Hi, guys. Welcome back. I'm Katherine MacGyver, and this is your lean six sigma green belt. So today we're going to doom or on control charts. I want you to be able to recognize a control tower in any of its iterations. I want you to know definitively why it control chart is different from a scatter plot. So why were drawing another type of graph?
And I want you to be able to read a control charts. So what exactly is this telling you?
So with that, the first thing that I want to reiterate is that control charts are time studies or run charts. You hear that? But the there different from scatter plots because you capture two variables. So Samos a scatter plot there. But because one of those variables is
always time and the other one is what we defined, the quality characteristic,
the quality characteristic is your project objective? So how long it takes to complete it? How many errors? How much variation? Whatever you are saying, is your project objective or your process Quality measure. If you were doing this outside of a domestic project,
that's gonna be your quality characteristics. So literally
what time. Did you measure this? And what are you measuring? That's it. So scatter plots is a measurement of relationship. This is purely an accounting or reporting exercise. On this day, we got this measurement. On this day, we got this measurement.
So they are different, but you can use either of the
to determine process stability. We're going to focus on control charts from here on out.
So this is a great control chart. So a couple of things that you're looking at on your X axes? We're looking at our timeline. So remember, we measure it in time. This one happens to be in weeks. Um, And then if you look at our why axes We're looking at our outputs.
This one happens to be in meters per week.
So with that, the reason why I like this control chart. So you see your individual data points. So you've got your jagged little blue line. You see your mean right down the middle as your daughter Gray line.
You see your target? So this is new. We haven't gone over this with our, um,
our former control chart. When we started looking at it, target is where you start adding your your metrics. And so if you remember in our stability, I said that you really need to be in kind of a metrics, data driven organization to use this target is where we're going to say we want 78% of
whatever meters per week is measuring.
So now you can see how did how frequently do you consistently perform at that target? You see that your mean is below your target, which means that there's an opportunity there.
You need to do something to push your mean up or get your entire process to perform higher. What you would I also like about this one is is you actually have to
spec limits. So your upper control limit this one is mathematically derived. This is three standard deviations from the mean, and then you also have your customer specifications. So control limit three standard deviations from the mean upper and lower and your customer specifications.
Thankfully, for this process, the customer specifications is outside of the control limit, which gives you a lot more room.
So if you are looking at this, the first thing you're going to see is this process is capable. Were functioning within our customer specifications. And you're also going to look at this and say your process is
stable. Your as as I'm looking at this, You There are a couple of areas that you have some instability in your process. But for the broad for broad states, we're going to say that this process is stable. You do have chunks above your mean, which is good, because in this particular situation you are pushing towards your target.
So ideally, you want to be having your data points above your mean as often as possible.
So when you are looking at your control chart, you will always have four elements. You will always have a mean. You will have a measure of time. So some unit of time you will have a quality characteristic and you will have your data points.
Then you're optional elements,
your upper and lower control limits or your upper and lower customer specifications. Go back to your understanding of the process. So sometimes you're not gonna have an upper control limit. This is when you're having more is better. We want bigger
measurements. Sometimes you're not gonna have a lower control in this, So this is when nominal is best
less is butter like your golf scores. Eso with that while you always wanna have at least one because you gotta know where you're functioning, that you may not have them. So really, the key takeaways for here are mean time units. And then you're gonna be looking at your data behavior itself
again. Upper and lower control limits is going to give you rejection criteria.
But you have to think about what your measurement is. Bigger is better, smaller is better or it needs to be within this specifications. So my three minutes to put my pants on or 20 minutes when my pizza's cold. So think through that as you're constructing in reviewing them, what is ideal for your process?
so we have spent a little We spent a lot of time talking about your normal distribution, your bell shaped curve,
the way that you take your normal distribution
A and apply it to a control chart
is you flip it 90 degrees. So
think of a normal distribution as an accordion. You squish it all together, you turn it sideways and you plug it in. That's going to be your normal distribution. You spread it all out. Where you're going to be looking at is a control chart. So with that, the same statistical rules apply. If you remember thinking about a normal distribution
one standard deviation from the mean and either direction gave us 68% of our data points.
We expect 68% of our data points to function within one standard deviation of the mean. If they do not, there is something going on. And you, as the amazing greenbelt practitioner you are, are going to do some root cause analysis and some hypotheses
Then two standard deviations from the mean, so above and below. You should have 95% of your data. If you do not see 95% of your data points within two standard deviations of the mean your process is out of control.
Root cause analysis. Hypothesis testing.
99.7% of all of your data points should be within three Sigma. This is the reason why we use it for our upper and lower control limits. If you have something that is outside of three Sigma, it is going to be that 30.3% of all data points.
Unless there's a lot of it, in which case you have some special cause variation going on
root cause analysis Hypothesis testing to root those out to get your data performing where you are, where you want your data to be closer to your average. Decreasing your variation.
So with that, control charts actually can teach us quite a bit. Of course, you know, processed ability and capability. We've beat that horse, but whether or not you need to stratify your sampling in your data collection plan and this will come up in our, um, shoe hurt rules.
But if you remember back to our sampling techniques, I said you could do random. You can do, um, stratified. Or you could do segmentation sampling.
If you see some unusual patterns in your control chart, it gives you an idea that you need to segment or stratify your data because random isn't covering all the things to be indicative and tell a good story, you can assess the results of your project solutions.
So remember Lean and six Sigma is about decreasing waste and variation,
so you should be expecting to see some appropriate movements on your control chart related to your project objectives. You can determine your needs for standardization if you're process is out of control or unstable. This tells you that you need to have some serious standardization activities
in your organization to get you back into control.
Back to be in stable, where you really can use your statistical process control to drive decisions. And, of course, it does help you identify your special cause. Variation. Shoot Hurts rules. It's our last module on SPC. I promise you it's going to be worth it.
So with that today we went over the elements of a control chart. You know that their time and quality characteristic not relationship. So not like a scatter plot. You know how to read them. You're going to be looking for your mean. You're gonna be looking for your time, your quality characteristic and your actual data points.
You know that your upper and lower spec limits
relate to what you are going to be looking for from a what is ideal for your project objectives or your customer requirements, and you know that you'll be able to find these on all charts with that work. Actually going to start talking about charts and the different types in our next module, so I will see you guys there.