Problems with Attribute Data

Working with attribute data alone will never improve a process. One reason for this is that attribute data is a results measure - that is it tells you what the outcome of a process is. Thus, while it will tell you how many of the outputs are defective, it doesn't provide information about how the process is operating in the way that variable data can. Whenever possible, variable data is preferable to attribute data within an SPC program.

Another problem with attribute data is that it provides a results measure of a number of different processes. As well as variation within the process itself, the process of recording attribute data is quite subjective (particularly in comparison with taking measurements) and can introduce further variation. For example, would two inspectors always give the same count as each other if they were given the same samples to check.

Strict operational definitions of what constitutes a 'defect' or a 'defective item' can go some way towards alleviating the problems of variation in the inspection process. However, a further problem with attribute data is the tendency to combine counts to produce a single figure for the number of defects or defective items. When this happens, vital information about the process is lost, and the resulting control charts tend to be less sensitive.

One way in which attribute data can be used more effectively is to use a chart which allows you to record multiple characteristics as well as the overall count. In this course you will be using a u-Chart designed for multiple characteristics, to record not only the average number of defects per unit, but also a breakdown of the number of occurrences of the individual characteristics which are collectively labelled as defects. This extra information can then be used as the basis for a pareto analysis to determine which problems need to be addressed first.