Chapter 3: SPC as Shewhart intended
Following is a set of four articles
written by
"SPC—Back to the Future"
Part 1 – An 11-year-old can do it!
SPC?
"SPC? Oh, Statistical Process Control. Nothing to do with us. We're
not in manufacturing."
"SPC? Oh yes, I think the shop-floor does some of that. Nothing to
do with me: I'm a manager."
"SPC? Oh, I don't understand anything about that. I just collect
the numbers and pass them on to Quality Control."
"SPC? Oh, that's not for me. I'm no mathematician."
"SPC? Oh, no way. I don't trust statistics."
"SPC? Oh yes, of course. I'm a professional statistician. It's
quite simple really. But you have to be sure that your data are normally
distributed, else it's not valid."
Six responses. Six sad responses.
11-year-old Patrick
A story which the American management
teacher Dr W Edwards Deming was fond of relating during his celebrated four-day
seminars concerned 11-year-old Patrick Nolan (Neave, 1990a: 393-395). Day by day,
Figure 1 shows
Figure 1
There were however two exceptionally late
arrival times, well beyond those limits, which
"If 11-year-old
My
school-bus
* I am indebted to
I also travelled to school by bus. The bus
would usually arrive at the bus-stop near my home some time between 8.25 and
8.35, though occasionally a little earlier or later. Now and again, like
To be fairly sure of catching the bus, I
had to be at the bus-stop by 8.25. To be really sure
of catching it, I had to be there by 8.20. But much of the time I'd then be
waiting 10 to 15 minutes, occasionally longer – often when the weather was cold
and wet. (Sympathy, please: this was a bus-stop, not
a bus-shelter!) So I would often be soaking wet
and/or freezing cold by the time the bus arrived. Not a good start to the day!
If only the (common-cause)
variation in those bus-arrival times could have been smaller! If only the bus
could have arrived within, say, one minute either
side of 8.30, rather than within 5 or even 10 minutes. Or, indeed, one minute
either side of 8.25, or one minute either side of
8.35 – or one minute either side of any suitable average time of arrival. Then I could have
arranged my mornings much more efficiently – and, with rare exceptions,
suffered no more than a two-minute soaking!
Thus I learned at an early age that variation affected my quality of life. The variation was
actually more important than the average time of
arrival. The greater the variation, the more I risked either missing the bus
altogether or getting wet through.
Variation is the enemy of quality.
How
it all began
In the early 1920s, people in the Western
Electric Company were hard at work trying to improve telephone technology and
associated equipment. For a while they made great progress. But then the rate
of progress slowed. They were still trying as hard,
if not harder than before. They were still pouring time and money – and
probably emotion – into the improvement effort, but somehow it just wasn't
working any more. They were experimenting and analysing and trying to interpret
data in just the same ways as before. Those ways had previously reaped great
rewards. But no longer. Increasingly, not only were they failing to improve:
they were beginning to make things worse rather than
better! That is when they invited
Now we'll let Dr Deming take up the story
(transcribed directly from a presentation to an audience in
"Part of Western
Electric's business involved making equipment for telephone systems. The aim
was, of course, reliability: to make things alike so that people could depend
on them. But they found that the harder they tried to achieve consistency and
uniformity, the worse were the effects. The more they tried to shrink
variation, the larger it got. When any kind of error, mistake or accident
occurred, they went to work on it to try to correct it. It was a noble aim.
There was only one little trouble. Things got worse.
Eventually the problem went to
1.
Treating a fault, complaint, mistake, accident as if it came from
a special cause when in fact there was nothing special at all, i.e. it came
from the system: from random variation due to common causes.
2.
Treating any of the above as if it came from common causes when in
fact it was due to a special cause.
What difference does it make? All the difference between failure and
success”
How did
Apart from introducing some of the basic
concepts, this early piece of history is important in emphasising that the
environment and purpose in and for which SPC was created was one of improvement. Shewhart invented the control
chart to provide guidance on the types of action most likely to bring
about improvement and warnings on the types likely to do harm.
Interestingly,
Learning
And Unlearning
We're already two-thirds through this
article. So I can probably now risk confessing my origins without too much fear
of frightening off those readers who have come this far!
I began my career life as a conventional
mathematical statistician. I learned conventional mathematical statistics as a
student, and then, as a Lecturer, I taught what I had learned.
In my own defence, I had some reservations. These led me to dabble in areas of the
subject regarded by the purists as slightly unconventional. But I had neither
the wit nor the courage to dip more than my toe in the water.
So then came my
stroke of great good luck! I was singularly fortunate in the early 1980s to
become involved with the British subsidiaries of the first American company to start
taking
FIGURE 2 The Deming Medal
But I was puzzled. I was told that
Despite that continuing perplexity, in
1985 I received an invitation to assist
I looked forward eagerly to that first
four-day seminar. Now I would at last learn the
truth about where my great knowledge of mathematical statistics would fit into
it all!
Wrong again! Granted, there was some stuff about collecting and analysing data. But it
really was rather disappointing. The only technique he ever seemed to use was
the control chart – and
he didn't even do that right! Where were the
probabilities, the normal distributions, the Central Limit Theorem, the action
and warning limits? You see, for years I had covered control charts in my
60-lecture first-year course at the University: I'd probably spend one lecture,
maybe even two, on it. Seemed rather dull, really – nothing more than a
slightly glorified significance test. And this was
all he was using – and, worse still, without any of the clever mathematics?
Yes, unlearning is much more difficult—and painful—than learning.
Words
of wisdom…
…from
"The fact that the
criterion which we happen to use has a fine ancestry of highbrow statistical
theorems does not justify its use. Such justification must come from empirical
evidence that it works. As a practical engineer might say, the proof of the
pudding is in the eating." (Shewhart, 1931: 18)
"Some of the earliest
attempts to characterise a state of statistical control were inspired by the
belief that the normal law characterised such a state. The normal law was found
to be inadequate: all hopes [for such an approach] are blasted." (Shewhart, 1939, p12)
…and from
"It would be wrong to
attach any particular figure to the probability that a statistical signal for
detection of a special cause could be wrong, or that the chart could fail to
send a signal when a special cause exists. The reason is that no process is
steady, unwavering."(Deming, 1986:
334)
"It is true that some
books on the statistical control of quality and many training manuals for
teaching control charts show a graph of the normal curve and proportions of
area thereunder. Such tables and charts are misleading and derail effective
study and use of control charts." (Deming, 1986: 335)
"It is nothing to do with probabilities. No, no, no,
no: not at all. What we need is a rule which guides us when to search in order
to try to identify and remove a specific cause, and when not to. It is not a
matter of probability. It is nothing to do with how many errors we make on
average in 500 trials or 1000 trials. No, no, no—it can't be done that way. We
need a definition of when to act, and which way to act. Shewhart provided us
with a communicable definition: the control chart. Shewhart contrived and
published the rules in 1924. Nobody has done a better job since." (Neave, 1990b: 4)
…and from the Japanese,
who both learned and unlearned…
"The ease with which [
"Prior to Deming's visits
in the early 1950s, Japanese quality control had been butting its head against
a wall created by adherence to difficult statistics theories. With Deming's
help, this wall was torn down." (Noguchi, 1995: 35-37)
…and finally, from
"The
control chart is no substitute for the brain."
References
(Some of the quotations are abbreviated,
but without losing their original sense.)
·
Deming, W
·
Deming, W
·
Neave,
·
Neave,
·
·
Shewhart,
·
Shewhart,
·
Statistical
Quality Control (August 1950), JUSE,
"SPC—Back to the Future"
Part 2 – Boring – or not ?
Boring charts
Remember 11-year-old
Let's look straightaway at the first six run charts:
Figure 1.
Stop! For the
moment, don't read beyond this paragraph. Just look at the charts. If you had
to write reports for the boss on the behaviour of the six processes represented
on these run charts, what sort of things would you say?
- - - - - - - - - - - -
Okay. That was almost a
trick question. It's difficult to write reports on these six processes. There's
very little to say. They're boring!
They all show some
variation (else the graphs would simply be flat horizontal lines). But none of
them have any "features of interest". There are no trends, no sudden
shifts, no patterns. Just the same boring behaviour day after day, month after
month.
Not only that. The same boring behaviour is shared by
all six processes! Were it not for the fact that the numbers on the verticals
are different, you might well suspect there were not six different processes at
all. They could have been six different sections of just one boring process!
Boring charts
are useful
But being boring has its uses. It is, of course,
useful in practice to be able to predict what our processes will do in the future. Is their performance likely to be suitable for our
purposes? Where should we prioritise our future improvement activities?
And…
…if the behaviour of a
process has been "boring", i.e. unchanging, at least for a while,
and
…if nothing occurs to
change that boring behaviour,
it's rather likely to continue that same boring
behaviour into the future!
Thus, subject to those two "if"s,
we can predict future behaviour. That's very useful
in practice. So we can write something positive in those reports!
Phrases used to describe processes which are
demonstrating such boring but useful behaviour are said to be
in
statistical control
or
exhibiting
controlled variation
or simply
stable
Never mind
"can". How?
The boss, reading our reports, will not be content
with seeing that we can predict future behaviour.
What is that prediction?
For each process we could draw some
"safety margins", one a little above the current data, and one a
little below. Our prediction (subject to the second of the "if"s) could then be along the lines of:
We predict that future data will
continue
to be comfortably contained between these safety margins,
and
show
no trends, shifts, patterns, etc.- we might express this as "behave
randomly".
Figure 2 shows the run charts with such "safety
margins" drawn in.
There are well-established methods for calculating from the data where the
safety margins should be, rather than just drawing them in "by eye".
If those well-established methods are used, the safety margins are then called control limits, and the graphs are called control charts. Be careful! Some people (including
"experts") sometimes get it wrong! A good method of calculating
control limits will be developed in the next article.
Interesting
charts
Now let's look at another six run charts: Figure 3.
How about writing reports on those?
- - - - - - - - - - - -
Now we have a different story—or, rather, six
different stories! Unlike the first six, these charts are not
"boring". Why are these processes
relatively "interesting"?
Because things happen in
them. The behaviour of each one of these processes changes
during the time covered. Sometimes the changes are gradual, sometimes abrupt.
For example, process (a) starts off fairly level
(centred around 13), then drops for a while to an average of around 6, and then
rises quite abruptly and exhibits more volatile behaviour. Yes, now we have
something to say! (Later in the article we'll see what these processes actually
were and what was going on in them: you'll then be able to judge how well your
reports fit the facts!)
Earlier we saw that processes which are
"boring" i.e. are in statistical control—have
the distinct advantage of being "predictable". Such predictability is
never 100% certain, but one can be pretty confident about it subject to that
second big "if".
It follows that the same cannot
be said of "interesting" processes! They are "interesting"
precisely because they are unpredictable: they do
surprising things, their behaviour changes unexpectedly. (I'm assuming that we
do not know in advance of any reasons likely to cause such changes.) Thus we
describe them with the opposite terminology. These processes are
out
of statistical control
or
exhibiting
uncontrolled variation
or
unstable
So, in terms of predictability,
"boring" is nice
"interesting" is nasty
Interpreting
process data
The time is now ripe for some crucial thinking about
trying to analyse/use/interpret data from the two types of process.
With processes which are in
statistical control, isn't it pointless to try to "explain" why any
particular individual value in the data is what it is?
For "in statistical control" means
"unchanging behaviour"—so there can be nothing to
"explain"!
Appreciation of this one fact could result in many
management meetings being cut to a fraction of their normal length!
I'll now begin to give the game away about what these
processes were. Figure 1(a) charts the number of spots showing in repeated
throws of four dice. Figure 1(b) is the number of Heads when 25 coins are
tossed. The number of Heads goes up and down – no kidding! But there's nothing
to learn from when it goes up and down, nor by how
much. It's just chance, random. Same for the dice,
surely. And, of course, this fits in with our earlier interpretations of the
data, before knowing what the processes actually were.
Don't think that all of the
processes charted in this article are this "trivial" – these simple
ones have been included for a very definite purpose, soon to be revealed! But
let's reconfirm that all the charts in Figure 1 are
"boring"—nothing exciting to report. So, what's sauce for the first
two geese is surely sauce for the other four ganders!
A good way of summarising this important message is:
If there is no evidence that the
variation in the results you're observing is really different from the sort of
behaviour you might get when tossing coins or throwing dice, the process is in
statistical control: don't waste time trying to find reasons for particular
results.
That's not to say there are no causes of the
apparently random variation. There are causes: as we
saw last month, we call them common causes. But it
is wholly illogical to think you can find out anything about them by just
reacting to individual values in the data. Deeper and longer-term study is
needed.
On the other hand, when you see evidence of a real
change (e.g. a result which is quite untypical of what you've been seeing
previously), of course it makes sense to try to discover why you've seen what
you've just seen. And the sooner the better, while the information is still
hot! The sooner you look, the more likely you are to find the special cause which has produced the change.
The great importance of this guidance was beautifully
summed up by
Controlled variability
It will not be profitable to try to
determine the cause of individual variations when the variability is
controlled.
Uncontrolled variability
It will be profitable to try to determine
and remove the cause of uncontrolled variability.
So what about the other four processes in that first
batch? We now know that it is not sensible to try to explain individual results
in those processes either. But what were they?
Figure 1(c) was one for those of you who know
something of
Figure 1(d) was a series of measurements of my
pulse-rate, recorded just before breakfast over a period of 24 consecutive days
(the final 24 days of October 1991, to be precise).
Figure 1(e) shows the total lengths recorded in the
first 24 samples from A Japanese Control
Chart.(Wheeler, 1984)
(This is a highly-recommended case study, to which I shall refer several times
in these articles. It is available both as a document and a video.)
And Figure 1(f) shows the monthly American trade
deficits (in billions of dollars) during 1988 and 1989.
Quite a selection of processes! And deliberately thus
chosen.
The sharp
double-edged sword
The facts that the processes in
Figures 1 and 3 were respectively in and out of statistical control were pretty
obvious just by looking at the run charts. Often the choice is not so
clear-cut. So then how do we decide?
It's the control chart
which helps us again.
If we calculate control limits for
the charts in Figure 3, using precisely the same method as before but now
applied to these data from processes which are out of
statistical control, we get Figure 4. Now each process has points outside the control limits.
So, the control chart serves us in two roles:
1.
When
the process is in statistical control, the control limits
predict the likely range of variation in the future—the near future, at least;
and
2.
The
control chart helps us diagnose when such prediction is
feasible and when it isn't. When points fall outside the control limits, this
is evidence that the process is out of statistical control, i.e. such prediction
is not feasible.
These two features make the control chart a supremely sharp,
double-edged sword :
It can diagnose.
It can predict.
Another phrase often used to describe the control
chart is the Voice of the Process.
It helps the process speak
to us, telling us e.g. what it is doing and what it
is capable of doing.
What of the six out-of-control processes (Figures 3
and 4)? What were they?
They were the same
processes as before! But these data came from later times, when special causes were in operation. In some cases, I know what the special causes were, for it was I that
created them!
§
Figure 4(a) For the first six
points, I used four dice as before. For the next six points I used only two
dice. For the rest, I used six dice. I was changing the
process, i.e. producing special causes. The control chart had already diagnosed
that something had been happening to the process,
but of course couldn't tell you what. It simply told
you that it was worth looking.
§
Figure 4 (b) Here I began by
tossing 25 coins as before. But over the final 10 points I added two extra
coins each time.
§
Figure 4(c) Here I did something to
artificially double the scores near the end of the sequence.
§
Figure 4(d) These data were from
November 1991. Near the end of that month, my doctor prescribed a beta-blocker
to reduce my blood pressure and pulse rate!
§
Figure 4(e) Here a fault developed.
It is evident from the chart when the fault was diagnosed and rectified.
§
Figure 4(f) These were the monthly
The control charts cannot
tell you the whole story—but they certainly tell you when there is a story to
be told!
Guidance for
improvement
As we know from the Western Electric story, Shewhart
created the control chart to provide guidance for improvement.
What kinds of interpretations of data, and what kinds of actions, are likely to
be fruitful? Just as important, what kinds of interpretations and actions are
literally likely to do more harm than good?
Here is a summary of guidance for improvement using
control charts:
If the
control chart judges the process to be in statistical control, improvement
effort should be directed at the process as a whole, using information over a
relatively long period of time. Do not be distracted by shorter-term data or,
even worse, individual data-points.
If the
process is out of statistical control, initial improvement effort needs to be
directed at trying to identify the special cause(s) of the instability and
taking appropriate action; in this case it is justifiable, indeed necessary, to
investigate shorter-term effects in the data, particularly as guided by points
which are beyond the control limits.
Appendix: Technical notes
1.
Control
charts are given different names, according to the types of data represented on
them. The charts here for "one-at-a-time" data are sometimes referred
to as X-charts. Why "X"? It's the mathematician's favourite
letter-I know of no better reason!
2.
If a
process is in statistical control, the control limits on its X-chart are
sometimes called its Natural
Process Limits - for
that's what they are!
References
§
Deming, W
§
Neave,
§
Wheeler,
“SPC – Back to the Future”
Part 3 – So, how do we compute those control limits?
The double-edged sword
How do we decide where to draw those control limits,
and why?
Recall from last month's article that we need them to
provide us with that "sharp, double-edged sword". Get it wrong, and
one or both edges of our sword will be blunt and useless.
What are the two edges of the sword?
- When the
process is in statistical control, the control limits predict
the likely range of variation in the future (the near future, at least);
and
- The
control chart helps us diagnose when such prediction is
feasible and when it isn't.
Let's illustrate (Figure 1) with a couple of the
control charts we saw last month: my early-morning pulse rates before and after
I was prescribed the beta-blocker!
Figure 1
(The control limits in Figure 1(b) were calculated
from the data on that chart. Since, in fact, these
were simply later data from the same process as in
Figure 1(a)—at which time the process was in statistical control—we could in
practice just have used the same control limits as in Figure 1(a). But when you
first saw the graphs, you didn't know they were from the same process, so these
control limits were computed reflecting that lack of knowledge.)
Two criteria
Control limits must satisfy two criteria:
- They must
be far enough apart to comfortably contain virtually all
the data produced by the process when it is in statistical control; and
- They must
be close enough together for some of the data-points to
lie outside them when the process is out of statistical control.
What do these criteria imply?
Simply that the control limits should
cover the extent of common-cause variation in the process: no less (else
we'll contradict the first criterion) and no more (else we'll contradict the
second criterion). That's it! That's the guiding principle.
So how do we do it?
- We'll need
some "indication" or "measurement" of the common-cause
variation.
And then
- We'll
designate the distance between the average (the Central Line of the control chart) and the control limits in
terms of that measurement of the common-cause variation. Specifically: the
distance out to the control limits will be a suitable multiple of the measurement of variation.
How can we measure variation?
Measuring variation is a less
familiar concept than just measuring weight, or time, or indeed calculating an
average. But that is what we're going to need. A
"measurement of variation" is what it says: it measures variation! If
the variation is large, the measurement is high; if the variation is small, the
measurement is low.
Mathematical statisticians have long had their
favourite measurement of variation: they call it the standard deviation.
It's not the only way of measuring variation—though
"standard" might give that impression! It is made to look even more
sacrosanct through traditionally being designated by a Greek letter : s( pronounced "sigma" ).
In fact, the standard deviation can be somewhat
daunting to non-mathematicians: it has a complicated formula. Fortunately, in
the control-charting context, we do not have to suffer that formula – for the
very good reason that it is not appropriate for
calculating control limits! But…here comes a very deep bear-trap—one into which
many of the unwary have fallen. "Scientific" calculators can compute
that formula for you. All you do is enter the data and press a button ( labelled
s, or s, or sn-1, amongst
various possibilities ). So, if you like pressing buttons, that can be very
tempting!
But it will
often give you wrong control limits!
Don't
do it!
In a nutshell, the trouble with the "conventional
formula" for s
is that if we use it on data from a process which is in
statistical control then we get sensible control limits; but if we use it on
data from a process which is out of statistical
control, we may get crazy limits! But how do we know
whether our process is in or out
of statistical control? By using a control chart, of course. Errr…, that
means we have to calculate control limits. But, you've just said that unless the process is in statistical control, the
conventional s may give us
crazy control limits. Right!
Call it a circular argument, begging the question, or
what you will. The conventional s will not
do the job. The edges it gives our sword may be extremely blunt.
We'll soon see why.
What is the
"conventional s
"?;
So what is the "conventional s ", the standard
deviation-just the general idea, not the detailed formula? It is, in
effect,
an indication of the
"typical" or "representative" distance between the
individual items of data and their average.
Let's see if you can guess roughly what s is just by looking at some data!
Figure 2 shows another couple of the run charts based on data we saw in the
last paper. I've changed the pictures slightly. To help you, I've inserted
their Central Lines (averages) and I've given them both the same vertical
scale.
Estimate roughly the value of s for both of these two processes. (
Try the first one, then check the answer below. Then come back and try the
second one. )
Figure 2
Figure 2(a)
The average is 14.1, and the lowest and highest values are 10 and 19. So the
largest distance away from the average of any of the values charted is about 5.
Several of the values are at a distance of around 3 or 4 away from the average,
with the rest being closer. A guess for s of anything between 2 and 3 would
have been very reasonable. The standard formula for s actually gives 2.5.
The variation in Figure 2(b) is obviously smaller.
Look back at it now and make a guess—preferably before reading the answer
below!
Figure 2(b)
A very acceptable guess here is that the variation is about half that of the
first chart, which would suggest that s
is around 1.2 or a little bigger. The standard formula in fact gives 1.5.
So, while it is true that the conventional formula for
s can be regarded as quite complicated
by non-mathematicians, it is possible to make a
reasonable guess at its value. One cannot expect to get very
close to its value by guesswork, but one can hit the general "order of
magnitude" of s, i.e. be "in the same ballpark".
Shewhart's "3 s -limits". Why
So how many ss away from the Central Line should
the control limits be placed in order to satisfy those two criteria above? In
Shewhart's own words:
"Experience
indicates that 3 seems to be an acceptable economic value."
That's it! He experimented. He found
that if he placed the limits much closer than 3s from the Central Line then he
embarked upon too many false trails for special causes which didn't exist; and
if he put them much further out then he started missing important clues of
special causes.
Figure 3 shows Shewhart's 3s-limits for the run charts of Figure
2.
Figure 3
As with all the control charts in Figure 2 of last
month's article, these control limits comfortably contain
the data from the processes.
The conventional s doesn't work!
But those processes are in statistical control. What
happens if we try the method on processes which are out of
statistical control? We know what we want to happen:
we want to get points outside the control limits to
signal that there are special causes around.
As I've already warned you, the conventional s doesn't do the job. To demonstrate
why, let's look at some of my pulse-rate data again. Figure 4 is similar to
Figure 1(b) except that it covers the 12 days before and after I started taking
the beta-blocker.
Figure 4
Who could doubt (even if I hadn't told you the
circumstances) that a special cause had changed the behaviour of this process?!
But suppose we were to try to compute control limits
from these data using the conventional s .
What was my average pulse-rate over those 24 days? It
was about 75: during the first half of the chart it was well above 75, in the
second half it was well below.
Now what was that conventional s ? You'll recall :
an indication of the
"typical" or "representative" distance between the
individual items of data and their average.
What's that here? It's
around 10! Virtually all the data lie between 80 and 90 or between 60 and
70—i.e. on average about 10 from 75! Where would that
put the 3s -limits? At
about 45 and 105!! Try drawing them in on Figure 4. I think you'll agree they're
not very useful! A special cause has massively changed the behaviour of this
process, yet none of the data reach even halfway to those limits. It's not the
data's fault: it's the fault of stupid limits!
What can we do instead?
Why has this
happened?
As pointed out earlier, the crucial requirement is
that the s used for
control limits must represent common-cause
variation. My pulse-rate process was in statistical control both before and
after taking the beta-blocker—but of course with very different averages. Common-cause variation is the "random" variation
around the local average. If you attempted a "by eye" assessment of s
separately for each half of Figure 4 you'd
probably come up with, say, 3 or 4—not 10!
Unfortunately it is not always the case that we can
easily split the run chart into such separate in-control sections. Often, for
example, changes in the process are much more gradual than here. So we need a
more automatic method of measuring localised
variation only. We'll still illustrate it with the pulse-rates data.
The 24 pulse-rates are listed in Table 1. Under the
data I have entered the day-to-day changes in the
pulse-rates. These are called moving ranges (MR) and
they represent the most localised variation available to us in our data.
Table 1
|
Pulse- |
81 |
80 |
82 |
82 |
85 |
86 |
88 |
78 |
89 |
81 |
87 |
76 |
66 |
69 |
64 |
64 |
64 |
66 |
64 |
69 |
67 |
63 |
61 |
|
|
MR |
|
1 |
0 |
2 |
0 |
3 |
1 |
2 |
10 |
11 |
8 |
6 |
11 |
10 |
3 |
5 |
0 |
0 |
2 |
2 |
5 |
2 |
4 |
2 |
A much more suitable s for control limits is thus based on
the average (mean) moving range, MR-bar. (The bar is
a common shorthand for "mean".) MR is, on average, a little larger
than s, and so needs scaling down appropriately. The scaling factor normally
used is 1.128; i.e.
s = MR ÷
1.128
If you'd like to do the arithmetic on Table 1, you
should find that the moving ranges add up to 90, giving
MR-bar
= 90 ÷ 23 = 3.91 and
s = 3.91÷
1.128 = 3.5
in accord with our "by eye" assessment
earlier!
You might also like to confirm that the average
pulse-rate is 74.7 and therefore that the 3s —limits work out to 64.2 and 85.2.
Draw these limits on Figure
4. Now we have something sensible! Not only do
several points lie outside these limits: the rest of the points can hardly be
described as "comfortably contained" within them! Now we have a
control chart which reflects the truth. This process is out of control: it has
been subject to a substantial change. In this case the change is, of course, an
improvement—exactly what the medication was intended to provide.
This method for computing control limits was used for
all 12 control charts last month and will also be used throughout the case study
comprising next month's article.
Appendix: Technical Notes
a) The
distance from the Central Line to the control limits may be computed more
directly as 2.66MR-bar : this gives the same answer
as dividing by 1.128 and then multiplying by 3.
b) When
computed from out-of-control data, MR-bar tends to
be a little larger than if the process is in control, as the special cause(s)
will tend to inflate one or more of the moving ranges. An alternative method
sometimes preferred to avoid this problem is to use the median
moving range, in which case the 2.66 should be replaced by 3.14. This method
will be employed in the case study described in Article 8.
c) A
phenomenon which seriously widens the control limits
is a "zigzag" pattern, caused e.g. by over-adjustment in the process.
If control limits computed from moving ranges seem too wide, check for this
effect.
d) Some
people prefer a different calculation of control limits for processes like (b)
and (c) in this article, giving a so-called np-chart.
But there's no real need.
SPC—Back to the Future!
4. Case Study: " ... And what's this saving cost you?"
"Elementary, my dear Watson "
The quotation which forms the title of this article
deserves immortality. It is a question that should be asked repeatedly at most
meetings of both management and politicians.
It comes from a video "The Short-Sighted
Boss" made for the DTI's 1986 National Quality Campaign. An electrical
goods manufacturing company is running into trouble. In a dream, its Managing
Director brings in
The question is also an ideal title for the case study
described in this article. The case study is not a dream—though it must have
seemed like one to many people: a bad dream. It is
originally reported in
Tips of icebergs
The case study tells the story of an improvement
effort in a section of a traditionally managed company. That section is
referred to as "Department 13". (For any superstitious readers, it
might be worth pointing out that it was Department 13's customer
that turned out unlucky, rather than Department 13 itself!)
At the beginning of the story, Department 13 had
material costs which amounted to 75% of its total production costs. So a
project team was formed and given the task of reducing those costs.
Over two and a half years, the project team made four
major changes. The first three were of a technical nature, while the fourth was
a change to a cheaper supplier. The details, and resulting material costs, are
summarised in Table 1.
|
Table 1 |
|
|
Period |
Average Material Costs |
|
(Project team formed) |
|
|
Jan–Aug, Year 1: |
$215.22 |
|
(First change: to improve
material utilisation) |
|
|
Sep, Year 1 to Feb, Year 2: |
$208.20 |
|
(Second change: further
process modification) |
|
|
Mar–Jun, Year 2: |
$205.37 |
|
(Third change: formulation
of material altered) |
|
|
Jul-Dec, Year 2: |
$201.22 |
|
(Fourth change: cheaper
supplier) |
|
|
Jan–Jul, Year 3: |
$198.46 |
The management were so pleased with the cost savings
that the project team and Department 13 received a special award in August of
Year 2.
At their regular Board meetings, the management had
been basing their discussions on the usual kind of monthly reports—described by
Wheeler as "several pages of tables of numbers, obtained from computer
printouts, and reduced down to a size that no one over 45 can read without
their bifocals"! Table 2 shows the part of the report for July, Year 3
relevant to Department 13.
|
Table 2: Monthly Report for July,
Year 3 |
||||||||
|
|
|
|
|
|
Year-to-Date Values |
This YTD as % Diff of last YTD |
||
|
|
July Actual Value |
Monthly |
% Diff |
% Diff |
Total or Average |
Target |
% Diff |
|
|
Production |
34.5 |
36.0 |
-4.2 |
-2.0 |
251.5 |
252.0 |
-0.2 |
-8.0 |
|
Material Costs ($ / 100 lbs) |
198.29 |
201.22 |
-1.5 |
-1.9 |
198.46 |
201.22 |
-1.4 |
-3.6 |
|
Manhours per 100 lbs |
4.45 |
4.16 |
+7.0 |
+4.5 |
4.46 |
4.16 |
+7.2 |
+9.3 |
|
Energy and Fixed Costs / 100 lbs |
11.34 |
11.27 |
+0.6 |
+11.3 |
11.02 |
11.27 |
-2.2 |
+9.2 |
|
Total Production Costs / 100 lbs |
280.83 |
278.82 |
+0.7 |
+0.9 |
280.82 |
278.82 |
+0.7 |
+0.4 |
The discussion would typically run as follows.
Examination of Table 2 shows a mixed bag of results. Production volumes are
down on the month, though pretty much on-target with respect to the
year-to-date figures. Material costs figures are down, confirming the project
team's diligent efforts. There is bad news regarding labour content (expressed
as "Manhours per 100 lbs") and energy and fixed costs, though the
latter is no great surprise because of inflationary effects, etc.. The good and
bad news seems to be mostly balancing out, since the total production costs are
virtually on-target. All agreed? Time to move on to the next area of the
report?
Just under the surface
But ... !
A point which Wheeler makes so strongly in his book is that such typical
discussion and "analysis" of the figures can only be the tip of the
iceberg at best. How can much more of the iceberg be revealed? By putting data on control charts. So let's do it.
First,
the material costs. Figure 1 is Wheeler's picture, covering the current month
and the previous two and a half years since the project team was set up. (We
have not reproduced the numbers here, since the pictures tell the story;
however, all the data are available in his book.)
Figure 1
Several matters deserve comment.
First, this is not a
control chart: it is five control charts! Of course.
The process has been changed four times. Remember that phrase: "Voice of
the Process"? If the process has been changed four times, there is a
succession of five different voices to listen to and compare.
Second, material costs have indeed dropped! But, more
than that, note that (with one or two possible exceptions) each portion of the
chart is distinctly lower than the previous one and
each portion is locally in statistical control. What does this mean? This is
not just a downward trend in costs. Each change made
by the project team has dropped the costs to a lower level. Figure 1, drawn
this way, identifies the cost reductions with the project team's actions. It
couldn't be just coincidence. The project team was doing precisely what had
been asked of them: they were reducing material
costs.
These are real data, and the picture is not quite
crystal clear. For example, unlike elsewhere, the third portion of the chart
does not contain any points below the previous Lower
Control Limit. However, those control limits in the second portion are rather
wide, and every point in the third portion is lower
than every point in the second portion—so there is
little room for doubt. (Remember: "The control chart is no substitute for
the brain"!)
Also, there is just one point outside its local
control limits (at the end of the fourth portion).
A final point worth mentioning is that
Further under the surface
Other figures are available for Department 13. Do you
recall some concern over the "Manhours per 100 lbs"? Would it be
helpful to put those figures on a similar chart?
Wouldn't it just?! See Figure 2.
Figure 2
Ouch! This picture is
absolutely crystal clear. Each portion of the chart is distinctly higher than the one before. And each portion of the chart
is locally in statistical control. No exceptions. No doubt. Again, this is no
mere trend. Each and every
change made by the project team raised the required labour content for the
product.
But, of course, that wasn't their
concern. Their job was to reduce cost of materials.
They did it—and indeed management rewarded them handsomely for it (remember
what happened in August of Year 2).
The original account in Wheeler's book contains
several other charts, most of which we shall omit here for brevity. One showed
the production volumes drifting down over most of the period—though they
appeared to have been moving upwards before the
project team became active. Energy and fixed costs mostly showed a
straightforward gradual trend upwards, confirming the suggestion that this was
mostly an inflation effect. Total production costs had nevertheless been gently
moving downwards, though interestingly they immediately increased
after the change to the cheaper supplier. Have you ever known a cheaper
supplier raise your costs?
Who paid?
But the worst is yet to come. The one important thing
not yet investigated is the quality of what
Department 13 produces. That, of course, cannot be judged until we see what happened
to Department 13's customer over the same period.
Department 13's customer was (logically enough) Department 14. Department 14
was in serious trouble. A major cause of scrap in Department 14 was that the
material from Department 13 "will not mould". Figure 3 shows the
terrifying picture (in five control charts, as before) of the monthly
percentage of material lost because of this problem.
Figure 3
Yet again, could there be any doubt? Each phase of the
process is locally in statistical control. But each portion of the chart jumps
to a higher level than the one before. Yet again, this is no mere rising trend.
These are step-change increases in the proportion of scrap coinciding with each
of the project team's actions. The unarguable conclusion is that each of the
project team's actions increased the scrap-rate. And not by small amounts! In
two years they had increased the scrap-rate eight-fold,
from an average of around 2% to an average of around 16%.
And who paid? One could say: the whole company—and its
customers. Certainly the people in Department 14 paid, and paid dearly. Some of
them had lost their jobs because of their department's poor figures. Whereas
Department 13 got a special award…
Where is quality made?
The trouble was that this company was not putting its data onto control charts. Traditional
management reporting in tables of figures, not graphs, had concealed the grim
truth.
Control
charts and the understanding of variation are not just for the shop-floor,
administration processes, paperwork, and the like. As
"Quality
is made in the Boardroom."
In his final book, The New
Economics for Industry, Government, Education (Deming, 1994: 36-38), Dr
Deming observed that maybe only 3% of the potential gains from the statistical
control of quality (i.e. from learning about, understanding, and dealing with
the causes of variation) come from such areas. He described its application to
overall business strategy and companywide systems such as personnel, training,
purchasing, legal and financial matters, etc., i.e. to the management
of the organisation, as:
"Here
are the big gains, 97%, waiting."
If the 3% is big (and it is), the 97% is massive.
References
§
The
Short-Sighted Boss (video) (1984), Video Arts (for the
Department of Trade and Industry).
§
Wheeler,
§
Deming, W Edwards (1994), The New Economics for Industry, Government, Education,
Massachusetts Institute of Technology, Centre for Advanced Engineering Study.
6.3. Operational Definitions:
Understanding and use for improvement
· What is an Operational Definition?
It
is a definition which reasonable men can agree on and do business with. Words
have no meaning unless they are translated into action agreed upon by everyone.
An operational definition puts communicable meaning into a concept. It is not
open to interpretation.
· Why is an Operational Definition necessary?
All meaning begins with concepts. A concept is
ineffable (i.e. unspeakable, beyond description) One can hardly do business
with concepts--use of concepts rather than operational definitions causes
serious problems in communication e.g. Reports, Instructions, Procedures which
are incomprehensible to all except those who have written them! Operational
definitions are necessary for achieving clarity in communication.
· Importance and use of Operational Definitions.
Deming
regards understanding and application of operational definitions as of supreme
importance. The Japanese paid great attention to use of operational definitions
and Deming realised that the benefits obtained were comparable to the benefits obtained by use of
concepts and tools of Statistical Process Control.
Shewhart
believed his work on Operational definitions to be of greater importance than
his work on the Theory of variation and Control Charts.
It
will be appreciated that use of operational definitions has a great deal to do
with reducing variation. After all if there is more clarity in defining and
communicating the needs of internal and external customers, variability is
bound to reduce.
Examples of Operational
Definitions:
·
Clean the table--clean enough to eat on
or eat off or to sell or to operate on?
We need to specify to make it an operational definition.
·
Satisfactory? – For what? To whom? What
test shall we apply?
·
Careful, correct,
attached, tested, level, secure, complete, uniform – all need
operational
definitions.
·
Definition of sales
and accidents often change due to
management pressure.
·
Zero Defects – What is a defect?
e.g. What is a surface with no cracks? What is a crack? Do cracks too small to
be seen with the naked eye need to be counted? Are these to be detected by a
magnifying glass? What magnification?
· Related Concept: "There is no true value of
anything." If so, what is there?
There is a number that we get by
carrying out a procedure--a procedure which needs to be operationally defined.
If we replace this by another procedure (also operationally defined ) we are
likely to get another number. Neither is right or wrong. If the procedure is not
operationally defined we are likely to get different numbers even with the same
procedure!
For example:
a) Percent iron content of iron ore mined by Yawata Steel Co.---
Old method : by scooping
samples off the top of trucks.
New Method: by taking samples off the
conveyor belt.
Neither method is right or
wrong. The question is--does either serve the purpose better? If so, use it.
b) Long term average number of red beads attained in the red beads
experiment.
Contrary to what one would expect, this figure is not directly related
to the proportion of red beads. Hundreds of usages of two different paddles
have produced averages of 9.4 and 11.3 respectively! The clear answer must be
10 as per the law of large numbers since the ratio of white beads to red beads
is 4:1. However we must appreciate that this would be true if sampling were
done by random numbers. But this is mechanical sampling--a different method.
c) Value of
(circumference / Diameter ).
Mathematical theory says it is
an irrational constant with the first few figures being 3.14159265……But the
conditions under which this is true are--
· The measurement method
must be of infinite accuracy for both the straight line and the circle.
· The circle must be a
perfect circle.
· Both the lines
(circumference as well as diameter) must be of zero thickness.
In practice the above conditions are not satisfied. So however we
define and measure
C/D we will never obtain the value
. e.g. if we measure up to 3
places of decimel--
C=6.237cms, D=1.985cms and C/D= 3.142065. The procedure must be defined,
there is no right or wrong.
The
relevant question is not whether an operational definition is right or wrong?
The question is does it do what we want it to do?
Chapter 4: Theory of Improvement
4.1 Introduction to the
The first
person to come
up with the
concept of a
cyclic nature of
operations was Walter
Shewhart . Although he came
up with the concept in
the late twenties , the first
recorded explanation of
this cycle was
not until 1939
in a series
of lectures which
he delivered at
the University of
New York . This series
of lectures was then
converted into a
book called “ Statistical methods
from a viewpoint
of Quality Control “ . This book
was edited by
One can see from
the figure that
Shewhart was suggesting
the turning of
the wheel to
facilitate improvements . The origins
of this wheel
can be found
in his book “ Economic Control
of Quality of
Manufactured Product “ which
he published in
1931 . Here he explained
how the Industrial
Revolution brought about
a change in
the relationship between
the manufacturer and
the consumer . Previously the
craftsman was in
touch with his / er
customers directly . S / he would
find out what
the customer wanted
and would go
ahead and make
it . The Industrial Revolution
created barriers between
the craftsman and
the Customer . There now
existed a "Chain" of craftsmen
that delivered the product
to the customer . Since the
Craftsman now lost
"direct" contact with
the Customer , there were
no longer "customised" products . Now items were mass – produced :
i.e. they were produced
in large numbers .
In order
to determine how
to meet the
general needs of
a number of
customers Shewhart suggested
the use of
the above cycle
by defining the
three parts of
the cycle thus :
Specification : A commitment
that has to be met
to satisfy requirements
Production : An effort
that is carried
out to meet
these requirements .
Inspection : An act
carried out to
assess the effectiveness
of the efforts
to meet these
requirements .
Shewhart called
this " The
Scientific Process of
acquiring Knowledge " .
He suggested the
use of Statistical
Methods to " break " the
barriers between the
manufacturer and the
customer . By the use
of Statistical Methods , one
could determine the
" average " needs of
the customers and
could hence be " ensured " of
meeting the needs
of a number
of customers . Also Statistical
Methods could be
applied at the
three stages in
the cycle to
improve the operations
within the company
thus : After going through
the cycle once , whatever learned
from the inspection
process – shortfalls / advantages
of the production and
specification processes – could be
used as a
starting point of
the second journey
through the cycle .
That is , the
specifications could be
found wanting from
a customer point
of view which
could be improved
further , alternatively , the
production process could
be found wanting
in some respects
as regards meeting
specifications - which could be
then corrected , and so
the cycle would
continue to rotate , every journey
through the cycle
making the outcomes
better and better , and
also giving the
manufacturer better and
better insights into
his manufacturing processes , his customers
and ultimately his
organisation .
Dr.
Deming understood exactly what
it was Shewhart
was trying to say
. When he got
an opportunity to
deliver lectures to
the Japanese on
Quality in 1950 , his
lectures carried a
lot of concepts
that were introduced
by Shewhart . The
lectures that Dr. Deming
delivered to the
Japanese were converted
into a book
called " Elementary
Principles of the
Statistical Control of Quality
" which was
published in 1950
and re - printed in
1952 .
Here
one can see the origins of the Deming Wheel as we know it today . Dr.
Deming was trying
to get the
idea of the
cyclic nature of
operations to the
Japanese . While doing so
he mentioned how
products used to
be designed before
the Industrial Revolution
and how the
same concept were
used even after
the Industrial Revolution .
He also
mentioned as to
how the “ OLD WAY “
was flawed in
this that the
manufacturers did not
pay any attention
to the Customer . He
went on to
explain the “ NEW WAY “
of providing Customer
satisfaction .
This idea
of Dr. Deming is
shown in the
figures above . Here the
difference between the
old way and
the new way
is evident – especially how
Dr. Deming has explained
Shewhart’s ideas .
The courses
that Dr. Deming conducted
for the Engineers
and Scientists in
Japan were 8 – day
courses . He gave two
different versions of
the cycle then . It is
interesting to note
how he made
the distinction between
the steps to
be taken on
the shop floor
and the broad
steps to be
taken by management . See the
insistence by Dr. Deming
to continue Research
continually and place
importance on Quality
of the Product - the
message was the
same to the
Engineers and the
Executives .
He basically
expanded on THE
NEW WAY of
manufacturing making the
steps in the
cycle more elaborate . He further
went on to
describe what he
called "THE BETTER WAY” . This
concept of a
"Helix" or "Spiral" of
Improvement was perhaps
the first ever
recorded . There are many
who claimed that
the first to
come up with
the concept of
the Spiral of
Quality was Dr. Juran . Dr. Juran did
not put forth
the spiral concept
of Quality till
1974 .
The steps
1,2,3,4 are the
same from THE
NEW WAY . The radius
of the helix
which goes on
increasing in size
indicates the improvement
in the product
due to an
increase in knowledge
of the process , materials , etc.
In his
book “Out of the Crisis” ,
He
called it the
Shewhart Cycle but also clarified
that this what
was the Japanese
called the Deming
Cycle . Although he did
not use the
acronym PDCA , the steps
in the cycle
do indicate that
it is a PLAN
- DO - CHECK - ACT cycle .
In 1990 , Dr. Deming professed
what he called - A
System of Profound
Knowledge . This was
the culmination of
his philosophy . As a
result of this , the
Deming Wheel further
underwent a change
and the word "CHECK" was
replaced with the
word "STUDY" . Dr. Deming
emphasised that you
need to "Study" the
results - in checking you
might miss something .
In
his last book " The New
Economics for Industry , Government and
Education " the Deming
Wheel appears as
shown .
As seen in the
figure , he has altered
the wordings of
the Cycle shown
above and made
it sound simple . He
has called it
the PDSA cycle
in this book . It
should be noted
now that the
cycle is no
longer called the
PDCA loop but
the PDSA loop . This
loop has become
the underlying philosophy
of the QS
9000 standard . Many have
gone on to
add different steps
under the PLAN , DO , STUDY and
ACT phase , but the
basic notion of
a cyclic nature
of operations has
remained the same .
4.2 The Origins of the Theory of Improvement
The
origins of this thinking go back to the early twentieth century . Walter Shewhart
was a great
follower of the
philosopher and epistemologist , Clarence
Irwing Lewis . Epistemology deals
with the science
of learning and
acquiring knowledge . Lewis had
written a book
in 1929 called “ Mind and
the World – Order - An
outline of the
Theory of Knowledge “ . In this
book he had
mentioned that we
must have a
theory to begin
with when we
want to acquire
knowledge . A theory could
be a hunch , a
set of principles , a set
of laws , etc . Actual testing
of the theory
and recordings of
the observations could
make us improve
the theory , change the
theory or , even abandon
the theory .
No theory is wrong – only effective
or ineffective . Every theory is good in it’s own world , but may be ineffective in another . Let us take an example
to explain this .
There
is a child aged 6 years who has been watching his father go to work everyday .
The father goes to work on a motorcycle . The child has seen something that
happens every morning and has drawn a conclusion – “My father kicks the lever
on the right hand side very hard ,and the motorcycle revs to life“ . His theory
is very simple, you just have to kick
the lever very hard to make the motorcycle start. One fine day , when no
one was at home, he decided to “start“
the vehicle himself . He climbed on the motorcycle and kicked on the
lever hard, but nothing happened . He kicked again, but nothing happened again
. Obviously, his theory was ineffective . There must have been something he
missed out . So, the next time he observed his father closely as he started his
motorcycle. He found that he did miss out something! His father turned a key on
top before kicking at the lever! So that was it ! His theory changed – now he realised
that one had to open a “lock“ before he kicked the lever. Now if he did not have a theory – he would not have anything new to
learn, he would not have had any questions .
So
now with his revised theory, he tried to start the motorcycle himself and
succeeded. Some time later, he saw his
father kicking away at the lever, but the motorcycle did not start . He then
heard his father tell his mother that he would take the bike to the neighbours'
to see if they could help out. The boy accompanied
his father to the neighbours house. The neighbour was an elderly gentleman who
listened to the boy's father very attentively . He then took the bike , ran
with it for some distance , then jumped and sat on it and lo and behold ! – the
bike started ! The boy was pretty confused by now . His
revised theory about the “key" and
"kick" needed some revision.
He asked his father "what did he do that yo weren't doing ?" . The father told the
boy that "he would understand when he grew up !"
When
the boy grew a little older , he learnt about
Soon
he grew up and inherited his father's motorcycle . Whenever he faced a problem of starting , he would just run with the bike and start it
! One fine day , he faced a starting problem again , only this time , even when
he ran with the bike , i did not start .
So , he approached the mechanic who had put up shop a few blocks away and gave
him the following explanation – “I tried to run with the bike and start it in
motion , but it wouldn’t start .“ . To which the mechanic just bent down and
put his hand under the chassis and removed a contraption which he called the
“Spark Plug“ . He then went on to show the boy “how dirty it was“ , cleaned the
tip and put the plug back in it’s place under the chassis . Now when the mechanic
kicked the lever the bike started . It was then that it dawned on the boy that
when he kicked at the lever or ran with
the bike , a spark was formed in the gap in the Spark Plug which then ignited
the fuel and started the bike . ( this explanation will make hard boiled
engineers laugh )
Now all
through his stages
of growth , the boy
lived in different
worlds , and , he had a
theory that was
useful until it
was replaced by a better
theory that “ explained”
things better and
more importantly – helped him
predict the outcome
of actions . So we
can say that
a theory is -
“A
sentence that relates
cause with effect ; fits
without fail all
observations of the
past , and , helps us predict
the future for
a similar set
of causes , with the
risk of being
wrong”
§
Information , though
easily available to
everyone , is not knowledge .
§
Theory is a
statement that relates
cause to effect
and helps us
predict the future .
§
Interpreting
information with the
aid of theory
leads to knowledge .
§
No theory is
wrong - just adequate or
inadequate .
Thus theory
has temporal spread . Theory is
a window into
the world . Without theory
there are no
questions to ask . Without
theory we will
have nothing new
to learn . All knowledge
advances theory by
theory . This is
the foundation of
the PDSA cycle .
To
sum up , the steps
in understanding a “ Theory
of Knowledge “ are :
·
Ask Questions
·
Formulate Theories
·
Carry out Experiments
·
Observe
·
Confirm Theories
·
Verify Theories
·
Act on any
differences
·
Make improvements if
necessary
·
Repeat the cycle
A pictorial
representation is shown
below .
Applied
science is more exacting than pure science . In other words , a scientist can
carry out experiments in a laboratory under controlled conditions - An
engineer or Manager on the shop floor does not have this liberty . By using a
theory of knowledge however , the engineer / manager can learn while carrying
out their daily activities . Thus the
job of a Manager nowadays can be defined as follows
The System ( Organisation ) consists
of People and Resources . The Job of a Manager is to work on the System – To improve
the system constantly with the help of the people in the system applying a
theory of knowledge .
4.3
Applications of the Theory of Knowledge
In order to encourage process thinking, breaking down barriers between
departments is very important. A practical way of doing this is for people of
different departments to get together and list out the inputs and outputs of
the process – this itself can lead to a much better understanding of the
process .
Inputs are a combination of a number of factors , these are inclusive
of but not limited to :
·
People
·
Machines
·
Methods
·
Materials
·
Measurements
·
Environment
The flow
diagram shall indicate inputs and outputs as a combination of these factors . A
typical flow diagram will look like one shown below
A Flow Chart is helpful in understanding a system and this
understanding is necessary to trace the consequences of a proposed change. The
very exercise of constructing a valid flow chart leads straightaway to some
considerable improvement. ( Valid here means what actually happens, not what is
supposed to happen. )
· Important points about a system:
a)
A system is unlikely to be well defined in practice unless
it is both suitable and adequate for the job intended and is written down in a
way comprehensible to all involved.
b)
Whether in the form of flow charts or textual a system does
need to be documented to indicate what actually happens.
c)
If a system cannot be written down, it probably does not
exist.
d)
The greater the complexity, the greater the indications of
trouble – there is a need for simplification.
Applying the theory of improvement to this
diagram we can get a figure as shown below :
Putting the Theory of Improvement
in Practice
The
following set of steps can be used to put the theory of improvement in practice
:
PLAN
Develop an
Improvement Plan / Projects as a part
of the Business
Plan focussed towards the
Customer
Step 1 : Identify
and Prioritise the
Step 2 : Document
the Present Process ( Include Process
Mapping )
Step 3: Create a
vision of the
Improved Process
Step 4 : Define
the Scope of
the Improvement Effort
DO
Carry out
the Improvement Plan
Step 5 : Pilot
Proposed Changes
STUDY
Study and
evaluate Results
Step 6 : Observe
what was learnt
about the Improved
Process
ACT
Adjust or
change Process Based
Knowledge gained
Step 7 : Institutionalise / Operationalise the
new mix of
resources
Step 8 : Repeat
Cycle / Steps if Improvement
not sustained / for new
Projects
Extending
this to the entire business of the organisation , this theory will now look
like this :
A further manifestation of the
theory of improvement can be used to approach problem solving :
Danger of sub-optimisation
·
There is interdependence between component sub-processes of
a system. Management of a system therefore requires knowledge of the
interrelationships between all the components within the system and of
everybody that works in it. The greater the interdependence, the greater is the
need for communication and co-operation between the components.
·
If the above is not taken into account and attempts are made
to improve the system (by alteration of procedures) without considering the
ramifications on all the outputs it may be a case where "we are being
ruined by best efforts" by local sub-optimisation. This optimisation of a
sub-process is often incompatible with optimisation of the system as a whole.
Management
action required to avoid the above danger.
- All activities
should be co-ordinated to optimise the whole system.
- Performance of any
component sub-process should be evaluated in terms of it's contribution to
the aim of the system, not for it's individual production or profit, nor
for any other competitive measure.
- Optimisation of a
system may be incompatible with optimisation of a larger system. A further
responsibility of management, therefore, is to look for opportunities to
widen the boundaries of their system for the purpose of better service and
profit i.e. win-win.
· Deming gives supreme
importance to both internal and external customers.
This is clear from some of his
quotes given below:
a)
"The consumer is the most important part of the
production line. Quality should be aimed at the needs of the consumer--present
and future."
b)
"The consumer is more important than raw material. It
is usually easier to replace the supplier of raw material than it is to find a
new consumer."
c)
"People on a job are often handicapped by inherited
defects and mistakes."
d) "A problem in any
operation may affect all that happens thereafter."
· Practical gains can be
made by sorting out problems in the upstream system and thus preventing
occurrence of downstream difficulties rather than trying to manage results.
Chapter 5: Joy in Work , Innovation and Co –
operation ( Win – Win )
5.1 Introduction
It is important to understand
·
The need for innovation of product and process in addition
to continuous improvement
·
The importance of joy in work in bringing about such
improvement and innovation
·
The need for creating a climate where joy in work,
continuous improvement and innovation can flourish
·
How such a climate can be created?
Deming
accorded a lot of importance to the above aspects in later years as is clear
from some of his quotations given below :
"Management's overall aim should be to create a system in which
everybody may take joy in work." ( At Denver in August 1988)
"Management's job is to create an environment where everybody may
take joy in work". (In the T.V. documentary 'Doctor's Orders'.)
"Why are we here? We are here to come alive, to have fun, to have
joy in work." (at the start of many of his Seminars.)
Chances
of joy in work are destroyed by faulty practices of management such as
performance appraisal, M.B.O and arbitrary numerical targets. These practices
deliberately introduce conflict, competition and fear which are the direct
opposite of Co-operation ( win-win ).
Important
aspects to keep in mind about Deming's controversial points concerning these
practices.
Abolition
of performance appraisal, fear, arbitrary targets and mass inspection are natural consequences of the development of
Deming's Principles and will be good
only in the appropriate environment. In the wrong environment they can do more harm than good.
For
example: Mass inspection becomes redundant when processes are brought into
control, moved much nearer to partnership ( than conflict ) and an environment
of continual improvement established. It would be crazy to abolish mass
inspection if this environment has not been created.
- Why are these practices so common?
The answer is that they are all examples
of making the best of a bad job. When the working environment is bad such
practices do (at least on the surface) make things less bad. The concept of joy
in work is irrelevant (even ridiculous) in this context. But Deming is
concerned with the far sighted objective of turning the bad job into a very
good one! And joy in work plays a large part indeed in this context!
- How on earth do we motivate people if appraisals,
fear, targets, incentives, threats and
exhortations are removed?
Deming's
answer to the above question: If management stopped de-motivating people they
would not have to worry about motivating them! If management are successful in
their job of enabling, engendering and encouraging joy in work there would be
no need for motivation. People with joy in work are fuelled by intrinsic ( rather
than extrinsic ) motivation – they become able willing and enthusiastic to
contribute to the 4 prongs of Quality.
Deming
suggests that only ~2% of Management and ~ 10% of shop floor people experience
joy in work – if this figure can be increased to, say, 25% or 50% it would mean
a dramatic change and lead to a transformation, which is what Deming is
talking about!
- Innovation and Improvement – the four prongs of
Quality.
In
addition to continuous improvement, Deming attaches a lot of importance to Innovation. This is clear from some of
his quotations from 'Out of the Crisis':
"Statistical Control opened the way to innovation."
"One requirement for innovation is faith that there will be a
future. Innovation, the foundation of the future, cannot thrive unless Top
Management have declared their unshakeable commitment to Quality and
productivity".
Deming
refers to the four prongs of Quality as :
a) Innovation in products
and services.
b) Innovation in process.
c)
Improvement of existing products and services.
d) Improvement of existing
process.
The above is the order of importance. The order of application has to
be the reverse – it is dangerous to invent new processes / products when the
present ones are behaving badly / unreliably and we do not know why?
- Where does innovation come from?
It
does not come from the customer. "How would he know?" The customer
does fine prior to the to the innovation. People can't miss what they do not
know about. But when good innovation does become available, the customer
discovers that he needs it! (e.g. word processors, microwave oven, felt tip
pens, synthetic fibres).
We
need to stay ahead of the customer. Innovation comes from the producer who
looks ahead to find answers to the questions:--What new products or services
will help the customer, be attractive to him, entice him, whet his appetite?
- How does Management create a climate where
innovation flourishes?
Deming's
answer :
"Concentrate on generating joy in work and the power of intrinsic
motivation will lead to innovation, thus enabling the organisation to grow and
prosper."
The
opportunity for innovation and the environment for it to flourish has to be
nurtured by Management. That is why Deming frequently says :
"Quality
is made in the Board Room".
"We have smothered innovation by
patchwork, appraisal, putting people into slots, rugged individualism",
Says
Deming.
- Why should people do a good job instead of only
time-serving and getting away with the minimum they can?
Three possible reasons could be :
a) fear
b) financial incentive or c) they
want to
d)
will obviously be the most effective. This answer emphasises the importance of
joy in work
5.2. Backbone of the New
Philosophy
Co-operation
(win-win) as opposed to Competition (win-lose).
·
Deming has brought out the pivotal role of co-operation in
his following statement in one of his Seminars in
"The very purpose of the 14 points is to help carry out
co-operation, to create joy in work."
·
Deming's description of the needed transformation as a
"new system of reward" in his following quote ( from a paper read at
"The transformation will be a new system
of reward. The aim will be to unleash the power of human resource contained in intrinsic
motivation. In place of competition for high rating, high grades, to be number
one, there will be co-operation between people, divisions, companies,
governments and countries. The result will, in time, be greater innovation,
science, applied science, technology, expanded market, greater material reward
for everyone. There will be joy in work, joy in learning. Anyone who enjoys his
work is a pleasure to work with. Everyone will win, no losers."
5.3. Basis of
transformation
Conversion from the old economics
based on conflict and competition ( I win – you lose ) to a new economics based
on co-operation ( win-win ).
This
means :
·
Conversion from the mistaken
belief that competition is inherently good for everyone to the realisation that
working together for mutual benefit and benefit of society at large has far
greater potential.
- Conversion from the current society in
which both cause and effect of there being winners is that there must be
losers to a future society in which there need be no losers and neither
are any desired.
This
is Total Transformation.
5.4.
What will such transformation do?
- Open the door for
optimisation rather than sub-optimisation-- to realise the full potential
of the system.
- Make possible the
changes required in climate i.e. joy in work, innovation and co-operation.
5.5.
Meaning of co-operation?
· It does not mean
mandatory co-operation or taking money from one wallet and giving to another.
It does not mean mere patience or forbearance.
For example if our flight is delayed and we are held up in the gate area
for hours, we may eventually be thanked for our co-operation. But: "That's
not co-operation. What can prisoners do?"
· The word is sometimes
used with evil connotations – for collusion, for dividing up markets to the
disadvantage of customers. This is not what we are talking about.
· It is win - win as
opposed to win - lose. This does take more time and energy but is superior,
more so when there is a common goal. The gains are much more than winning at
the expense of the other.
· Win - win style implies
building trust, gaining commitment and managing opposition.
· Winning in the
collaborative ( win – win ) style means fulfilling your needs consistent with
your beliefs and values, finding out the real needs of the other side and
showing them how to get the same while you get what you want.
5.6.
Why is competition regarded as good and important for progress?
·
In the absence of a system of co-operation (i.e. adoption of
co-operation as a Theory, a Principle and an objective) a system of competition
is appealing and is seen as effective.
·
A system of competition is easy to create and easy to manage
(e.g. lowest tender, performance appraisal, bonuses and incentive plans, man of
the month schemes, M.B.O, arbitrary numerical targets, grading and ranking disregarding
knowledge from the experiment on red beads.)
·
In management it is often easier to develop a strategy whose
direct aim is to choke a competitor rather than create improvements for the
benefit of customers and society as a whole..
·
The current apparent need for competition, rather than being
an automatic 'fact of life', indicates the large scale of the transformation
needed. It makes the best of a bad job whereas the real task is to transform
the bad job into a good one so that the bad practices become redundant.
·
Confusion arises because 'competition' and 'improvement' are
often used synonymously---& consequently competition is considered as
'good'. This is not correct because competition only focuses attention on the
need for improvement. Real improvement comes about only by a change in the way
the organisation is managed, a quantum jump to a better system, learning the
alternative system and work towards improvement each day.
5.7.
Meaning of competition – competing against vs competing with.
· The word competition is
used in the following senses:-
a)
Win - lose as in a contest: we may call this competing
against.
b)
Existence of alternatives.
c)
Competitive meaning as good as.
d)
Competing with e.g. as in a game of basketball with friends
in which both sides enjoy developing their skills and players even change sides
after each game.
When used in sense a), meanings b) and c) are often implied.
·
Competing against is very different from competing with as
will be clear from a comparison of the characteristics of each of the two:-
|
Competing against |
Competing with |
|
Stress on winning |
Stress on having fun,
improving skills,
acquiring
mastery—winning / losing is unimportant.
|
|
Focus on other team /
person / group |
Focus on improvement
of own work,
skills teamwork. |
|
Little room for
co-operation |
Co-operation an
integral part of process. |
|
Fear of losing |
No such fear |
|
Fight for market share
– be no.1 |
Focus on customer –
improve standard of living. |
|
Judge own performance
and compare |
Continual improvement
of product, process with other people and systems. Open to new ideas. Does
not ignore what others do but focus is not on beating them. |
'Competing
with' will give you superior product or service but it is wrong to call this a
benefit of competition as understood in the usual sense ( i.e. competing
against ).
5.8. Harmful effects of
competition – co-operation vs. competition
·
Business / Industry is essentially a co-operative situation.
A competitive reward system, merit rating etc constitute an 'imposed structure'
which changes this into a forced competitive situation. This causes conflict
and destroys the desire to help. It creates an artificial scarcity of winners,
no matter how excellent the people involved.
·
It is the 'system of beliefs' which determines how people
act, not the way they are rewarded externally. For example if one holds the
belief that success is not outscoring anyone else but the peace of mind
obtained by doing one's best, intrinsic motivation will drive him towards
excellence in work. Co-operation encourages such intrinsic motivation while
competition kills it.
·
Competition between departments leads to maximising results
of each department which in turn means sub-optimisation due to the reality of
interdependency of departments. For maximising results of the organisation as a
whole each department should spend time and effort to ensure that the product
or service is optimal for the next (i.e. it's internal customer). This can only
be achieved by co-operation.
·
In a competitive environment, greater effort may only mean a
greater attempt to get credit – instead of getting together, people will fight.
If no one cares who gets the credit, amazing results can be achieved. This is
only possible through co-operation.
·
Win - lose implies much destruction. By definition one
person / group's 'gain' is at the expense of another person / group i.e. the
other person / group suffers a loss. The destruction is actually much greater
because the struggle to win in the short term (for survival) uses up resources
which could otherwise be used for long term advancement.
5.9.
What should Management do?
·
Learn to practice co-operation as a system, as a principle,
as a strategy and as an objective – instead of the easy management system of
competition. In the words of
"One of the first steps in the transformation will be to learn
about co-operation: Why, What and How?"
What the above implies is that co-operation should not be incidental or
an accident: it should be the aim – the new management style!
·
Provide the type of leadership which nurtures teamwork. What
does lack of teamwork mean in practice? This is very aptly described in
·
Remove the various
manifestations of the win-lose philosophy which hold us back from co-operation
(Win-Win)and are barriers to real progress viz.
·
Merit system.
·
Performance Appraisal constrained by a forced
distribution, where it becomes formally necessary to put somebody else under in
order to get a higher rating. This creates an artificial scarcity of 'winners'.
This is unavoidable in a game of tennis where somebody wins, somebody else
loses--but does life in general and work in particular have to be played like a
game of tennis? Surely a better way is to have everybody working for the
Organisation rather than against each other!
·
Managing departments in
a system of competition. In an inner competitive environment, what matters the
trouble you may cause to other departments? (Indeed it can actually be
advantageous!) The tremendous harm that this causes is clear when we appreciate
that every department is a supplier or customer or both to every other
department.
As
the industrial world expands, considerations of efficiency & economics move
us to greater and greater interdependence. It is better for this
interdependence to be based on co-operation and partnership rather than
conflict and distrust.
Co-operate
with competitors for mutual benefit.
"Compete? Yes, but
in the framework of co-operation first. Everybody wins."
Examples
of such co-operation :
·
Possible areas of co-operation could be – increasing the market,
improving industry image, using interchangeable parts, agreeing on standards,
working with common vendors and improving the processes for producing products
and services.
·
2 service stations on same corner, taking turns to stay open
late night – both make greater profit and it benefits customers also!
·
Deming's car refused to start. The garage-man arrived ( to
tow away the car for repairs ) driving his competitor's pick-up truck – two
competitors had just one truck, resulting in considerable savings to both.
5.10. Immense financial
loss due to competition and conflict
Harm
that is caused by internal competition and conflict, the fear that is thereby
generated and the good that is brought about by internal co-operation and
team-work, is of massive proportions. For example :
·
A purchasing manager, under pressure to reduce his figures,
changes to a cheaper source, even if he buys poorer products and service as a
result.
·
Engineering design imposes unnecessary tight tolerances to
compensate for the fact that manufacturing never reaches the standards asked of
it.
·
Departments performing better than budget start spending
near the end of the year because they know that otherwise their next year's
budget will be reduced.
·
Towards the end of the month, salesmen start doing
everything to meet their quotas, with scant regard for the problems caused to
Manufacturing, Administration and Delivery, let alone the customer.
5.11. Huge Financial Advantages of Co-operation: An
example
Consider an organisation with just 3 areas or departments. In the left
hand column of the following tables are listed options available for each area
to adopt or not to adopt, according to their choice. The remaining columns show
the potential effects of adoption of the options. The effect of each option on
the individual areas and hence (by summation) on the whole organisation is
examined. For the sake of ease of
illustration it is assumed that each + or -corresponds to gain or loss of the
same amount of money
throughout.
Options which are locally beneficial to one area may well be quite the opposite
for other areas-- suppose each areas gains and losses are as indicated.
Now
let us examine what would be the likely gains and losses under 3 different
scenarios :
Scenario 1.
Internal
competition and conflict between Departments / Areas, hierarchical management
style, MBO etc.
In
such an environment there are barriers between areas and each area naturally
adopts options which are beneficial to itself. The likely gains and losses to
different areas and the net effect on the organisation are tabulated below:
The
net effect on the organisation happens to be zero in this case, i.e. equal to
the effect of doing nothing at all (which would be a rather easier way of
achieving the same result!) Depending on the details, the net effect could have
been positive, Zero or negative.
|
Areas
and their options |
Effects of Options |
|||
|
Effect
on Area A |
Effect
on Area B |
Effect
on Area C |
Net Effect on the Company |
|
|
Area
A |
|
|
|
|
|
i |
+ |
- |
- |
- |
|
ii |
+ |
- |
+ |
+ |
|
iii |
+ |
- |
- |
- |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Area
B |
|
|
|
|
|
i |
- |
+ |
- |
- |
|
ii |
+ |
+ |
- |
+ |
|
|
|
|
|
|
|
|
|
|
|
|
|
Area
C |
|
|
|
|
|
i |
+ |
+ |
+ |
+++ |
|
ii |
- |
- |
+ |
- |
|
iii |
- |
- |
+ |
- |
|
|
|
|
|
|
|
|
|
|
|
|
|
Net Effect of Adopted Options |
++ |
-- |
0 |
0 |
Scenario 2.
Management environment improves so that areas become aware
of the effect of their actions on other areas and the old inter-departmental
rivalries
are replaced by teamwork for mutual
benefit and benefit of the organisation.
In this changed
environment only those options are adopted which produce net benefit to the
organisation. Hence only 3 of the 8
options are now adopted and the likely gains and losses would be as tabulated
below :
|
Areas
and their options |
Effects of Options |
|||
|
Effect
on Area A |
Effect
on Area B |
Effect
on Area C |
Net Effect on the Company |
|
|
Area
A |
|
|
|
|
|
i |
+ |
- |
- |
- |
|
ii |
+ |
- |
+ |
+ |
|
iii |
+ |
- |
- |
- |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Area
B |
|
|
|
|
|
i |
- |
+ |
- |
- |
|
ii |
+ |
+ |
- |
+ |
|
|
|
|
|
|
|
|
|
|
|
|
|
Area
C |
|
|
|
|
|
i |
+ |
+ |
+ |
+++ |
|
ii |
- |
- |
+ |
- |
|
iii |
- |
- |
+ |
- |
|
|
|
|
|
|
|
|
|
|
|
|
|
Net Effect of Adopted Options |
+++ |
+ |
+ |
+++++ |
Thus, by this more
judicious and restricted choice of actions (i.e wise choice of both action and
inaction) the organisation is much better off and in the new management environment everybody gains.
Scenario 3.
The environment improves further
and options which never saw the light of day previously are now considered. These
are options which are locally disadvantageous to the area which can adopt them
but which give benefit to other areas.
Amongst
this greater range of options, again the ones to be adopted or nor adopted are
now chosen according (respectively) to whether they are or are not of net
benefit to the whole organisation. The likely gains and losses in this
situation are tabulated below :
|
Areas
and their options |
Effects of Options |
|||
|
Effect
on Area A |
Effect
on Area B |
Effect
on Area C |
Net Effect on the Company |
|
|
Area
A |
|
|
|
|
|
i |
+ |
- |
- |
- |
|
ii |
+ |
- |
+ |
+ |
|
iii |
+ |
- |
- |
- |
|
iv |
- |
+ |
+ |
+ |
|
v |
- |
+ |
+ |
+ |
|
vi |
- |
- |
+ |
- |
|
Area
B |
|
|
|
|
|
i |
- |
+ |
- |
- |
|
ii |
+ |
+ |
- |
+ |
|
iii |
+ |
- |
+ |
+ |
|
iv |
+ |
- |
+ |
+ |
|
Area
C |
|
|
|
|
|
i |
+ |
+ |
+ |
+++ |
|
ii |
- |
- |
+ |
- |
|
iii |
- |
- |
+ |
- |
|
iv |
+ |
+ |
- |
+ |
|
v |
+ |
- |
- |
- |
|
Net Effect of Adopted Options |
++++ |
++ |
++++ |
+++++ +++++ |
The
bottom line results speak for themselves.
5.12. Analysis and advice
of
Deming
is of the opinion that scenario 1 describes most of the world. He mentions
a company which is divided into 4
companies, each supplying all the others. Each is rated on it's own profits!
Such foolish practice is common.
Deming
further summarises the lessons to be learnt from the above exercise as follows :
"Options for various companies to consider, or for managers of the
various divisions within a company, must be stated, and enumerated. Approximate
effects of action on each option stated, (yes, no or leave it alone) for every
staff area involved, must be computed.
A new set of options might turn out to be better for everybody.
Ingenuity is required to generate the options to consider. This is management's
job.
Divisions should not operate without recognition of their
interdependence. They will otherwise almost with certainty make decisions which
are far from optimum for the company as a whole, and everybody will lose."