Donald is correct. My observation, however, is that few service employees use any form of a control chart. Having attended many Institute for Healthcare Improvement and National Association of Healthcare Quality conferences and looked at hundreds of improvement posters, virtually all still use a bar chart for time series data.

First, let's get people using control charts, even if it's an XmR used when it isn't the best choice, then let's get them to understand rational subgrouping. With the Excel-based SPC software available today, there's no excuse for using bar charts when control charts are so readily available.

Good article.  A quick “trick” to help identify the warnings that Wheeler gives about incorrect subgrouping is to look at the range chart.  For subgroup sizes less than 7, in theory it is likely that if you have rational subgroups you can have a subgroup range of 0 from time to time.  Even though there might not be a 0 on the R chart, you can expect that it is possible to see a 0 from time to time over the long haul.  In both examples provided, Figures 4 and 8 show ranges that look more like Xbar charts – distributed consistently above and below Rbar, but more importantly no range values remotely close to 0.  That alone should be a strong indicator that the subgrouping is not so rational.

Why is subgrouping used, taught and learned in almost all books & seminars? The article tells us when not to use it, but how to know when we must use it?