WHAT IS A CONTROL CHART?

A control chart is used to monitor a process variable over time. That variable
can be in any type of company or organization - service, manufacturing, non-
profit and, healthcare. It provides a picture of the process variable over time
and tells you the type of variation you are dealing with as you move forward
with continuous improvement. This understanding of variation is the key to
using control charts effectively.

A control chart is one of many process improvement techniques. It is not the

answer to all your problems. Nor should a control chart be used alone. There
are always other process improvement tools that should be used along with
control charts.

WHAT DOES A CONTROL CHART LOOK

LIKE?

Let's take the simple example of driving to work. A possible control chart (X
chart from the X-mR control charts) is shown below.
The process variable (the time to get to work) is plotted over time. After
sufficient points, the process average is calculated. Then the upper control
limit (UCL) and the lower control limit (LCL) are calculated. Nobody sets these
values- they are determined by the process and how you sample the process.
The UCL is the largest value you would expect from a process with just
common causes of variation present. The LCL is the smallest value you would
expect. As long as the all the points are within the limits and there are no
patterns, only common causes of variation are present. The process is said to
be "in control."

WHAT ARE CONTROL CHARTS USED FOR?

We will continue with the driving to work example. There is one "out of control"
point on the control chart. It took much longer to get to work that day. You had
a flat tire - a special cause of variation. In this example, the objective is to
remove the special cause of variation to keep it from ever coming back into
your process. It is not part of the way you designed the process. Can you totally
remove the possibility of getting a flat tire on the way to work? No, but you can
significantly reduce that probability with proper tire maintenance, checking the
tires regularly, etc.

This is one purpose of a control chart - to monitor a process for those

special causes of variation that can occur and remove them so they don't
occur again. Being in statistical control is not a "normal" state. It takes work
to get a process there.

If you know the reason for the special cause, you can remove it from the
calculations. This gives the following control chart:
This process is essentially in statistical control. You found the reason for the
out of control point and recalculated the control limits. The control chart now
tells you the average of the process and the spread in the data. The average
time it takes to get to work is 25.8 minutes. And as long as the process stays in
control, that time will vary from about 18 minutes to 33.5 minutes - and it is
all due to the normal variation in the process.

Thus, another purpose of a control chart is to estimate the process

average. Another purpose is to estimate the variation (the spread in the
histogram). It is important to remember that a calculated average or standard
deviation has no meaning unless the process that generated it is in statistical
control. The average for the first chart shown above with the special cause of
variation was 27.7 minutes compared with an average of 25.8 without the
special cause - about a 7% difference.

With the exception of the day where you got the flat tire, the process is in
statistical control. You have a consistent and predictable process. Now you can
ask yourself the following question:
"Is my process capable of doing what I want it to?"

With some processes, there are specifications and process capability (Cpk)
values can be calculated to help judge whether the process is acceptable. With
others, the objective may be to continuously reduce or improve a variable over
time.

If you want to improve a variable that is in statistical control (e.g., reduce the
time it takes to get to work), you must fundamentally change your process.
That doesn't mean re-training someone or telling them about their mistakes. It
means fundamental change. In this simple example, it could be getting up
earlier, driving faster, taking a different route or moving closer to work.

Suppose you decide to get up 30 minutes earlier and leave the house 30
minutes earlier to avoid some of the rush hour traffic. You do that for two
weeks. The resulting control chart is shown below.

The control chart now shows a run below the average. It is obvious that leaving
the house earlier has reduced the time it takes to get to work. The control
limits may now be recalculated to reflect the new average and new variation in
the data. The resulting control chart is shown below.

Another purpose of a control chart is to judge the impact of your process

improvement efforts. In this example, the process changes worked, new
control limits were calculated, and the process can be monitored into the
future for the appearance of any special causes.

Note that not all special causes are "bad." Suppose a point occurs below the
lower control limit and your objective is to reduce the variable. This is a good
thing, but it is still a special cause. Something happened that is not part of the
normal process. If you can find out what happened and start doing it all the
time, then the process will be improved.
WHAT IS PROCESS CAPABILITY?

Process capability compares the process output with the customer’s

specification.

The purpose of a process capability study is to compare the process

specification to the process output and determine statistically if the
process can meet the customer’s specification.

A capable process is:

 Stable and not changing.

 Can fit within the customer’s specification with a little extra room
(usually 25%) to spare.

The less variation there is in a process, the more capable it will be of

meeting the customer’s specification.

Capable AND Centered

 When using process capability as a means for measuring how well our
process is producing compared to the customer’s requirements, it is not
good enough to just make sure the process is capable; it must also be
fairly well-centered.

MEASURING PROCESS CAPABILITY

The primary measures of process capability used are the process capability ratio, the
process capability index, and the Cpk.

 Process Capability Ratio (Cr)—Describes what portion of the specification the process
is taking up. The Cr should be no greater than 75% in order for the process to be
considered capable.
  Process Capability Index (Cp)—The inverse of the capability ratio. The Cp should be at
least 1.33 in order for the process to be considered capable.
 Cpk—The capability measure preferred in most industries because it indicates whether
it is capable and how well-centered the process is. The Cpk is the smaller of the Cpu
(capability of the upper half of the process) and the Cpl (capability of the lower half of
the process). Cpk = smaller of {Cpu, Cpl}

WHAT ARE CONTROL CHARTS?

Control charts give a statistical signal that a process has gone out-of-control so the
operator can make the necessary adjustments to bring the process back into control.

 The chart itself can’t control the process. The control chart provides a signal so the
operator can take action on the process.

A process that is stable is in-control. If the process changes, we get a signal from the
control chart that it has gone out-of-control.

Adjusting a process when it is not statistically out-of-control is considered over-

adjustment and results in increased process variation.

Control charts can be used to measure product characteristics or in-process

parameters.

 Measuring an in-process parameter is the best way to monitor a process because it

may provide a signal that the process is going out-of-control before bad product is
made.

When to Use a Control Chart

 When controlling ongoing processes by finding and correcting problems as they occur.
 When predicting the expected range of outcomes from a process.
 When determining whether a process is stable (in statistical control).
 When analyzing patterns of process variation from special causes (non-routine events)
or common causes (built into the process).
 When determining whether your quality improvement project should aim to prevent
specific problems or to make fundamental changes to the process.

Control Chart Basic Procedure

1. Choose the appropriate control chart for your data.
2. Determine the appropriate time period for collecting and plotting data.
3. Collect data, construct your chart and analyze the data.
4. Look for “out-of-control signals” on the control chart. When one is identified, mark it on
the chart and investigate the cause. Document how you investigated, what you learned,
the cause and how it was corrected.

Out-of-control signals

 A single point outside the control limits. In Figure 1, point sixteen is above the UCL
(upper control limit).
 Two out of three successive points are on the same side of the centerline and farther
than 2 σ from it. In Figure 1, point 4 sends that signal.
 Four out of five successive points are on the same side of the centerline and farther
than 1 σ from it. In Figure 1, point 11 sends that signal.
 A run of eight in a row are on the same side of the centerline. Or 10 out of 11, 12
out of 14 or 16 out of 20. In Figure 1, point 21 is eighth in a row above the
centerline.
 Obvious consistent or persistent patterns that suggest something unusual about
your data and your process.

Figure 1 Control Chart: Out-of-Control Signals

5. Continue to plot data as they are generated. As each new data point is plotted, check
for new out-of-control signals.
6. When you start a new control chart, the process may be out of control. If so, the control
limits calculated from the first 20 points are conditional limits. When you have at least
20 sequential points from a period when the process is operating in control, recalculate
control limits.

Defects:
We define Quality as conformance to customer specifications; so, is
nonconformance the same as a defect? Well, not exactly.

A defect is a glitch that prevents your product or service from being acceptable
to the customer.

Nonconformance means that you didn’t match the customer or product

specifications. They are not exactly same. Every defect means the product
doesn’t conform to specifications but not every nonconformance is a defect.

There are two types of variation in all

processes:
 Common Cause Variation
 Special Cause Variation
Common cause variation is the result of predictable errors. For example, an
admitter is going to select the wrong insurance plan every 100 admissions.
Some admitters may have better accuracy, but no admitter will be able to
always perform an admission with no errors. This is simply not possible even
with highly trained and experienced admitters. Errors that result from
common causes will predictively vary around the target performance, e.g.
admitters should have a 98% accuracy rate. An important point is that to
reduce the common cause variation the process must be improved. It will
serve no purpose to tell the admitter who chooses the wrong insurance plan
once every 100 admissions to be more careful.
Special cause variation is the result of unpredictable errors. For example, a
new admitter without proper training is put on the midnight shift of a busy
inner city emergency room. Clearly the number of admitting errors is going to
be very high until she obtains more training, coaching, and experience. How
many actual errors she will make is highly unpredictable. In this situation,
the root problem is not the process but one of the admitters. This chart will
help to clearly distinguish between special cause and common cause variation.
Notice for the month of January that accuracy rate is between 100% and 94%
except January 22. For that day the accuracy plunged to 81%. Through our
investigation we find that for a number of hours during that day Medicare days
could not be checked because the system was down and the admitter was not
aware she can call the intermediary to check days.
The point is common cause variation must be addressed by improving the
process and special cause variation must be addressed by fixing directly
that which caused the variation Furthermore, most of the variation is due
to common cause variation.
If, however, all variation is addressed as though it is special cause variation,
which is generally the case, the following undesirable conditions will be
created:
1. Processes will never get fixed. Since nearly all variation is the result of
the process variation, treating all variation as the result a special cause
problem results in processes never being fixed.
2. Employees can be inappropriately blamed for problems. “Fixing” the root
cause of problems then becomes ordaining a “culprit” and holding him or
her accountable for a problem he or she didn’t cause. This creates a
negative environment because employees are afraid they will be blamed
for problems, problems that are not their fault. This is referred to as
“management by exception,” and it implies that if employees did their
jobs as they should, no errors would occur. This is obviously false.
3. Management’s time will be wasted “firefighting,” i.e. fixing daily the
common cause problems that occur as though they are special cause
 problems. A busy manger doesn’t mean he or she is an effective
manager. In many cases it is just the opposite. This is known
pejoratively as “management by last data point.”
We have found this situation can last for a very long time if management is in
denial how badly their processes are functioning and find it more convenient to
blame employees. Since processes are 100% the responsibility of management,
to acknowledge that process are ineffective is to admit management may be
ineffective. True leadership is needed in this situation.

Common-cause versus special-cause variation

When trying to understand the variation evident in a run chart, it is useful to
begin with the idea of stable, background variation which appears random and
is called common-cause variation. (Further variation may be superimposed on
top of this.) Common-cause variation is present to some extent in all processes.
It is an inherent characteristic of the process which stems from the natural
variability in inputs to the process and its operating conditions. When only
common-cause variation is present, adjusting the process in response to each
deviation from target increases the variability. \Looking locally" for causes
of common-cause variation is fruitless. This variability is an inherent
characteristic
of the way the system operates and can only reduced by changing the
system itself in some fundamental way.

On the other hand, for variation which shows up as outliers or speci¯c identi
¯able patterns in the data, asking questions like, \What have we changed
recently?" or \Did something particularly unusual happen just prior to this
and what was it?" very often turns up a real cause, such as an inadequately
trained operator or wear in a machine. This latter type of variation is classi¯ed
as special-cause variation.5 Special-cause variation is unusual variation, so it
makes sense to look for \something unusual" as its cause. The investigation
should take place as soon as possible after the signal has been given by the
chart
so that memories of surrounding circumstances are still fresh. If the cause can
be located and prevented from recurring, a real improvement to the process
has been made. In summary, control charts tell us when we have a problem
that is likely to be solved by looking for \something unusual" as its cause.

Reducing common-cause variation

The reduction of common-cause variation is also very important, but control
charts are not designed for this task. Quite di®erent tools and ways of thinking
are required. For example, it may be possible and worth while to control
variability
in some inputs to the process or aspects of the operating environment.
In order to do this, however, we must ¯rst ¯nd out what types of variability in
the inputs and operating environment are most important as causes of
variability
in the end-product. It may be possible to modify some internal parts of the
process. Making changes without knowing what e®ects they are likely to have
on the product constitutes tampering, with all its ill e®ects. Making informed
changes requires planned investigation. Methods include observing the e®ects
of experimental interventions, and also performing observational studies which
relate \upstream" variables (such as measurements on aspects of incoming
raw
materials, operators, procedures, machines involved, etc.) to characteristics of
the product. Regression methods are often useful for this. But let us return to
control charts.

Subgroup Selection:

Subgroup selection is crucial to the proper use of the control chart. Subgroup
should be chosen rationally. The choice should minimize the variation within the
subgroups. This will allow isplation of special variation between the subgroups
while capturing the inherent process variation within the subgroups. The
frequency with which sub groups are selected will help minimize the amount of
variation between the subgroups by isolating batches, shifts, production runs,
machines, or people. This enables the special variation to captured, isolated,
observed, and analyzed. Decisions concerning rational subgroups often require the
combine knowlwdge of those directly involved with the process and an experienced
statistician. Rational subgroup selection may require a trial and error solution,
may produce several false starts, and may very well require a great deal of
patience.

Example-1:
Example:

Mean values and ranges of data from 20 samples (sample size = 4) are shown in the
table below:

Mean Mean Mean Mean Mean

S. of Rang S. of Rang S. of Rang S. of Rang S. of Rang
N Sampl e N Sampl e N Sampl e N Sampl e N Sampl e
e e e e e
1 10 4 5 9 5 9 10 4 13 12 4 17 12 4
2 15 4 6 11 6 10 11 6 14 12 3 18 15 3
3 12 5 7 11 4 11 12 5 15 11 3 19 11 3
4 11 4 8 9 4 12 13 4 16 15 4 20 10 4
Sum of mean of 20 samples = = 232

Average of mean values of 20 samples = = 11.6 (Center Line of X bar Chart)

Average of Ranges of 20 samples = = 4.15 (Center Line of R Chart)

Upper Control Limit of X bar chart = 11.6 + A2 4.15 (A2 = 0.729 for sample size 4)

= 14.63

Lower Control Limit of X bar chart = 11.6 - A2 4.15 (A2 = 0.729 for sample size 4)

= 8.57

Upper Control Limit of R chart = D3 4.15 (D3 = 2.282 for sample size 4)

= 9.47 9.5

Lower Control Limit of R chart = D4 4.15 (D4 = 0 for sample size 4)

X-Bar Chart

Sample data at S.N 2, 16, and 18 are slightly above the UCL. Efforts must be made to find
the special causes and revised limits are advised to calculate after deleting these data.
R Chart

All the data are within the LCL and UCL in R Chart. Hence variability of the process data
is not an issue to worry.

Control charts for Attribute type data (p, c, u charts)

c-charts counts the number of defects in an item. c-charts are used only when the number
of occurrence per unit of measure can be counted such as number of scratches, cracks
etc.

c-chart formulae

= centre line of c chart

Example: c-chart

Data for defects on TV set from 20 samples (sample size = 10) are shown in the table
below:

Sample No. of Sample No. of Sample No. of Sample No. of

No. Defects No. Defects No. Defects No. Defects
1 5 6 4 11 6 16 5
2 4 7 5 12 5 17 4
3 5 8 6 13 4 18 6
4 6 9 8 14 7 19 6
5 4 10 7 15 6 20 6

CL =

c-Chart

None of the sample is out of the LCL and UCL. But the chart shows cyclic trend.