No graph, no meeting! Part 3
In the following blog posts, I will occasionally show an example of commonly used graphs. For this purpose, I will use a data set that I have built up myself during my 'process improvement' courses. This data can serve as an 'easy to understand' example to teach my trainees the material I discuss in this series. The dataset contains over 2,500 body lengths, the participant's gender, his or her nationality and year of birth. After all these years of collecting, I have one regret: I should have also asked whether the participants were left or right-handed. This file is certainly not suitable for real scientific research, but as an example it has already inspired hundreds of people to start working with data.
The first graph, the "Pie Chart", is the simplest of all. This example shows the ratio of women to men in my data set. It immediately shows the strength of the Pie Chart: it is able to display a limited number of categories perfectly. However, as soon as the number of categories rises above 3 to 4, you better switch over to the "Pareto Chart".
With the right software, you can make very nice stratification pictures. The stratification parameter here is the decade in which people are born. You see, for example, that the proportion of women in a process improvement training programme is steadily increasing. By the way, you can ignore the category of people who were born in the 1940s, because the sample is too small to come to the right conclusions.
As far as I am concerned, this is the most universal chart of all. It can be used in just about any company and in just about any business context. This chart is a structured version of a bar chart. It shows the number of times a certain category occurs in the dataset and sorts these results from large to small.
Some of you will no doubt have heard of the 80/20 rule. This chart visualises this rule perfectly. In the example, you can see that 80% of the people in the dataset come from a very limited number of countries. 4 nationalities are responsible for 80%+ of the participants in my courses. Knowing that there are 74 nationalities in the file, the 80/20 rule certainly applies here.
So the Pareto chart is best used if you want to visualise more than a handful of categories and if you want to show very clearly the relationship in numbers and percentages of these categories. It is also an ideal chart to use in a tiered accountability process. You can use this graph to visualise almost all KPIs of a company, or at least the reasons why the KPI does not score 100%.
To conclude, a little anecdote. It is over 10 years ago that a trainee presented me with the following challenge. He had been given the task of determining the root cause of the remaining failures in a high-tech process. This involved a small number of failing devices on a daily basis, so it took him several weeks to collect about 150 devices that could be used for his research. The research itself was not easy and it took a month before all the results were in. These 150 defects split into more than 20 categories, all of which were more or less equally frequent. "How do I sell it to my management that we can't actually do anything about these failures?" I suggested writing the probable cost of the structural solution for each bar in the pareto chart. The cheapest solution cost EUR 150k. Based on this thorough documentation, the company has learned to live with this limited failure rate and no one will ever return to this problem again!
Next time I will discuss some charts showing the frequency of occurrence, this time not of categories, but of counts and measurements.