Don Reinertsen – Is It Time to Rethink Deming? [the broken link was removed] AGILEMinds
I feel very strongly about the value of Deming’s ideas. I am glad people challenge those ideas and try to push forward management thinking. Helping us manage organizations better (to get better results and allow people to better enjoy their jobs and lives) is why I value Deming’s ideas. To the extent we find better ideas I am very happy. I understand I will disagree with others on the best ways to manage, and believe healthy debate can be productive.
What Don Reinertsen discusses in the video, about special and common cause is not the best way to look at those ideas, in my opinion (though I would imagine it is the most common view). For data points that are common cause (within the control limits and not a special cause pattern) it is most effective to use common cause tools/thinking to improve. For indications of special cause (points outside the control limits or patterns in the data, such as continually increasing results that indicate a special cause) it is most effective to use special cause tools to improve.
This does not mean that a point outside the control limits is caused by a special cause (also know as assignable cause). It is just best to use special cause tools and thinking to address those data points (and the reason this is true is because it is most likely there is an assignable cause). The control limits do not define the nature of the point, they define the type of improvement strategy that should be used.
Don also says repeatedly that you don’t “respond to random variation” in Deming’s view. That is accurate. But then he implies this means you don’t address system performance, which is not. You work on improving systems (that are in control) by improving the system, not by responding to individual common cause data points (random variation) as if it were assignable cause variation.
The purpose of the control chart (that Shewhart developed) was to help you most effectively take action (knowing if special cause thinking, or system improvement, was the best improvement strategy). The control chart shows if the results are in control and tells you that the system is preforming consistently (and identifies a special cause so special cause tools can be used immediately, this is important because special cause improvement strategies are time sensitive). It tells you nothing about if the results are acceptable.
Continual improvement was also central to Deming’s management philosophy (based on the business value of the many improvement options available in every organization). For Deming this meant working on improving the system, if the results are in control, instead of trying to deal with finding a specific assignable cause for one data point and acting on that. If the issue is one of the system performance (no indication it is a special cause) the most effective strategy to get better results is to improve the system, rather than approach it as a special cause issue (examining individual data points, to find special items in that event to be improved). You can use special cause thinking, even where system improvement thinking would be better. It will work. It is just not very effective (improvement will be much slower) compared to focusing on system improvement.
I agree with Don that the United States mentality, not only in nuclear plants but everywhere, is to apply special cause thinking as the strategy for process improvement. This is one the areas Deming was trying to change. Deming, and I, think that setting your improvement strategy based on a common cause (system improvement) or assignable/special cause (learn what is special about that one instance) is the most effective way to achieve the best results. We believe in continual improvement. We believe that the effective way to improve, when a system is in statistical control, is by focusing on the whole systems (all the data) not assignable cause (special cause) thinking where you look at what is special about that bad (or good) individual result.
The economic consideration of whether the costs of improvements are worth the benefit is sensible (and I do not see Dr. Deming arguing against that). That is separate from the best method to improve. For Deming the best method to improve means using special cause thinking for assignable cause issues and common cause thinking for systems issues.
The idea of where to focus improvement efforts is not something Dr. Deming made as clear as he could have, in my opinion. So I see the argument of Deming not prioritizing where improvement should occur voiced occasionally. This is a weakness in Deming’s content, I believe, more than his philosophy (but I can understand it causing some confusion).
A huge reason for this miscommunication, I believe, is that Dr. Deming wanted everyone involved in improving. This is not a radical idea today (though it is still uncommon in most companies). It was radical when Deming was trying to get people to adopt this change in mindset. And for Dr. Deming the priority was that all these people are engaging their minds and working on improvement all the time. I believe his main focus was trying to get management away for the idea that our supervisor can only focus on a very limited number of improvements. Dr. Deming wanted everyone, everyday, to see how the systems they worked with could be improved. And within that context prioritizing the important business areas, best returns on investment, etc. is perfectly consistent with Deming’s ideas.
Dr. Deming wouldn’t want effort that could be better directed to a more important improvement spent on less important improvement. But the waste he was most interested in was in failing to have every mind in the organization constantly improving the systems they work in. This is a different focus than seeing a very limited improvement effort and therefore making sure only to work on the improvements that the CEO has decided are priorities.
The idea that Deming wanted to treat people like robots is extremely inaccurate. More than most any management expert Deming’s focused on the importance of treating people as people. Nevertheless it has been a fairly common reaction to creating systems that produce repeatable results, to claim that such a thing means treating people as robots. It just isn’t a sensible reaction at all, if you know much about Deming’s management system, in my opinion. The 2 pillars of the Toyota Production System (lean manufacturing) are directly from Dr. Deming’s ideas: respect for people and continuous improvement
He also makes a standard claim that Deming is primary (almost exclusively) focused on quality control. That isn’t accurate. That may be what MBA are told in business school, but it is not what Dr. Deming taught, ever. Not in 1950, 1970 or 1990. I have written about many of the other management concepts Deming discusses many times on this blog. One very simple example that Deming understood very well that quality control was not the only or even the main factor, Dr. Deming on innovation, from page 10 of New Economics:
And what is required according to Deming? Innovation. Deming did not believe, or propose, only the control chart. It was one tool to help improve performance. He provided a huge amount of material on much more comprehensive management strategies. It isn’t a huge deal when managers applying some of Deming’s ideas don’t know about most of his ideas. But giving a talk about “Is it time to Rethink Deming” and then stating that Deming’s ideas were limited to control and sampling and giving the impression this is an accurate representation of what Deming proposed I do have a problem with. I don’t want people to accept this impression and therefore avoid looking into Deming’s ideas because they believe they are so limited. I would definitely agree we need to move far beyond Deming’s ideas, as presented in this talk. But I think Deming moved far beyond those ideas decades ago.
The idea Deming didn’t believe in experimentation is also wrong. The Plan-Do-Study-Act improvement cycle is central to Deming’s management philosophy. It is focused entirely on learning via experimentation. It is a very effective way to experiment and improve in organizations. Deming referred to the PDSA cycle as the Shewhart cycle (again having learned of it from Shewhart – and included it as a central part of his management philosophy from the initial formulation until his death) and many have referred to it as the Deming improvement cycle. PDSA is fundamental to applying Deming’s ideas. To read more on PDSA read the Improvement Guide.
I have no problem adopting to kanban even if the system is not in control. I agree with the speaker on that point.
Related: Good Process Improvement Practices – Management Improvement – Bad Advice on Management – Performance without Appraisal – Find the Root Cause Instead of the Person to Blame – Stratification and Systemic Thinking – Blame the Road, Not the Person
The red bead experiment is a more a fable than a simulation. It is certainly a very big over-simplification of reality. The psychology of the game is important to the message we are meant to take away, and showed how Deming understood the importance of psychology in managing people. That so many people do see parallels to their workplace, I think is a sign that the fable works for them.
Shewhart set the control limits at 3 standard deviations based upon empirical evidence. The purpose was to determine at what level which improvement strategy (common cause improvement or special cause improvement) was most effective. It is true that given this genesis, it is possible for other systems (other organizations) to have other values (2.6 or 4 standard deviations, or whatever) that would be more effective. An organization could determining that a different value turned out to be more effective for them. But, I believe almost any organization that tried this is making a mistake (I believe 3 works very well and has been shown to in many environments, and so I need strong evidence to believe an organization can effectively pick a better value – but I do believe it is consistent to state that it can be done). I would say unless your organization has a large amount of world class knowledge and experience with control charts and improvement I believe it is a mistake to try. And I am not aware of any organization that has decided to empirically evaluate to find a value that provides them better results.
But I really think that worrying about this is an “academic” exercise as 3 standard deviations has proven to be very effective (even though it is based on empirical evidence rather than a mathematical principle). It also ironic, to me, in that this somewhat flips the normal “academic” argument. Where normally an issue is seen as “academic” when it is that abstract “book knowledge” (say like, math) instead of “real world” evidence – such as empirical evidence.
Don seems to make (or come close to making) a mistake many in the six sigma community make. They believe the sigma limit is somehow telling you how much you care about improvement. If you really really care you go to 6, if you just kind of care you go to 2. That isn’t what the sigma level tells you. It tells you if special cause or common cause thinking should be used. If you really really care then you spend lots of resources focusing on improving in that area. The special-cause/common-cause signal just tells you whether to focus on finding assignable causes of variation or working on systems improvements.
I do believe that there will be fewer control charts in a knowledge work environment than a manufacturing environment. This is due to the nature of knowledge work not being as easy to assign data to. Deming well understood that many useful metrics are not available to manager (and not even possible). He would not be surprised that you can’t have control charts for many important processes. Deming strongly favored understanding the principles and figuring out what the best way to apply those principles was in a specific system (company). Deming never advocated treating a knowledge work environment as identical to manufacturing.
I do think the principle that you apply special cause thinking best to assignable cause problems transcends the type or system (or workplace). And use system improvement thinking to address the vast majority of improvement efforts. Without the control chart to help falling into the trap we tend to fall into (treating system problems as assignable cause problems) I think we need to understand the principle even more than in manufacturing (where the tool, control charts, help us avoid the error). Some may argue that knowledge work has more or less assignable cause situations. Don seemed to argue there are at least more cases where assignable causes are due to an individual (for example heroic software engineers) than in a manufacturing environment. I would agree that is probably a true guess at a difference in those two broad types of systems made up of people.
I still feel the trap of searching for assignable causes to specific data points instead of applying system improvement thinking is a big problem in knowledge work organizations. And understanding the ideas of common cause and special cause improvement strategies are just as relevant for knowledge work. I would say that it is harder because there will be more instances where the control chart tool can’t be relied on to help keep us on track. We have to understand the concepts and apply them as we run into new improvement efforts.
Going back to a different topic, I believe the reason Dr. Deming didn’t focus on cost benefit analysis is that was pretty well understood. He didn’t need to convince people to focus on those things that would have the biggest payback. His main focus, in this area, was to caution against short term thinking (focusing on, for example, the initial price tag versus the long term cost and benefits). Dr. Deming certainly believed you prioritized limited resources based on the long term value of benefits you could expect.
As with so many things, this again touches on more of his ideas (he expressed his management philosophy as an interconnected system composed of 4 interrelated components: systems thinking, an understanding of variation, psychology and theory of knowledge). He understood that many important factors were unknown and unknowable. So while he would support the idea of making judgements about what is the best place to invest in he would be skeptical of the belief that the projections you create are “true.” The projections (cost benefit analysis) are a way to guess and help make decisions. Hopefully you are practicing PDSA and learning about the strengths and weaknesses in your projections – so they improve over time.
@agilemanager David J Anderson
Your post included a very interesting video and commentary about Deming’s thinking. I agree with most of your comments about the video. Following are some additional observations.
In terms of product development, Deming suggested that people often don’t know what they want. Management cannot simply rely on market surveys—and the ensuing statistical analysis—to determine what products and services to offer. In this respect, Deming shared a philosophy of product development similar to that of Steven Job’s.
Although I agree with Don that the wrongful use of statistics in new product development is problematic, I believe that Deming never advocated the use of particular statistical method—in product development—in the manner that Don suggested. Deming advocated the application of statistics pertinent to on-going processes such as the manufacture of springs—which is quite different from new product development. See http://www.businessweek.com/magazine/content/07_24/b4038406.htm
Don made one statement I do take issue with: “Deming’s methods tend to be focused on maintaining the status quo.” Quite the contrary, Deming once stated that “the entire management system had to be thrown out.” He once passionately asked the rhetorical question: “Do you think that the tellers were the ones responsible for the failure of Continental bank?” It would be interesting to hear Deming’s take on the “root causes” of our current financial crisis.
Thanks for a great blog post.
by Tim Higgins
Understanding variation as Deming offered in a system of profound knowledge is incomplete without appreciation for a system, psychology and theory of knowledge.
Seems as if Deming saw a huge part of understanding variation as the ability to distinguish between common and special causes. And as you indicated the reason to distinguish is to take ACTION appropriate to the type of variation. Change (design / re-design) the system or find cause. He offerred that finding a cause is not an improvement. Finding the cause of the fire does not improve the building. Returning the building to its former (pre-fire) state does not improve the building. Finding and eliminating negative special causes returns things to a predictable and thus manageable state. Having a predictable state is beneficial (and likely necessary for improvement) but in and of itself is not an improvement.
The distinction between the types of causes (and thus the appropriate action) needs be made regardless of whether or not we have control charts and whether we are in a manufacturing or other kind of business. Agreed, having only common cause variation present does not mean do nothing. Agreed, that prioritization of investments in improvement is needed – organizations and individuals have finite resources and cannot afford to improve everything everywhere all the time – there is a new economics which includes cooperation and I believe and understanding that the benefit we seek is improvement in the system (as opposed to improvement in a part taken separately).
Your reminder that the control limits are empirically based might help some buy into Don Wheeler’s (with Chambers’) observation (captured here without looking it up in Understanding Statistical Process Control: SPC is ultimately a way of thinking with the charts acting as a catalyst for the thinking process. In the same book he offers that without an understanding of Production Viewed as a System and without the ability to employ PDSA “the techniques [SPC] are of little value.” Shewhart’s empirical base then confirms Wheeler’s SPC is not about statistics and it’s not about process control.
Industry, government and educational organizations as well as individuals need to recognize the appropriate actions to take based upon the kind of variation present. It seems fundamental. AND we need to be able to decide on appropriate actions in response to events whether or not we have control charts.
I find Deming’s observation that the most important application of the distinction between variation types is in the management of people an interesting way to bring at least 3 of the SoPK elements together – pyschology, understanding variation, appreciation for a system.
Thanks for noting that understanding and applying Deming is a useful prerequisite for moving beyond him. Attempting to move beyond something we have misunderstood will generate useful insight only accidentally. If we understand, then seek to move beyond, we may purposefully generate continued new insight.
[Wheeler and Chambers have a section in their book that shows six distributions, including a triangular and an exponential, and demonstrate how well 3 sigma limits for action apply to each one of them – still not a statistical basis for the action limits, “merely” a practical one. Deming I sense sees them as the economically based limits for action another part of the new economics.]
Thanks for the article.
What was your thoughts on the 94% argument that Don also made?
There was some good discussion on this subject here http://www.coregroup.no/2011/10/when-systemsthinkers-collide/ (comments at the end).
Great, great post, John.
I wish I had a good explanation as to why those focusing on Lean, Deming, or the Quality movement in general seem to only be interested in understanding the tools and how to apply them. Very rarely do they seek to understand the purpose of those tools.
Yes, controls are there to identify problems that need to be addressed – but this stops short of blowing the “Why?” chain all the way back to the root…and the root is NOT about better quality of product, reduced cost, increased market share or anything else.
What Deming discovered is that all these tools for improvement exist to achieve 1 thing: Reducing the mental, emotional, behavioral burden that we have come to believe work MUST be. (Lean calls this overburdening of people “Muri”)
Deming stated that the worker should be able to approach his work with great joy. That joy is enabled and facilitated by the elimination of wasteful, useless, unproductive and/or panic-stricken work environments. If you can uncover the reasons for outliers (uneveness of processes, or what Lean calls “Mura”) you can reduce the Muri. The less Muri, the less worry and strain to deal with the urgent problems of the day, the less urgency, the more time for experimentation, more experimentation yields more learning, and learning results in innovation.
All too often, I think something gets forgotten – that the tools of the quality movement and the process improvements they generate are not the end. They are the means to an end – greater enjoyment of our existence, which includes both working and living as simultaneous experiences.
David,
I agree with the idea that the 94% figure is questionable to begin with, and may be even more questionable in knowledge work. I do still think that most organization should shift to looking for solutions based on the entire system and not looking at individual people and special causes (even not just people).
So I agree that the 94% figure is not of much value. Saying it is 94% or 80% really make no difference. Even saying it is 50% doesn’t change anything. It doesn’t actually have any impact on what action you are suppose to take. For an improvement effort a special cause or common cause strategy will be used (my believe is certainly far more often (above 75% of the time a common cause strategy will be most useful in knowledge work organization. And in general special cause issues are of lower scope (therefore just counting the number of occurrence probably underestimates the importance of system issues (versus special causes). But, in any event I certainly an see reason to question 94%.
I believe Deming focused on in largely as a counter to the continual barrage of explanations that it is “Joe” and “Hank” that are the problem. Dr. Deming saw over and over you replace “Joe” and “Hank” and nothing changes. The system continues to produce the same results.
Now you have to be careful – often we can be mislead by “regression to the mean” into thinking a system that stays the same got better after some change because we take action when these odd events happen and the system would have “regressed to the mean” without a change or a random change too.
I do believe certain people (leaders, technical masters, insightful engineers) have an inordinate ability to take heroic action to overcome systems (which is another argument for reducing the 94% figure). They can both work miracles and be extremely difficult to replace and maintain success. The best possible result, in my opinion, is to capture the remarkable benefits of these exceptional people and create systems that are robust. This means creating processes to track and manage success, use check sheets, monitor in-process measure, use kanban and agile methods (to have system that are much more robust)…
While I accept the arguments that knowledge work is different (and thus the 94% figure shouldn’t be considered) I have found I don’t share a frequent desire of those making that case. Which largely amounts to “let heroic knowledge workers” be free to creatively build success without all your controls making them into robots that can’t find inspiration for their gifts to flourish.
I believe designing process for knowledge workers doesn’t kill innovation it allows it to flourish. Those processes are not going to be the same as manufacturing, though.
I’m not sure I really am doing a good job of explaining my thoughts on this question. Here are some previous posts on innovation and process improvement.
Maybe I am missing the point of your question. Most of the time when I see knowledge worker types arguing against the large system focus (or lack of focus on difference among individuals) the issue seems to really be that they believe knowledge work needs to be free and you can’t use ideas of process improvement (and other quality tools). They argue that those things are just for the factories or white color factory type work (call center…). I don’t believe that is true. I believe the same concepts are applicable to knowledge work, but there are differences that must be considered (likely fewer control charts, likely more variation, likely more variation between people’s performance…).
I did read the post and comments you linked to (twice). I am still not sure I really grasp the full question though. At the same time I may just thinking there is more to the question than there is.
Knowledge workers work in different environments:
Before the concept is developed,
before commit to produce,
before initial production,
before initial stable production,
before delivery,
after delivery,
common service bureaucracy,
…
What is measurable and what is “mistake proofing” varies wildly in different environments.
Pingback: Why Use Lean Manufacturing Concepts if So Many Fail To Do So Effectively » Curious Cat Management Blog
This video is so filled with mistakes, misinterpretations and confusion that I would declare it virtually useless insofar as it purports to be about Deming.
He says Deming was focused on Inferential Statistics. Not true. In fact Shewhart specifically distinguishes inductive methods from control charts in “Statistical Method from the Viewpoint of Quality Control”.
He makes no mention of the importance of random variation within the limits as one of Shewhart’s criteria of control. In fact, he confuses it.
He says the control limits are calculated by adding and subtracting the Std. Dev. to and from the mean. He makes no mention of rational sub-grouping and efforts to make process limits to be limits of common cause variation. Rational subgrouping is central to the use of the statistical control chart. I suspect he doesn’t know anything about it.
He (incorrectly) states that the probability of being out of control is .27%. No. The idea of probability doesn’t apply. Nor does the idea of inferential statistics. There is no population and even if there were, those probabilities are *conditional* probabilities. So not only does he not understand Deming, he mis-states classical statistical theory.
He says that the idea of what is or is not a common or special cause is “purely mathematical” and has nothing to do with the underlying cause system (whatever that means). No. The idea of what differentiates common and special cause is economical. Shewharts book is titled “Economic Control of Quality of Manufactured Product”.
He contradicts himself saying at one point that the amount of variation has nothing to do with whether or not a process is in control in the ‘Deming World’. He later says, “It’s just a belief that high variation is going to be caused by strange (??) events”. This causes me to wonder if he has even a basic concept of statistical control.
If his talk was sold to his (apparently) Belgian audience based on his authority on Deming, he should refund their money. Deming called people who taught these methods without knowledge “Hacks”. He said that students, particularly beginners, deserve a “Master”.
Pingback: 94% Belongs to the System » Curious Cat Management Improvement Blog
I had a long debate with Don a few years ago. He’s a brilliant person; hence debating with him is really hard. In the end, it was clear to me that he really didn’t understand variation. His original video has been removed but in summary, he implies that Deming’s Common/Special Cause variation is simply looking at anomaly events (special) or normal events (common). Sort of like what’s the big deal. That view of Deming/Shewart’s view of variation couldn’t be further from the truth. I wrote my own version on this topic a few years ago here… ( https://medium.com/@johnwillis/misunderstanding-variation-is-the-root-of-all-evil-b204cd7f6474 ). Someday I’ll corner Don and finish our debate.