What is a control chart? How can you use it to monitor processes?
What is the UCL, LCL and Center Line (CL) of a control chart? How are they calculated? How can you most easily draw control charts?
KnowWare International Inc./QI Macros
2696 S. Colorado Blvd., Ste. 555
Denver, CO 80222
p: 303.756.9144 - 888.468.1537
This is Jay Arthur. People often ask me, what is a Control Chart? What does it do for me? Why do I need one? Can't I just draw a plain old bar chart or something like that?
The answer is, a Control Chart will help you find a lot of things that are going on in your process that you wouldn't be able to see otherwise.
So, how does this work? Let's say we're tracking something like errors. Well, it turns out we're going to plot the number of errors every day or period of time. That will create what's called a line chart. It's also known as a time series chart.
Now, with the line chart, you can't really tell if anything is good or bad or stable or predictable. It's just a chart. So, the first iteration of this was to change this to what's called a Run Chart. We added an average line and the average line here actually divides all this up.
There are some ways to analyze runs in this type of chart and figure out if there's something unusual going on, but with the advent of modern software like QI Macros, it's unnecessary—we can just jump right over to a Control Chart.
A Control Chart begins and we start looking at the variation, how far is each point from the center line? If we take that bit of data, it turns into what's called the standard deviation and that allows us to calculate an upper control limit and a lower control limit.
This type of chart would be known as a Levi Jennings chart. There are other ways we can calculate variation so for an XMR or X Moving Range Chart, here you can see there's a difference between each point, and that's what we use as our variation. That helps us figure out how to create a control limit.
In an X-Bar R Chart, we may be sampling for data points or five data points, and the range between those data points is what we use to start to figure out the variation.
To calculate upper and lower control limits (each Control Chart, once it's created, has one, two, and three standard deviation lines), the upper control limit is Three Sigma and the lower control limit is minus Three Sigma. That allows us to divided up into zones here.
The first zone called Zone C has almost 2/3 of all of your data. The next zone up Zone B that includes 95.5% of your data, and the total outside would be 99.7% of all of your points should fall within the control limits. This enables us to do some analysis and figure out is there something potentially out of control with our process.
This is one of the first rules that came out. These were developed at Western Electric in 1920, so you have to understand these have been around a while. There's no magic. This has all been tested, and once you have this out of control point, then we might want to come in and do some root cause analysis or special cause analysis.
We'd ask the 5 why's—why was this point out of control? And that's what we want to do in each one of these types of conditions.
Now, given that we have all these other lines, we can figure out some other potentially unstable conditions. Here, you can see Zone B, we should have 95% of our data points in there, but if we have 2/3 above 2 standard deviations or low to Sigma then that's 2/3 of your points. So, 2/3 are outside. That seems to be out of whack.
Similarly, if we have 4 out of 5 points above one Sigma or below one Sigma, that's 80% of your points are outside of zone C here, and 2/3 should be in here. So, that seems like an out of whack kind of condition.
We could also have 8 points above or below the center line. This is what's known as a run. One wouldn't expect there to be that many points above the center line. We can also see things like a trend.
Another type of out of control condition is what's called hugging. If you have 15 points in a row right here in Zone C, well guess what? We should have some points outside of here so that means that for some reason the data is maybe being manipulated and needs to be investigated.
When you do have a stable Control Chart as this one is where there's just points and there's nothing out of control, then we can start to make improvements. So, we'd want to reduce the variation.
That's our goal—that's what we want to create when we start making our improvement projects going forward. And that's what a Control Chart is. I mean it's not that hard and with modern software like QI Macros, all you need is some data here.
Download a 30-day trial of QI Macros from qimacros.com. You can also take my free Lean Six Sigma Money Belt Training online at lssmb.com. Get out there and start making some improvements!
Statistical Process Control: Control Charts for Proportions (p-chart) - YouTube. Create Control Charts (X-Bar & R Chart) in Excel. What is a Control Chart?. Statistical Process Control | R-Chart (Control Chart for Ranges). Create a Basic Control Chart. Control Charting Explained (SPC). Control Charting Rules & Interpreting Control Charts. سلسلة أدوات الجودة: 08 - خرائط المراقبة. Control Charts | 7 QC Tools in Hindi | 7 QC Tools | Lean Six Sigma. Nelson rules - Wikipedia.