Editor’s note: The first installment of Scott Cornish’s series on process improvement appeared in the July issue of Newspapers & Technology. In this, the sixth installment, Cornish begins to discuss the seven classic quality tools supporting process improvement.
 

Last month, we began our process improvement project with the “define” stage. We went over project management, the project charter and problem definition. We also noted and briefly described the seven classic quality tools. This month, we begin to look at the tools in depth.

Before beginning, I want to cover the fundamental component of any process improvement project: data.

According to our friends at the American Society for Quality (ASQ), data are quantitative or qualitative facts presented in descriptive, numeric or graphic form.

As we progress in this series of articles, we will use all three types of data. When we begin to use various types of control charts, we will need to distinguish between attribute data and variable data.

Variable or continuous data result from measurements on some continuous scale such as length, weight, temperature or ink density. These scales are called continuous because between any two values are an infinite number of other values. Attribute or discrete data result from counting the occurrence of events. Examples might include counting the number of web breaks per week or defective plates per issue date.

Tools in wide use

One other note: These tools are not new. They were in wide use when I became involved with the quality profession more than 20 years ago. I’m certain that many of you have seen them before. I have used them selectively for small projects and day-to-day troubleshooting. For example, I’ve used the combination of a check sheet, together with a Pareto chart, a number of times.  These simple but powerful tools can be used on a day-to-day basis.

So let’s start with the check sheet. As I noted last month, it is a tool used to gather data on the frequency of occurrence of particular events or defects. The type of data collected in this case will be attribute data.

I mentioned last month that the objective of our test project is to examine what it would take to guarantee a home delivery time of no later than 6 a.m.

As part of that project, let’s assume that we would like to gather some data related to home delivery. In this case, let’s take six routes over four weeks and count the number of days per week that carriers met the 6 a.m. deadline. The table would look like this:

From this data, the team will note some items that could warrant further review and investigation:

* Something happened during Week 3 for two issue days.

*The highest frequency was five days for all the carriers. The team should investigate the cause, which could have been a breaking news event, production equipment problems, etc.

*Carrier C, with exception of Week 3, delivered the last paper by 6 a.m. every day of the week. Carrier F showed the next best performance.

*Performance for Carrier B steadily improved during the 4-week period. This should be investigated and might show the effect of a new carrier progressively learning his route.

Pareto chart basics

The second tool I’d like to cover is the Pareto chart. As noted last month, the chart is based on the principle that 80 percent of the variation in a process is caused by roughly 20 percent of the variables. A Pareto chart graphically shows this, but we first need to gather some data for it. This will involve another form of a check sheet that has been slightly modified.

For this example, let’s say that the team turned its focus to the pressroom. The team thinks press off-time contributes to the time that the newspaper is ultimately delivered.

From pressroom reports, the team decided to categorize the causes of late press off-time and count the frequency of each type of occurrence. The results of that research are summarized and ordered by most to least common below:

Microsoft provides an Excel template to prepare Pareto charts. I used it to prepare the chart below:

As you can see, 19 of the 34 occurrences of late press off-time were caused by two of the categories: prepress system problems and press mechanical problems. As the project progresses, the team may choose to investigate these more closely.

The final tool we’ll cover this month is the histogram, which provides a graphical picture of the frequency distribution of data. The histogram allows detection of distributions that do not demonstrate a typical bell-shaped curve and show how process spread and central tendency relate to process specifications.

Back to our 6 a.m. project. The team has access to data that notes the exact time the same six carriers delivered the final paper on their route for the 4-week period. All the data is grouped between 5 a.m. and 6:40 a.m.

You can use Excel to create a histogram using this data.  I used it to prepare the chart below:

This data is interesting when shown graphically. As we will discuss in later articles, this looks like it might be a normal, or “bell-shaped” distribution. From a statistical standpoint, that is usually good. But in our case, the requirements are that we should have all deliveries to the consumer complete by 6 a.m. Therefore, in this case, we want to see a “one-sided” distribution, which means no data should appear in the columns after 6 a.m. Thus, this histogram, in combination with the check sheet, gives the team additional information to investigate.

One more thing. How did we determine to use five categories for the cells in the chart?

Remember the rule of thumb that states you should use the square root of the number of data points you have. In this case, we had 24 data points. The square root of 25, which is very close to 24, is five, so that’s the number we chose.

Next month, we will continue to explore the other basic quality tools in more depth.

Scott Cornish has more than 20 years’ experience in production and quality assurance at newspapers large and small. He welcomes comments and questions on this and previous articles.  Scott can be contacted via e-mail at scott@practicalprocess improvement.com.