Some Principles of Statistical Thinking
The purpose of collecting data is to make predictions
Since all decisions are predictions, data is needed for decision making
Measurements and measurement methods must be defined and agreed upon
The accuracy, repeatability and precision of measurements must be estimated
Graphs provide more information than tables
Data portrayed in time sequence provide context
All data has noise associated with it
Improvement strategies depend on the nature of variation encountered
The Salary Theorem states that “Engineers and Scientists can never earn as much as business executives and sales people.”
This theorem can now be supported by a mathematical argument based on the following two bases:
Postulate 1: Knowledge is Power
Postulate 2: Time is money
We know that Power = Work/Time. Since Knowledge is Power and Time is Money, and Knowledge is Power (Work/Time) by substituting we have:
Knowledge = Work/Time (Time = Money)
Solving for Money we can see: Money = Work/Knowledge.
Thus, as Knowledge approaches zero, money approaches infinity no matter how much work is done.
Conclusion: The less you know, the more money you make.
The following is part of a response I sent to Mike Stoecklin who had posted in the Deming Institute’s blog.
The first question is what does he mean by ‘Understanding Variation’?
Understanding variation can be very broad. There is the sense of understanding variation as a kind of ontology – I sometimes call it The Ontology of Flux (coined I think by Mark Wilcox) – it is a way to view the world (e.g. Heraclitus oft quoted (perhaps mis-quoted) expression that “…one cannot step in the stream twice..” is an example of this broad thinking; “All is Flux” – also attributed to Heraclitus. It is a fundamental principle of eastern thinking that everything is in flux all the time. Alan Watts entitled an excellent book, “The Wisdom of Insecurity” in which he talks a lot about ‘impermanence’ (variation). Understanding variation and its ubiquity changes the way one views the world. I used to tell students sometimes that “It’s like putting on a new pair of glasses. The only difference being that once you put the glasses of variation on, you cannot take them off again.”
Understanding variation can be more narrow in the sense that Dr. Deming and Dr. Shewhart have discussed in many of their works. This includes, but is not limited to, the idea of Common (Un-assignable) Causes and Special (Assignable) Causes and the implications of this distinction for management practice. If one essence of the ‘Toyota Way’ is continuous improvement, how does one improve? The strategy for improvement depends on the type of variation detected in the process and using the wrong strategy may well make things worse. The implications in the management of people of understanding variation in this way are indeed profound including the quote from Dr. Deming that is cited in his post. I think the Deming’s extension of Shewhart’s understanding of variation to management of people is one of his principle gifts to knowledge.
Further, the elements of Deming’s System of Profound Knowledge interact. For example, understanding variation shows the silliness of trying to meaningfully measure a person’s performance. The interaction of the system with the actions of the individual is another example of these interactions. Heero’s example is good. I saw this type of thinking first from Ed Baker in Dr. Deming’s 4 day. I would say the example of Heero’s equation might be more complete if it showed that over all performance (P) is the combination of Individual contribution (I), System contributtion S (factors over which the individual has no control – built in) and the interaction of the individual with the system in which he or she finds him/her self. Thus P = I + S + (S x I) which is how I’ve always shown this equation.
Several speeches are mentioned in the various biographical sketches of
Deming and his impact on Japanese management. In particular a talk
he gave at the Mt. Hakone Conference center in 1950 seems to have
been important. Attendees included the top industrial managers in
Japan at that time. In fact, it was subsequently estimated by Ishikawa
that upwards of 75% of the heads of companies representing the
corporate wealth of Japan were in attendance.
I wrote to the organization that sponsored the event (JUSE) seeking
information, and they were kind enough to respond. They sent me a
transcript of Deming’s talk recorded in Japanese. I had it translated
into English in this country and what follows is the translation of that
transcription. – jsd – 1995
Dr. W. E. Deming
Presidential Adviser on Sampling Methods for the US Treasury
The opportunity to speak with all of you is my greatest honor. I will not give
a sermon on statistical techniques. I leave that to the statisticians. Henceforth
I shall speak of the truly important problems of manufacturing and sales, the
statistical techniques which are helpful in the solution of these problems, and
how all of you can use these techniques. Afterwards, I will answer your
The problems and methods I will now discuss are extremely important to
both Japanese and American manufacturers, as well as those of England,
New Zealand and other countries around the world.
As all modern-day manufacturers are striving to make their business
prosperous in the long term, the following issues are necessary:
1. Better design of products to improve service
2. Higher level of uniform product quality
3. Improvement of product testing in the workplace and in research
4. Greater sales through side [global] markets
First of all, I am not a miracle worker. I am not an economist, or an expert in
business sales or research. Neither am I a manufacturer. However, I firmly
believe that statistical techniques and associated practices are useful to all of
you. As a statistician, while being in close contact with the problems that all
of you encounter, I have had the honor of doing research about those
problems with statisticians, salesmen and other experts from diverse groups.
Such groups include manufacturers, economists and technicians, as well as
university professors and the world’s largest, most successful trade
We all have our different perspectives, but I believe that we can solve the
various problems facing modern Japanese manufacturers through
Of course, you all have heard about statistical product quality administration
before. However, if you do not understand it, or if you do not support your
sales researchers, statistical product quality administration is not useful and
your business will not be able to expand.
I must point out that I do not mean it would be good if you were simply to
sell factory products. These days it is not enough only to sell goods. Next
year, if you have that concept you all probably will want to sell a lot, or at
least as much as you can. The English word “marketing” does not just mean
sales; it is in fact a science, the science of knowing such things as: what the
people who buy these products month after month think of them, whether
they buy them again, and the reasons why. Afterwards, I will return to this,
but now, I would like to continue my discussion of statistical product quality
Statistical Product Quality Administration
Statistical product quality administration is a splendid new tool. It is being
applied in every industry, beginning in modern Japan and America as well as
England, New Zealand, and various other countries. Whether it be on a
large scale or a small scale, it is being researched and executed in an
extremely large number of manufacturing plants. Some results of this are:
1. Costs go down
2. Producers can economize on raw materials
3. Production levels increase, and waste decreases
4. Product quality becomes more uniform
5. Producers and consumers gain the ability to agree on product quality
6. Quality is improved, so inspections may be reduced
7. Appliances and techniques can be used to a higher degree
It is already twenty-five years since statistical product quality administration
was implemented in America, but it was not developed much until 1942, and
the extent of its use was narrow. However, for the following two reasons, in
about 1942, we developed statistical product quality administration.
In that year, a pamphlet commenting on the principles of statistical product
quality administration in simple language, easily understandable by general
technicians, was published by the American Standard Association.
Secondly, beginning in 1942 an eight-day course on statistical product
quality administration opened in the US, and several hundred technicians
were trained. I ran this course in Tokyo for 220 Japanese technicians this
July through the sponsorship of the League of Japanese Scientific
Techniques; and its content was similar to that of one I ran again in August
at Fukuoka’s Kyushu University.
In modern Japan many technicians, mathematicians, and statisticians are
researching statistical product quality administration. Furthermore, since
coming to Japan I have learned that the splendid achievement these people
have made in statistical product quality administration is already apparent.
The results from all these countries is surprising. They have demonstrated
superb ability in the statistical sphere.
The knowledge and brains applied to statistics by the Japanese are an
essential national resource; it is important in the same way as water power,
forests, and railroads. And that statistical knowledge, much like water
power, is not useful at all unless it has an impact on work opportunity and
work. With water power, if one were to get rid of turbines and generator
machines, no power would emerge.
Similarly it may be recognized that without effective use, all of modern
Japanese statistical knowledge would not be helpful to the advance of
products, product quality, or product uniformity. You all must look for
people who have both statistical knowledge and excellent experience with
technical knowledge and employ them in factories. In addition, to aid your
technicians you must seek mathematical statisticians as consultants.
However, no matter how excellent your technicians, you who are leaders,
must strive for advances in the improvement of product quality and
uniformity if your technicians are to be able to make improvements. The
first step, therefore, belongs with management. First, your company
technicians and your factories must know that you have a fervor for
advancing product quality and uniformity and a sense of responsibility for
Nothing will come of this if you only speak about it. Action is important. If
you demonstrate enthusiasm for the improvement of product quality, your
product quality administration will certainly advance. Responsibility for
product quality means guaranteeing one’s own factory’s products to the
utmost degree possible. The greatest guarantee of product quality is not
words, but executing product quality administration. When you effect
product quality administration, show administrative charts or methods as an
indicator, as there is no better way to guarantee product quality to the
By showing enthusiasm toward product quality and uniformity and
responsibility to product quality . You will gain the opportunity for your
technicians to put product quality administration into effect.
However, product quality administration at the factory cannot be
implemented in a single day. It requires along time. At first do it on a small
scale, and once you think that has value, then expand. The process of
product quality administration, consisting of the combination of statistical
and manufacturing techniques is long and tiring.
Well, I think you must share an interest in how I define statistical product
quality administration. Product quality administration is most useful, and
moreover it illustrates the methods by which producers can economically
produce the goods that buyers indicate they want.
All of the words in this definition have their respective meanings.
“Economic production” means low price production: in other words,
elimination of waste; faster production; fewer defects in products, raw
materials or machines; the practical use of techniques; improved uniformity
in product quality; and the opportunity for producers and purchasers to agree
on product quality.
“Most useful” means the design and product quality must suit the purpose of
the good; raw materials, mechanical manufacturing techniques, transport,
and products must be the best as we consider them from the viewpoint of the
marketplace. If you do not conduct market surveys about what quality or
what design will be in demand, your products will not succeed in being
The product must match the market. The product quality must be adapted to
that market. If product quality is too high, or too low, it w ill not be right. If
product quality is too good, the price is very high, and only a quite limited
group of people can afford to purchase it. Even so, if product quality is too
low, despite low prices, people will not repeat the purchase, and before long
the business will slow down. Therefore, in order to make the most useful
product, you must conduct market surveys about what kind or product
quality and what kind of design are required. Additionally, through market
surveys, you must make sure of the price buyers can pay for the product.
On this subject, there are other reasons why continuing market surveys is
useful to economical production. Every month you must make roughly the
right amount of product, or you cannot achieve economical production. If
you are left with 10%, or 25%, of your goods unsold, your profits will
disappear. Further, if you cannot respond to a large number of orders, y au
will not maximize your profit. As a result, in order to achieve economical
production, it is necessary to conduct market surveys. Based on market
surveys, you will he able to stand on a stable foundation of monthly product
sales. Then you can find out what kind of people (e.g. wage workers,
farmers, businessmen) demand that product; or whether, having bought the
product once, they will not but it again; or whether the people who bought it
are satisfied with the product; or again if they are unsatisfied, why they are
not satisfied. Thus by means of market surveys, you can effect
administration of advertising, product design changes pricing, and
No matter what kind of manufacturer you are, if you have long-term plans,
you must work to implement product quality administration in the wider
meaning, as in the definition which I laid out a moment ago.
In the last ten years there has been no scientific method which has
experienced such rapid expansion as has statistical theory. In the context of
today’s Japan, the most useful thing for manufacturers could be nothing but
the appropriate application of statistical techniques.
At the end of my discussion of market surveys, I would like to explain my
thoughts on the problem of statistical product quality administration with a
diagram. This diagram not only makes clear my thoughts on product quality
administration and market surveys, but I think it is extremely easy to
understand. Below I have drawn a pie graph “wheel” divided into four
Concepts regarding product quality
Sense of responsibility for product quality
This wheel rolls along the line of “concepts regarding product quality” and
“sense of responsibility for product quality.” The fact that the four stages of
the wheel are connected one to the other with no beginning and no end is
very important. This is the reason why I drew a circle. You must not stop
product design or testing. When your products emerge into the real market,
after having inquired into how the product is useful to people, and what they
think of it, you redesign it. There is no end to product quality administration.
Using product quality administration, producing goods continually being
improved, I want you to make more and more adapted items that buyers will
want, designing, redesigning, and then finding cheap, better ways to make
them. While this certainly benefits the purchasers, it benefits you as well.
Finally, I would like you to understand that the latest stage of statistical
product quality administration for factories lies in pertinent market surveys
Now I will add to this two or three explanations of market research.
Special Advice Concerning Market Surveys
This year is 1950. The current markets and manufacturing techniques are
considerably different from those of l935. In Japan the recent changes may
possibly be greater than those in any other country. However, businessmen
from everywhere are facing the necessity for new methods. In the past, in
1935 and before, businessmen simply made things they wanted to sell, and
sold those things they had made. They tried to sell these products both in
domestic and international markets. There were those who succeeded, and
those who failed. There was no way to determine how much their enterprise
would improve if they made products of a different design, better adapted to
the buyers’ needs, with better methods. The stages of manufacture and sales
of that time were as follows:
1. Product design (shoes, cotton materials, silk materials, magnetic
products, electrical appliances)
Today, in 1950, we all must design, manufacture, and sell in the same way,
nut science has expanded. Rather than following the example of these three
steps, four stages, including market surveys which I have just highlighted
have become necessary.
These days we cannot compel people to buy. When, selling a product,
whether it be on domestic or international markets, we must deliver the
things people truly want. Moreover, we must manufacture at a price that
invites purchase, furthermore, when markets are far away, at a price which is
competitive. As I mentioned before, if the product quality is either too high
or too low, it is no good.
In the past two or three years, market surveys have come to be extremely
successful in America. However, I cannot say their evolution is complete.
Market research by modern manufacturers themselves, or through market
research specialist companies, is progressing swiftly. I think if it were not
for that progress, many of those businesses would not be prosperous now.
I firmly believe that if product quality administration and market surveys,
prudently and scientifically, were used in a correct manner you would be
able to create a market for Japanese goods overseas, and the Japanese
standard of living would greatly rise.
Market surveys give answers to the following kinds of questions: “What
kind of goods would it be profitable to produce that people would demand
and be able to buy?” I would probably say this about that subject: the
techniques of statistical product quality administration count the methods of
making cheap and uniform high-quality products for you. The
manufacturing and market surveys that comprise the four stages of product
quality administration consider not only the manufacturers, but also the
The process of sales is not something that finishes simply with transporting
the products to the marketplace, and receiving money. In today’s sales, after
selling the product, the businessman must think about whether he has
satisfied the customer, and how improvements can be made from then on.
Market surveys accomplish better service for the consumer, and moreover
they serve the purpose of increasing profit for management and workers.
As I have said before, I am visiting here as a statistician, but I strongly
believe that, as in the four parts of that wheel diagram, we must not fail to
use statistical techniques in today’s industrial world. If it is the case that
businessmen try to perform a great service to the public, I emphasize that
they should find and put excellent statistical techniques into each of those
To conclude, I want to offer you one or two things I have realized about the
statistical techniques that are currently being used. Number one: at the first
stage, design, you must conduct market surveys and inspections, applying
the statistical techniques for experimental and planning methods and
inspection of samples. Furthermore, you must perfect the manufacturing
process, using the technical skills of factory workers and machines, and
utilizing these techniques to conduct inspections of raw materials.
Number two: in manufacturing in the second stage good product quality and
uniformity of product quality are important. This means that by the time of
shipping, you can transport quality products through the use of statistical
techniques of inspection of samples, experimental planning methods and
marketing surveys. Number three: at the third stage, sales, the statistical
technique of the market survey is used. Number four: at the fourth stage of
service research the techniques of sample inspections, experimental planning
methods, and market surveys are used.
After you have reached the fourth stage, you return once again to the
beginning stage in wheel. Here, then, putting together thoughts from your
previous results, you may begin to implement product design changes.
I repeat myself, but the necessity of statistical techniques must not be
ignored. Without them I deeply believe that businessmen cannot long
sustain their prosperity. What I just spoke about now is true not only in
Japan, but equally in Chicago, Manchester, or Amsterdam. Every
businessman around the world is now facing these same problems.
By the way, I believe you all probably have many questions to ask. You
probably want to know more about market survey techniques, or about how
they are managed. Anticipating one of your questions, why don’t I give an
answer now. What they are doing to carry out market surveys in America
may be answered as follows: it is fine to request the services of a
trustworthy market research company on your own.
On this topic, I would like to ask for your care on one more issue. Even if
you hear developments from professional buyers in New York, they are not
necessarily all correct. Therefore, because you do not know what American
shop for in a basic product, and you only respond to orders, you may be
driven out of the market. This is due to lack of knowledge about what kind
of design or product quality Americans are demanding, or what markets may
prosper in the future. You, Japan’s skillful technicians, can manufacturer
items that businessmen of American and the rest of the world cannot. You
don’t know whether the things you want to sell are what Americans want to
buy, but you know the way to find this out. That method is statistical
research in the realms of manufacture and market surveying.