A Pareto chart is a type of chart that contains both a bar chart and a line graph. The bar chart identifies the relative frequency of occurrence of the causes arranged in decreasing order of frequency and the line graph shows the
cumulative frequency of each of the bars. The purpose of the Pareto chart is to identify the most important or the vital few causes from a list of all possible causes or the trivial many. This chart was named after Vilfredo Pareto
who used this type of analysis to analyze the distribution of wealth in Italy and found that 80% of the wealth was controlled by 20% of the people. Later, this rule was found to also apply in other areas and is called the 80-20 rule.
For example 80% of the revenue is contributed by 20% of the customers or 80% of the problems are caused by 20% of the causes. If the Pareto 80-20 rule applies, then it is beneficial for a company to focus on the vital few causes first
to solve most of the problem.
Applications of Pareto chart
Pareto chart is usually used when the output is continuous or discrete, but the input causes are discrete. Pareto chart can be used to identify the vital few from the trivial many. For example, if you analyze a problem and find that
there may be 15 causes, it may not be important to work on all 15 causes. You can analyze the data to identify which are the most important causes that you need to focus on first. A Pareto chart helps you identify which of the causes
are the most important (or the vital few).
When you use the Pareto chart it may be helpful to create a before and after Pareto chart. Once you address the most important causes on the before Pareto chart, you will find that there may be other causes that now dominate the Pareto
chart – so any further improvements you need to make you will have to use the new Pareto chart to determine the causes to work on.
Whenever you need to focus your data collection, causes, or solutions, you can use the Pareto chart. Pareto chart is usually used in Analyze phase of a Six Sigma project. Pareto chart it also commonly included as one of the seven basic
tools of quality.
Pareto chart example
We were interested in address an issue of defective parts that were being created by a particular piece of equipment. These defective parts were causing rework at the end of the line and needed to be addressed. A discussion with the
people working in the area yielded a number of causes for the defective including operator skill, machine breakdown, thermal temperature variations in the shop floor, tool wear, and incoming part quality. In order to address this
problem, it was decided to use a Pareto chart. However, no data was available for this chart. Data was collected for a period of 1 week and the causes of defects were monitored using a check sheet. The table below shows the data
that was collected.
Using this data, a Pareto chart was created using Sigma Magic software and is shown below.
The Pareto chart contains a bar chart and a line graph. The bar chart shows the frequency of occurrence of the causes. The bar chart is sorted in descending order with the most frequently occurring value shown first and the next more
frequently occurring value shown next and so on. You can either read the raw count of the frequency of occurrence on the left vertical axis or the percentage of frequency of occurrence of the right axis.
The cumulative line graph, usually shown in red, shows the cumulative frequency of occurrence – which means the first data point is the frequency of occurrence of the most important cause, the second data point is the total frequency
of occurrence of the first two important causes and so on. The cumulative line graph reaches 100% when all the causes have been covered. You can use the cumulative line graph to identify how many causes you need to work with to solve
80% (or any given percentage) of the problem. Draw a horizontal line at 80% on the right vertical axis and determine where it intersects the cumulative line graph. Then draw a vertical line at this point of intersection. The causes
listed to the left of this line indicate the top most important causes that contribute to about 80% of the problem.
For this example, 80% of the problem is caused by 2 causes: Supplier Issues and Tool Wear. If we had limited resources, we would focus on addressing these two issues first and this would solve a majority of the problems we are facing.
If it takes too much time for fixing supplier quality issues, we could focus our efforts on incoming quality inspection to sort the good and bad incoming part quality and we could review and improve the tool change procedures to
ensure tool wear is not causing the part quality problems. These two short term actions can provide a band-aid until we delve into the root causes of poor supplier quality and tool wear to determine permanent fixes for this problem.
Limitations of Pareto chart
Pareto chart is not applicable in all situations. In this section, we will cover 5 situations where the Pareto chart would be less useful or at least we need to be watchful for these situations when we use the Pareto chart to
solve our problems.
Situation 1: It is not necessary that 80% of the problems are always caused by 20% of the causes. In fact, this is only a rough guideline. It is possible that for a given situation, 80% of the problem may be caused by
35% of the causes. This is also okay as it helps you identify a subset of the causes to work on. However, what if the Pareto chart shows us that all the causes are equally important – for example if we have 20 causes and each cause is
only contributing roughly 5% of the problem. In these causes, a Pareto chart will not help us determine the vital few and we may need to focus on a lot of factors to address the problem we are facing.
Situation 2: A second type of problem where Pareto chart is not that useful is when the problem is non-stationary. When you collect the data for one month, the Pareto chart indicates a different set of causes that are
important but when a different month of data is collected then it indicates a totally different set of causes. In such cases, Pareto chart is not that useful as it does not have much predictive ability, by working on the causes
identified by the Pareto chart you might end up working on causes that would have solved most of the problem in the past but new problems may arise in the future and hence this analysis is of not much value.
This type of problem may also occur if the data is invalid. It is possible that the data for the Pareto chart was collected over a long time ago and the process or the people involved have changed so the older data may no longer
be valid. Especially, if you are creating a Pareto chart based on very old data, then the analysis may not be of much use as it does not have much predictive value. Make sure that the data you use for creating the Pareto chart is
current and represents the problem that you are currently having.
Situation 3: A third type of issue we may face with using Pareto charts is when the causes that we want to use to create the Pareto chart are so difficult or expensive to measure that it would make it impractical to
collect the data to create the Pareto chart. For example, if we needed an expensive X-ray diffraction machine or need to perform destructive testing to determine the type of defect. In such cases, we may need to think about other
ways to address the issue rather than spending a lot of time or money to determine the causes. The causes to create the Pareto chart should be easy to collect.
Situation 4: A fourth type of problem where we may encounter when we use a multi-level Pareto chart to address an issue. For example, the first level Pareto chart is created to identify the defect category which
shows that cause A amounts to 80% of the problem. We then create a second level Pareto chart by region to find out which region (N, S, E, W) is causing defect A. This Pareto chart then shows that a majority of the problem occurs
in the N region (say 70%). If we now address the defect A for the N region, then we are effectively only addressing (80% * 70% = 56%) of the problem. If we create one more level below (say the type of vendors causing defect A in
the N region), then the contribution of that cause would further drop down. We need to be cognizant of the dilution of the problem as we use a multi-level Pareto chart.
Situation 5: We need to be aware of the fact that if we use the Pareto chart based on frequency of occurrence of a cause sometimes, we may make the wrong decisions. For example, if the first example we discussed,
we stated that part quality and tool wear were the most important causes to address. This was based on frequency of occurrence of these issues. What if the defects caused by the supplier issues is very inexpensive to fix while the
defects caused by a machine breakdown is very expensive to fix? In such a scenario, we may make the wrong conclusions by looking just at the frequency of occurrence alone. We could potentially plot the Pareto chart not for the
frequency of occurrence but the cost of fixing the problem and we may end up with a totally different looking Pareto chart.
In summary, a Pareto chart is a powerful tool that enables team members to focus on the most import causes to address a problem they are tasked to solve. This approach provides a faster return on investment by addressing the most
important problem first. However, as the article points out, we need to be aware of a few pitfalls. Here are our recommendations when you plan to use Pareto charts.
Make sure you use recent representative data to create the Pareto chart so that your conclusions and recommendations are valid.
Ensure that the causes for which you want to create the Pareto chart are easy to measure and data can be collected quickly and affordably.
Don’t use too many levels when you create the Pareto chart otherwise you will be diluting the problem you are set to address.
Consider both the frequency of defects and the cost of fixing the defects when you create and use the Pareto chart.
Finally, it is not necessary that 80% of the problem is caused by 20% of the causes. If your problem does not follow the Pareto principle, you cannot force your data to follow the 80-20 rule. You will need to approach your
problem from a different angle.
Follow us on LinkedIn to get the latest posts & updates.