Having once worked for a company that embraced six sigma principles, I know that the answer to you question is Yes. And No...
It is a quality improvement regimine that strives to develop precesses that reduce defects to the lowest posible level (six sigmas, or standard deviations, refers to defect rates well below 1%).
In some processes that is desirable. If you produce medical electronics you want 0% of your products to electrocute patients, for example. But it can be mis-applied: do you care if every pretzle in the bag weighs exactly the same as every other pretzle?
It is useful for focusing an organization on quality. And they do use sound techniques for evaluating and prioritizing process improvement efforts.
Defects are not the only measure of quality and customer satisfaction. It can be over done, which is when it begins to resemble a cult. It is a technique, not a magic incantation. Embracing six sigma does not automatically improve quality. That takes hard work no matter what method/philosophy you use.
It is a quality improvement regimine that strives to develop precesses that reduce defects to the lowest posible level (six sigmas, or standard deviations, refers to defect rates well below 1%).
In some processes that is desirable. If you produce medical electronics you want 0% of your products to electrocute patients, for example. But it can be mis-applied: do you care if every pretzle in the bag weighs exactly the same as every other pretzle?
It is useful for focusing an organization on quality. And they do use sound techniques for evaluating and prioritizing process improvement efforts.
Defects are not the only measure of quality and customer satisfaction. It can be over done, which is when it begins to resemble a cult. It is a technique, not a magic incantation. Embracing six sigma does not automatically improve quality. That takes hard work no matter what method/philosophy you use.
MRC_Hans
Penultimate Amazing
- Joined
- Aug 28, 2002
- Messages
- 24,961
I don't think it is a fad. Well, it can be overdone, or more appropriately, misapplied. As patnray says, if you make every pretzle weigh within 0.1%, but that is not what it should be about. Then you have set your control range wrongly.
Basically, six sigma is about normal distributions. Most things that you make enough of show a normal distribution. That is a bell-shaped curve. Now, at +- one standard deviation, most of the units are within the limits (I can't look up the exact figure here, and I forget it, but its to the tune of 60%). Thats +- one sigma.
Two standard deviations (+-) encompasses around 95% of the units (correct figure, anyone?). That's +- two sigma.
When you produce to +- six sigma, for all practical purposes, ALL units are within the limits. So if this is a controlled process, you will not need to test the result (=lotsa money saved).
In the case of the pretzels, lets say we want them to weigh 20g +- 10g. So if we make a six sigma production line, we make sure that practically no pretzels weigh less than 10 or more than 30.
Hans
Edited to add: Well, I suppose that was not very helpful.........
Basically, six sigma is about normal distributions. Most things that you make enough of show a normal distribution. That is a bell-shaped curve. Now, at +- one standard deviation, most of the units are within the limits (I can't look up the exact figure here, and I forget it, but its to the tune of 60%). Thats +- one sigma.
Two standard deviations (+-) encompasses around 95% of the units (correct figure, anyone?). That's +- two sigma.
When you produce to +- six sigma, for all practical purposes, ALL units are within the limits. So if this is a controlled process, you will not need to test the result (=lotsa money saved).
In the case of the pretzels, lets say we want them to weigh 20g +- 10g. So if we make a six sigma production line, we make sure that practically no pretzels weigh less than 10 or more than 30.
Hans
Edited to add: Well, I suppose that was not very helpful.........
Nyarlathotep
Philosopher
- Joined
- Mar 26, 2003
- Messages
- 7,503
I used to do QC work for a foundry. They were starting to use 'six sigma' methods. They sounded okay in theory but the company I worked for seemed to want to use project B.I.F.F. instead (Big Improvements For Free, where they want improvements but don't actually want to spend any money or change anything) instead. Beyond a bit of shuffling around of equipment and making a lot more paperwork, their changes seemed to consist of a daily meeting telling everyone not to make defective castings (of course, how simple, just don't make defective castings). As a QC inspector, I saw no noticable change in the number of defective castings coming through. It may be my own bias at work, here, but it seemed it was getting worse when I left. Though I did here that the Six Sigma techniques were working pretty well in some of our other plants.
The moral, I guess, is that the efficacy of Six Sigma is dpendant on it's implementation as much as anything.
The moral, I guess, is that the efficacy of Six Sigma is dpendant on it's implementation as much as anything.
SRW
Master Poster
- Joined
- Jul 25, 2001
- Messages
- 2,903
Sigma Six has it origins in Japan when Dr Juran first started teaching Total Quality Control and Statistical QC in the late 40's early 50's. It really took off in the 80's and 90's when Motorola and GE started to apply the concepts. It can work but takes lots of buy in.
From what I understand, it is a modern 'improvement' upon Shewart's and Deming's (and others') quality control work.
Personally, from what I know about quality control and statistics, I think 3.4 defects per million for any process or procedure is pretty much wishful thinking, as different processes and procedures would, I imagine, have a different 'defect structure'. Maybe it is possible, I don't know for sure. Either way, I think that improving quality is more than eliminating defects.
Six Sigma is a great approach (minus its jargon), it is just not a new approach statistically speaking, or equivalently, it is an old approach (statistical techniques to study processes and procedures) in a new package, but well worth looking into.
Personally, from what I know about quality control and statistics, I think 3.4 defects per million for any process or procedure is pretty much wishful thinking, as different processes and procedures would, I imagine, have a different 'defect structure'. Maybe it is possible, I don't know for sure. Either way, I think that improving quality is more than eliminating defects.
Six Sigma is a great approach (minus its jargon), it is just not a new approach statistically speaking, or equivalently, it is an old approach (statistical techniques to study processes and procedures) in a new package, but well worth looking into.
davefoc
Philosopher
I am somewhat familiar with the ideas associated with improving quality through quantitative testing to achieve very low defect rates.
The idea is that if one just does simple go/no go testing one is missing a great deal of information that can be used to predict what the real defect rate is before one has built the massive number of widgets to determine it.
For instance in the case of your pretzels, one might just have a machine that measured their weight and kicked them out if they were too small or too big. This is fine, but you might be able to significantly reduce the defect rate if the actual weight of each pretzel was measured and you adjusted your process when the average weight wasn't centered between too low and too high.
The problem with this theory for some means of production is that some kinds of production gradually drift out of nominal condition but the drift may not be enough to substantially affect the defect rate. Great expense can be expended shutting down production to readjust the process so that out put is centered between the acceptable range when just doing nothing would have produced perfectly reliable products.
I have never heard the term six sigma, black or green belt so I hope my answer was at least somewhat relevant to the discussion.
The idea is that if one just does simple go/no go testing one is missing a great deal of information that can be used to predict what the real defect rate is before one has built the massive number of widgets to determine it.
For instance in the case of your pretzels, one might just have a machine that measured their weight and kicked them out if they were too small or too big. This is fine, but you might be able to significantly reduce the defect rate if the actual weight of each pretzel was measured and you adjusted your process when the average weight wasn't centered between too low and too high.
The problem with this theory for some means of production is that some kinds of production gradually drift out of nominal condition but the drift may not be enough to substantially affect the defect rate. Great expense can be expended shutting down production to readjust the process so that out put is centered between the acceptable range when just doing nothing would have produced perfectly reliable products.
I have never heard the term six sigma, black or green belt so I hope my answer was at least somewhat relevant to the discussion.
Hellbound
Merchant of Doom
MRC_Hans said:
*sniff*
That's about the saddest thing I've ever heard. No idea what a pretzel is, you poor, poor man.
Hey, PM me a mailing address and I'll mail ya a bag of em
