Definition of Six Sigma
Six Sigma – iSixSigma › Forums › Old Forums › General › Definition of Six Sigma
 This topic has 47 replies, 17 voices, and was last updated 19 years ago by Darrlyn.

AuthorPosts

November 4, 2003 at 4:22 am #33767
Is six sigma classified as three or six sigma away from the mean?Hence, Is this statement correct?When a product’s quality measurement falls three sigmas away from the mean in the positive direction, then the product is said to be six sigma or in the top 99.7 percent, because the quality measurement is six sigmas away from the opposite side of the curve.Thanks :)
0November 4, 2003 at 8:20 am #92018
Mannu TharejaParticipant@MannuThareja Include @MannuThareja in your post and this person will
be notified via email.Hi Nicole,
6 sigma started off with 3 sigma away from the mean, i.e. total of 6 sigma window (+/ 3 sigma). This covers 99.73% aera. Process capability is also calculated with denominator as 6 sigma.
But now a days, we aim at achieving 6 sigma at on side of the mean (12 sigma window).
Friends correct me if i am wrong.
Regards,
Mannu0November 4, 2003 at 8:58 am #92024
Jani AtelsekParticipant@JaniAtelsek Include @JaniAtelsek in your post and this person will
be notified via email.We are still calculating the capability using the 6 sisgma distributions. We have only changed the goal of the Cpk (long term capability), which is 2,04 now days in six sigma systems.
0November 4, 2003 at 11:17 am #92025Mannu’s definition is correct, but does not mention the “1.5 Sigma Shift”. Six Sigma is equated to 3.4 defects per million opportunities; 3.4 DPMO actually equates to Z=4.5, due to an assumption that processes can naturally degrade up to 1.5 sigma in the long term. So “Six Sigma” is not a purely statistical term.
Search this site for information on the 1.5 Sigma shift. Posters here are divided on whether it is a valid assumption. I don’t think it is valid.0November 4, 2003 at 11:54 am #92028This is the best definition I’ve found for Six Sigma, from a statistical standpoint:
https://www.isixsigma.com/library/content/c010101a.asp
0November 4, 2003 at 12:17 pm #92029
Mannu TharejaParticipant@MannuThareja Include @MannuThareja in your post and this person will
be notified via email.That is what i meant to say.
In the Cp formula we take 6 sigma in the denominator .
Now a days for 6 Sigma we take Cp as 2 (12/6).Regards,
Mannu0November 4, 2003 at 12:21 pm #92030
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.The definition of six sigma is as follows:
It is a forward thinking strategy to change the way of doing business and improving profitability. Some people call is continual reengineering, some called it a company’s DNA or others call it a cultural thing. The six sigma has three components: Strategy, methodology and measurements.
Statistically, Six Sigma means having twice the normal capability (+/ 3 Sigma), i.e. 6 sigma on each side. The 1.5 Sigma shift is considered based on the assumption that a process can be controlled with in +/ 3 Sigma using control chart type devices for sample means. This is explained in Juran’s Handbook.
With that, the defect rate equates to 3.4 PPM with a shift of 1.5 sigma. Remember the 1.5 sigma shift does not occur all the time, it is the maximum value. Any shift more than 1.5 sigma will lead to the process shut down.
I hope it helps. Welcome comments.
PRaveen0November 4, 2003 at 1:25 pm #92032
Banking EngineerParticipant@BankingEngineer Include @BankingEngineer in your post and this person will
be notified via email.Tanya,
That may be a good explanation, but the illustration is wrong. They are showing what sure looks like a 3sigma shift of the six sigma process, and calling it a 1.5sigma shift. The way it is drawn right now, both distributions have the same number of defects – one is centered, and one is shifted, but both are approximately (visually, since there are no actual sigma lines drawn) 3sigma from the specification limit.0November 4, 2003 at 4:28 pm #92041Praveen,
I have a few comments/questions.
You said: Statistically, Six Sigma means having twice the normal capability (+/ 3 Sigma)
Is it normal to have a capability of +/ 3 sigma? Is a Cp = 1 a normal capability?
You also said: The 1.5 Sigma shift is considered based on the assumption that a process can be controlled with in +/ 3 Sigma
If the process is controlled within +/ 3 sigma, wheres the shift? By definition of control, there is no shift. Are we talking process shift here or sampling error because they are not the same thing.
And last but not least: Any shift more than 1.5 sigma will lead to the process shut down.
Is this based on fact or speculation? Does this mean that we will never see a process that has a long term variation greater than the entitlement variation that exceeds a 1.5 shift in the Z value? How can this be true with no consideration of the ratio of the short term variation to the tolerances? Wouldnt a process with a larger ratio of the short term variation to the tolerance range naturally allow more degradation than one that has a small ratio?
Every time I hear an explanation from you Six Sigma Gurus, I just see more confusion piled on. Wouldnt just simplify this whole discussion with this explanation:
For variables data, the sigma value is 3 times the Cpk. A process that has achieved Six Sigma has a Cpk of 2.00. The traditional goal for capability was a Cpk of 1.33; Six Sigma has raised the bar to 2.00
Statman0November 4, 2003 at 5:10 pm #92042
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Statman:
Thanks for your comments and questions. Let me try answering them as follows:
Is it normal to have a capability of +/ 3 sigma? Is a Cp = 1 a normal capability?
Answer is yes, as it includes 99.73% (practically everything) of the process output.
If the process is controlled within +/ 3 sigma, wheres the shift? By definition of control, there is no shift. Are we talking process shift here or sampling error because they are not the same thing.
– Answer is yes, the 1.5 sigma shift is derived from sample to sample variation. Actually when we say 1.5 sigma shift, it means 1.5 sigma shift in the process means.
Any shift more than 1.5 sigma will lead to the process shut down.
Is this based on fact or speculation?
– This is assumed based on the control chart theory developed by Shewhart in 1920s. It means if the process means shift by 1.5 sigma, the process has shifted a lot, i.e., an assignable cause has been introduced, therefore it must shut down.
Every time I hear an explanation from you Six Sigma Gurus, I just see more confusion piled on.
– Thanks for assuming me as a Guru. Unfortunately, I am not a Guru. Instead, I am just a practitioner or a student of Six Sigma, an old one since 1986, and always learning. Your comments have been quite helpful.
Wouldnt just simplify this whole discussion with this explanation:
For variables data, the sigma value is 3 times the Cpk. A process that has achieved Six Sigma has a Cpk of 2.00. The traditional goal for capability was a Cpk of 1.33; Six Sigma has raised the bar to 2.00
– If it is your way of understanding six sigma, great! Reviewing different interpretations in different words test our understanding of the intent of Six Sigma, not of a formula. As to the Cp and Cpk goals, if I remember in early eighties, original acceptable values were 1.0, that turned out to be the borderline case and insufficient. Since then due to increasing customer expectations, Cp values have been rising to 1.33, 1.67 and now to 2.0. I am sure you would recollect some of these values.
Regards,
Praveen Gupta0November 4, 2003 at 6:14 pm #92045Hi Praveen,
Sorry about referring to you as a Guru. That was not polite. I made assumptions based on your book bio.
Let me comment on your comments. (This is fun, finally an interesting discussion this week)
Is it normal to have a capability of +/ 3 sigma? Is a Cp = 1 a normal capability?
– Answer is yes, as it includes 99.73% (practically everything) of the process output.
I think you are confusing process capability with the process distribution. When a process demonstrates stability, the natural behavior of the process and its ability to meet specifications is called its capability. If you are defining the normal capability as +/3 sigma, then you are saying the specs are +/ 3sigma.
The natural behavior (free from special causes) of a process that can be modeled as normal will demonstrate 99.73% of the process output +/ 3 sigma. But this is not its capability it is its expected distribution.
If the process is controlled within +/ 3 sigma, wheres the shift? By definition of control, there is no shift. Are we talking process shift here or sampling error because they are not the same thing.
– Answer is yes, the 1.5 sigma shift is derived from sample to sample variation. Actually when we say 1.5 sigma shift, it means 1.5 sigma shift in the process means.
Once again, you are talking about sampling error not shift because if the mean is shifting, this process is not in control it has special causes. I do fully agree that there is sample to sample variation (which for as long as I can remember, has been defined as sampling error) and that there will be a risk due to the sampling error of underestimating the true process sigma as well as a risk that a process deemed in control is not free from special causes. This is not shift, it is sampling error.
Are you implying that only a process that is in control will have shift? Or will shift occur in processes that are not in control?
Any shift more than 1.5 sigma will lead to the process shut down. Is this based on fact or speculation?
– This is assumed based on the control chart theory developed by Shewhart in 1920s. It means if the process means shift by 1.5 sigma, the process has shifted a lot, i.e., an assignable cause has been introduced, therefore it must shut down.
If you are basing this on control chart theory, then a shift of any sigma will be eventually detected and determined a special cause. Remember, a special cause is not detected only when a result is outside the 3 sigma limits. There are a lot of processes in the world that have degradation greater than a 1.5 sigma shift. The question is always economic not statistical. If the short term variation is small relative to the tolerances and the cost to replace/repair is high, the process will be allowed to degrade much more than a process that the short term variation is large relative to the tolerances and the cost to replace/repair is low.
Mathematically, the simplified explanation comparing the sigma value to Cpk is accurate. Why would any other definition be different than the mathematical form?
Highest Regards,
Statman0November 4, 2003 at 6:35 pm #92046
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Statman:
It is great to have some academic discussion! That’s what is forum is all about. However, I think our discussion scope is growing and may confuse people about these definitions, that we want to keep it simple. Actually, it may scare people away from Six Sigma!
It might be better to have a direct telephone discussion rather than confusing people in the forum. If you agree, feel free to call me at 847 903 4575 (cell) or email discussion at [email protected].
Regards,
Praveen Gupta0November 4, 2003 at 6:47 pm #92047
John J. McDonoughParticipant@JohnJ.McDonough Include @JohnJ.McDonough in your post and this person will
be notified via email.Praveen
I think your sensitivity is good, but I am enjoying this discussion. It’s good to see a technical debate devoid of the emotionalism and name calling that so often accompanies online discussion. I suspect there are others who are also learning from this.
But, I don’t know the distribution of folks on this forum. I could see where the discussion could be annoying or a turn off to some. I would hate to see some folks turned off from one of the most powerful tools to come along in a while because the experts (that would be you!) can’t agree.
–McD
0November 4, 2003 at 6:58 pm #92050Hi Praveen,
I will email you if you would like to take this discussion off the forum.
However, I would hope (and believe) that the strenght of Six Sigma is at a level that it is robust to disagreement and debate. I once heard a measure of the health of an organization is based on the difference between the private conversations and the public conversations. The closer these are to the same, the healthier the organization.
I believe this is also true about a movement like six sigma
Highest Regards,
Statman0November 4, 2003 at 7:16 pm #92053
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello McD:
I am afraid that I may enjoy this forum too much! Of course it takes time to write and can become an addition. Especially having a statistical discussion after a while (may expose my vulnurabilities!). Statman and I are not really disagreeing about the Six Sigma concepts, instead, I see the difference in terminology that we learn from our places of employment or teachers with varying background. I thought it would be appropriate to take care of symantics (hopefully spelled correctly) off line. Otherwise it will take a lot of writing. I do see Statman loves writing too.
Thanks for your participation.
Regards,
praveen0November 4, 2003 at 7:20 pm #92054
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Statman:
I saw a lot of agreement between our understanding of concepts. Instead the difference is in out terminology. To address those differences it will take a lot more writing (my boss may be watching! just kidding.), and I think it might be a lot easier to align off line. Of course, then if you like we can get back to the forum! I have never participated on any forum like this. I am really having fun at this forum.
I guess iSixSimga is a great place for having an exciting disussion on six sigma.
Regards,
praveen0November 4, 2003 at 7:57 pm #92057
GabrielParticipant@Gabriel Include @Gabriel in your post and this person will
be notified via email.Praveen, please keep it online. I find it insightful.
Maybe I am getting scared about Six Sigma, but it is not because of the discussion itelf. It is because of the concepts that these discussions uncover.But I prefer to be scared with the reality and not to be happy with the ignorance.0November 4, 2003 at 8:40 pm #92059
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Well, it sounds like people want to continue this discussion on the forum. Here is my reply. I like writing papers, etc. I think this discussion could become a paper, something to think about to collaborate!
Here are my some responses. As a disclaimer I am not a statistician, however, I have practiced and taught a lot of it for many years. For any theoretically erroneous statements, I apologize in advance.
———————
Sorry about referring to you as a Guru. That was not polite. I made assumptions based on your book bio.
– Which book you are talking about? Go ahead answer for me!
Is it normal to have a capability of +/ 3 sigma? Is a Cp = 1 a normal capability?
– Answer is yes, as it includes 99.73% (practically everything) of the process output.
I think you are confusing process capability with the process distribution. When a process demonstrates stability, the natural behavior of the process and its ability to meet specifications is called its capability. If you are defining the normal capability as +/3 sigma, then you are saying the specs are +/ 3sigma.
– Yes, I have learned to define capability as +/3 Sigma. No way it represents specs because specs are defined by the customer or the design engineering. You might have meant control limits.
The natural behavior (free from special causes) of a process that can be modeled as normal will demonstrate 99.73% of the process output +/ 3 sigma. But this is not its capability it is its expected distribution.
– Statistically distribution means display of accumulated probability density. While the capability implies process spread, i.e., +/ 3 sigma.
If the process is controlled within +/ 3 sigma, wheres the shift? By definition of control, there is no shift. Are we talking process shift here or sampling error because they are not the same thing.
– Answer is yes, the 1.5 sigma shift is derived from sample to sample variation. Actually when we say 1.5 sigma shift, it means 1.5 sigma shift in the process means.
Once again, you are talking about sampling error not shift because if the mean is shifting, this process is not in control it has special causes. I do fully agree that there is sample to sample variation (which for as long as I can remember, has been defined as sampling error) and that there will be a risk due to the sampling error of underestimating the true process sigma as well as a risk that a process deemed in control is not free from special causes. This is not shift, it is sampling error.
– Error or shift both represent variance, that is what we are all concerned about.
Are you implying that only a process that is in control will have shift? Or will shift occur in processes that are not in control?
– Every process will have a shift whether in control or out of control. It is only when the shift becomes excessive or unacceptable process becomes out of control.
Any shift more than 1.5 sigma will lead to the process shut down. Is this based on fact or speculation?
– This is assumed based on the control chart theory developed by Shewhart in 1920s. It means if the process means shift by 1.5 sigma, the process has shifted a lot, i.e., an assignable cause has been introduced, therefore it must shut down.
If you are basing this on control chart theory, then a shift of any sigma will be eventually detected and determined a special cause. Remember, a special cause is not detected only when a result is outside the 3 sigma limits. There are a lot of processes in the world that have degradation greater than a 1.5 sigma shift.
– To understand the 1.5 Sigma shift in the process mean, one needs to understand Standard Error of Estimate, again a measure of variation. For a sample size of 4, std. Of process mean of 1.5 sigma may be equal to about 3 sigma shift for each measurement. Thats why Xbar and R charts are more powerful than X and R charts. Again, Jurans handbook explains this better than I can explain it here with the help of graphics.
Mathematically, the simplified explanation comparing the sigma value to Cpk is accurate. Why would any other definition be different than the mathematical form?
– If the interest is just to play with the number, you are right! However, it is more important to understand concepts behind the numbers and their practical application. I believe having a statistical thinking (random vs. assignable) is more valuable than knowing all the statistical tools in this world, including Cp, Cpk.
0November 4, 2003 at 8:56 pm #92060Praveen,
Just wanted to let you know that I’ve enjoyed reading this thread also. I only have a chance to read the forum nightly, but I have enjoyed this discussion so far. Please do consider sharing the rest of your thoughts publicly … it benefits my thought process.
Regards,Carol0November 4, 2003 at 9:29 pm #92062Hi Praveen,
OK the ball is back in my court.
Based on your replies, I dont think it is a matter of semantics. However, I do agree that we do not differ that much in our view of process performance and process variation. Also, you do not need to be mathematical statistician to have an input into this discussion as this is practical application we are concerned with.
I think that there are four interrelated subject lines here so for the sake of brevity and so I dont have to keep copy/paste to longer and longer replies let me see if I can capture them.
For a variables output, how do we define capability?
How do we define shift?
Do processes have a shift greater than 1.5 and allowed to continue to operate?
Is the definition of 6 sigma any more than raising the standard of acceptability for the process Cpk?
I hope that captures it.
1. True, process spread will affect process capability. However, process spread does not define process capability. Without out the context of tolerances, there is no way to assess capability. If I have a process that has a process spread three times the size of the process spread for another process that has different tolerances, can I say that the one with the smaller spread is more capable? Not without the context. If I have a process with a ratio of short term variation to long term of 1.8 and another with a ratio of 1.2, I can say that the process with a 1.2 is better controlled but I cant say that it is more capable.
2. This is just a point of confusion for me. There is the practical explanation of the 1.5 shift that is simply based on shift happens. This is intuitive to anyone that has spent time with process analysis and is clear from the second law of thermodynamics. A process in statistical control is an unnatural state and rare. However, every time I hear the justification of the 1.5 shift from a statistical perspective it is based on sampling error of either a chisquare of normal distribution. So my question is simply this: are we compensating for worst case for sampling error with the 1.5 or are we assuming process degradation of 1.5. And dont tell me it is both. You can not give a practical definition in terms of degradation and a mathematical justification due to sampling error. I just cant logically allow it.
3. I believe that it would be far more valuable for a process engineer to understand reaction principles concerning the process not the sampling error (standard error). If I have long term data of my process, I can quite easily through components of variation analysis extract the entitlement (short term) variation of the process and compare it to the total variation of the process. If this ratio is large (large total variation vs short term) and the process is stationary and delivering to the specification, this would tell me that the process has been allowed to drift (degrade) without correction. There is a very good chance that the reason for the drift is economic. The cost of adjustment out weighed the cost of the drift. So the sampling error argument just gets in the way of good process research.
4. No interest in playing with the numbers. The interest is being able to explain these concepts to the people that can make improvements. Since there is no mathematical difference between the sigma value and the Cpk except for a constant. I believe that it is less confusing to explain that the Cpk defines the ratio of the process tolerances to the process output and the new standard for world class is a ratio of 2.00. Bringing in the sigma level only adds confusion because they dont know if you are talking about control limits or what.
Ball back to you.
Regards,
Statman0November 5, 2003 at 12:00 am #92068Man, you definitely have drank too much of the koolaid.
0November 5, 2003 at 1:07 am #92075Praveen, Statman,
Thanks a lot for sharing your viewpoints like gentlemen. The rest of us are learning in the process.
With SS having been around for more than a decade I feel at least one organisation should have data to confirm the 1.5 sigma shift. Can someone from Motorola/Allied respond please?
That would take the “assumed” out of it. As a wag said it makes an ASS out of U and ME.
Is there, anywhere in the textbooks or software, one comprehensive example with numbers (albeit concocted) of all the major concepts of SS? M***t*b has different examples for different examples for different topics.
Regards
mu0November 5, 2003 at 5:51 am #92086
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Mu:
‘i’ think my understanding would be as close to the originator as it could get as ‘i’ learned directly from the inventor Bill Smith while helping him prototyping the six sigma. My responses have been based on what ‘i’ have learned during my training at Motorola.
‘i’ don’t know if you can get any different responses from anybody at Motorola, my belief!
I hope that answers your question. It appears that Statman is pretty consistent in his inquiry. That’s great and salute to him. Since it is not converging, that’s why I wanted to handle it off line first!
There was a hypothesis in this Six Sigma model and that was the technical assumption and it brings Six Sigma closer to the reality, instead of ASS U ME’ing.
Regards,
Praveen0November 5, 2003 at 5:59 am #92087
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hi Carol:
I am glad you liked the thread. Thanks to Statman for starting this. He just loves to write. I do not think I can keep up with him. I think I already have spent a lot of time today, that I enjoyed though, today on this Forum. I better do some work otherwise else…
I appreciate your thoughts about sharing our experience. It is a winwin for everyone. So stay engaged continue to share your ideas!
Regards,
praveen0November 5, 2003 at 6:01 am #92088
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Stan:
Be my guest. You can also have some Six Sigma CoolAid.
Praveen0November 5, 2003 at 10:57 am #92090
HemanthParticipant@Hemanth Include @Hemanth in your post and this person will
be notified via email.Hi Nicole,
I generally read all the responses before posting a message here, but I didnt had the patience of going through all the messages (euphemism…) posted here. To put it simply the statement which you made is not right because the definition of sigma level (three sigma, four sigma or six sigma…) is defined as the number of standard deviations you can fit between your process mean and the closest specification limit.
To clarify further, any process which follows a normal distribution contains (theoretically) 99.73% of population between plus and minus 3 std deviations (read, sigma) band. This is true for any process following normal distribution and is entirely delinked from “sigma level” calculation.
So if you wish to define six sigma level then you could say,
“When the mean for products quality measurement falls 6 standard deviations (read, sigma) away from the closest tolerance limit then your process is a six sigma process.”
The underlying assumption is your distribution is falling between your tolerance limits and not outside…:o)
This is even better explained on this site under new to six sigma heading..
Hope this was helpful…
0November 5, 2003 at 6:17 pm #92107Hi Praveen,
Due to the long delay in you reply I have had to ask for a ruling by the judges and according to rule 3 section 2.1.1 of the isixsigma forum standard practices for debate the judges have ruled that you have not replied within a reasonable time window which allows me to post a followup to my last post. So here it is.
(I just made that up)
There are just a couple of points that you made in your last post that I have to challenge.
First, you said: Error or shift both represent variance that is what we are all concerned about.
Actually, sampling error represents uncertainty in estimating the process parameters. There is always uncertainty associated with inferences about the process parameters based on a sample. This uncertainty is due to the variation not part of the variation. Shift is instability in the process and will increase the total variation of the process (which will affect the level of uncertainty).
Secondly, you stated: To understand the 1.5 Sigma shift in the process mean, one needs to understand Standard Error of Estimate, again a measure of variation. For a sample size of 4, std. Of process mean of 1.5 sigma may be equal to about 3 sigma shift for each measurement. Thats why Xbar and R charts are more powerful than X and R charts.
The standard error of the estimate of the process mean is dependent on the total number of data points not the size of the subgroups. For example, (assuming a stable normal process) if you control charted 100 individuals and calculate the standard error of the process mean, the standard error will be:
Sigma/sqrt(100) = sigma/10
If you take the same data and subgroup to 4 the standard error will be:
(Sigma/sqrt(4))/Sqrt(25)= (Sigma/2)/5 = sigma/2*5 = sigma/10
So an IMR chart is not less precise in estimating the process mean than a XbarR chart. It is less precise in estimating the process level at each individual point, but we do not use a control chart to evaluate the difference of individual points, we use it to assess the stochastic behavior of the process.
Statman0November 5, 2003 at 7:51 pm #92109Hi Praveen,
Lets make this simple since you feel we are not converging to consciences (although I dont think that convergence is ever the goal of debate):
Choose the most appropriate answer (in your opinion) to the following. I would encourage all how visit this forum to provide their opinions of the most appropriate answers to these questions. We can then see if there is convergence
1. Process capability for a variables output is defined as:
A. the natural behavior of a stable process and its ability to meet specifications
B. +/3 Sigma (capability implies process spread, i.e., +/ 3 sigma)
2. It is appropriate to include a 1.5 Sigma shift in the process sigma because:
A. Over the long term, the short term process mean tends to shift and drift and the 1.5 shift is to compensate for these potential instabilities in the process.
B. The average calculated from a sample is an estimate of the process mean subject to uncertainty; the 1.5 shift is used to compensate for the worst case sampling error in this estimate.
3. Any shift more than 1.5 sigma will lead to the process shut down so it is highly unlikely to ever see a process that has total variation greater than a 1.5 sigma of the short term.
A. True
B. False
4. A Six Sigma process (variables output) is equivalent to a Cpk = 2.00 process. Since both of these metrics only concern themselves with the estimate of short term variation, achieving Six Sigma is measuring designed in variation and totally disregards process management variation. Therefore, neither Sigma level nor Cpk by themselves, are appropriate metrics for process improvement.
A. True
B. False.
Statman0November 5, 2003 at 8:42 pm #92110
GabrielParticipant@Gabriel Include @Gabriel in your post and this person will
be notified via email.Sorry Statman. With the following I will only add more noise to the noise:
1. Process capability for a variables output is defined as:
A. the natural behavior of a stable process and its ability to meet specifications
B. +/3 Sigma (capability implies process spread, i.e., +/ 3 sigma)
Both. This is from AIAG’s SPC handbook:
Process capability: The 6 sigma range of a process’s inherent variation, for statistically stable processes only, where sigma is ussualy estimated by Rbar/d2.
Cp: Process capability INDEX defined as the tolerance width divided by the process capability.
You know, I can say that my process is capable of producing parts within 0.2mm, as I can say that my gage has a repeatbility of 0.01mm. Put tolerance/0.2 and you have the capability to meet a certain specification (Cp). Put 0.01/tolerance and you have r&R. The difference? It is nice to know what your process is capable to deliver regardless of the specification, so you can know forehand if it is capable to meet a given specification.
Another interesting thing is the requirement of “statistically stable process only” (the underline is from AIAG).
2. It is appropriate to include a 1.5 Sigma shift in the process sigma because:
A. Over the long term, the short term process mean tends to shift and drift and the 1.5 shift is to compensate for these potential instabilities in the process.
B. The average calculated from a sample is an estimate of the process mean subject to uncertainty; the 1.5 shift is used to compensate for the worst case sampling error in this estimate.
None. It is not appropiate to include the 1.5 sigma shift. But if I had to choose one answer, I would choose:
C. Due to the uncertainty when calculating the subgroup average, you may miss shifts of up to 1.5 sigma even if Xbarbar was the exact stable process average. So the process may actually shift and you may not nocie that. To cover for that, you assume that the 1.5 sigma shift actually happened.
3. Any shift more than 1.5 sigma will lead to the process shut down so it is highly unlikely to ever see a process that has total variation greater than a 1.5 sigma of the short term.
A. True
B. False
B. False. Come on! A shift of 1.52 sigma can pass undetected with about the same chance that a shift of 1.48 sigma. Yet, you may want to shut down the process with a smaller shift or you may be willing to continue with the process even if the shift is larger than that.0November 5, 2003 at 10:43 pm #92117
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Gabriel:
Thanks for your responses. I am glad you are continuing with this thread. I believe there is an issue of interpretation and application of these concepts.
I wish I have an intellect to keep up with Statman’s challenges and tests. I have explained my understanding of the subject matter in my earlier responses. I am unable to add more to the discussion.
Regards,
praveen0November 5, 2003 at 11:36 pm #92120Gabriel,
I agree with all you have said (as usual – who did you learn from?), except your answer C. The 1.5 is only useful in the absence of real data. And when applied to attribute data – you want to assume that the shift not only happened, but it stayed there? I don’t think so.0November 5, 2003 at 11:44 pm #92121Mr. Gupta,
You think quite highly of your credentials, but Bill Smith was an engineer, not an inventor. He observed process behavior and shared it with his peers at Motorola. He was a good guy, but the real work went on all over Motorola and there is absolutely no evidence that the achievement of those around Bill is any better than those at the Automotive group, Semiconductor Group, …. Where were you when this was being rolled out to the world? It appears to me that you just showed up to cash in. You were not at GE, AlliedSignal, …Were you still at Motorola? – if you were that is not as good a credential as it was in the early 90’s.0November 6, 2003 at 12:10 am #92123
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Dear Stan:
I appreciate your anger at me using a little i, instead of big I. [It’s ok to be lightheaded sometime!] I am a practitioner and work hard at what I do. I respect everyone’s contributions, including yours.
I think you need more CoolAid than ever before! I think you are the one who jumped in this thread with KoolAid. Aren’t you?
Your comments reflect you very well, I do not think I am qualified to reply to them.
I only responded to a comment when some one said Motorola.
BTW: I had been a Six Sigma instructor at Motorola and Motorola University since 1987. FYI. If you have a job for me, I would be happy to send you my resume with references! I am sure you may have a lot more experience than me, I respect that.
Don’t take it personnaly man, Stan. We don’t have to make a thread on this topic either. That would be a total waste of time for people.
Take care,
praveen0November 6, 2003 at 12:15 am #92124
Stan copycatMember@Stancopycat Include @Stancopycat in your post and this person will
be notified via email.Mr. Gupta,Like I said earlier: Where were you when this was being rolled out to the world? It appears to me that you just showed up to cash in. You were not at GE, AlliedSignal, …Were you still at Motorola? – if you were that is not as good a credential as it was in the early 90’s.
0November 6, 2003 at 12:18 am #92125
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Statman:
I think you are asking some tough questions! Let me try to select the best answers from your choices:
1. B
2. A
3. A
4. B (Cp = 2, not Cpk)
Statman, I have passed ASQ test for CSSBB, pls. don’t prove ASQ wrong!
Have fun!
Praveen0November 6, 2003 at 12:25 am #92126
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hey Stan:
You sound like a cool guy and a tough cookie! I guess I have to answer your question since YOU have asked. I am ok to be a fool, at least for once. I know how it feels to be a fool now. Thanks for your reminder.
I was at Motorola from 1981 – 2000 in different roles. I certainly was not at GE and Allied as you might have been. However, I might have trained their people at Motorola University.
I do not know you at all. I would appreciate if you could shed some light on yourself. I think people in this forum may love to hear that too. I am sure I might have missed your earlier introductions.
I suppose that this kinda of discussion make this Forum a little more interesting!
Regards,
praveen0November 6, 2003 at 12:26 am #92127Praveen,
I don’t think you would find these questions on the ASQs test. For one thing, I don’t think there is a right answer. I just want to determine what the common understanding of these question are.
Also, if they were ASQ BB exam questions I would have to fill them full of double negatives and obscure terminology so that I can really test your ability to figure out the question rather than test your knowledge.
Cheers,
Statman0November 6, 2003 at 12:30 am #92128
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Hello Statman:
If there is no right answe, you must asked all the wrong questions! I do agree that to pass these multiple choice questions, one needs to learn how to answer the questions, instead of learning the material.
I would love to know a little more about your background, if it is ok. I think we have interacted well at this Forum. I have shared my background in some earlier messages at the instructions of STAN. Maybe, there is something we can do to make the world a better place to live, or make this Forum better sounding!
Regards,
Praveen0November 6, 2003 at 5:01 am #92139You flounders are going to get me worked up again! Didn’t we just chase off a yahoo talking smack about 3.4 dpmo and how that was what six sigma was all about (Reigle)? Wasn’t he a motorola / Harry guy also?
While all of this debate is “interesting” aren’t we missing the TRUE defintion of six sigma. Using a problem solving methology and an analytical tool set to improve the performance of our product and processes to a level that is appropriate to the business (or something similar along those lines…)
I WAS at AlliedSignal when this was first rolled out and spent numerous years working as a BB and a MBB at that company and I guarentee you that Six Sigma was NOT defined in terms of dpmo, cpk, cp or any of these other metrics. Six Sigma was defined in terms of achieving substantial business results (this may be in terms of quality, cycle time, process or product performance, inventory reduction, or whatever…).
I looked through all of these posts and I only found ONE that even tried to talk about it in these terms.
This is not helping our cause….. Someone who is trying to learn more about what Six Sigma is all about is going to see all this non sense and decide that they really are not all that interested anymore.0November 6, 2003 at 6:59 pm #92193
John J. McDonoughParticipant@JohnJ.McDonough Include @JohnJ.McDonough in your post and this person will
be notified via email.Bravo Jim!
I think I was probably not the only one tempted to jump in, but instead just sitting back on the sofa, Bud in hand, enjoying the Statman and Praveen show. (and learning a thing or two in the process).
Of course you are right. There have been countless quality initiatives in the past, each focused on some statistical tool or another. The thing that has really differentiated Six Sigma is the focus on business results. Of course teaching people to use a broad range of tools to make data based decisions significantly helps, but at the end of the day, it’s the focus on business results that sets Six Sigma apart. And after all, wasn’t that what the original question was about?
–McD0November 6, 2003 at 11:21 pm #92205
Praveen GuptaParticipant@PraveenGupta Include @PraveenGupta in your post and this person will
be notified via email.Here is a definition of Six Sigma:
qA longterm and forwardthinking strategy to fundamentally change the way of doing business for generating immediate improvement to profitability.
Praveen0November 7, 2003 at 3:50 am #92218
HemanthParticipant@Hemanth Include @Hemanth in your post and this person will
be notified via email.Hi Jim
I am sorry I got you worked up. Honestly, I agree with you, six sigma is all about improving your processes in a structured and scientific manner. I was just answering a specific query posted here. I agree with you to the extent that people do get bogged down by too much of dpmo, sigma etc. But is it not true that six sigma methodology needs to be supported by these statistical jargon? and I am only helping some one who is attempting to understand this.
Hemanth0November 7, 2003 at 4:20 am #92219Hi Hemanth,
It is interesting to note how quickly people jump to say that of course six sigma is all about process and product improvment etc… but that sure doesn’t stop them from spending a tremendous amount of time talking about nonsense like this thread…
It is absolutely true that six sigma is SUPPORTED by statistical jargon, but not this kind of jargon. The critical thinking comes first and the stats are there to support the thinking! I have not seen a lot of critical thinking in the posts so far.
And, look at the title of the thread – Definition of Six Sigma! All of those posts defining six sigma without hardly even a thought of the definition we just agreed upon. Just a bunch of hoohaa about the 1.5 sigma shift.
Disapointing…0November 7, 2003 at 4:53 am #92220
HemanthParticipant@Hemanth Include @Hemanth in your post and this person will
be notified via email.Hi Jim,
The objective of six sigma is process improvement. But lot of things go into achieving this goal.
Let me ask you, “Do you know how to calculate sigma level or calculation for % population under a normal distribution curve?” Believe me every black belt is taken through these stuff.
Does it mean this is all unnecessary??? As a black belt I was taught to think statistically and not just by mean or average. So lets not discourage some body who’s putting in effort to understand this.
I dont know what were your experiences, but I have found this knowledge of process improvement (both methodology and stats) very much applicable and useful.
As far as the title goes..I say “Whats in the name..!!”
Dont you think our discusssion is under a very wrong thread..(may be somebody might pose an objection on that..)
People like you can help a lot..these type of messages will only discourage them to post their queries…
I am sorry I am being a bit harsh here..
Hemanth0November 7, 2003 at 1:33 pm #92225Praveen,
Thank you for the response. Do you know how many people on this forum claim to be the “absolute” with regards to Six Sigma? I know of at least 10. I know most of them and they definitely “enhance” their real role, I have to assume you are one of them. I know Motorola U in 2003 does not mention your name. I am not one of them, but I do claim to be a competent practitioner who does not buy all of the hype.
I think you add value in the responses to well thought out questions. I have picked up your book and I will say it is in the top third of the books related to Six Sigma. Just understand that I think that books that sell Six Sigma as the end all, be all, are mostly hype.
The real issue facing companies is growth and efficiency, the tools packaged as Six Sigma help to move in that direction. All of the collective intelligence focused on improvement helps move toward that. But so does Lean, so does TPS, … And none of it is worth anything without a solid agreement within an enterprize of how business is to be done – sometimes known as a Qualtiy System.
So please Praveen, continue to help, but stop the hype. Nobody cares that you worked next to Bill Smith except you. Most people who wave the banner of Six Sigma do not know the tainted history, nor should they care. We are moving forward, not looking back.0November 7, 2003 at 4:57 pm #92234Amen – I could not have said it better.
0November 7, 2003 at 5:05 pm #92235The previous post is from a Stan copycat. I wish I had seen it before reading Praveen’s response to it . I understand the “tough cookie” comment.
Mr Stan Copycat – find your own name to post with. I do enough damage to myself without help.
“that person called Stan”0December 9, 2003 at 8:26 pm #93387
DarrlynParticipant@Darrlyn Include @Darrlyn in your post and this person will
be notified via email.Thank You for the website it helps a lot!!!!!!!!!!!!
0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.