Pareto: The Ferrari of charts

While the name of Vilfredo Pareto, the Italian economist and sociologist, may not ring a bell in quite the same way that Enzo Ferrari’s name does, it is nonetheless true that Pareto’s contribution to data analysis may be commensurate to Ferrari’s impact on auto sports—at least in some circles.

Pareto analysis is an often-overlooked way to examine measurement (or count) data in categories. A Pareto diagram is a powerful but simple bar chart that ranks related measures in decreasing order of occurrence. Pareto himself studied wealth and poverty in Europe in the last century, and determined that wealth was concentrated in the hands of the few, while poverty lay in the hands of the many. Based on the unequal distribution of things in the universe, this analysis reflects the law of the significant few versus the trivial many.

According to this law, 80 percent of the effects are generated by 20 percent of the causes. By understanding which is which, organizations can identify ways in which improvement efforts will generate the greatest impact. Reducing the largest bars will clearly produce a more significant effect than diminishing the smaller bars.

An example might be represented in an organization that identifies 50 occurrences of one problem—perhaps insufficient postage in billing customers—and three of another, identified as inaccurate billing of customers. The first problem costs an average of $2.00 per occurrence, or around $100, while the second generates average costs or losses of $200, or $600 total for the three occurrences. While the temptation might be to attack the problem that occurs most frequently, the cost data illuminates the ways in which savings can best be generated. A Pareto of costs would clarify this difference:


Figure 1: Pareto chart of costs of errors

A counts Pareto chart, on the other hand, would reflect the disparity in number of occurrences, but would not give information that might lead to greater cost savings.


Figure 2: Pareto chart of categories of errors

Of course, Pareto analysis is not appropriate for every kind of data. Two questions must be answered, in order to determine the usefulness of Pareto diagrams: “Can the data be categorized?” And, “Is the rank of each category important?”

Examples of data that can be categorized include cost centers for products or services; sources of greatest challenge in learning a new concept; plant sites that generate greatest productivity figures; types of surgeries in a clinical environment; sales of specific product lines; drop-out rates for training or educational programs, etc. Another area where Pareto is especially useful in analyzing issues lies in satisfaction indices generated by customers in retail or service environments.

A hotel guest response card, for example, might generate a Pareto such as that in Figure 3:


Figure 3: Guest approval ratings

Knowing that guests find the recreational facilities most satisfactory, and room service least satisfactory, demonstrates the ways in which simple categorization of data can enlighten analysis and provide positive feedback to service centers in the organization. Category data can provide impetus for, say, the dining room staff to undertake improvement efforts, in order to improve its guest satisfaction ranking.

Going further with the data, one might use Pareto analysis to chart operating costs for each of the categories, in order to demonstrate where improvement efforts might generate the greatest cost savings. One can imagine, for example, that the costs of providing room service do not justify the service, especially since it is not one of the services most valued by guests. This, of course, would be an extreme course of action, taken when concerns about cost outweigh a simple desire to improve in every area.

A college advancement office might want to determine where to hold regional alumni events, in order to attract the largest group of alums. Here, category data can show where alumni reside (Figure 4):


Figure 4: Geographical distribution of alumni

A second chart will show the kinds of events and venues that draw the greatest numbers of alumni (Figure 5).


Figure 5: Alumni events and attendance

The same office might be interested in knowing how much has been generated in alumni contributions in past regional gatherings (Figure 6):


Figure 6: Alumni contributions by region

It is important to note that Pareto analysis does not automatically determine the appropriate response, but instead offers clear comparisons among categories, to facilitate decision-making related to the categories. The data above has been stratified, rendering a multi-perspective analysis that will facilitate problem-solving. The college, for example, may respond to the Pareto charts in a variety of ways, with decisions to:

  1. Cultivate the areas where contributions have been traditionally lower, in order to increase participation;
  2. Spend more to create an event in a region that seems to draw the biggest donors;
  3. Offer small-scale events that seem to attract biggest donors;
  4. Plan large events to attract the greatest number of alumni;
  5. Or a variety of other approaches (a chart of types of cars driven by alumni, for example, might illustrate the fact that Ferrari drivers are the school’s biggest supporters!)

Pareto data may come from a variety of sources, including rich surveys that generate a great deal of information. Appropriate sampling techniques must be used in order to assure that data is accurate, and there is a real science to asking questions that can be answered quickly, while still producing a variety of areas of analysis.

Despite its simplicity, Pareto analysis is among the most powerful of the problem-solving tools for system improvement. To get the most from this tool, it is important to gather data in a systematic way, as noted above. Other ways to derive the greatest source of analysis include creating subdivisions and planning for repeat analyses.

Subdivisions are useful when data has been first recorded at a very general level, but problem solving needs to occur at a more specific level. A retail chain manager might create a Pareto diagram for all the customer returns of furniture by store in his district. Once he or she has identified the store which contributes the greatest number of returns to the total, the next step might be to analyze that store’s returns by furniture type. If “chairs” is the biggest category of furniture returns for the store in question, yet another Pareto of chair returns might help to discover whether dining room chairs, occasional chairs, wooden chairs, or upholstered chairs were being returned more frequently. Because the Pareto principle holds for subgroupings of data, teams can perform such successive analyses to target small elements of a large problem.

Once improvement efforts have been initiated and applied, it may be important to repeat the original analysis, using Pareto charts to facilitate understanding of the impact of those efforts. If a store manager has worked with delivery staff to reduce the number of fine dining sets being damaged and returned, it would be useful to repeat Pareto analysis, using more current data, and then compare the two.
Of course, Pareto analysis must be used in the context of other sound process analysis. If, for example, a system is not stable, but instead reflects wildly fluctuating data, a Pareto chart may misrepresent the ranking categories, ultimately leading to false understanding and inappropriate improvement efforts. It is generally important to use control charts to assure that a system is in control before undertaking Pareto analysis in improvement efforts.

Pareto analysis, as useful as it is, does not apply to every situation, nor give adequate information about all processes. It is, nonetheless, a powerful part of the larger stable of tools that will advance process improvement. After all, though your Ferrari may be the most powerful and sophisticated automobile on your block, you wouldn’t use it to haul lumber for your weekend project.

Neither Enzo Ferrari nor Vilfredo Pareto would approve of that.

Reference: Graham, Jackqueline D., and Michael J. Cleary. Practical Tools for Continuous Improvement, Vol. 1. Dayton: PQ Systems, Inc., 2000

Originally published in the October 2008 edition of Quality eLine, our free monthly newletter.

Subscribe now!