Data Patterns and Trend Analysis

Video Activity
Join over 3 million cybersecurity professionals advancing their career
Sign up with
Required fields are marked with an *

Already have an account? Sign In »

9 hours 53 minutes
Video Transcription
Hi, guys. Welcome back. I'm Katherine MacGyver and the seizure lean six Sigma green belt. So in this module, we're gonna pull together what we learned about statistical process control. So we understand that we like it because it's a time study. It gives us a view of our process
over time, your clear or you have a great understanding of what process stability is.
So you know that that's really what we're looking for. Your familiar with control charts and how to read them. And you know that the X Bar are X bar, M R and attribute charts are going to be your get down. So now we're gonna pull all of this together and those shoe heart control chart rules that I with hinting at we're going to go through, which will give you an opportunity
to read these data patterns
and understand the implications of each of them.
So we like it daughter patterns and control charts, because this is ultimately what we're trying to build them for. I mean, they do make very pretty pictures, but you want to use control charts to keep your process in control.
If you remember back one of the assumptions on statistical process control is that the practitioners will be able to identify special cause variation.
That's what the shoe heart rules do, and that's what we're gonna learn in the rest of this model. So if you remember from our history of quality and continuous improvement lectures, Shoe Heart was one of the great quality Cooper's. So he was. And he was a contemporary of Deming and Juran, and his contribution to
how we use the world today really was statistical process control. With that, he did quite a bit of research and studying into How can you use this
on developed his rules. So there are eight of them. They are considered the universal standard for recognizing out of control or unstable processes and special cause variation. You may also hear of these rules called the Western Electric Rules, So shoe heart works for Western Electric when he was developing these.
But you will notice if you, you know, say, Google them.
There is some variation, so the patterns themselves are always going to be the same. But different practitioners have different ideas of how many data points you need in that pattern to call it special cause variation for you.
I went a little bit conservative, so I went with the original shoe. Heart control rules
or control chart rules. Some of the newer ones are a little bit more liberal on. Then we start wandering into that area of statistical significance versus operational significance. If you see three data points, you may be inclined to say it's a trend. But for this a lecture. We're going to be a little conservative,
so use your discretion and what's appropriate for your organization.
So rules one through for your first rule and the one that's going to be like the Whoa, Everybody hold the whole stop. The press is going to be one point outside of control limits. Remember, your control limits come from two clan come from two different places,
your customer specifications or three standard deviations from the mean eso. It depends on whether or not your customer was explicit.
They will be explicit if you ask the right questions. I promise you you'll get there. But let's say that for some reason you haven't gotten there yet. Three standard deviations from the mean If you have a data point that is outside of that that's telling you that if you remember from how we break up our normal distribution
three standard deviations from the mean accounts for 99.7% of your data points one outside of their says it's that 10.3. We know this is special, cause
we know we need to stop and do some root cause analysis and figure out what the heck happened. Um,
four nomenclature for you guys. You'll notice we say Zone A, B and C. This is an older way of describing standard deviations from the mean This is how shoe heart did it sues Own A is the space between two and three?
Um, zone B is the space between one and two standard deviations
and zone C is the space between average and one standard deviation. So rather than saying the standard deviation lines which are denoted in red here, it says that there is the area in between. So rule number two,
two of three consecutive points greater than two sigma on the same side s Oh, this is going to be hanging out in your zone. A on this tells us that you're looking at a shift.
So for some reason you have these points that are actually pretty far away from your average line. That's going to be in that 68 to 95% of your data. So those 30 or so in between that's going to give you where, um,
do you condone this? Indicates that there is a shift going on in your process.
This may be a good thing if you've say just done a process improvement in project. So you've seen this dramatic change or it may be a bad thing, which indicates that you need to do some root cause analysis and figure out what the heck
for five consecutive points greater than one signal on the same side, same sort of thing you're looking at a process shift. This can be a good thing or a bad thing, but you want to pay attention to it because now you have data consistently performing above your average
or below your average. It can go either way for this graph we're looking at above, but it can also be below,
um, excuse me
seven consecutive points above or below average. It does not matter where in the zones these land. This indicates a shift eso Maybe you've done, You've implemented some solutions and now you're seeing the benefit of those. So in this example there, below the average line.
So let's say we wanted Teoh
May. We wanted to decrease the amount of time it takes to complete this process than our shift is good. We like what we're seeing. We validated that the solutions are beneficial. But then let's say that perhaps this is going to be our revenue scores
below the average line is bad. We want to try and figure out what caused that shift
so back to doing your root cause analysis. This really helps you hone in on the time frame when that shift occurred, which is different than your scatter plot, where all of the data is clumped together with no sense of time.
Um, rules five and six. I really like these rules. Um, Rule number five is our trend rule, so we're gonna say six consecutive points. Increasing or decreasing. This is a little bit conservative. Generally speaking, I've heard five. I've heard as few as three. We're going to be conservative for you,
but you can take a peek at this and say yep, we're seeing a trend up and down.
This is different than a process shift, because what you are looking at is an incremental change. So perhaps you are doing ongoing quick hits and PDC A's. And so every time you measure, you're seeing a little bit of improvement. Or you could be seen regression towards the mean. Or if you remember, when I talked about how
lean Six Sigma projects have an unfortunate tendency of failing after 3 to 6 months
because people tend to go back to doing work the way they used to. If you see that shifting that direction you make may ask yourself whether or not you're controlled, plan is sustainable.
Friends are the reason why I wanted to teach you guys control plans in your green belt because I want you to be able to recognize when you are shifting, either in a good way because you have a culture of kaizen or in a bad way, because you're maybe not necessarily, um,
complain with the process or the solutions. Or maybe the processor solutions weren't as effective as they initially looked.
Um, next one up, a rule 6 14 points in a row, alternating up and down. This indicates mixture. If you remember, they are by model distribution. Where I was like Oh my God, I love this. When we were talking about a typical distributions because I said that I mean, do you have two different processes?
What you are seeing in your measure is you have multiple different processes. So perhaps you are including all of the data
from Team A and Team B or from day shift in night shift. But what you're gonna be looking for here is that you need to segment your data. Eso remember, through our sampling, we said that you need to segment. Sometimes this would be one where you have some logical slices. Take a peek at it. If for some reason
you cannot determine two different processes where you are looking at here is in our operator Air Inter operator variation. So if you remember when I talked about variation when I said we're doing the same process, But I'm getting 100 errors and you're getting five errors,
so now you're gonna have to ask yourself, what do the operators of the people doing this process do differently? So there's some opportunity for internal best practices.
There's also some opportunity for retraining and maybe go back and ask some questions about, Really, Why is it that you do it this way? So going back to kind of your current state process map toe, identify non value. Add in those wastes the next rules up
15 consecutive points within one sigma of average. What this tells you
is that you have some stratification in your data. So stratification in your data indicates that everybody is starting to pull together very nicely. Eso it's the opposite of your mixture where we said you should probably segment or stratify this data.
This one says that you have it segmented or stratified.
The question with this and why we want to pay attention to this is we want to make sure that we're not excluding meaningful data points. So if you see this and you know that you have measured everything in your process, we're going to say this is good. This is a really in control process
and this is a pretty high signal level because of how your operate how close you're operating to your average.
But if you are cherry picking the measures that you want to measure now it's starting to reflect that it's really good when it may not be really good. So stratification requires you to be a little critical and say, Are we really looking at all of the data
on the last rule? 14 consecutive points, alternating above or below the average within one sigma. So what you're looking for here is a saw tooth pattern.
In this case, we have three above three below a chunk above. We're gonna have a chunk below. This is telling us that we have over control. So one of the things with process controller process tampering is you see a few above and you're like, Whoa, whoa, We might be starting a trend. Let's go ahead and change something and drop it back down.
So if you tamper with your process to frequently,
you start to see these shifts in your process. Eso For this one, you need to tell your operators Whoa! Hold your horses. We want to see a longer run of it. It may not be a trend. It might just be where we want it to operate.
So with that pop off my camera when we go through the rules we're gonna look at What is the magnitude of those? So rule number one out of control limits Large number 22 out of three points above standard or above two. Standard deviation. Medium.
Um, when we're looking at for a five points Rule number three,
that's going to be small. Um, we're looking at seven points above or below. This is gonna be sustained. So this tells us that if we did an intervention, it's functional six points, increasing or decreasing. This is a trend. 14 points alternating. That's that mixture. So multiple processes air being measured
15 consecutive points. That stratification.
So that's telling you that, um,
you may not be measuring the entire process and then 14 consecutive points. Alternating up and down in that saw tooth pattern is over control or processed tampering, where you're trying to respond to something that may not be special cause variation.
So with that today we went over shoe hearts, eight control chart rules. You know how to read your control charts. So you've built them. Now you know how to read him, and you understand what those data patterns mean for you. So now we're going to switch into control plans, so I will see you guys there
Up Next