amazing video, this should have way more views. I've been searching around for 30 minutes self teaching Statistical Process Control and this video is gold, far better than any of the related videos. Thank you so much for sharing your knowledge and especially in such clear/concise/refined form.
Hey everyone! Thanks for watching! What'd you think? There's a few more things I want to mention... 🙂 First of all, I do my best to reply to ALL comments so feel to ask questions. I'm happy to help. Second, If you have anything you want me to cover, let me know! This channel is for you so let me know what problems YOU are facing right now...and maybe i have a tool to solve it. Finally, If you enjoyed this video please subscribe! This is a small but growing channel and every subscription, view, like, share, is HUGE for me. Thanks! ~Evan
Hi, that's the difference between a signal within a process that is "in control" and an indication that the process is not in control? It looks as though an out of control process is one that has lots of signals. Is the response to signals if both "out of control" and "in control" not the same between these two states i.e. that you should react to signals and not to noise? How is 'out of control' specifically defined? Thanks
This is a good question…In control vs out of control is really a moment in time characterization. So short answer is if you have a signal, the process is out of control and ideally you should take action to investigate why. A process is in control when no signals are detected in the most recent sample. A process is OC when the most recent sample shows an out of control signal. So over a long run, it’s very possible that a process has been both in control and out of control at different time periods. Knowledge of the process, the amount of risk associated with OC datapoints, and resources available to react to them will dictate how you respond to OC data points. If a process is frequently OC, it just means it’s generally a process that goes out of control and needs more attention.
@@TheEngineeringToolboxChannel Thanks so much for your fab answer. This makes a lot of sense and makes this method feel much clearer now :) Great video (and enviable Excel skills). I've subscribed.
Great question! I say don’t focus on the noise because looking at each datapoint individually is often a waste of time when the process is in control. If the process is control, it is predictable and attempting to reduce variation on a datapoint by datapoint basis can actually lead to more overall variation. Reducing variation in the overall process requires process or system level changes. Things like new equipment, new procedures, etc. You need to ask yourself if the process is capable of meeting customer demands. If it is, then I would suggest leaving it alone. If it isn’t, then process level chances are where you need to focus.
Yes. Example: I arrive to work consistently between 8am and 8:15 am every day, but my scheduled start time is 8am. I am in control (predicable and consistent) but I am not meeting my boss’s specification.
Shawn Chavis Thanks for pointing this out. I said there are seven OTHER rules...so the first rule (discussed in this video) plus the other seven makes eight total. :) Be sure to find the video where I go into each rule in depth for more info on this.
Questions... why do the control limits not change over time? It seems the control limits could be recalculated after each new subgroup of data is added. Furthermore, when a control chart is initially created, how do you know how many subgroups are needed to have sufficient data sets to generate a control chart?
The purpose of a control chart is to prevent the process from changing. If the control limits or mean change over time (as more data is collected) then it defeats the purpose of using the chart in the first place. The idea is to measure the process, make sure it is in control, then lock the control limits so you can detect if the process has shifted.
@@TheEngineeringToolboxChannel This makes sense thank you! However, this part still seems like a chicken and egg situation to me. When you say "The idea is to measure the process, make sure it is in control, then lock the control limits" ... how can you make sure it is in control, if you do not yet have a control chart with locked limits?
You would collect the data for a period of time until you have a good sample size. (usually 10-30 datapoints is enough). The goal of the initial data collection should be to capture the variation you would expect to see in the process under normal operation. Then use that data calculate control limits. If all points are within control limits, lock the limits (do continue not use new data to recalculate mean and control limits). If you have datapoints that are outside the limits, you need to investigate those events to see if you can determine what caused them. If you CAN determine what caused those out of control points, then you would remove them from the control limit calculation, then lock your control limits. If you can’t, you do not have a process that is in control and there is no value in locking limits yet. Process should be mostly in control and/or you can assign a cause to the out of control points before locking limits.
Process A has a lot of MORE variations in the data points as compared to the data points in Process B. Can this indicate that process A is not in control compared to process B.
You are correct in that process A has much more variation. That doesn’t mean it’s out of control though! Thanks for watching! Hope you made it to the end where I explain why. 😀
Good thinking, but actually that is not the case! The size of the increments alone don’t matter…it’s the consistency of those increments that matters! I.e. do they all vary the same amount or are there outliers that vary more than others. Hope that helps!
Thanks for the feedback! I have very basic audio and editing capabilities because I am a small channel and have no income from doing these videos. I’ve made some changes to my editing process since this video. By the way some feedback for you on how to give feedback....your tact is terrible and why are you complaining about free information? Hopefully this helps! :)
amazing video, this should have way more views. I've been searching around for 30 minutes self teaching Statistical Process Control and this video is gold, far better than any of the related videos. Thank you so much for sharing your knowledge and especially in such clear/concise/refined form.
Glad it was helpful!
Hey everyone!
Thanks for watching! What'd you think? There's a few more things I want to mention... 🙂
First of all, I do my best to reply to ALL comments so feel to ask questions. I'm happy to help.
Second, If you have anything you want me to cover, let me know! This channel is for you so let me know what problems YOU are facing right now...and maybe i have a tool to solve it.
Finally, If you enjoyed this video please subscribe! This is a small but growing channel and every subscription, view, like, share, is HUGE for me.
Thanks!
~Evan
The explanation is so clear! Thank you so much!
I love you my brother! I wish you were my instructor. I have said it before and I will say it again you are God sent!!!
Thanks very much, very informative.
Very helpful, thank you!
So glad to hear it. Thank you for watching :)
hay, thanks
Your voice is beautiful!
Bravo!
Hi, that's the difference between a signal within a process that is "in control" and an indication that the process is not in control? It looks as though an out of control process is one that has lots of signals. Is the response to signals if both "out of control" and "in control" not the same between these two states i.e. that you should react to signals and not to noise? How is 'out of control' specifically defined? Thanks
This is a good question…In control vs out of control is really a moment in time characterization. So short answer is if you have a signal, the process is out of control and ideally you should take action to investigate why.
A process is in control when no signals are detected in the most recent sample. A process is OC when the most recent sample shows an out of control signal.
So over a long run, it’s very possible that a process has been both in control and out of control at different time periods. Knowledge of the process, the amount of risk associated with OC datapoints, and resources available to react to them will dictate how you respond to OC data points. If a process is frequently OC, it just means it’s generally a process that goes out of control and needs more attention.
@@TheEngineeringToolboxChannel Thanks so much for your fab answer. This makes a lot of sense and makes this method feel much clearer now :) Great video (and enviable Excel skills). I've subscribed.
Thanks for the video! Question, you say you don’t want to focus on Noise, but what if my intention is to reduce variation?
Great question! I say don’t focus on the noise because looking at each datapoint individually is often a waste of time when the process is in control. If the process is control, it is predictable and attempting to reduce variation on a datapoint by datapoint basis can actually lead to more overall variation. Reducing variation in the overall process requires process or system level changes. Things like new equipment, new procedures, etc. You need to ask yourself if the process is capable of meeting customer demands. If it is, then I would suggest leaving it alone. If it isn’t, then process level chances are where you need to focus.
Thanks!
can a process be in control and yet not meet the specification? can you please give me an example
Yes. Example: I arrive to work consistently between 8am and 8:15 am every day, but my scheduled start time is 8am. I am in control (predicable and consistent) but I am not meeting my boss’s specification.
Good video. But at ~6:30 you say there's 7 rules for control charts, but then show 8 on the screen?
Shawn Chavis Thanks for pointing this out. I said there are seven OTHER rules...so the first rule (discussed in this video) plus the other seven makes eight total. :)
Be sure to find the video where I go into each rule in depth for more info on this.
Questions... why do the control limits not change over time? It seems the control limits could be recalculated after each new subgroup of data is added. Furthermore, when a control chart is initially created, how do you know how many subgroups are needed to have sufficient data sets to generate a control chart?
The purpose of a control chart is to prevent the process from changing. If the control limits or mean change over time (as more data is collected) then it defeats the purpose of using the chart in the first place. The idea is to measure the process, make sure it is in control, then lock the control limits so you can detect if the process has shifted.
@@TheEngineeringToolboxChannel This makes sense thank you! However, this part still seems like a chicken and egg situation to me. When you say "The idea is to measure the process, make sure it is in control, then lock the control limits" ... how can you make sure it is in control, if you do not yet have a control chart with locked limits?
You would collect the data for a period of time until you have a good sample size. (usually 10-30 datapoints is enough). The goal of the initial data collection should be to capture the variation you would expect to see in the process under normal operation. Then use that data calculate control limits.
If all points are within control limits, lock the limits (do continue not use new data to recalculate mean and control limits).
If you have datapoints that are outside the limits, you need to investigate those events to see if you can determine what caused them. If you CAN determine what caused those out of control points, then you would remove them from the control limit calculation, then lock your control limits. If you can’t, you do not have a process that is in control and there is no value in locking limits yet. Process should be mostly in control and/or you can assign a cause to the out of control points before locking limits.
@@TheEngineeringToolboxChannel I see, makes complete sense! thank you so much!
Process A has a lot of MORE variations in the data points as compared to the data points in Process B. Can this indicate that process A is not in control compared to process B.
You are correct in that process A has much more variation. That doesn’t mean it’s out of control though! Thanks for watching! Hope you made it to the end where I explain why. 😀
Yes I did ! Thanks for the reply
prosses b is clearly in more control the a , reson being the increments of change are lower granted theres no time scale but still
Good thinking, but actually that is not the case! The size of the increments alone don’t matter…it’s the consistency of those increments that matters! I.e. do they all vary the same amount or are there outliers that vary more than others. Hope that helps!
Volume terrible and why the background music? Fix that and it's thumbs up.
Thanks for the feedback! I have very basic audio and editing capabilities because I am a small channel and have no income from doing these videos. I’ve made some changes to my editing process since this video.
By the way some feedback for you on how to give feedback....your tact is terrible and why are you complaining about free information?
Hopefully this helps! :)
@@TheEngineeringToolboxChannel Spot on!! Some people just wont stop complaining. Thanks for the free information. All gold dust.