Hi, guys. Welcome back on Catherine MacGyver and this is the leans, your lean six sigma green belt. So
we just have been hanging out in theory of constraints land. But really, we have been tap dancing the line between analyze and improve. And I do want to call out that even though in yellow. But we talked about having these really fine night toll gates to move your project forward.
We do want them because we don't want this to be a zombie project that just drags on and on and on.
But, um, there is quite a bit of overlap between the analyze and improve phases. So where you're trying to figure out what's going on with your process is what are the root cause is in the driver's, and then where you are testing your improvement.
If for some reason you don't see the response that you were expecting, you're gonna end up going back to the drawing board, which means,
ah, potentially going back to your analyzed phase as well.
So with that, we're going to keep tap dancing the line between those two phases in this module on, we're going to go over a design of experiments. So we're going to do an overview. Um,
we're going to talk about a what factorial design means we were going to go over the full factorial and the fractional factorial. We did talk about this in Yellow Bell, and one of the things that I called out in Yellow Belt
is you're probably never gonna see this. However, as a green belt, it's really important for you to know how these air done because you'll be able to modify them and continue to do your experiments. These
design of experiment principles will become factors in your pilot design was you get to your improve phase.
the reason that design of experiments is important is because it allows us to test for cause Ality. So the last few modules and all of our root cause I was like, Whoa, see why a This is correlation only it's not cause ality
design of experiments allows us to actually test four cause ality does X change. Why that being said it really hyper focus is our improvement efforts because we're basically testing our solutions before we that were rolling them out or piloting them.
So that's where you get you make sure that you're paying attention to the most impactful
solutions for your organization.
So now we're talking about design of experiment. We're gonna talk about factorial design. This is actually what your design of experiments are in fact world. If you remember
back to middle school, math means factors. In our case, our factors were going to be the inputs or the independent variables we're looking at. So all of those root causes out to their each one of those is going to be a factor.
And we're gonna look at switches so on off on. Remember, we talked a little bit about two answers when we talked about the binomial distribution. This is not one of the times that you would use this. However,
when we're talking about settings for green belt, we're going to talk about on off as you get into Black Belt and you start doing more with sectoral design,
you may get into multiple different settings. So, for example, high medium low or,
um, shifts or, like, you know, weekday days, weekend days, weekday swings, weekend swings. Those were sitting so you can get many different factor or many different settings or levels within each of those factors. But for our case, we're going to do Yes, no, on off
Because we have so many root causes identified from our root cause analysis.
the next item is your outputs, these air, you're dependent variables or your why Factors. So if you remember back to what is a process and we said, why equals f of X or why is the result of the functions of X or independent variables these air going to be our outputs?
So there are two types affect Toral designs. There is a full fact, Orel which will study all combinations of factors and levels. Um, this is great. This is absolutely fantastic. I've only ever done one of these in my career
because they're very time consuming and expensive.
But we love them because this is absolutely conclusive. The most diagnostic test we can do. We can do a full factorial design and make sure we test every single combination of factors we're looking at. But that's not really all that feasible.
So instead, we're going to spend some time in fractional
fact world, designed which fractional factorial looks at key relationships between factors and levels. so this is much faster. However, there is a big opportunity to miss interactions. So if you have variables that play off of each other like distance and weather,
those will play off of each other. You may miss the influence of one on the other
by doing a fractional doctoral design. So that's something to keep in mind as you are going through and designing and selecting your key relationships.
So for a full factorial experiment, we're only gonna be looking at a two by two eso Your base is how many different levels you can have and your exponents is how many different factors you're studying.
So ah, full factorial design isolates each hypothesized factor independently. So when you set this up,
your force first experiment is going to be your baseline or your control. This is your no hypothesis. So when we think back to our hypothesis development, we still apply these those same ideas of Nolan alternate
Your next layer experiment because we're only doing two by two
is going to be a positive with one variable a negative or no change on your second variable, and you're gonna measure
next one negative with your first variable positive with your second variable. So now we're testing each variable independently, and then measure. And then you're last experiment with this sequence. Turn on all of your factors
and then measure. And then what you're gonna look for from there is which one of these experiments performed best to your project objectives?
what I coach all of my greenbelts who do
fractional factorial is set up all of your experiments in tables like these. So if you have, you know, five different independent variables or root causes, I want to see five different columns for those variables and be very explicit and neat and orderly because these can get really confusing
when we start looking at our fractional factorial.
So our faction are fractional factorial Z. The first thing that you do is because we're saving time and money is you're gonna omit your control.
You know what your baseline is because you measured it in the measure phase of your Jamaica project. So you're not going to run a control,
then you're going to choose
which ones are. Do you guys think are going to be most impactful to your process and measure those then
the last experiment that you will run, um,
or I recommend that you run
is turning on all of your variables. And what that tells you that gives you is a range of magnitude. So your baseline you have from your, um the measure phase of your project and then turning on all of the variables tells you what the absolute best case scenario
could look like. So it gives you a range of improvement.
Unless, of course, the variables interact with each other and they cancel each other out, in which case this is a useless test and you should test the test, the variables that are most important. So if you remember back to our root cause analysis and I said that especially with Issue Carol Was
and affinity diagrams, it's really important to kind of get a weight of what do we think are the most important variables? Not as much in five wise, because five wise drills down to those
That is where the fractional factorial is, where that's going to be an input, because we're gonna look at those and say, OK, we think variable one is more important than variable, too. So we're going to test that
your fact. Your design of experiment is also or can also be your pilots or done in conjunction
with your pilots and you're testing. So when you go through and you do, um, hypothetical runs of your solutions with your departments in your clothes settings, you're gonna want to set this up and record your results to compare. So get really comfortable with the idea of factorial design.
we're actually going to wrap up our design of experiments. He was a greenbelt. Now, know why we do factorial design and you know how to set up your experiments and measure them? The key difference between full fact oral and fractional factorial is fractional factorial. You make some practitioner discretion judgments
and select your tests. Where is full?
You test everything regardless how significant or insignificant it may be.
We're staying in our improved phase for our next Montel, but we're going to switch over to a more hands on unless thought based experiment where we're going to talk about our future state process mapping. So I will see you guys there