Improving a Process

                    David and Sarah Kerridge

There is not one way to improve a process, but many. These are
not alternatives. Used with understanding, all contribute to the
continual improvement of every process, and the whole system.
Each makes other methods more effective, and so they should be
used together. To illustrate this, we concentrate on the
practical problems of using the Deming Cycle, and show how other
actions help it work.

     NB Deming called the cycle the Shewhart cycle: others call
     it the PDSA cycle, and the Japanese call it the Deming
     Wheel. It was certainly given its present form by Deming, so
     it seems fair to name it after him.

A unified approach

This combined and unified approach to improvement is typical of
the Deming Philosophy. Instead of learning one technique, and
applying it as much as we can, we take a system view.

   This fits in with the way we tackle anything complicated. An
automobile is simple compared to most processes. If we want it to
run well, we do not spend all our time on the electrics, and
ignore the fuel supply, or concentrate on the tyres and forget
the brakes. If there is a break-down, it probably affects just
one part, and naturally we find out which, and work on that
first. But for trouble-free motoring, we make sure that all the
essential parts are regularly serviced. We do not wait for
something to go wrong.

Seven ways to improve a process

We express these as a list of actions and questions. Any one may
produce dramatic improvement on its own. For example, improvement
in the measurement process, even though it does not directly
affect the process, may reduce tampering. More often it is the
interaction between these approaches that produces results. What
is more, we must see the investigation of this one process as
part of the transformation of the whole organisation.

   Without overall change, it is hard to improve an individual
process, and the improvement, even if we achieve it, seldom
lasts. But equally, working on a process can make some of the
ideas behind overall transformation more concrete, and fix them
in people's minds.

1  Study the customers' needs. Is the output of our process the
   most helpful that could be given to them? Is it causing
   problems in a later process? There is no point in improving a
   process until you know what a good result really means.

2  Flow-chart the process. Are there unnecessary stages, or
   examples of rules 2-4 of the funnel? Have you identified all
   the internal and external customers and suppliers? Do you
   listen to them?

3  Improve the training of the process operators. Introduce
   Operational Definitions.

4  Study ways to measure outputs and inputs. What measures are
   most relevant to the success of the process? Check that the
   measurement processes are under statistical control before
   attempting to use the measurements to study the process.

5  Reduce variability of the inputs. The inputs include every way
   in which the rest of the system affects the process. Can you
   reduce the numbers of internal or external suppliers to the
   process? Do the suppliers understand your process?

6  Plot the outputs and inputs on SPC charts. Remove special
   causes. Eliminate tampering.

7  Collect suggestions for improving the process, and test them
   using the Deming Cycle.

   There are more ways to improve a process, but these are enough
to make the point. The Deming Cycle relies on checking the
results of a change, using measurement. When the process itself
varies less, and measurements on it are more accurate, it is easy
to see the effect of a change. Besides which, the understanding
of the process which comes from all these different ways of
studying it will suggest changes that should be tried.

What should we do first?

For an individual process that has not been studied before, the
order given above is reasonably good. This does not mean that we
finish one before going on to the next: we usually do several at
the same time.

   Even if the process suffers from a major problem which must be
solved, do not neglect the general, systematic, approach. There
is a good reason for this. If the cause of the problem had been
obvious, such as something broken, it would have been put right
immediately. So we expect the investigation to take time.
Occasionally a problem disappears, still unexplained, as a part
of overall improvement. Most often the cause is easier to trace
when systematic improvement is under way.

   It is so natural to rely on trouble-shooting that we give some
examples of this. One process had consistently bad results over
many years. Every so often a trouble-shooting team went out from
head office, found a problem and fixed it: but things were soon
just as bad as before. Then control charts were plotted for
inputs and outputs, and the process improved without further
specific action. In another case, a long standing problem dis-
appeared after a change to a single supplier.

Applying the Deming Cycle

Because it relies on stability, and on good measurements, the
Deming Cycle is most effective when the other six approaches to
systematic improvement are under way. It is not a recipe, but a
system: in other words, it does not tell us what to do, but how
make what we choose to do systematic and effective.

We must first decide on which change we should try. This is part
of the "Plan" stage of PDSA. There will usually be plenty of
ideas to choose from, provided everyone understands that learning
about the process is more important than guessing the "right"
answer. So no-one is blamed for making the wrong guess. After
all, even if a change does make things worse, it will suggest
ways to make things better.

   We can only test one change at once, so if there are many
suggestions, we need ways to choose between them. Here are some
key questions:

1  Can it be tested on a small scale?

2  Will the effect be seen reasonably quickly?

3  Will the effect be easy to measure?

4  Does the test require new measurements, or will existing
   measurements be sufficient?

5  Has the measurement already been studied and shown to be
   stable?

6  Is the test simple to do?

7  Will it take long to do?

8  Can the test be done without disturbing the ordinary running
   of the process?

Obviously if the answer is "yes" to all these questions, the
change should be tried as soon as possible. We will seldom be so
fortunate, but the number of "yes" answers gives a crude order of
priority among different possible changes. Often practical
considerations, like the need to maintain the enthusiasm of the
team, will provide the deciding vote.

   There is, however, one general principle which might be
overlooked. In the long run, a change that reduces variation,
without making the average worse, is more desirable than one
which improves the average, leaving the variation as great as
before. This is because reduced variation makes other improve-
ments easier to find.

Be systematic

Once a choice has been made, make sure that you do not waste any
of the information from the experiment. Keep systematic records
of each stage. Do not rely on memory: it plays too many tricks.

   Using the Deming Cycle does not just help to improve the
process. It has great educational value for everyone who takes
part in it. It develops team-work. Ideas at the "Plan" stage will
often be wrong, and many cherished ideas disproved. Members of
the team will learn not to trust guesswork, but to use theory,
and yet rely on facts. They will also see the practical value of
theory, measurement and Operational Definitions. These things
will affect their whole attitude to transformation.