I don’t know much about control charts (but I know what I like)

I’ve never been trained in using control charts, never passed a test or worked with anyone who’s knowledgeable in them.
Everything I’ve learned has been off of the Internet

image

Cue an angry mob of disillusioned blog readers

Or books, books as well, they’re famously good for learning.

Cunninghams Law says that..

“the best way to get the right answer on the Internet is not to ask a question, it’s to post the wrong answer.”

So I’m throwing this out there, the following may be all wrong, which is a good opportunity to get it right. This is why I’ve never much bothered with typing about control charts, leaving that to the more patient Inspector Guilfoyle.

Regardless, here is  3 Things I Think About Control Charts (that are probably wrong).

1 They’re only a heuristic (but they’re the best heuristic we’ve got)

Red sky at night? Shepherd’s delight.
Red sky in the morning? Shepherd’s warning.
That’s a heuristic.
A heuristic is a mental shortcut that allows people to solve problems and make judgments quickly and efficiently.

Back to the shepherds..if you live in a place like the UK where the weather comes from the west mainly, it means that at dusk with the sun setting in the West a red sky will be a harbinger of good weather the next day. This is because red sunsets are caused by dust and particles being trapped in the atmosphere due to high atmospheric pressure. A high pressure weather system coming in will bring settled and calm weather.
If there is a red sky in the morning however, ie in the East with the rising sun, it means the high pressure system has moved on and a low pressure system is more likely to follow with rain and wind etc. Red sky in the morning, shepherds become sad.

KNOWING that something is a heuristic, rather than a cast iron truth, is important. Shepherds needn’t become sad all the time just because of the colour of the sky. It’s just a rule of thumb
When I got my hands on control charts I thought they were the secret weapon. They’d show me THE SECRETS OF THE UNIVERSE. They look so technical.
And they do, but reading them and saying “LOOK! THERE!” and jabbing a finger at the screen and arguing over tiny little points is a ROAD TO DISASTER.

Deming said that they are a heuristic, something designed pragmatically, using trial and error. So they are!

Which leads to number 2 on the list….

2 A blind man on a galloping horse

Don’t bother with to-the-millimetre-accuracy, any changes in data should be huge enough to see by a blind man on a galloping horse.
Now, I’m no six sigma. Got nooo idea about using control charts in manufacturing, where millimetres matter. But I’ve noticed that if a service has never had anybody look at its performance data by separating out signal from noise, to look for any actual signal, then nothing much has probably changed for the better for years.
How could it? If you are seeing noise in your data, but seeing it as “improvement” then how could you manage?  If this years KPI is 2% higher than last years, but really-so-what-it’s-just-noise but you don’t know that, then you’ll be taking action when you don’t need to take action, or taking action when you don’t need to.
In essence, Managing By Thrashing About.

u0hpYmD

“The management team thrashed the issue about, then thrashed just for fun”

So if you’ve never bothered taking action based on true real actual signal, the first time you actually do that there will be true real actual improvement. You’ll be able to see it on a control chart FROM A MILE OFF. No need for being the maths expert. Read Controlling Chaos, it’s all in there

3 You’ll never know enough to know enough (so just know enough for now).

When I started this control chart larning, I didn’t know where to stop. Each bit I’d learned seemed to rely on something else that I didn’t know anything about. So I’d google that and the same thing would happen. Something would only make sense based on something else that made no sense whatsoever. So I’d google the next thing, and then another. And another and another. Out of fear! Cos each mention of a word or a technique made me think I HAD to know it otherwise I was a fraud who should go back to comparing averages to see which one was higher, and announting it with a red or a yellow coloured block. Ie performance management for toddlers.
I didn’t want that, so kept on reading. I say reading, not learning. Cos after a bit it gets really really hard! Do control charts depend on normally distributed data? How many pieces of data are needed to start a chart off? What happens if there are some huge outliers when setting a chart up, do you ignore them? Keep them in, leave them out? Then you google stuff and look what comes up….

Capture

Hence the inequality holds! It HOLDS. Got that? Good.
No of course that makes no sense whatsoever.
Just set up a Shewhart control chart (the bog standard one, XmR?) and put some process data in it and see what happens. It will always be a more accurate and useful way to model reality than taking an average of it all and sticking a nice colour next to it.

(Oh, and download a free excel template off of the internet or if you’re flush, buy Winchart. It’s a bit 90’s in it’s demeanour but it does the basic SPC stuff well without drowning in features and complexity. Usable! The best compliment there is for stat software)

Advertisements
This entry was posted in data, deming, information, learning, measures, statistics, systems thinking and tagged , , , , . Bookmark the permalink.

One Response to I don’t know much about control charts (but I know what I like)

  1. Pingback: How to be hopelessly untrendy | thinkpurpose

Comments are closed.