Pareto charts and Pareto-derived Paynter charts
are important statistical tools in quality improvement
programs. Six Sigma and other methodologies
rely upon Pareto analysis to define and focus
continuous improvement efforts.
Although the term Pareto principle (or
Pareto effect) has become a standard
term in business and quality improvement jargon,
it is often not clear when to use Pareto charts,
or how to use them effectively. At Statit,
Inc., Pareto and Paynter charting capabilities
are built into our Statit e-QC product. Combined
with the ability to "drill down" dynamically
through layers of data, Statit e-QC enables
you to bring these powerful tools to your Continuous
Quality Improvement (CQI) program. In this article
we define the Pareto principle, the Pareto chart
and Paynter charts. Further, we illustrate when
to use these tools and how to maximize their
use. References for further study are provided.
The Pareto Effect
The term Pareto Effect, or the Pareto Principle,
is based upon the research of Italian economist
Vilfredo Pareto, whose 19th century study established
that 80% of the land in Italy was owned by 20%
of the population. The findings of Pareto's
research are observable in many phenomena and
the principle has become generalized into the
80/20 Rule. It turns out the 80/20 rule
can be observed and measured in many disciplines
and fields of study, and often characterizes
variations in processes and products for which
cause can be assigned. The term Pareto effect
is widely used in analyzing many business processes.
To illustrate how commonly used this principle
has become, one often hears observations such
as 20% of our products account for 80% of our
sales, or 80% of our customers only utilize
20% of the features in our product. Although
the term Pareto principle has become widely
used, it is not always clear how to specifically
use Pareto analysis techniques in a CQI program.
When to Use a Pareto Chart
The Pareto chart is a specific tool of statistical
analysis used to find one or more areas in which
to focus process improvement efforts. It is
important to note that you do not need to discover
80% of your variances can be assigned to 20%
of your causes. Your results may reveal that
indeed, one defect, which is only 20% of total
defects, results in 80% of product rejects.
Use Pareto analysis when you want to determine
the "vital few" causes that are responsible
for the majority of defects in a product. Pareto
analysis can also be "weighted" so
that you can determine which defects result
in the most costs, or which defects result in
the most customer complaints. The Pareto chart
allows you to focus your efforts to achieve
the greatest improvements by identifying the
largest issues facing the process, and helps
you in determining priorities in maximizing
return of investment of time and expense.
Your criteria or weighting of factors may vary
(such as "cost to fix", critical defect,
customer impact, etc.), but the goal and results
of Pareto analysis remain the same: to provide
focal points for process improvement.
When to Use a Paynter Chart
The Paynter chart is an extension of a Pareto
chart that provides further definition to a
Pareto chart. The Paynter chart is used to indicate
what items make up the count for each Pareto
reporting period or count. By drilling down
through Pareto and Paynter results, Statit e-QC
allows you to see different views and slices
of your data that gives a more complete picture
of your process variations.
How to Use Pareto Charts
The first step in a Pareto analysis is to define
groups (categories, bins) for the defects you
wish to track. Whether you are sampling process
measures in a continuous flow, count data, or
taking measurements of discrete components,
you decide on the categories for grouping defects,
and then collect sufficient data to create a
sample for analysis.
The next step is to determine counts or frequency
of each occurrence. In the example above of
the physical characteristics, you have several
categories and the total number of defects for
each category summing to 100% of the rejected
parts. Statit e-QC will automatically create
the Pareto chart based on your data. In the
Pareto chart, the left vertical axis is labeled
with frequency (the count), the right vertical
axis is the cumulative percentage. The horizontal
axis of the Pareto chart is labeled with the
group names. The Pareto chart is a histogram
or bar chart ordered in descending frequency
magnitude, with a graph line representing the
cumulative percentage total of each bar adding
up to 100% as it approaches the upper right
of the chart. The results clearly display categories
having the highest frequencies on the left side
of the chart.
An Example Pareto Analysis
A part you are manufacturing may be sampled
for physical characteristics and measures. You
measure length, width, thickness, density, and
other physical specifications. For the sake
of illustrating this example, the Pareto chart
shows that "length" is the most frequent
reason for part rejection. Again, you don't
need to have 80% for the Pareto analysis technique
to be effective. Possibly, you have 10 different
categories or bins for rejects, and length rejects
account for 50% total of all rejects. Although
not quite the 80/20 rule, most quality experts
would agree that further investigation into
the "length" defect is suggested by
these results, certainly if you are trying to
reduce the total number of defects.
An Example Paynter Analysis
The Paynter chart is used to further analyze
the results of a Pareto chart. Continuing with
this example, let's agree that rejects due to
"length" are either grouped as "very
shorts", "shorts" or "longs".
The Paynter chart will show these sub-groups
as separate bars within the "length"
bar of the Pareto histogram. In this example,
length rejects are shown as: "very shorts"
= 5%, "shorts" = 15%, and "longs"
= 80%.
At first glance, it appears that you would
want to focus on fixing the "longs"
problem because they cause 80% of rejects due
to length. But, in order to determine where
to focus your process improvement efforts, you
decide to factor in, or weight, these results
by the costs associated with each type of "length"
defect. It turns out that "longs"
are easily fixed at a cost of 0.5 units of labor
and materials. "Shorts" and "very
shorts" must be scrapped at an estimated
loss of 10.0 units of labor and materials. "Very
shorts" have the additional liability that
should one of these parts make it to final assembly,
it could cause failure of the entire engine,
resulting in production stoppage and loss of
customer goodwill that cannot be determined
through cost accounting techniques. Thus a weighted
Pareto and Paynter analysis of this manufactured
part reveals that, for the "length"
defect, 20% of the defects, result in over 70%
of the costs associated with that defect. It
would appear that by eliminating the class of
"short" defects, substantial cost
savings and improved customer satisfaction would
result. Eliminating the "long" defect,
although it would reduce the total number of
defects, would not achieve as great a cost savings
or customer benefits as eliminating the "short"
defect.
Statit e-QC Gives you the Tools for Pareto
and Paynter Charting
As you collect data and process results in
Statit e-QC, you can use drill-down techniques
in the software to analyze differences due to
shifts, days-of-week, production lines and many
other production factors. Often, the clues needed
to solve manufacturing variances are hidden
in the details. The tools that Statit e-QC provides
enable you to unlock those mysteries and solve
real problems.
You can view a live demo example of Pareto
and Paynter charts at http://live.statit.com.
The chart below is an example of a Pareto chart
from the demo example.
The live
demo allows you to click through several
scenarios and example with actual manufacturing
sample data.
Listen to our Pareto/Paynter Charting
Webinar
For more information about Pareto charts, Paynter
charts, drill-down techniques and weighting
factors, watch
and listen to the webinar presented by Guy
March, VP of Professional Services, Statit.
