These simple, visual tools have been around for a long time. And, I've observed, those with keen numeric skills have made them more and more complex.
Yet the simple is good. And handwritten is even better for communicating the state of a process to those involved in that process.
Think about it. It is a simple graph. Time is plotted along the bottom. It can be hours/shifts/days/months. It doesn't matter what but the period needs to be appropriate for the data. Then, each period, a person places a dot to measure the parameter in that period.
The 3 horizontal lines are a mean or target level for the paramater and upper and lower control limits (UCL and LCL). Typically, these lines are placed 2 standard deviations above and below the mean.
This recognizes that there is inherent varability in a process. If the varability stays within bounds, the process is working. If a point exceeds the bounds or shows a trend within the bounds, there is un-natural variablility.
In the first case, we say the variation comes from common causes. In the second, we call is special cause. To mess with common causes is called "Tampering". To ignore special causes is called "Neglect". Don't tamper. Don't neglect.
It's that simple.
Yet, the beauty of the control chart is not the dots or the lines or the statistics. It is in the conversation the chart data provokes. The chart focuses attention on the right thing...is the process stable? If not, what causes the instability and how do we fix it so it stays stable, longer? It allows the group to avoid finger pointing and talk about issues that matter.
It happened again for me this morning. It never gets old.
If you are not using this simple tool, try it.
Click here to subscribe to Learning about Lean by email.
8 comments:
Wow, I found another advocate of simple measurement charts.
It's all to easy to fall into the trap of over complicating and and over measuring.
I like them handwritten and filled in by the operators of the process/equipment.
For pen & paper version. What data you use as basis for UCL & LCL?
@Chris-- I prefer hand drawn as well and almost all of ours are. I couldn't use a photo of one of them, for company reasons, you understand. Yes, simple is good.
@Panu -- The data is 30 or so previous points. Find the standard deviation and that lets you quickly plot the UCL and LCL. Keep it simple
Great post. It also focuses discussion on the right systemic point. Not why some individual result is bad but if the result is within the control limits focusing on improving the system. If you don't like any result inside the control limits then improve the process so that the control chart indication of process capability matches desires (instead of pointlessly worrying about expected results).
Is it okay to dismiss values when you calculate -CL's?
I mean, if process varies too much, wouldn't it make sense to take out abnormal values to calculate control limits that would be more in line with what they should be rather than what they actually are.
@Panu-when calculating the CLs, you must use some judgement. If you have clear outliers, throw them out.
At the same time, an important part of Lean is to "document reality". If the process varies, capture that, whether you like it or not.
If it is important for you to reduce that variation, try some countermeasures and then do the SD again.
On the other hand, if you are reducing variation just to make your statistical self feel better, then you are tampering. Better to focus your efforts on something that matters.
Don't overthink this, Panu...get started, involve your team, start learning.
Now.
I know I'm probably stating the obvious, but control charts are the basis for any capability study in Six Sigma. There are some specific rules on the outliers and trending that shows issues within the process (like continuously increasing values even within the limits) but I believe the commenters nailed with, use some judgement, keep it simple, and get started. I've seen too many control or production charts created at quarter end with metrics that are either too late or don't have any actionable data.
A favorite jazz musician once said “Making the simple complicated is commonplace; making the complicated simple, awesomely simple, that's creativity" Charles Mingus
I appreciate the simplicity of control charts, how they encourage participation from the team, and, as was mentioned, document reality, taking variability of process outcomes into account. Thank you for sharing concise directions for how to set up a control chart!
Post a Comment