LEARN HAVENHEALTH HAVEN
WORK HAVENEARTH HAVEN
DHARMA HAVENSCIENCE HAVEN
Part of the variation you detect will result from actual differences among the instances you are comparing, and part it comes from variation in your measurement process. Your process of observation or measurement is a natural system that is just as subject to intrinsic variation as any other system. Oddly, human observers add another dimension of complexity into the situation, by sometimes systematically and automatically suppressing variation, detecting no difference when differences are actually present.
Variation is often troublesome, for those who don't expect it; but it also can be a blessing. How to work with variation successfully depends on ones purposes and on the types of variation that are present.
One interesting example comes from history of the manufacturing of firearms. When skilled craftsmen built an entire finished piece, there was no need to be concerned about whether a particular part of one rifle was the same as the corresponding part of another: What mattered was that the various parts of each particular rifle fit each other, so that they would work together successfuly.
When Eli Whitney developed the practice of building rifles with interchangeable parts, a new standard was required: All instances of a certain part had to be nearly identical. In modern assembly-line manufacturing, interchangeable parts are required, and variation from one instance of a part to another causes all sorts of problems.
Modern methods of quality improvement systematically reduce that variation, increasing quality and reducing costs.
A similar change in methods has occurred in education -- with much less happy results. When students of different ages and levels of mastery of various subjects were schooled together in a small local facility, differences among the students in the class -- for example differences in the rate of learning some particular subject -- were no real problem. There were always students working at quite different levels in any particular subject.
When modern educational systems were developed, with much larger school populations and students assigned to classes matched by age, with everyone in the class studying the same level of the same topic at the same time -- when, in other words, the schooling system came to resemble an industrial assembly line -- differences among students became problems for the system. In this case, efforts to reduce variation among students succeed, but with tragic consequences. We "fail" students who can't keep up, for whatever reason, and we keep on failing them again and again until they "drop out." That makes them no longer the problem of the school system -- they become the problem of the courts and the probation officers and the welfare system.
Contrasting the amazing successes of modern quality management systems in industry with the shocking failures of modern school systems reveals a difference that is one of the main causes of the current educational disaster. Successful industrial quality improvement systems treat all the human beings involved in the process as individuals worthy of respect, and marshal everyone's work toward reducing variability in their physical products. Modern hideously unsuccessful schooling systems treat human students as if they were defective parts when they fail to keep up with the learning rate expected for the class; and when teachers refuse to give up on slower students, those teachers are treated as if they were malfunctioning machinery.
Some schools are breaking out of this trap, by applying the quality improvement systems developed for industry. Rather than being treated as (potentially defective) products of the system, the students are recognized as the workers who actually accomplish the learning that is the whole point of the schooling system. The community pays the school system to help students learn. Each student has a unique task, which only they can accomplish: to learn whatever they can about whatever they want to learn about. Teachers become managers, who are there to do whatever they can to help the workers do their jobs.
Here we will discuss four very different examples.
Walter Shewhart and W. Edwards Deming taught that reacting to random ups and downs as if they mean something always degrades the process. The appropriate procedure is to let the process run -- and study it. By doing so, one can learn how to improve processes so that quality is increased and costs are decreased.
Betty Edwards found that drawing realistically requires
shifting to a different way of seeing. This slightly altered state of consciousness
which enables you to see accurately also enhances creativity. The left
hemisphere's analytic mode of processing sees stereotyped shapes -- but
the right hemisphere is interested in the actual shapes.
Top of Page
Top of Page
All processes and all results of processes are intrinsically variable. Measure any characteristic of anything, and the result will vary from instance to instance.
Variation confounds simplistic strategies
Variation constrains quality, with materials and products randomly deviating from their designers' and users' intentions in many ways. Some deviations make no difference; some cause problems. Often, the larger the deviation, the worse the problem.
Sometimes variation can be reduced, by understanding what is causing it in a particular situation -- vibration can be damped or prevented, tools can be sharpened more often, worn parts can be adjusted or replaced, and so on. In other cases, such as variation in incoming raw materials, in customers' requirements or in students' interests, variation can be accomodated or even welcomed. But in any case, in order to deal appropriately with variation, one must first understand it.
Most of us tend to use the terms "accurate" and "precise" as if they had the same meaning, something like "close to the goal or target." However, people in technical fields often use these words to distinguish two very diferent aspects of quality.
Imagine a student of a martial art, like archery. There is a target set up at a distance, with some indication of the point of aim. An arrow is released, and flies to strike the target. We can measure the distance from the arrow to the point of aim, and use it as a crude estimate of the quality of the shot, or the skill of the archer (or the archer's teacher).
If only one shot is made, there is no basis for distinguishing between precision and accuracy. Suppose that there were several shots -- say 25 -- and the paper cover from the target now has a cluster of holes in it. Now we can define precision and accuracy: Precision is a measure of the dispersion of the group of results -- low variability equals high precision. Accuracy, then, is the deviation of the center of the group from the point of aim.
Marksmanship trainers teach students to distinguish these two aspects of quality. If you aim a rifle at a target and clamp it in position, you can, by firing it several times in the same position, measure the average deviation of the shots from the target, and adjust the sights so that the next cluster of shots is centered on the target. This is called "sighting in" the rifle. Then the marksman herself goes through a similar process, trying to use the sights in a consistent way, to learning produce as small a cluster as possible. Once she understands how her clusters deviate, on average, from the target, she can adjust her aim to bring the center of her shot pattern in line with the center of the target.
These improvements in accuracy are only possible by holding the same aim for a group of shots, and using the center of the resulting cluster to estimate the deviation of the entire process of aiming, firing and the rifle and ammunitions' functioning from the target. Because of the intrinsic variability of the process, trying to adjust the aim after each shot just makes things worse.
FUNNEL ZIP Funnel experiment simulation software
Deming teaches that managers tend to meddle with their processes -- that they have, in fact, been trained to do so, reacting to each rise in performance as if things were improving and reacting to each dip in performance as if things were beginning to fall apart. Yet, typically, nothing has changed at all. What the manager sees is usually random variation that is intrinsic to the process--the process as it is currently designed, set up, operated and managed.
Two different sources of variation must be distinguished: Common causes of variation are causes intrinsic to the process. Special causes of variation are relatively rare events that may require immediate action: If a machine has started producing consistently defective parts, for example, it should be taken off the line and repaired. But eighty to ninety percent of the variation in the product is due to common causes--causes due to the nature of the process itself. The only way to reduce that kind of variation is to change the process.
When managers react to variation due to common causes as if it were due to a special cause, two very unfortunate outcomes are inevitable. First, such "tampering," as Deming calls it, degrades the performance of the process; the product will be more variable than if the process had been left alone. Second, tampering makes it impossible to understand the common causes of variation, leaving no information about how the process could improve.
Dr. Deming shows his students (including CEOs) how tampering (in the technical sense of reacting to random ups and downs as if they meant something) always degrades the process. A manager who understands that tampering never helps will let the process run -- and study it.
The number of examples of different kinds of tampering that have become the standard operating procedures in nearly all of our organizations is truly staggering. The most destructive tampering is tampering with people: Blaming individuals for problems that are intrinsic to the system they work in, and rewarding them for successes that also are intrinsic to that system, is tampering.
Tampering with people causes both of the problems that any tampering causes--degradation of the process and obscuring the common causes of variation. Moreover, tampering with people undermines intrinsic motivation. Without intrinsic motivation there is no pride in workmanship and no joy in work; and without pride in workmanship and joy, there is no quality.
The manager who understands the destructive effects of tampering has no choice but to stop; however, in order to be able to stop tampering, that manager requires the support of higher-level management. (Until now, tampering has been a major part of the manager's job description.) Once management can agree to stop tampering, they can allow the system to vary naturally. By studying that variation, they can begin to understand what causes it; and, with the new knowledge, they can improve the system, raise quality, reduce costs and stay in business.
To study the processes that make up their system, management must involve those who actually use those processes--the people who actually do the work. The people who build the product or provide the service are the only people who really understand the processes that management has assigned them. The role of management changes from giving orders and giving out punishments and rewards, to leading and supporting the workers in improving quality.
If the workers are to succeed in studying the processes they use and in creating ways of improving them, they must have several kinds of support. For example, they must be given the understanding that change is possible and that management is committed to supporting them in studying and improving the system; they must receive training in their new job--training in process improvement and in other skills they will need, as well as in Deming's philosophy of management; and their suggestions must be put into effect and the results studied. (An inappropriate suggestion must be discussed openly--also, further training may be indicated.)
Top of Page
"Variation and Accuracy"
Quality Gamebox CD (For Windows): Simulations and experiments demonstrat Deming’s red bead experiment and funnel experiment; plus McConnell’s dice experiment illustrates the benefits of reducing variation; central limit theorem shows the effect of increasing sample size, and the quincunx explains the normal curve. $39.00.
Dharma Haven Home Page
Support Dharma Haven
While you Save on Books and Music
Your Comments and Suggestions
Revised on April 14, 1999
Copyright © 1999 Dharma Haven
Top of Page