Six Sigma?

[Kevbo]

@Stranger_On_A_Train

The classic case is Motorola putting this methodology in used in its semiconductor microprocessor manufacturing to identify the key areas to reduce the incidence of defects to less than 3.4 per million units.

Which is a failure rate of 4.5 sigma. Better to have a catchy name, than to be mathematically precise in naming your statistical quality improvement program, obviously.

It is worth noting that Motorola moved on from Six Sigma starting about 1991. GE, under Jack Welch, was an adopter, and many Jack Welch worshiping CEOs followed suit. Post Welch, GE is using Six Sigma…not so much.

And that is just the start of the bullshit. My experience in three different companies is that it has never been anything more than busy-work eyewash. When Larry Bossidy (a Jack Welch protege’) announced his Six Sigma implementation initiative via a video-taped message to AlliedSignal employees, the ONLY reason he gave is that he hoped it would impress Wall Street and bump the stock price. Neither product quality, nor yield were apparently any sort of goal.

There was a mention up-thread of of “design of experiments” (but not by that name.) In every case this requires that you start by deciding which process variables matter, and which can be ignored. In other words, you have to understand how the variables impact the process. In which case you aught to be able to analyze the process by traditional means, instead of the DoE methodology that often resolves to either a local minimum/maximum, or a boundary condition. It is an essential truth that if you understand something well enough to get good DoE results, you don’t need DoE.

One great proponent of DoE was Genichi Taguchi, and “The Taguchi Method” is a widely used DoE technique. But Taguchi’s work does not withstand scrutiny. His most used example is that of optimizing a voltage regulator. But it turns out his “optimization” results in a circuit that doesn’t regulate at all, because he started out by assuming constant supply voltage, and constant load current…which are the very things, variations of, that a voltage regulator circuit is intended to counter. If you hold those constant, then the entire circuit can be replaced by a single resistor. Further, Taguchi’s optimization was to replace precision resistors with transistors with tightly controlled “Beta” parameters. Beta in a typical transistor has a -50 +100% spread. It is really hard to control, and EE’s work around that. If you want a tighter value, you test individual components and select for what you want…but Taguchi claimed that his technique would result in so much savings that you were taking food from the mouths of widows and orphans if you didn’t follow them.

My personal hero, Bob Pease explains it in detail here, but you probably need to be a EE to follow him.

What's All This Taguchi Stuff, Anyhow?

What's All This Taguchi Stuff, Anyhow? (Part II)

[filmore]

@ivylass

We don’t use it at my work, but in my quest for a new job I keep coming across jobs that, at the least, have an interest in my Six Sigma level.

If you’re seeing a lot of job postings where they require the cert, then it seems $79 might be worth it. It may help you get past the automated resume parsers and HR gatekeepers to get an interview in the first place. If the job uses it heavily, I doubt if you’ll be able to jump right in with the knowledge you get from the course. But at least the cert will make you familiar with the concepts.

[stillownedbysetters]

We use six sigma in my workplace and have achieved a fair amount of success with it, but this is probably because it hasn’t taken over the corporate culture. A group of people from various departments were trained all the way up to the black belt level. They now make up a Continuous Improvement department that periodically audits our manufacturing, logistics, and a few other departments, looking for inefficiences and areas for improvement. When they find one, they develop a project plan and if approved by management, they implement it. None of our rank and file employees are trained as ‘belts’ on any level. When a six sigma project is implemented in your work area, you are temporarily managed by the appropriate member of the CI team (someone with work experience in your department’s processes).

Our CI group has saved the company a goodly amount of money and has steamlined quite a few of our processes. It was of particular value in our company because we made a large amount of acquisitions in a short time and were running multiple systems throughout the company that did not sync with one another and were, as a result, being money drains.

We’re told the CI team will continue to exist as long as it it financially justifying its existence. Presumably that means we’ll eventually get another management flavor of the month, but six sigma is not a program without value, if your circumstances warrant it.

As advice to the OP, I wouldn’t go whole hog spending money on the training. But I agree that a basic course/certificate would probably be a big plus to any company who utilizes the system.

[Lightray]

(Disclosure: I was certified as a Six Sigma black belt, many years ago.)

@ivylass

So maybe I don’t have to fret that I don’t have a white belt in Six Sigma.

If you get the job, they’ll be much more interested in your statistical prowess, than in your familiarity with Six Sigma jargon.

However. People trying to fill job openings nowadays are often buried under resumes, so much so that you need some sort of screener to weed it down to a manageable number. Checking “is this person Six Sigma certified” is a simple task that even HR can handle.* In this case, your resume doesn’t pass the screening and doesn’t get you into the interview where you can impress them with your command of statistics.

If you do get into an interview, discuss times you’ve used root cause analysis and statistical analysis (t-test, ANOVA, DOE, etc.) and you should be fine.

• I kid! I kid! There is no task simple enough for HR to handle.

[GargoyleWB]

I did up to Green Belt in Six Sigma. It’s greatest value has been in socializing a bunch of statistical shortcut methods and tools to help plan projects. Like violence, it is most useful when you can choose not to use it. It’s been an occasional useful resume bullet point for new jobs also.

The worst part of Six Sigma is the zealot culture, it really is a corporate cult and it is impossible to have any objective criticisms, analysis, or productive discussion about any of the techniques. The user of the process is always blamed for any shortcoming, never the process itself. E.g. “Darryl’s group didn’t show the same productivity, but that is due to incomplete follow-through on blah and blah and blah…”

Our megacorp went through the grinder several years ago, and has mostly abandoned it as a waste of time that is better accomplished by existing alternative methods (Toyota system, Lean, Deming, et al).

[AuntPam]

I can’t speak for the manufacturing sector, but in sectors where a high degree of human interaction and knowledge of highly flexible products (a basic design with numerous options; several basic choices with numerous options; a multitude of product choices) is the norm, Six Sigma is worthless. It’s not even worth zero: it actively impedes clear thinking and productivity by employees.

The amount of time wasted on this crap by people who should be doing real work is amazing. If there’s a problem, find the people who are blaming it on the “other guy/department/division” or “the supplier”. Put them in a room together. Regulate discussion (no yelling). Make each side explain what it’s trying to do and why the Other Side is not helping.

Remind everyone that they have a common goal (and what it is) and ask for their help in solving the problem.

That’ll work.

If you get a a rageaholic in the group, fire him or her immediately. Then you’ll get a really fast solution.

[snowthx]

Here’s the article from a few years ago about a consulting firm stating that companies adopting Six Sigma saw weak stock performance compared to market trends…

*Since announcing the adoption of Six Sigma on July 1, 2001, Home Depot shares are down 8.3% compared with a 16% rise in the S&P 500 over the same period. The stock rose more than 2% yesterday on the New York Stock Exchange, to $41.07, after Mr. Nardelli’s resignation.

Honeywell shares are down 7.2% since its Six Sigma announcement in early January 2000, compared with a 3.6% fall in the S&P 500. Shares of 3M are off about 1% since late December 2003 versus the S&P 500’s 29% climb. GE shares rose sharply in the 1990s, but they’re down 16% since July 2000, when the company adopted Six Sigma, compared with the 2.6% fall in the S&P 500.*

I am not sure if this analysis has any value, or if the study has been recently updated.

For the OP, I agree with the posts suggesting gaining some familiarity with the concepts and even going as far as the bottom-level cert, just to help get gigs.

[Busy_Scissors]

@Kevbo

There was a mention up-thread of of “design of experiments” (but not by that name.) In every case this requires that you start by deciding which process variables matter, and which can be ignored. In other words, you have to understand how the variables impact the process. In which case you aught to be able to analyze the process by traditional means, instead of the DoE methodology that often resolves to either a local minimum/maximum, or a boundary condition. It is an essential truth that if you understand something well enough to get good DoE results, you don’t need DoE.

One great proponent of DoE was Genichi Taguchi, and “The Taguchi Method” is a widely used DoE technique. But Taguchi’s work does not withstand scrutiny. His most used example is that of optimizing a voltage regulator. But it turns out his “optimization” results in a circuit that doesn’t regulate at all, because he started out by assuming constant supply voltage, and constant load current…which are the very things, variations of, that a voltage regulator circuit is intended to counter. If you hold those constant, then the entire circuit can be replaced by a single resistor. Further, Taguchi’s optimization was to replace precision resistors with transistors with tightly controlled “Beta” parameters. Beta in a typical transistor has a -50 +100% spread. It is really hard to control, and EE’s work around that. If you want a tighter value, you test individual components and select for what you want…but Taguchi claimed that his technique would result in so much savings that you were taking food from the mouths of widows and orphans if you didn’t follow them.

DoE in chemistry went through a phase of respectability - I think it probably did conform to your assessment of It is an essential truth that if you understand something well enough to get good DoE results, you don’t need DoE, but it didn’t seem like complete bollox. I never did it myself, but had a few PhD students who used it on industrial placements - the results were pretty average, tbh, but it was more a training exercise to do something different. We’re interested in discovering new chemistry, not a deep refinement of a process that more or less works, so it didn’t seem like the right tool for that.

What would be seriously good is if DoE delivered in parameterising the actual ingredients of your reaction, like the ligand or catalyst, using principle component analysis [as opposed to continuous variables like temperature or concentration]. You certainly can do this, according to some DoE proponents I’ve seen speak, but in reality no one does. Whether it’s because it’s too time-consuming to set this up accurately, so it’s actually quickler to run 100 reactions and see what happens, or because DoE is just a giant cauldron of shite, I’m not sure.

[ChockFullOfHeadyGoodness]

@ivylass

What is the benefit of Six Sigma?

It keeps Six Sigma trainers and consultants employed.

Also, it looks like it could be useful in a manufacturing environment.

But I work at a company that had a raging hard-on for Six Sigma in the mid-2000s, that insisted every single salaried employee should have at least a green belt and it was an item on the annual performance appraisals. But my department wrote software, and it’s completely worthless in that context. I got sent to a two week class on the company dime, but never did my project (which was to be in addition to our regular work assignments) or got my green belt. After about 2 years, the CEO retired, and that was it for Six Sigma. For the next few years we threw tons of money at Gallup to tell us how to be more “engaged” because Employee Engagement™ was going fix all our problems. This year we’ve moved on to some other consultant to tell us how to be more Inclusive™!

[Nava]

@Stranger_On_A_Train

(long reply to me pointing out that there are a bunch of statistical functions)

Did you read what I said?

Allow me to rephrase:

Like any other function, the Gaussian is not always applicable.
Lots of people learn only the Gaussian and insist on applying it where it is not applicable.
If you insist on applying a function where it is not applicable, you will get crap results.
I never said that “statistical analysis is bad” (my undergrad thesis was in stats for a teacher whose specialty is recursive DoE, basically what people tried to boil down in Six Sigma), what I’m saying is that too often it’s done wrong by people who don’t know their stats from their ass.

[Poto]

It’s a usefool tool if you can measure something related to a process. Any process. But the softwawre, Minitab, is expensive for an individual to buy. If you have a job and do research of any sort, ask your boss to buy it. http://www.minitab.com/en-us/lp/free-lean-six-sigma-tools.aspx?WT.srch=1&WT.mc_id=SE4014&gclid=Cj0KEQiA0-GxBRDWsePx0pPtp4sBEiQACuTLNnZ7-1TuRHMwbOjlH-2oJ-tW3-LaqHURM9PCO0panmEaAn2l8P8HAQ
there are week long training classes available.

Many employers tie it to some $$ goal you have to make in cost savings. Not the best idea from my point of view. You can optimize a process without making all those $ goals dumped on you. “Hey boss, I saved us some money.” “Good, carry on.”

[Napier]

A mutual friend described to me how Bill Smith, the initial promoter of this at Motorola, got drunk one night and cried on his shoulder about how he never meant it be taken this seriously - that would be just stupid. But so many people were making a fuss about it, he didn’t know how to calm them all down.

This is hearsay, but it’s a friend I trust. For what it’s worth.

[Stranger_On_A_Train]

@Nava

Did you read what I said?

I appear to have read more of what you wrote than you did of my response.

Stranger

[Voyager]

To use it in manufacturing you have to produce in high enough volumes to be able to measure at that level. In non-manufacturing environments, which is where probably most of the training was done back when it was popular, it became a problem analysis tool. When you ask how they get quantitative data and they say they do a survey you know the concept is being misused.
I’m actually quite surprised that any non-manufacturing company even thinks about Six Sigma anymore. I checked the date when I opened this thread since I was sure it must be a zombie.

If you remember back when this stuff was hot, getting a Baldridge award was a kiss of death like being on the cover of Sports Illustrated. And the Quality Trainers in Bell Labs (on PQMI, not Six Sigma) came from a division that designed a computer - which was a piece of crap.

[Mangetout]

@Mangetout

In my experience, effective use of any of these popular methodologies (Six Sigma, PRINCE2, ITIL etc) typically involves cherry-picking the bits that work and assembling them into a procedure that properly fits the business context.

That being the case, many employers are looking for experience of using the methods, rather than qualifications.

@Stranger_On_A_Train

Although I agree with your first statement, the sad fact is that many employers are looking for those specific qualifications as some kind of proof of expertise in the methodologies.

When used appropriately in the correct context, Six Sigma methodologies can be useful in using quantitative methods to put experience and data into a framework where root problems and corrective actions become evident. The classic case is Motorola putting this methodology in used in its semiconductor microprocessor manufacturing to identify the key areas to reduce the incidence of defects to less than 3.4 per million units. Because of the high volume of manufacture for this type of product, statistical analysis was used to identify specific processes and procedures which were resulting in unnecessarily or unexpectedly high numbers of defects which could be addressed easily. At lower volumes, or in the design phase prior to manufacturing, other more qualitative methods are used to assess the sensitivity and likelihood of a process or design feature to result in defects so the problem can be fixed before finish tooling is made or a production line set up. ‘Lean’ Six Sigma methods, designed to be used by non-statisticians in more informal contexts, were created to give the system more flexibility.

Unfortunately, since Six Sigma was so successful for Motorola and a handful of other companies in a few areas, big corporations like General Electric, Hewlett Packard, Northrop Grumman, and Johnson & Johnson, just to name a few, decided to implement it as a cure all for all quality and process control problems whether it made sense or not, and ‘train’ the entire workforce in an abridged version that was simple enough for managers and secretaries to understand. And of course, everyone has to work on a “Six Sigma project” created specifically to meet the certification. The result: a jumble of nonsense make-work that rarely does anything to solve a real problem and often creates more useless documentation or reporting. Even the US government has gotten in on the act, and the results that I’ve seen have been spectacular (for ironic values of ‘spectacular’), suggesting that they should abandon long-used inventory and configuration management systems in favor of unworkable crap that was an improvement only in the sense that passing out after getting drunk is a relief, if but temporary.

For the o.p.: Six Sigma can be a useful methodology when applied by people who understand the methods and can gauge the applicability to a specific problem pertaining to process control or something similar. It is not a catholicon for all business problems, even legitimate ones such as poor market research or misgauging production quotas, and does nothing in the case that your chief director on a major program is a blithering idiot or your CEO is committing a bit of ‘light treason’. (I always felt that Arrested Development could have made a good running gag out of a Six Sigma like improvement process, which of course would be headed by G.O.B.) The reality is that Six Sigma methodologies only work for process improvement if the people using it are already competent and are just using it in a conscious, directed way to identify root cause and correction opportunities. Using the process blindly will achieve the same kind of results as driving down the highway at eighty miles and hour in blizzard conditions. The online course may introduce you to the basic terminology but will not equip you with any of the skills you’ll need to learn how to appropriately use it. Your prospective employer should be more interested in your applicable skills, with Six Sigma certification as a “nice to have” (at best) but something they can provide as training if it is really necessary. Otherwise, if they just have a box to tick in order to assure that X percentage of the workforce is Six Sigma certified, well, you know what kind of den you’re walking into.

Don’t get me started on project management certification from PMI. Jesus, there is an entire industry that exists to convince complete idiots that success is just following some flow diagram and brainlessly regurgitating acronyms that should remain meaningless to anyone who enjoys sanity.

Stranger

Agreed - I would say there are two different approaches then:
Work for a company that acknowledges they use a blend of cherry-picked techniques (they shouldn’t be looking for anything more than a good working knowledge of the principles)
Attain the qualification and work for a company that can demonstrate measurable success in formally implementing the methods.
(And treat with great caution anything that falls in the middle)

‘The middle’ happens to describe the place I currently work - we supposedly use Six Sigma in our production/ops, and PRINCE2 in our admin, but in reality, it’s like nobody here even understands the definition of the word ‘plan’. We just heap up all of the possible tasks in a big messy, unsorted pile and smash ourselves against them until we are broken.

[msmith537]

@Mangetout

‘The middle’ happens to describe the place I currently work - we supposedly use Six Sigma in our production/ops, and PRINCE2 in our admin, but in reality, it’s like nobody here even understands the definition of the word ‘plan’. We just heap up all of the possible tasks in a big messy, unsorted pile and smash ourselves against them until we are broken.

Six Sigma as well as Kaizen, Taguchi methods, QFD, and Zero Defect are quality control frameworks.

PRINCE2 along with ITIL, PMI PMBOK, Agile, Scrum, Lean, and Extreme are project management frameworks.

But speaking as a certified project manager and professional management consultant, I am confident that all these frameworks are bullshit. They all represent idealized approaches to managing projects or business processes that don’t necessarily work in the real world. Why don’t they work?

Companies are typically unwilling or unable to invest the time and money required to implement a project properly using the methodology. i.e. scrimping on planning or QA.

They lack the systems or infrastructure to properly track these projects or processes.

Lack of training or knowledge.

People simply don’t work like robots where you can accurately predict how long it will take to perform tasks. Particularly if those tasks have some element of uncertainty or variation like designing or writing code.

Corporate politics tend to get in the way of actually getting anything done.

When pressed for time or budget, companies will almost always abandon formal processes, thinking that planning and documentation just “waste time” and just plow ahead.

[Nava]

@Stranger_On_A_Train

I appear to have read more of what you wrote than you did of my response.

Stranger

Your response involved using multiple statistical functions. My gripe is about the obsession with the One and Only Function.

[Lightray]

@Busy_Scissors

DoE in chemistry went through a phase of respectability - I think it probably did conform to your assessment of It is an essential truth that if you understand something well enough to get good DoE results, you don’t need DoE, but it didn’t seem like complete bollox. I never did it myself, but had a few PhD students who used it on industrial placements - the results were pretty average, tbh, but it was more a training exercise to do something different. We’re interested in discovering new chemistry, not a deep refinement of a process that more or less works, so it didn’t seem like the right tool for that.

What would be seriously good is if DoE delivered in parameterising the actual ingredients of your reaction, like the ligand or catalyst, using principle component analysis [as opposed to continuous variables like temperature or concentration]. You certainly can do this, according to some DoE proponents I’ve seen speak, but in reality no one does. Whether it’s because it’s too time-consuming to set this up accurately, so it’s actually quickler to run 100 reactions and see what happens, or because DoE is just a giant cauldron of shite, I’m not sure.

I work at a chemical plant, in process chemistry, and I use DOE all the time. It is a useful tool for identifying a local optimum point of the process. I’ve done PCE, but it’s an utter chore, and is not really giving you the same information, anyway.

You absolutely have to understand the process you’re working with. We have a plant statistician (PhD) who is utterly useless, because she doesn’t know a damn thing about chemistry. If she’s going to do something on one of my processes, I have to tell her where to find the data, have to tell her which variables to include or discard, have to tell her what parameters to study, and have to interpret any results she gets. All she does is format the data and drop it into a statistical package – which is by far the easiest part of the whole deal.

Someone saying that they know statistics or Six Sigma is not impressive to me; it’s what they’ve done with either and how knowledgeable they are in talking about it that I’d be interested in. However, if we’re looking for someone with quality or statistical analysis background, having Six Sigma [whatever]belt on their resume will at least get it looked at.

[Stranger_On_A_Train]

@msmith537

But speaking as a certified project manager and professional management consultant, I am confident that all these frameworks are bullshit. They all represent idealized approaches to managing projects or business processes that don’t necessarily work in the real world. Why don’t they work?

Companies are typically unwilling or unable to invest the time and money required to implement a project properly using the methodology. i.e. scrimping on planning or QA.

They lack the systems or infrastructure to properly track these projects or processes.

Lack of training or knowledge.

And that, in a nutshell, is the rub. There is nothing particularly novel about Six Sigma or any of these other frameworks; they take pre-existing statistical process control, design for quality/manufacturing/reliability, and other methods and bundle them up with a bunch of color diagrams and cute acronyms. The methods themselves are workable, provided that the people using them understand both the intent behind the methodology and have a thorough knowledge of the process or product being improved. The problem is that when they are implemented in a “corporate-wide initiative” without any specific goal other than to be “100% Six Sigma Certified!” they end up with a bunch of ill-trained people trying to misapply methods to problems that are easily solvable by a simple application of common sense. Six Sigma, SPC, and other statistical methods for measuring and improving processes have a fairly narrow range of applicability to problems with sufficient data to be statistically meaningful, and one has to understand the nature of the problem and the variables affecting it in order to correctly apply the methodology, which is unfortunately rare.

The same is true with systems engineering; in theory and expert practice, it is a great framework for generating and validating product and process requirements, flowing them down to a detail design level, and then verifying them by appropriate testing and measurement. When system engineering becomes its own separate department that does some random shit at the beginning of a product design process and throws a jangled mess of requirements over the wall, it is worse than useless and often dooms a program at its inception, because the requirements are internally inconsistent, physically nonsensical, or impossible to functionally verify, a point many major manufacturing companies (especially promoting their “systems of systems integration capability”) have failed to adequately grasp.

@Nava

Your response involved using multiple statistical functions. My gripe is about the obsession with the One and Only Function.

Which, again, is not correct. Even in my “green belt” class, the fact that distributions beyond the Gaussian exist and can be applied was acknowledged, and Six Sigma and other SPC methods can use non-Gaussian statistical distributions. That this is not presented in detail in the normal training is due to the required scope to teach these methods in depth and the fact that for most non-time series random failure type of problems, a Gaussian distribution is a good first approximation of variability. When you move to multiple parameter distributions the analysis and verification becomes far more complex, which is well beyond what can be presented in the typically abbreviated training, and frankly beyond the scope of most of the people teaching the course.

Stranger

[LSLGuy]

ETA: **msmith537 **was the last post when I started. I’m not quibbling with Stranger’s just above mine. Though I will be curious to see if he has feedback more learned than I. …
I’ve been spared the pain of working with/under six sigma. In my office roles it’s been more like msmith537 says: “ignore all planning and simply flail away at the problems”. So I’m speaking from some ignorance here. But …

Several posters have said in effect “six sigma doesn’t apply to us because we don’t have millions of repetitive actions; only thousands.” While that might be strictly true as part of the definition of standard deviation, that says nothing that amounts to “statistical process control is useless unless carried out to six standard deviations.” Which is IMO what those posters are saying.

The practical objections to cult-like management fads are well-founded. I have worked through a few of those. The practical objections to “Everything is Gaussian; even things that by inspection are clearly not” are well-founded. The practical objections to “a clerk who struggles with a calculator can do this well if they simply follow the cookbook” are well-founded.

But saying that applying statistical methods is pointless in the absence of millions of data points is silly. IMO.