behavior. • Residual charts.
• A different vendor begins to supply the tool- • Autoregressive charts.
ing or material. The manual describes and identifies the use of
each and provides references for further study but
Other Types of Control Charts provides how-to instructions only for the basic
The new edition of SPC Manual discusses the charts.

QUALITY PROGRESS I JUNE 2006 I 31

STATISTICS

Warnings GREG GRUSKA, a Fellow of ASQ, is the vice president of

It is important to understand the real meaning product development for Omnex Systems and a principal
of “random.” Selecting a random sample requires consultant in performance excellence and a Six Sigma Master
specific techniques. Black Belt for Omnex, Ann Arbor, MI. He directed the devel-
In practice, many people think blind selection opment and initial implementations of Comprehensive
is random selection. In reality, this may be hap- Process Control Planning, a book published by Omniface
hazard or convenience sampling. Using haphaz- Corp. Gruska is a writing member of the measurement sys-
ard or convenience sampling when random tems analysis, SPC and failure mode and effects analysis
sampling is required can lead to biased and
manual subcommittees of the Automotive Industry Action
therefore erroneous conclusions.
Group supplier quality requirements task force, which is part
SPC is useful and necessary for continual
improvement, but many applications do not of the international task force governing ISO/TS 16949. With
realize its full benefit because they lack knowl- master’s degrees in mathematics and engineering from Michi-
edge of the tools and the processes to be ana- gan State University and Wayne State University, Gruska is
lyzed, maintained and improved. also an ASQ certified quality engineer. He has been a mem-
Although basic control charts, covered in all ber of the board of examiners and a judge for the Michigan
introductory SPC courses, have a wide applica- Quality Leadership Award since 1994.
tion to random processes, there also are many
stationary processes or ones with predictable
ECs that need advanced charts or the application CHAD KYMAL, an international trainer and consultant, is
of a basic chart in a manner reflecting the actual chief technical officer and founder of Omnex Inc. He wrote
process nature. the ISO/TS 16949:2002 Implementation Guide and the
If the goal is to eliminate the need for charting Auditor Handbook to ISO/TS 16949:2002—A Guide
by building knowledge of and confidence in the
to the Automotive Process Approach to Audits, both
process, SPC charts need to be used so organiza-
published by Paton Press. He has served on the Malcolm
tions can increase their understanding of the
Baldrige National Quality Award board of examiners and
common causes and special causes affecting their
processes. Then they can replace SPC charts with as an RABQSA certified lead auditor. Kymal has a master’s
robust policies and techniques governing process degree in industrial and operations engineering and an
control. MBA, both from the University of Michigan.

REFERENCES AND NOTES

1. SPC Manual, second edition, Automotive Industry

Action Group, 2005.
2. The SPC Manual, second edition, includes discussion
and references to many of these advanced charts.
3. “Improving Heart Failure Care Through
Education,” www.ihi.org/ihi/topics/improvement/
improvementmethods/improvementstories/
improvingheartfailurecarethrougheducation.htm.
4. W.E. Deming, The New Economics: For Industry,
Government, Education, second edition, MIT Press, 2000. Please
5. There are several classifications of stochastic
comment
stationary processes. This example is just one type. If you would like to comment on this article,
6. The manual does not maintain these are all the
please post your remarks on the Quality Progress
possible charts that can be used. The ones discussed do
Discussion Board at www.asq.org, or e-mail
cover the majority of situations.
them to [email protected].

32 I JUNE 2006 I www.asq.org