SPC on the rocks*

[David Kerridge <dfkerridge@mac.com>]

From time to time we apply SPC a bit loosely. It doesn't necessarily
do harm. It is not the calculation that is the problem, it is the way
we interpret the results.

We must be careful if we apply SPC to a process that *shouldn't*
be under statistical control. The number of messages to the DEN
seems to me to be an example. But I have seen many others in
applications to administration. Backlog of orders, staff illness,
and so on. No doubt DEN members can think of others.

As we know, the commonly taught idea that observations have to be
normally distributed before we can apply SPC is wrong. But they
do have to be independent. With some processes, of which messages
to the DEN are an example, one message stimulates another, so we
get a "thread" of messages on the same topic.

That ensures that the process is not stable in the short run. If it
should, for some period, be stable, that is the surprising thing,
not the existence of bursts of activity or inactivity.

I would not expect it to be stable in the long run either, for
deeper reasons. Few of us are completely free to apply the
Deming philosophy at any time. Only the happy few who
run their own business can be sure of that, or those who work
in a fully transformed company.

In times of economic stress, the long-term focus of the Deming
Philosophy is hard to maintain. Survival may be all that we can
think of. The basic principles may help us to achieve that survival
with the least long-term damage, but damage there will be.

I can think of many other reasons for long-term instability. This
doesn't make the analysis useless. We should expect instability
in most non-manufacturing processes, and the pattern of that
instability can help us understand what is going on. So I must
repeat, it is not wrong to do the calculation, but we must be
aware that we could draw wrong conclusions.

--
Best wishes

David

dfkerridge@mac.com