The S in SPC stands for “statistical”. And for a lot of us it's the “statistical” bit that's the most intellectually-challenging, time-consuming and daunting aspect of SPC.
The S in SPC means having to give serious thought to which statistical method to use with which data so that we can draw the “correct“ control chart. The S in SPC means we have do the actual—often time-consuming—calculations using the right formulae. The S in SPC means that the finished charts are more of a challenge to explain to laypeople. We have to be ready for the question we dread: "Can you please just remind us again how you calculated those control limits?"
But it sometimes occurs to me that—paradoxically—the S in SPC is almost incidental to SPC’s power and resonance. It’s as if the main selling point of SPC is not so much its statistical backbone but simply that it’s a visually compelling way of presenting data that positively insists on showing context. SPC pretty much prohibits you from looking at data points in isolation. And it sometimes feels like this enforcement of context is more important than the underpinning statistical method.
With SPC, you’re simply not allowed to say: “In April 2016 there were 321 attendances, and in March 2017 there were 342 attendances, so that’s a 7% increase”. No. You have to look at all of the ten intervening points—the values for May and June and July and so forth—before you can form a view about whether it was an increase or not, or whether it was just the normal bumping around of monthly numbers that you might expect to get when you are watching a process that's actually stable.
I’ve been dipping recently into Donald Wheeler’s classic book Understanding Variation: the Key to Managing Chaos, and it’s interesting that in his first chapter he lays down what he calls his First Principle for Understanding Data, which is: No data have meaning apart from their context. He goes on to say that the three consequences of this principle are:
So yes: context is everything. I think we can all agree with that.
But does that mean that the S in SPC is actually superfluous? Would run charts and control charts possess the same descriptive resonance if they existed merely as time-series charts? Just jagged lines with blobs on them: no median lines, no overall process average lines, no control limits?
And the answer to that is no, they decidedly wouldn’t. There is actually some special magic conferred by the lines whose values we calculate using the appropriate statistical techniques. These lines, these band-widths, they point us in the direction of the P in SPC. They are a signal that the thing we’re measuring and describing with a run chart or a control chart is a process.
So it turns out that the point of the S in SPC is to make the P in SPC more visually clearer to the audience.
[30 August 2017]