Capability Statistics - Cp/Cpk vs Pp/Ppk and Sigma level

[@quality_bus]

Dear Paul. Of all the videos I've watched so far, this one has helped me the most and reminded me of checking on the basics from time to time. Thanks a lot.

[@paulallen5321]

You're very welcome!

[@ShantanuTiwari-t6l]

Hi Paul- I passed the PE exam this week. Thanks to your Cp/Cpk videos.

[@paulallen5321]

Well done! glad I could of some assistance...

[@ZenS56]

Great explanation. Currently no available training at my employer. As a budding engineer and new to SPC this is super helpful. Thanks mate!

[@paulallen5321]

Thanks Zen - remember I have a book written in the same style as my video's. It's called 'drink tea and read the paper' you can get it from LULU.COM

[@jeffreychang6620]

Preparing for an interview in 2 days, this was a great explanation of the difference between cp/cpk and pp/ppk. Thank you!

[@paulallen5321]

Jeffrey - hope it went well, If you think you'd like to support the making of these free video's take look at my donation page. buymeacoffee.com/osId2T0BL Many thanks...

[@serkanaygun6234]

Needless to say how much valuable your videos are. Thanks!

[@paulallen5321]

Serkan - If you like the video's maybe my book 'drink tea and read the paper' might be of interest. You can get it from LULU.COM

[@dianas8001]

this is wonderful. I work for a large medical device company and we overuse CpK all of the time instead of using PpK. I wil share this video!

[@paulallen5321]

Diana hope the video shring worked well for you? If you think you'd like to support the making of these free video's take look at my donation page. buymeacoffee.com/osId2T0BL Many thanks...

[@johnHiggins-j2l]

Hi Paul, great videos as usual which gets us all to think more deeply. Here are my pennies worth. Surely the overall objective is to get CPK and PPK to equate to 1.33 or higher. This implies getting the short-term variability to be the same as long term variability. This also implies getting the within (short term) to equate to the in between variability (long term). Achieving this status implies reducing the overall variability because overall variability is equal to short term plus long-term variability i.e., the square root of their average squares of the short- and long-term variability.

[@paulallen5321]

Hi John - Good additional comments as always. It's been a while since I made this video. However, I think the intention was just to explain what Cpk and Ppk are i.e a different type of calculation. I hope viewers read your comment as what you are adding is how they should be using these statistics. Many thanks...

[@internationalleansixsigmai6828]

Paul Allen is back up and running!!

[@nave2n]

Nice words Paul. Be practical and use your wit. Great way of explaining

[@paulallen5321]

Thanks for the poisitve comments..

[@josecuencatorres7064]

extraordinary explanation. Congratulations Paul !!

[@paulallen5321]

Cheers Jose..

[@factorygeek]

If we refer to the statistic textbook, standard deviations formulas of Cp/Cpk and Pp/Ppk are almost the same, but Bessel's correction is the use of n − 1 instead of n in the formula for the sample variance and sample standard deviation. This method corrects the bias in the estimation of the population variance. If the samples are collected from well controlled and monitored process, we can use Cp/Cpk. If the samples samples are collected from the process we don't know or including uncertainties, we can use Pp/Ppk. So basically, Cp/Cpk means the utmost performance of the process or machine under the best and well controlled conditions. Pp/Ppk means the real world performance when there is uncertainty we don't realize yet. I think the simplified caculation on the standard deviation for Cp/Cpk really misleads many people.

[@paulallen5321]

It may do factory geek - but it is the way it is worked out. Not my rules just an explanation of what is done for CPK vs PPk

[@zlatkomatic8741]

Wow, what a great explanation! I have just became huge fan, thank you for brilliant straight-forward approach to clarify these confusing matters!

[@paulallen5321]

Zlatko - thatnks for the comments, it helps enormously. If you think you'd like to support the making of these free video's take look at my donation page. buymeacoffee.com/osId2T0BL Many thanks...

[@ClintStevenson]

@@paulallen5321 Ditto to what Zlatko said! You rock!

[@paulallen5321]

Cheers Clint - thanks for your support...

[@deanopenn]

Hi Paul, Another comment on your videos haha. i have seen a scenario where If PP and PPK fail to achieve the minimum requirement e.g. 1.67 but CP and CPK passes. The big factor is the tolerance is so small it doesn't leave much room for any process variation as PP and PPK is long term. What are your thoughts on this due to tight tolerances?

[@anonmortal]

Great explanation and nice Les Paul, lefty.

[@paulallen5321]

Anon - Thanks for the positive comments...if you like the style of the video's you can get my book written in the same way. It's called 'Drink tea and read the paper' you can get it direct from LULU.COM...

[@andrewrivett9889]

Thanks Paul - I like your explanations

[@paulallen5321]

Andrew - glad i could help...

[@ahmadsafiuddinabdrahim9909]

Thanks Paul, it helps a lot to understand the differences between cp/cpk, pp/ppk, cm/cmk. I thought that If the samples is less than 50 we call it pp/ppk, >50 is for cp/cpk. I was wrong, not only me but the entire organization. :P

[@paulallen5321]

Ahmad - glad I could on this subject, If you think you'd like to support the making of these free video's take look at my donation page. buymeacoffee.com/osId2T0BL Many thanks...

[@abhideepsingh4484]

is there any way to calculate cp and cpk values for attribute data?

[@paulallen5321]

Hi Abhideep - you can convert to a cpk value using the defects/million. But personally I'm not a fan of it. Look up DPM to cpk converter on the internet and you'll find it...

[@Vs17Oct72]

Very well explained, with practical application. Thank you!

[@paulallen5321]

Vijay - If you like the video's maybe my book 'drink tea and read the paper' might be of interest. You can get it from LULU.COM

[@l_t_truth_seeker]

Thank You So Much Mr. Paul ... The BEST always

[@paulallen5321]

Truth Seeker - thanks for the positive comments. if you like my video's you can get my book 'drink tea and read the paper' from LULU.COM. You'll find its written in the same common sense style...

[@l_t_truth_seeker]

@@paulallen5321 Thanks sir, I am strictly working in manufacturing process making mechanical parts, I get worried about books not relating to my specialties.

[@paulallen5321]

Then you should find it ideal. As it's all about manufacturing. Although I would recommend that you read books completely different to manufacturing as you learn so much more than can be used in your situation...

[@l_t_truth_seeker]

@@paulallen5321 I understand your point, my main issue when the instructor or the book brings examples about other industries, I like your videos because you're making my life easier when you talk about machines, parts, products! through your teaching. It is helping me to connect the dots, Mathematics, Statistics and parts/products samples! Sounds great! Please keep it up the informative teaching, soon I am going to order that book as well.

[@paulallen5321]

Totally understand the route you're on. 25 years ago when I first learned a lot of the principles I didn't understand how applicable it is to all processes, but as time goes on and you see situation after situation you'll begin to see how these tools and principles apply everywhere. Let me know if you need any specific help. I'm more than happy to make a video about something you need explaining...

[@xxelaxela333]

Great video Paul Allen, but let's see your business card.

[@keithboone4800]

Great stuff. Thank you for this video!

[@paulallen5321]

Glad you enjoyed it!

[@arshdeepkhaira939]

Hey paul, I got a question from your book statistical process control for small batch production, page 33. Last line says we have S.D value of 1 and .5 . How did you get that ? Because in table about std dev shows .12 and .08. Can you please clarify this for me. Thanks

[@paulallen5321]

Arshdeep - good to hear from you. Those values refer to a the diagrams below on page 34. And those values were just chosen to work out an example that would be easy to understand the normalisation process. They do not relate to the table on page 33...does that make sense?

[@NTB1492]

Very very good explanation

[@paulallen5321]

Nitin - thanks for the positive comment. If you ever want a subject covering drop me a line. Otherwise look up my book 'drink tea and read the paper' on LULU.COM

[@geraintjones7192]

Thanks for the video Paul, I notice that you don't mention control prior to using the indexes. In actual fact, when the process is in control the indexes will all be the same. I understand the need to get a quick sighting and a commonsense view of the leeway , I also understand that if you sample for a longer or shorter period you will expose different variables. You mentioned about predictability but you will not be able to achieve that without stability in the process, the quicker you get to work on this the quicker you can enjoy the day to day predictability. The way we compute the indexes is irrelevant if the process is in control and that's where we need to get to isnt it? Its true to say that if I had a flow chart for the use of these and I just went ahead and computed the Cp and Cpk and these were better than I expected it would give me a warm feeling. If the flow chart directed me to knuckle down and stabilise (assuming I had to) I would agree, if this is what you meant I agree with you. There are lots of people out there comuting the summary stats and "cashing them in" before assessing the behaviour of their process, after all, who wouldn't want to get on with making money by producing. A stable process is quite an achievement and not a natural state to be in.

[@paulallen5321]

Hi Geraint - the aim of the video is to explain the calculations and that the difference between long term and short term is the Calc of the SD not the data collection periods. The is aim is not to tell them when to use Cpk. When I do tell them when to use Cpk I say if the data is approximately normal then the Cpk defect rate is a good estimate of what will happen in the next few weeks/Months. Personally I tell them to only use SPC if the process is in control. Interesting how I advise the opposite of the Wheeler approach, but there is no 'right way' there are just good approaches that work. What you have to avoid is mixing them and getting in a mess!

[@loveforever5687]

How can you use Ppk to look at with confidence if you only have 30pc sample

[@paulallen5321]

Good to hear from you, because Ppk has nothing to do with collecting lots of data, you are using std deviation to make a prediction of tomorrow and can happily be doen with 30. That is the power of Ppk, not looking at observed stats you're looking at predictions....

[@MA-jx9um]

Really like your videos, thankyou. I used to visit heat treatment suppliers, who didn't apply SPC to time and temp processes for multiple parts in ovens, saying it wouldn't benefit them, have you got a comment on this and if possible refer me to another you tube video of yours. Additionally do you do a version SPC for senior leaders quoting examples of successful applications in industry. Many thanks

[@paulallen5321]

MA - I'll make a video about an annealing process and how it behaves when you take your hands off it! As far as Seniors are concerned, it's a shame but I really don't have a great example of SPC use. Maybe it's because all my clients are a disaster!! The good companies never need my assistance...

[@hughduong19]

Dear, Paul Allen expert. Can you help me with Data Collection plan for SPC and PCIs, PPIs indexes?

[@muthupandi830]

Such a great explanation

[@paulallen5321]

Muthupandi - thanks for the supportive comments, if you'd like to know more I have a book called 'drink tea and read the paper' that you can get from LULU.COM. If you need anything more drop me a line...

[@peterblack3918]

hi paul, in the short term, i understand R to be range. What is dsubscript 2?

[@paulallen5321]

Hi Peter - D2 is a constant that you look up from a table. If you simply google Shewhart constants and look at the images you'll see the look up table...

[@markreddington3818]

I am confused how u obtain the d2 value for r over d2 calculation

[@paulallen5321]

Good to hear from you Mark, the d2 value comes from a table of constants. If you google Shewhart constants you should get it. You'll see the d2 column on there.

[@letslearntogether8199]

Sir, Please share any video on cost of poor quality, PEST analysis, SWOT, Cost benefits analysis, Balance scorecard.

[@letslearntogether8199]

@@paulallen5321 link not working

[@qualitytalks5409]

Can you please help to solve What is the Ppk of a process with a spread of 24 units, an average of 68, an upper limit of 82 and a lower limit of 54?

[@paulallen5321]

Prabhat - If you want more info look up my book 'drink tea and read the paper' on LULU.COM. Thanks

[@suvatmg]

Hi Paul, really appreciate your video series. I have few questions if you could kindly help me understand. 1. For a sample size of 4 would Shewhart constant be 2.0588? I can see d2 value increases as you increase sample size. how does this affect Cpk? 2. If Ppk takes wider standard deviation, would ppk give better indication of process capability if used for small sample sizes? Thanking you in advance.

[@paulallen5321]

ST - yes the d2 value is correct. The reason the d2 value increases with sample size is because the more you sample the bigger the range will get. As far as question 2 is concerned, my understanding is that for smaller numbers range is a better indicator of variaibility. That is why when the sample size gets to 10, shewhart switched to x bar s chart. as it's only when the sample size gets bigger that s can be used...

[@giuliabartolini652]

Hi Paul, very good explanation! I always got it wrong. Anyway, I have one question: how can we modify the number of data collected when we have low volumes (for example 100 parts per year)?

[@paulallen5321]

Good to hear from you Giulia - to be honest you can work out the Cpk with 25 data points, so I'm not sure you need to modify. If the volumes go lower then there is a good chance the cpk value is no longer a valid measure to use...

[@vinayakarp6751]

How to use for pre control chart in CPK and PPK

[@paulallen5321]

I'm not a pre-control fan, SPC charts are much better. If you think you'd like to support the making of these free video's take look at my donation page. buymeacoffee.com/osId2T0BL Many thanks...

[@rakshithagowda8171]

Why we selected 6 as an std how it derived ? Can u explain why we r using only 6

[@nave2n]

I'm not a stats expert but sharing what I know. In a normal distribution curve (bell curve), we can capture almost all the events within 3 standard deviations(SD) from the mean in one direction; in 2 directions (6 SD) you capture 99.997% of the area under the curve. In other words, 6sigma (SD) means your process is 99.997% defect free. In today's world, 6sigma is a pinnacle achievement for any industry. That's why we use 6. Maybe in future, 6sigma might be the base level and the pinnacle might be 12sigma. We will be using 12. lol

[@paulallen5321]

Yes because very few data points will ever show up, beyond +/- 3 sigma..

[@varunn5316]

Hello Paul, one of best explanations for Cpk and Ppk. Though I have a query1. Is there any difference in way of data collection in Cpk and Ppk 2.Any difference in way calculation of mean in both process

[@varunn5316]

@@paulallen5321 thank you

[@paulallen5321]

No problem...

[@derby121]

i wish mmy business analysis tutor could taught like this.. Pareto diagram - (including the cumulative “Lorenz” curve) if you can help sir .thanks

[@paulallen5321]

Syed - If you like the video's maybe my book 'drink tea and read the paper' might be of interest. You can get it from LULU.COM

[@RAVIKANT-yt9jf]

Sir, i want to know how should we collect the data for pp, ppk. Pl more explain the pp, ppk and when we do pp, ppk.

[@PaulAllenAPL]

Ravi - apologies, I have made a video to explain the answer to your question, however my channel has been hacked and I cannot load it currently....hope to fix the problem this week...

[@KuldeepSingh-zt7cj]

Thank you for the valuable explanation Paul. In addition to what i have learnt from your video. I would like to learn more on the method of collecting data for Cp/Cpk and Pp/Ppk. below are my few queries: 1. Can we mix samples from different families i.e different machine for the same samples? 2. Can we calculate Cp/Cpk instead of Pp/Ppk for development of new samples.Is it advisable? and if it not,why it is? Please help me out with your lucid explanation.

[@paulallen5321]

Kuldeep - you cannot mix data. But when you introduce a new process you should alwys do a CPK analysis. It predicts the defect rate when you run the process at full speed...