#12 데이터 제공하기 - Presenting Your Data
In This Chapter
▶ Investigating variation
▶ Understanding natural deviation and special variation
▶ Using run chart
This chapter introduces the importance of understanding and identifying variation. If you can identify what type of variation you're seeing in your process results, you can determine whether action is needed or not, and avoid taking inappropriate action and wasting effort.
Control charts can be used to identify types of variation in your process and in the various materials, goods and parts coming into your process, for example. This chapter covers how to use these powerful data displays. We focus on the most commonly used type, the X moving R, or individuals and range, control chart.
Later in the chapter, we refer to some other data displays and ways to assess variation, looking at histograms and hypothesis tests.
Delving into different types of variation
Things are seldom exactly the same, even if at first glance, they appear to be so. Variation exists in people's heights, in the many shades of the color green, in the number of words in each sentence of this book and in the time different people take to read this book.
Variation comes in two types - common cause and special cause:
- Common cause or natural cause variation is just that - natural. You should expect it, you shouldn't be surprised by it and you shouldn't react to individual examples of it.
- Special cause variation isn't normally what you expect to see - in the context of your processes, something unusual happened that's influencing the results. But it can result from a process change you've made to improve performance. In this instance, it will be the evidence you're looking for to confirm that the improvement has had an impact on performance.
You can use Statistical Process Control (SPC) and control charts to identify and define variations in your business processes.
Defining the type of variation is important as it ensures you take action only when you need to. Confusing one type of variation creates problems.
Understanding natural variation
Natural variation is what you expect to see as a result of how you design and manage your processes. When a process exhibits only natural variation, it's in statistical control and stable. Being in statistical control doesn't necessarily mean the results from the process meet the customer CTQs(refer to this link for CTQ: beho.tistory.com/6) but it does mean the results are stable and predictable. If the results don't meet your CTQs, you can improve the process using DMAIC. (here is the link for DMAIC: beho.tistory.com/3)
To determine whether the variation is special or natural, try the following simple experiment with some colleagues.
First, write down the letter 'a' five times. This in itself forms the basis for an interesting discussion on giving clear instructions so that everyone understands the requirement. You may find that some people write their 'a's across the page and others down the page. Some use capital letters, and others lower-case. One or two may even write 'the letter "a" five times'!
Now look at your own letters and ask whether they are all the same. Each 'a' is probably slightly different, but generally, they're likely to be pretty similar and at least each one can be identified as a letter 'a'.
The difference between your letters is natural variation, and your process for producing the letter is stable and predictable. If you repeat the exercise, you're likely to see the same sort of variation. To reduce the variation, you need to improve the process, perhaps by automating your writing or introducing a template. We continue this exercise in the 'Avoiding tempering' section later in this chapter.
Spotlighting special cause variation
Special cause variation is the variation you don't expect. Something unusual is happening and affecting the results. Special cause variation may occur if you don't identify an important 'X variable', which influences your process results, or if you don't manage the variable appropriately. The Xs will include a range of variables - for example, the accuracy and timeliness of the inputs to your process that you receive from your suppliers, or the level of reworks within your process.
When a special cause exists, the process is no longer stable and its performance becomes unpredictable. You need to take action to identify the root cause of the special cause, and then either prevent the cause from occurring again if it degrades performance, or build the cause into the process if it improves it. (Remember, not all special causes are bad. Sometimes they provide that an improvement has worked.)
Distinguishing between variation types
You need to be able to tell the difference between the two types of variation. If you think something is special cause variation when in fact it's natural, you may inadvertently tamper with the process actually increase the amount of variation. Likewise, if you think something is a natural variation when it's a really special cause, you may miss or delay taking an opportunity to improve the process.
Avoiding tempering
In the 'understanding natural variation' section earlier, we asked you to write down the letter 'a' five times as an example of natural variation. We suggest that to reduce the amount of variation, you need to review and improve the process. We will show what happens if you temper with the process by reacting to an individual example of common cause variation.
As an example, imagine that your manager doesn't understand the importance of distinguishing between natural and special cause variation. He wanders through your work area to see the output being produced. He feels that your letter 'a's show too much variation and asks you how you produce them. As you begin to demonstrate, your manager asks you to stop writing and points out your other hand is much better - after all, this is the hand he uses.
If you try writing with your other hand, your results probably show increased variation, and chances are you take longer to produce the output. Now imagine the output goes through an optical scanner - depending on the quality of your letters when you write using your other hand, you might see further problems. Your manager then provides some unuseful ideas to solve this problem, too.
Unfortunately, tampering happens all the time in many organizations. Managers often feel their role is to tamper.
Another example of tampering is a pointless discussion. You may often see reports comprising pages of numbers that somebody expects you to understand and perhaps base a decision on. The below figure shows a typical set of information that is practically meaningless to all but the person who created it.
Figures relating to sale activity often provide good examples of pointless data. You may hear statements such as, 'This week's figures were better than last week's, but not as good as those of the week before that' or 'It rained last Thursday, but the team did a great job this week' - almost certainly the differences in the weekly figures are a measure of the natural variation in the process and not due to a special cause.
Using control charts can help you make sense of the figures by enabling you to distinguish between natural and special deviation - but you may need to change the way you think. The different thinking needed is described as you work your way through the data from the above figure, eventually using it to create a control chart shown in the below figure.
Displaying data differently
The above figure doesn't tell you much. But if you present the data in a more visual form, you may begin to understand them.
Instead of giving the figures for only one month, a more useful method is to plot a graph, called a run chart, using figures for a series of months. A run chart plots the data in time order - it is a time series plot that makes it easier to spot any trends. A run chart doesn't tell you whether the variation is natural or special - to know that, you use a control chart to see whether any changes are part of the natural variation of the process or whether they're unusual and need a second look.
In the below figure, we use the figure for Location A and Product 3 to create a run chart that presents data through to the following March.
※ 참조: 본문은 책, "Dummy들을 위한 린 식스시그마(Lean Six Sigma)"와 The Toyota way를 정리한 내용을 기반으로 제 경험들을 조미한 글입니다. 현재는 원서 기반으로 내용을 정리하고 있는 중이며, 정리가 완료되면 한글로 번역을 추가할 예정입니다.