Introduction to SPC

Section 1: A Short History of SPC

Section 2: Types of Control Methods
Section 3: Run Charts vs. Control Charts
Section 4: Types of Control Charts
Section 5: Constructing a Control Chart
Section 6: Interpreting Control Charts

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 1 of 40
Introduction to SPC

Module Objectives

 We’re going to start with a short history of SPC, statistical process control.

 Then we’re going to introduce the various types of control methods.

 We’ll compare run charts to control charts.

 We’ll show you the different types of control charts.

 We’ll show you how to construct a control chart.

 And lastly, how to interpret your control chart.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 2 of 40
Introduction to SPC

Section 1: A Short History of SPC

SPC - A Short History

Before we get into the nitty-gritty of how to use SPC, let’s take a look at a short history. Many of you may know
that SPC stands for “Statistical Process Control.”

SPC was developed by Dr. Walter Shewhart of Bell Labs. In 1924, he developed a theory and called that theory
statistical process control. During the ‘20s, Dr. Shewhart actually presented his theory in a series of lectures that
were published in a book. That book is called Economic Control of Quality Manufactured Products. Shortly
thereafter, many manufacturing companies started applying his theory, and SPC came into wide use during the
1940s.

Shewhart was called the father of SPC. The fundamental concepts he established were used by quality gurus yet
to come—specifically, individuals like W. Edward Demming, Joseph Juran, and Philip Crosby. Let’s take a closer
look at the principles of Dr. Shewhart’s SPC ideas.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 3 of 40
Introduction to SPC

Shewhart's Fundamental SPC Ideas

Let’s take a closer look at four of Dr. Shewhart’s fundamental ideas of SPC. The first one is that all processes
display variation. That may be obvious to you, but it is an important concept that he built on.

Since all processes display variation, Dr. Shewhart suggested that this variation can be broken down into two
components.

One component he called the common cause component.

The common cause component of variation is one that is inherent to the process. It’s a steady component to
variability that is always there. It is caused by random or undiscoverable phenomena part of the process.

The second component of variation is something he titled assignable cause variation.

Assignable cause variation is intermittent to the process; meaning, sometimes it’s there, sometimes it’s not
there. It is due to special causes that are discoverable and removable.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 4 of 40
Introduction to SPC

Shewhart's Fundamental SPC Ideas (Continued)

Now that we understand that there are two components of variation, we can take a particular process
characteristic, sample that characteristic over time, and estimate the parameters of the distribution that this
characteristic came from. Such parameters might be the mean or the standard deviation.

Lastly, changes in this distribution can be revealed by plotting the samples versus time.

Notice here, if we were to plot our data over time, we would discover that the first group of data looks different
from the second group of data.

That is exactly the point of using SPC charts—to be able to take process variability, analyze it versus time, and
discover any assignable causes that may help you improve the process.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 5 of 40
Introduction to SPC

Shewhart's Improvement Cycle

Shewhart created an improvement cycle to facilitate the usage of these SPC charts. The improvement cycle had
four steps. They are called Plan, Do, Study, Act.

 In the plan step, we would plan a change or a test aimed at improving the process.
 In the do step, we would actually carry out the change or the test, preferably on a small scale.
 In the study step, we would actually study the results obtained using an SPC chart, and answer the
questions, what did we learn? What went right? What went wrong?
 And lastly, we would adopt the change or abandon it.

This cycle could be run repeatedly if appropriate and therefore, facilitating an environment for constant
improvement using SPC charts with plan, do, study, act framework.

Almost 100 years ago, the framework of plan, do, study, act seemed revolutionary. Today, we use define,
measure, analyze, improve, and control as our framework for process improvement, and we use SPC charts in
the control phase once we’ve discovered the X’s that are driving our assignable causes.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 6 of 40
Introduction to SPC

Section 2: Types of Control Methods

What is a Control Method?

Before we can explain how to utilize SPC charts, we first must understand what is meant by a control method. Let
us assume for one minute that we are looking at the process that creates animal crackers. That process is
obviously very big, but for this example will be encompassed in the black box called “the process.”

Coming in to that process are various inputs—different flour vendors, different bakers, different ovens, different
cracker shapes, various line temperatures, line speed, the amount of sugar, mixing time, and many other key
process input variables.

Coming out of the process are outputs that typically our customers care about such as cracked or broken
crackers, burnt crackers, generally good or bad visually appealing crackers, along with some continuous outputs
like weight and softness of our crackers.

During DMAIC, we will determine which of these inputs affect these outputs. In the control phase of a project, our
ultimate goal is to use control methods to make sure our inputs stay within acceptable limits. More commonly,
because our inputs are very difficult to control, we monitor our outputs. So we can also use control methods on
monitoring our outputs. Let’s take a look at the various types of control methods.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 7 of 40
Introduction to SPC

Error Conditions vs. Defects

Before we can define the most common ones utilized in industry today, we need to understand the difference
between an error condition and a defect.

An error condition is actually the root cause of a defect.

If you think about an iron being left on a shirt, the error condition is actually too much heat, and the defect is the
burnt shirt. So with that context in mind, let’s define some control methods.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 8 of 40
Introduction to SPC

Ranking and Definition of Control Methods

Let’s take a look at some of the stronger, more effective, control methods.

The first and best type of control method available to us is called a countermeasure. That’s an improvement made to the
process which will eliminate the error condition from occurring. The defect will never be created. This is also called a mistake-
proof, or a design, change.

The second-best type of control method is called a flag. A flag is an improvement made to the process which will not
necessary eliminate, but detect, the error condition from occurring. The flag will shut down the process so that the defect
cannot more forward.

The next-best type of control method is actually an SPC chart. An SPC chart applied on X’s or Y’s with fully trained operators
and staff who respect the rules is the third-best type of control method. Once a chart signals a problem or a change in the
process has occurred, everyone should understand the rules of SPC and agree to shut down for special cause identification.

The fourth-best type of control method is inspection. Inspection should be implemented as a short-term containment only. This
is when audits or 100 percent sampling of a process are implemented so that we can likely detect the defect caused by the
error condition.

The fifth-best type of control method is SPC implemented on X’s or Y’s with fully trained operators. Our folks have been
trained and understand the rules, but management will not empower them to stop for special cause investigation. Notice this
second type of SPC usage is much less effective than the first time we saw SPC on this chart.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 9 of 40
Introduction to SPC

Ranking and Definition of Control Methods - Continued

Now, let’s investigate some of the weaker types of control methods.

The third-least favorable control method is a standard operating procedure, or an SOP. This is simply a
procedure implemented to attempt to detect the defects. This action is not usually sustainable short term or
long term. This type of control method should be avoided if possible.

The second-worse type of control method is a warning signal. This is used also to attempt to detect the
defects. This is some kind of visual notification, but operators will be directed to ignore or disconnect these
warning signals if they aren’t actually utilized as flags. Flags actually shut down the process; warning signals
simply give you a little nuisance notifier.

And lastly, SPC appears as the worst type of control method if it’s applied to X’s or Y’s without proper training.
In this case, it’s sometimes called wallpaper. We’re only doing it for show; we’re not really doing it to help us
control or monitor our processes.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 10 of 40
Introduction to SPC

Control Method Examples

It is most common that countermeasures and flags are applied to the control of input variables, not output
variables. For example, if a temperature on an oven exceeded some specification, we could have an
automatically adjusting vent that would open to let out some of the air.

SPC and inspection are typically applied to output characteristics, not input characteristics. In our animal cracker
world, it is not uncommon that we might imagine an inspector actually inspecting animal crackers as they’re
coming out of the oven for cracks or burns or broken pieces and parts.

It is common that continuous output variables like weight or softness readings are characteristics measured on
SPC charts. Therefore, we may measure them and plot the results on an SPC chart. These are just a few
examples of which control method might work better for X’s versus Y’s.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 11 of 40
Introduction to SPC

Section 3: Run Charts vs. Control Charts

Run Charts

The basic chart that plots a performance characteristic over time is called the run chart. run charts have been
around for a very, very long time. In a run chart, the X-axis is time. The Y-axis is then the performance
characteristic of interest. In our case here, it’s the cracker softness. Assume that observations occur in time order
and that we’ve sampled our process one observation at a time every hour or so.

The first observation is 109 mils, so we plot 109 on our run chart. The next observation is 93, so we plot 93 on our
run chart and we connect the first observation to the second observation. We can then do this for the rest of the
samples that were collected. This run chart becomes a running record of a process characteristic over time. A run
chart also gives us a sense of whether or not this performance characteristic is in statistical control.

In a run chart, any cycles, trends, freaks or special causes in our data would make themselves very obvious. run
charts that have no cycles, trends or freaks in the data give us a sense of a stable process, and at some level,
they give us some predictability of our process performance characteristic.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 12 of 40
Introduction to SPC

Smith vs. Jones

(No narration on this slide.)

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 13 of 40
Introduction to SPC

Smith vs. Jones (Activity)

The Y-axis represents the percentage of individuals who were polled that favored Smith winning the election in
that particular point in time. And notice, the end of the graph has three missing observations. That’s because the
election is going to be held in three weeks.

So for each scenario coming up, you will be asked to answer a couple of questions. First, who do you think will
win, and secondly, do you see any outliers, special causes or trends in each of these graphs.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 14 of 40
Introduction to SPC

Smith vs. Jones - Explanation - Part 1

Let’s analyze the results of these three scenarios. In the first scenario of Smith vs. Jones, you should know that
this data was randomly generated from a distribution that has a mean of .5 and a standard deviation of .1. There
are absolutely, positively no cycles, freaks or outliers, or any kind of special cause variation in this data set.

It’s so easy to think that observation number 10, at about 35 percent, was a special cause, but in fact, it’s not. It’s
also easy to think that observations 11 through 16 were a trend in a downward direction, but in fact, that is false
as well.

This is one of the major limitations of run charts, is that our eyeballs are not trained well enough to be able to
discern between what is special cause variation and what is common cause variation, and personally, I would not
feel comfortable selecting either Smith or Jones to win the election, because there’s just simply not enough
information here to answer that question.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 15 of 40
Introduction to SPC

Smith vs. Jones - Explanation - Part 2

In scenario number 2, we are definitely working hard to play tricks on your eyes. We can see that Smith has
roughly 70 percent of the vote through the first 18 observations. On the 19th observation, he dips below 60
percent.

We notice that in the first 18 observations, the variability of the observations is pretty small and on the last
observation he drops quite low. Our eyeballs want to believe that in fact, Smith has been fooling around and that
the percentage of people who are going to vote for him has taken a nose dive.

The truth is, with five observations in a row all heading in one direction, this does not constitute a statistical trend.
So again, although our eyeballs really want to believe that in three weeks Smith is going to lose the election, there
simply is not enough data here to say that with confidence.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 16 of 40
Introduction to SPC

Smith vs. Jones - Explanation - Part 2 (Continued)

In this last scenario, we call this the rising star scenario, we can see that Smith is becoming more popular over
time, and he’s really reached the 50 percent range. And again, we want to believe that in three more weeks he
will certainly win the election.

Our untrained eyeball tells us that the mean of each group of two or three observations tends to be moving higher,
but in the end, we don’t have any statistical basis for making that conclusion off a run chart.

Hence, we find that run charts have huge limitations, because our untrained eyeballs cannot discern what is
common cause variation from what is truly special cause variation.

If, in fact, Smith is a rising star, then there would be special cause variation that would make you want predict that
he will in fact win the election in three weeks. But without any statistical analysis, we just don’t know.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 17 of 40
Introduction to SPC

Run Chart Limitations

So let’s just graphically make sure we understand the limitations of run charts. If we plot a performance
characteristic over time, our eyeballs want to recognize any downward possible trends in the data, and we ask
ourselves, “Has there really been a process change here?” We might also see an upward trend in the data, and
we ask ourselves the same question, “Has there been a change in the process in the upward direction?”

Not only can we be confused about a downward trend or an upward trend, we might also see a bunch of
observations very close to each other. We wonder if these five observations belong to the rest of the data set, or
only to each other.

Run charts are very limited because they don’t answer these questions for us. They don’t tell us if something has
actually changed in the process, or if this is just random variability. This leads to two major types of errors when
interpreting run charts.

The first is that we take random variation and misinterpret it as a special cause. That would force us to over
adjust our processes.

The second type of error would be to incorrectly interpret something as a common cause, when really there
was special cause variation there, and that would mean that we failed to take action when we should have.

You will be very, very, very excited to learn, as I was, that we can address the limitations of run charts by using
control charts. I know all of you are just jumping out of your seats to understand what we mean, so let’s move
forward.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 18 of 40
Introduction to SPC

Control Charts

Let’s come back to our favorite, animal crackers. A control chart, just like a run chart, has time on the X-axis and
the performance characteristic on the Y-axis. But it also adds a line that represents the average of your
observations. For an individual control chart, this line is called an X-bar line. It represents the average, or the
mean, of all of your observations.

Control charts also add what are known as control limits. Control limits are drawn at three sigma limits from the
mean. Notice, if we take the standard deviation of our data set, the average of these 25 observations is 103, and
the standard deviation is 10.3.

On our control chart, we can place upper and lower control limits at three sigma from the mean. If the standard
deviation of our data set is 10.3, our upper control limit is going to be 103 plus three times 10.3. This will give us
an upper control limit of 133.9.

In the same manner, we can figure out our lower control limit, which is 72.04. This tells us that if nothing changed
in our process, we would expect to observe observations between 72.04 and 133.9.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 19 of 40
Introduction to SPC

Benefits of Control Charts

There are many reasons why control charts are a favorable tool. I’d like to mention my top four favorites.

 From a high perspective, the first reason why control charts are of great benefit is because they support a
data-driven decision-making culture. Six Sigma is about using data to drive improvement in business
decisions.

 Secondly, control charts are a relatively low-cost control method to implement. With some paper and
some data, you can basically go a long way. Unlike mistake-proofing or expensive countermeasures or
flags, control charts don’t require the purchase of additional equipment or higher-end IT solutions to
implement.

 Third favorite reason for why control charts are a great benefit is because they use statistics to determine
when special cause variation exists. You don’t have to use your eyeballs to figure out if a process has
changed or not; statistics will help you through a control chart. We’ll take a closer look at that in a little
while.

 My last favorite reason for why control charts are a benefit is because they’re an easy-to-use tracking tool
for measuring process inputs and outputs. And that kind of goes along with my second reason as well.

So now that you have some background on control charts, we can get into some more of the nitty-gritty about how
to implement them, how to use them, how to chart them, and how to interpret them.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 20 of 40
Introduction to SPC

Section 4: Types of Control Charts

Types of Data

Before we show you how to select the right control chart, let’s spin through a quick refresher of the types of inputs
and outputs available to us. Recall our animal cracker world.

We have categorical inputs, continuous inputs, categorical outputs, and, lastly, continuous outputs.

Certain types of control charts apply to certain types of characteristics—specifically, categorical or continuous.
Let’s take a closer look at deciding on which chart to use and when.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 21 of 40
Introduction to SPC

Chart Selection Flowchart

There are many types of control charts available. The ones in the flow chart below are the most common and
most familiar to utilize. They were developed to be used by everyday individuals and require little advanced
mathematical skills. They’re easy to teach, easy to learn, and can be implemented with a simple calculator.

As we move forward, we will demonstrate to you how you select which of the following charts is appropriate.

The first question we ask ourselves when selecting which type of chart to use is the following:

Is the X or Y measured on a continuous scale?

 If the answer to this question is yes, then we will utilize the left-hand side of this flow chart, which
represents control charts for variable measurement.
 If the answer to this first question is no, then we will utilize the right-hand side of this flow chart, which
measures attribute X’s or Y’s.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 22 of 40
Introduction to SPC

Attribute Charts for Defectives

Once we’ve determined that the input or output is of an attribute nature, we ask the next question:

Does this X or Y measure defects or defective?

You may not recall the different between a defect and a defective, so let’s remind you. We know we measure
defects when a single unit in the process could have multiple flaws or errors on it. Conversely, we know we
measure defective when a single unit is, itself, either all good or all bad.

If you know you are measuring defectives, the next question you ask yourself is:

Does your sampling plan allow for constant sample size? Meaning, every time you want to track your level of
defected, are you able to pull the same number of samples each time?

 If the answer to that question is yes, then you have the flexibility of using an NP or P chart. An NP chart
would plot the actual number of defective in each sample of size N. A P chart will plot the proportion
defective in your sample.
 If the answer to your question of constant sample size is no, then your only alternative for control charting
is a P chart, which, again, will plot the proportion defective.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 23 of 40
Introduction to SPC

Attribute Charts for Defects

Let’s take a look at the defects branch. If you know you’re plotting defects, you ask the same question relative to
sample size.

Specifically, does your sampling plan allow for constant sample size N?

 If the answer to that question is no, then you would utilize what’s known as a U chart, which actually plots
the defects per unit, also called DPU.

If you are able to have constant sample size, then you can use either a C chart or a U chart, where C stands for
the actual number of defects in your total sample n; or U, which, again, stands for the defects per unit.

It can get a little bit tricky, so in a few minutes, we’ll take a look at how to apply this in the real world.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 24 of 40
Introduction to SPC

Variable Charts for Sub-grouped Data

If the X or Y we are measuring is a continuous variable, the next question we ask ourselves is:

Do logical, or rational, subgroups exist within our data? Meaning, when we sample, are we able to sample three,
four, five, or even eight or 10 samples at a time that logically belong together?

 If the answer to that question is yes and we have relatively large samples available to us and the
computation of a standard deviation is easy (meaning we have software to do it for us), then we would
probably implement an X-bar and S chart kind of combination. The X-bar would measure sample-to-
sample central tendency, and the S would measure sample-to-sample variation, or sigma. An X-bar chart
measures the average of each sample over time, and an S chart would measure the standard deviation of
each sample over time.

 If our subgroup size is very small or it is very difficult to compute a standard deviation (meaning we have
an operator doing this manually with a calculator), then it is more common to implement something called
an X-bar and R chart combination. Again, the X-bar chart measures sample-to-sample central tendency
of the averages of the samples over time, and the R chart measures the range. Notice it’s much easier for
an individual to look at five readings and determine what the range of those readings are than it is for an
individual to measure the standard deviation of five readings. An X-bar and R chart is the most common
type of chart utilized when rational subgroups are appropriate.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 25 of 40
Introduction to SPC

Variable Charts for Individual Observations

If your data does not allow for rational subgrouping and it is measured on a continuous scale, the likely chart for
you to utilize is something called an Individuals chart.

An Individuals chart measures sample-to-sample location where each individual observation stands on its own
and doesn’t logically belong to any other individual observation.

Along with an Individuals chart, which measures location of your performance characteristic, it is common to use
a moving range chart to measure the sample-to-sample variability, or the moving ranges, of your observation.
That is called an MR chart.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 26 of 40
Introduction to SPC

SPC Charts for Animal Crackers

(No narration on this slide.)

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 27 of 40
Introduction to SPC

Section 5: Constructing a Control Chart

Utility Customer Service Center

Let’s define a scenario to help you learn how to construct an X-bar and R chart.

We have a company by the name of Buffalo Gas & Electric. BG&E provides 1 million residential and business
customers with electric power.

Unfortunately, customers calling in to their service center experience, on average, hold times of 80 seconds. Our
task will be to create an X-bar and R chart to plot hold-time data for our customer service center callers.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 28 of 40
Introduction to SPC

Constructing an Xbar and R Chart

There are three important steps to constructing a control chart.

 Step 1 involves gathering the raw data.

 Step 2 involves plotting the raw data.
 And Step 3 involves calculating the control limits for the chart.

Let’s take a look at how we would do each of these steps.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 29 of 40
Introduction to SPC

Xbar and R Charts - Gather the Data

The first step in constructing a control chart is gathering the raw data. There are three individual parts to this step.

 The first part is selecting the sampling frequency, otherwise known as, “how often will we collect
samples?”
 Secondly, we need to select the size of sample that will be sampled during each frequency.
 And lastly, we need to, for obvious reasons, record the observations.

In our Buffalo Gas & Electric world, the sampling frequency will be hourly because it is thought that hold time
varies throughout the day. Thus, hourly measurements are of interest to our leadership team.

How many samples per subgroup will we collect? It was decided that five calls collected each hour from one of 20
customer service reps will be used for this study. And lastly, the observations collected from Bernie, our CSR, are
shown below.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 30 of 40
Introduction to SPC

Xbar and R Charts - Plot the Data

The second step in constructing a control chart is to plot the raw data that we’ve just collected.

This involves, first, calculating the average and range for each subgroup.

Secondly, plotting the subgroups’ means and ranges on each of the two appropriate charts—specifically, the X-
bar chart and the R chart.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 31 of 40
Introduction to SPC

Xbar and R Charts - Calculate the Control Limits

The last step of constructing a control chart is calculating the control limits and placing them on the graph.

We do this first by computing the average of the means, called X-double bar.

Then we compute the average of the ranges, called R-bar.

Next, we calculate the upper and lower control limits for the means and the upper and lower control limits for
the ranges.

And lastly, we place these quantities on the appropriate control charts.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 32 of 40
Introduction to SPC

Control Limits vs. Spec Limits

Let me leave you in this section with an extremely important concept comparing control limits to spec limits.

The definition of a control limit is:

A natural process limit based on the mean and standard deviation of observations measuring whether or not a
process is in or out of control.

These limits (again, called control limits) have no connection to customer requirements, otherwise known as
specification limits.

Specification limits, on the other hand, are customer-defined allowable tolerances or requirements that have no
connection to process data.

The most common and gross mistake made in the application of control charts is to confuse spec limits with
control limits. I want you all to raise your right hands and take an oath that you will never confuse control limits
with spec limits.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 33 of 40
Introduction to SPC

Section 6: Interpreting Control Charts

Points Outside the Control Limits

Now that we understand a little bit more about control charts, we can figure out how they can sense special cause
variation.

One way, the most common way, that a control chart tells you if something has changed in the process, is if you
observe one point outside of the control limits.

Notice on this control chart, we had a softness reading from one animal cracker at time interval eight that fell way
above the upper control limit. This would suggest to anyone running the process that something has changed,
and you are going to make cookies that are very, very soft, so you should stop for special cause investigation and
try to get your process back in control.

In a run chart, we do not have upper and lower control limits, so we couldn’t tell if that high observation just
occurred by random chance, or if something actually changed in our process.

The likelihood of observing an observation so much higher than the control limit is very, very small to have
occurred by random chance. Therefore, we are going to say that this happened because something special
caused it to happen and that is why this control chart is identifying special cause variation in our process.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 34 of 40
Introduction to SPC

Groups of Points On One Side

In this control chart, we see that we have zero points outside of the control limits. But does that necessarily mean
that we do not have any special cause variation? No, it doesn’t.

In fact, statistically, if you have nine points in a row all on the same side of the center line, you have a special
cause in your process, and in this case, starting with point 7, and ending with point 15, you will observe nine
points in a row on the same side, below the center line. This constitutes a statistical anomaly.

We can say that because observing nine points in a row on the same side of the center line is so unlikely to have
occurred by random chance, that we’re going to assume something special happened, some change in the
process shifted our mean down for those nine observations. And here, special cause variation does exist.

Without the control limits and the center line guiding us, if we were simply using a run chart, we wouldn’t know
what decision to make for sure.

What may have caused us to observe nine points in a row for cracker softness?

What if we had an operator change between point 6 and point 7? Operator 1 ran the process at the beginning
of the day, and Operator 2 came in just after point 6. Operator 2 may have adjusted the process to cause our
crackers to be less soft. And, he finished his shift right after time period 15. In this particular scenario,
Operator 2 is the special cause contributing to harder crackers.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 35 of 40
Introduction to SPC

Increasing and Decreasing Trends

In this chart, we’re not looking at observations outside the control limits or nine points in a row on one side of the
control limits; we’re looking at trending information.

Remember back to our Donna Rice effect slide from our pollster example; that was Scenario 2. In that example,
we saw five observations, all heading downward. Unfortunately, five observations do not constitute a statistical
trend.

It takes six, and in this particular slide, we see that there is special cause variation from observation 2 to
observation 7. These six observations tell us that something has changed in the process and special cause
variation does exist.

If we relate this back to cracker softness, think of the situation where you just turned your oven on. You turned
your oven on at time period 2. When you measured that cookie, that cookie had a softness of almost 130 units.
As the oven heats up, over the next five observations, notice that our cookies are getting harder and harder and
harder, or less soft. In this case, the oven warm-up cycle is the special cause that may have contributed to this
downward trend.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 36 of 40
Introduction to SPC

Zone Definitions

We have mentioned that a given control chart can be separated into sigma limits. These sigma limits correspond
to zones of a control chart.

The area encompassed within plus and minus one sigma of the center line is commonly referred to as Zone C.

The section of a control chart that extends from the end of one sigma limit to the end of two Sigma limits is
referred to as Zone B.

And lastly, the area of the control chart that extends from the end of two sigma limits to the end of three sigma
limits is referred to as Zone A.

With these zones defined, we can now educate you on the eight most common rules for identifying out-of-control
conditions on control charts.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 37 of 40
Introduction to SPC

The Eight Rules

We can summarize the eight rules as follows.

 Rule 1: One point beyond Zone A, or, outside of the three Sigma limits.

 Rule 2: Nine points on one side of the center line, regardless of the zone they’re in.

 Rule 3: Six points steadily increasing or steadily decreasing, as you saw earlier.

 Rule 4: 14 points alternating up and down every other time.

 Rule 5: Two of three points in the same Zone A or beyond.

 Rule 6: Four of five points in the same Zone B or beyond.

 Rule 7: 15 points in either Zone C. That will show us very, very tight distribution.

 Rule 8: Eight points, with none in Zone C, on both sides of the center line.

The point with these rules is that if any of these conditions are identified in your control chart, it is very likely that
something special has occurred. And remember from Shewhart, the point of control charts is to identify special
conditions. These eight rules will help you make good, statistically based decisions about out-of-control or special
circumstances that demonstrate and identify process changes.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 38 of 40
Introduction to SPC

The Point

If there’s only one thing you remember about this whole entire topic, it should be the point of SPC charts.

The point of SPC charts is not to plot data.

The point of SPC charts has nothing to do with monitoring your process.
The point of SPC charts is not to educate your employees on statistics.

The whole point of SPC charts is to drive action for process improvement.

We can drive action on special-cause variation and we can drive action on common-cause variation. Actions that
drive special-cause variability are based on increasing our process knowledge through root cause identification. If
a control chart signals an out-of-control condition, but action is never taken to investigate this, then the SPC chart
has absolutely positively no value and should not be used further.

Taking actions on common-cause variation to reduce the natural process variability is oftentimes more difficult
because it cannot be identified through simple use of SPC charts. Other, more comprehensive, DMAIC tools are
the key to identification and reduction of common-cause variation. When contributors to common-cause variability
are identified, we should take appropriate actions to drive the reduction of this type of variation.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 39 of 40
Introduction to SPC

Module Review

We’ve shown you a brief history of SPC, introduced you to the different types of control methods, compared run
charts to control charts, showed different types of control charts specifically.

We’ve also shown you how to construct a control chart and, lastly, how to interpret a control chart—a powerful
tool that I hope you’ll use in the future.

© Lean Methods Group. All rights reserved. No portion may be copied, rewritten, reproduced, or published.

Page 40 of 40