…Well, isn’t it?
Not much of a debate presented, here. Did you want this moved to IMHO? Or did you wish to set forth some actual reasons and arguments to support your opinion?
I don’t know what it is, frankly.
If the defects are out at six sigma, then yes, I’d say that the six-sigma indicates rubbish.
Anyone who has gone through quality training probably has a lot of points of discussion to offer.
My view - six sigma is only rubbish if it is misapplied - which I’ve seen happen way too often. If you are dealing in an environment with enough operations or transactions to make statistical analysis feasible, six sigma etc. it not only a good idea, it is a no brainer. (I’m not talking the specifics of a quality initiative, only in general.) When you try to apply it to something like design which doesn’t have many similar operations, you get into trouble. I’ve seen such a push from the top that managers of operations where it doesn’t make sense feel (or be pushed) into adopting it. It doesn’t work, it gets a bad reputation, and it gets silently dropped.
Two examples from long enough ago so I won’t get into trouble. When I was at Bell Labs, one of our factories adopted quality (not six sigma, something else.) There was an actual improvement - in fact they won the Baldridge award (and went bust, like so many Baldridge winners.) It worked for them. On the other hand, when I got training, it was from someone in our computer division. Our confidence was not increased when a group got the computer these guys had designed - and it was a hunk of junk.
Anyone else with quality experience?
Oh, quality control. And me with an MIS degree. I’m quite honestly embarassed at my lapse.
Aren’t these theories the business version of Miraculous!!! Diet!!! Pills!!! ?
It’s rubbish only if you throw it in the trash.
This is going to be hard to do without graphics, though here is a link. Say you are making a part. If you make the part too big or too small, it can be considered defective. Now your cutting process has variance. If it is in control, you can make almost all your parts within the needed limits, and you have good quality. If it is out of control, a lot of your parts will be out of spec. Six sigma is when you have only 3.4 defective parts out of a million (which is actually 4.5 sigma, but who’s counting). You can consider this as a Gaussian distribution of sizes centered on the correct size. If it is tall and narrow more of your parts will be in spec than if it is short and fat. If you draw lines around the center point, representing the points at which the part is too big or too small, the part of the distribution outside of those lines represent defects.
Now, if you look up Six Sigma you’ll find DMAIC and other quality improvement processes which have been tied into six sigma, and which get used for things you can’t really do statistics on (like programming.)
On preview - never mind - though maybe this will be useful for someone else.
The links I found, like the one above, don’t really give the basics, and are six sigma babble.
Might well belong in the other forum Tomndebb. I guess my reason for posting here was to try to spark a bit of a debate. We’re starting to get some six sigma methodology filtering through my workplace (Finance) and it’s got me a little, well bemused might be the right word.
One of two things have started to happen. Either we take a process that is mind numbingly clearly inefficient and rather than fix it, we measure it for a couple of months first, or we take another process that really could stand some measurement, and we measure the kind of things that can easily be worked around to give false good numbers. Those concerned will say, “Oh, you’re measuring the amount of time between receiving an email request and opening it?”, then set about opening and printing out all the emails in the queue so it looks like the emails have been dealt with and the numbers look good. Those running the programs are two levels above those doing the work, and they go around patting themselves on the back re the improvements without realising that things have actually got (a) worse, and (b) less efficient.
The clincher for me is that I’ve just been asked to do a House of Quality analysis on a proposed featureset, and I’m looking at this thing and I’m just not too sure that assigning a subjective value to something, then performing some mathematics that result in an answer to two decimal points (it was a subjective value remember), is going to get us out of the wilderness.
I was hoping the savvy Doper community could kind of give me an external perspective!
M
I really don’t understand it at all. So you’re saying if you improve the process and reduce the number of defects, you save money? Is that suppose to be some kind of secret? So fewer defects == good, fine, apparently my parents’ generation didn’t understand this, but why 3.4 per million, or whatever the mystic number is? What the fuck does “six sigma” bring to the table, other than telling me the manager that I should reduce the number of defects? Does that guy tell me exactly what I’m going to do on the widget assembly line to reduce those defects? 
To a certain extent measuring things makes sense, because that is the only way you can figure out if something is out of whack. I think your experience demonstrates the principle that people optimize what they’re measured on. If you don’t measure and reward the important things, you can get into trouble.
It sounds like you have another case of people using statistical techniques in places where they don’t make sense.
#1 sign that you’re not using quality where it makes sense - if the answer to how to measure is that you’re going to do a survey.
To be fair, there is one useful part of the quality improvement process, which is understanding the process by which you do things. Sometimes things are done out of habit and tradition, and charting the process makes it clear which things are redundant or too complicated. But for 15 years my salary came from factory guys. When their processes get screwed up, it becomes a lot more obvious than when the design or office process gets screwed up. The inventory you trip over is kind of a sign.
Six Sigma was just Motorola’s name for the Total Quality program. It’s a lot more than just eliminating defects. You start out with capability studies of your processes and then make sure you are controlling critical parameters during the process itself and not trying to inspect in quality at the end. But there’s a whole methodology and management technique that comes along with it. The wikipedia article is OK, but just outlines the program at a pretty high level.
At any rate, no it’s not “rubbish”. It’s more difficult to implement in some processes than others, but it (or something like it) is a proven technique used by most world class manufacturing operations.
Thanks for the feedback so far.
John Mace - are you in a position to assess its worthiness in a non-manufacturing environment or does it boil down to the skillfulness of the execution?
Thanks again.
M
It’s how you reduce the number of defects. And the savings comes from increased efficiency. However, when Motorola says they saved 17 billion dollars from six sigma, I say bullshit. I’m sure that when all the savings claims are summed it comes to seventeen billion, but these estimates are notoriously overblown, and almost never measurable or confirmed. The execs who are pushing six sigma don’t want to question it, the lower level managers who get bonuses for saving money don’t, and the grunts who know it is crap see no reason to take the heat from saying the emperor is naked. The six sigma saving calculation process is a textbook example of an out of control process without feedback!
BTW, one of the best Dilbert sequences ever had him teaching quality to the Elbonians. In the last strip, as he was leaving, one of the Elbonians praised him because the number of defects had gone down by 20%. Another Elbonian noted that they were in quality meetings 20% of the time. Dilbert thinks “Uh oh, they’re on to me.” I couldn’t find it anywhere to link to it.
The rubbish part comes in when people try to apply manufacturing quality methods to processes that are not repetitive and which do not provide enough data for statistically significant analyses.
The six sigma level may or may not be useful, also. Any fab that has 3.2 defects per million is way out of date! It also depends what you’re measuring. 3.2 defective transistors out of a million is zero yield.
Standard deviations. Six sigma is not in itself crap, only when misapplied. The whole point of such quality initiatives is to reduce waste, both of time, and materials.
Example:
I make wigets.
My widget machines make 1,000 widgets per hour.
100 of those widgets per hour have defects making them unsuitable for sale.
My effective output is 900 widgets per hour, plus loss of materials, time to sort out the bad ones, time to fix some of the widgets to salvage materials, etc.
So I have my crew sort the widgets by defect type.
Rather than spend money on fixing and sorting bad widgets i decide to shut down the line and fix/replace/adjust the part of my production line causing the most defects.
Result, my line now puts out 960 good widgets an hour and the bad ones are not as bad as they used to be.
Once again sort the problems, identify the parts causing problem, and correct.
Now maybe I am generating 988 good widgets/hour.
I decide to crank up the speed of my line to 1,500/hr, more speed usually equals greater potential for problems, reanalyze errors, identify problem area, solve problem.
Lather rinse repeat.
Eventually my accuracy level is so great I have little need for quality checks except as spot checks to watch for developing problems. High quality = less quality assurance and recovery labor = more profit.
When you are talking about small expensive items like ipods or xboxes, 1% line defects could easily be hundreds of thousands of dollars a day in lost or damaged product that may or may not be able to be recovered or repaired. Numbers like that are hard to ignore even in huge businesses. Dropping a couple million in lost production could easily offset itself in a few weeks if defects could be reduced to .5%. Fewer problems, less waste, more product available to ship on time, at a lower cost.
Ask an easy one next time, willya?
Six Sigma is… well, over-emphasized would not be too strong a word. It’s currently making the rounds for all sorts of QA procedures, including in R&D, finance, call center operation, and who knows what else; honestly, I think it’s best at making money for the people ‘training’ it. But anyway.
In theory, it’s not bad. The idea is to reduce the variance of your ‘output’, whether that’s manufacturing Left Handed Widgets or handling customer calls in your call center. You do this by
(1) identifying areas of improvement
(2) identifying and measuring your variance (what you want vs. what you get)
(3) instituting procedures designed to eliminate the source(s) of said variance
(4) collecting data for an arbitrary but statistically useful period of time
(5) determining if you’ve hit your goal w.r.t. variance
(6) going back to step 1, 2, or 3 depending on the outcome of step 5
Repeat ad infinitum.
Now, while it sounds like a plan for making overpriced consultants rich (and is), it CAN result in improvement. Most people’s problems with it are that they do not understand what it is to begin with (a process, not a solution); they do not understand statistics, what is statistically significant vs. what isn’t, and what should be considered part of the data set in the first place; and finally, what they really want to begin with.
Suppose you make Left-Handed Widgets in huge amounts, with the expected side effects of manufacturing (wastage of various sorts, defects in products, staffing issues, etc). Your company has, for wise and mystical reasons, determined that you need to implement Six Sigma methodology on your assembly line. Where do you begin? What do you analyze?
If you’re like most, you hire outside consultants who charge humongous amounts and turn your business upside down. In six months to a year, you’ve reduced your wastage by half! Wow!
After all is said and done, you’ve gone from 10 defective widgets out of a million to 3 defective out of a million (be patient, I’m generalizing).
Now the question is, do you actually SAVE anything by doing this? Or are the procedures you’ve instituted and the people you’ve hired/fired more expensive than the revenue you can generate from those 7 extra widgets out of every million?
And that really illustrates the problem. Most businesses don’t understand that there is no need to use six sigma practices unless it gains you more than you lose. And many of them don’t seem to know how to DO this sort of analysis, and of course, the commercial trainers of S.S. aren’t going to speak up about it.
Fair disclosure: I am not, nor have I ever been, affiliated with Six Sigma in any form. I have attended several classes on the process, geared towards the call center industry (but the theory applies to any large-scale repeatable process). I know whereof I speak in that limited fashion only… but candid discussion with an ex-trainer leads me to believe I am correct.
I’m not John Mace but I may be able to help. Fair disclosure: I work at a high level in the call center industry, in a position that involves a LOT of data analysis, and I have taken Six Sigma classes (but do not have one of the ‘belts’ or degrees for it).
Six Sigma is currently all the rage in call centers, which, as you may guess, are not manufacturing operations in any sense. Typically, they do involve repeatable (and thus measurable) operations, but often these are overshadowed by the non-relevant stuff. In short, the repeatable stuff is statistically insignificant when compared with the non-repeatable stuff.
That’s okay, though. What you do is ignore the stuff that doesn’t fit and concentrate on what does. :rolleyes:
I could go into a lot of detail here, but I think I’ll sum it up by saying: S.S. ain’t what it’s cracked up to be when you take it outside the factory. It’s not meant for something that involves individual interactions instead of repeated operations. It can make a difference, and a good one, if properly applied, however. The key is not to focus on end-user quality (ie, what the customer experiences) but instead focus it on the internal operations of the business. That means standardizing practices such as training, recruitment, operations, etc, and then being brutally honest in recording and documenting variations from those standards. If you do it right, in time improvement will come. Whether or not your customers ever see it is a different matter. But if you start saving money on operations, that’s not a bad thing.
As everyone knows, though, it’s really, REALLY hard to get everyone in a company to 'fess up when they screw up. It’s even harder to effect change on entrenched practices. Therefore, I have to conclude that S.S., while nice in theory, is pretty much just a way for high-priced consultants to make a living (outside the manufacturing world, I mean).
So, upon preview: yeah, it all comes down to the skillfulness of execution. Oops. I got a bit carried away.
Heh, just got a class on project management, not an expert so this is my impression so far:
In a nutshell, six sigma is a system that improves a process by eliminating defects, the advantage is that it points to flaws and improvements that can be applied to a system already in place, problem is that many companies paradoxically end up concentrating on improving the process but not the product itself. Another issue is seeing companies applying the system to a process that has not been established yet. (and then it runs into accusations of it being a fad or a fraud)
Although not much sourced, the commentary in this manufacturing blog is to the point:
http://kanban.blogspot.com/2006/11/pointy-haired-boss-adopts-six-sigma.html
Six sigma is a tool, but like any tool, it can be misused or ignored.