hey everyone and welcome to the engineering tool box Channel today I'm going to be talking about a very powerful tool known as a control chart but what exactly are control charts and why are they so useful my goal is answer those two questions in this video before I get into things I want to pose a question here we have two charts where I've plotted output data from two different processes all the time both data sets have the same scale 0 to 100 my question to you is are these processes in control or is one in
control on the other out of control let me know what you think and why in the comments below and by the end of the video I'll tell you the answer so the first piece of understanding control charts is understanding variation there are two types of variation common cause variation and special cause variation to help us understand these concepts let's use an example think about the time that you get to work every day is it exactly the same time or what you say you usually arrive between a certain range of times say you usually arrived at
work between 7:45 and 755 with an average time of 7:50 this range represents common cause variation it is predictable in other words your boss can count on you to arrive in this time range as long as the general process stays the same that is but what if your car breaks them you definitely aren't going to make it to work within the usual time range this type of scenario represents special cause variation where there is some atypical cause that leads you to arrive at work way later than usual luckily your boss uses control charts so as
soon as 756 hits you know something is up any calls after you tell him what's happened he sends a cab and a tow truck can you make it to work even though you arrived much later than usual alright I'm only joking about that hopefully your boss isn't using a control chart to monitor when you arrive at work if that were the case I'd say he has some serious micromanaging issues but you get the point so control charts help us separate and visualize common cause variation and special cause variation and that's really what separates control charts
from basic run charts for time series plots but how exactly do we determine what is common caused from special cause in other words where do we draw the line what amount of variation is considered common because as we talked about there's always going to be some amount of variation but key is determining how much variation is abnormal so this is where things get into this statistics but don't worry I'll keep things somewhat light for now basically some people that are much smarter than me have come up with ways of determining where we should draw the
lines between common and special cause variation in a process those lines are called the control limits the upper and lower control limits are determined first by calculating an estimated short term standard deviation or Sigma value based on the data set used the standard deviation is just a measure of how spread the data is then we add three standard deviations above the mean and subtract three standard deviations below the mean to set our control limits for a process that is generally under control we can then say with some degree of confidence that if a data point
falls outside these control limits then there must be some special cause driving that excess variation it is very important to note here that we are not simply taking the standard deviation of all the data points to get our Sigma value that would not be correct the Sigma values we want to use our derive using different statistical constants and equations specific to the type of control chart being used the simple or maybe not so simple explanation for this is that these equations and constants allow us to calculate the within subgroup variation or short term variation of
the process using the short term variation allows us to get a better measure of the amount of common cause variation in the process to understand this idea let's say I were to go to a machine and ask an operator to make me ten samples of a part in a roll without changing any settings because the samples were taken back to back over a small window of time and we told the operator not to change any settings any variation between those ten samples would be from common cause variation that is normal to the process but now
I'll say I waited for the next day and a different operator was running the machine and I asked him to make me ten samples without changing his settings either there's a very good chance that there is some variation between this operator samples and the first operator samples because it's a new date a different operator material may be different etc there are tons of tiny variations that could add up to a significant variation between those two sample sets so because of this if I were to take the standard deviation of all 20 of those samples the
standard deviation would be larger because that calculation would be including the variation within each sample set and the variation between each synthesis so again just to drive this point home with control charting we are interested in the short term variation of a process not the overall variation of the process for this reason the standard deviation and control limits are estimated from a measure of the short term variation and not that overall variation another concept that comes up when dealing with statistics and control charting is the idea of signal and noise so I want to address
that now signal and noise are really just terms for how we categorize the data on the chart where signals are points that are outside the control limits and noise are points that are inside the control elements so basically special cause variation is what leads to a signal and the common cause variation is what just creates noise and you may also see these terms used interchangeably so knowing what we know now about control charting let's go back to the question that I posed at the beginning of this video are these processes in control and the answer
is yes both of them are in control I know that because I actually put these to the test with my control charting template that I built in Excel which I'll link to that video in the description and when I did that all the data points are in control in other words the variation is all common cause for just noise but we can actually almost see this just visually the data and both plots are up and down and all over the place but the amount of variation is consistent around a mean and therefore predictable in both
cases so in control versus out of control is not determined by how much variation from the mean there is in a process but instead it is determined by whether the amount of variation from the mean is normal or not and this is a good segue into my last point about control charge which is that control charts by themselves tell us nothing about the performance of a process it is common to mistake control limits for specification limits specification limits tell us what is acceptable whereas control limits determine process control you can easily have a process that
is out of control and still meeting specifications and conversely you can have a process that is in control and not meeting specifications so I said that was my last point but I have one more thing that I briefed to talk about there are other rules that can be used to determine the special cause variation or signal in a process other than just looking to see if a data point falls outside of the control limits I won't go into detail in this video but there are a total of seven other control charting rules that can help
us understand process control the other seven rules help us detect patterns that can emerge in the data that represent certain special caused phenomena so now let's talk about the usefulness of control charts the first thing that control charts can help us do is to just understand if we are dealing with a process that is in control or out of control just understanding this is very important because a process that is in control it's much easier to improve or change than one that is out of control if we monitor a process or gather historical data analyze
it with a control chart and we find that it is out of control our first step should be to start diving in and understanding why once we understand why we should work to get the process to a state of control before working to improve it until we do this improvement efforts will be very difficult and it will be very hard to measure or detect any improvements that have made also if we don't have a process that is generally in control we won't be able to use the control chart to detect a signal from noise in
the future so if we do have a process that is generally in control we will be able to determine signal for noise and that is very important because we do not want to react to noise again this is because noise is caused by all of those variations that are natural to the process if we tried to reduce this variation it would be at best a waste of resources and at worst we could actually add more variation into the process on the other hand we do want to react to the signals because if there is a
signal that means that there is an assignable cause behind it that is more than just normal variation signals tell us when to take action signals either tell us something good happened or something bad happened in the cases where we don't want the process to move if we see a signal that means something bad happened in these cases we should investigate the root cause and try to correct them but what if we do want the process to shift in one direction if we were measuring something like defects we would want to see defects going down so
if we have a process that is in control and we get a signal that is below the lower control limit that means something good happened either we made a change on purpose and now we have an indication that the change is working or there was a change that we didn't know about it and we should investigate it so we can or hopefully this video has given you a better understanding of what a control chart is and why they are so useful I'm hoping to continue this topic in future videos where I'll break down each type
of control chart discuss those other seven control charting rules and probably do some more Excel tutorials as well so be sure to subscribe and turn on notifications so you don't miss those so with that thank you all so much for watching and I'll see you next time
