February 2006
"Quality Focus" - monthly Genesis column
Arithmetic versus Statistics
Many think there's not much difference between the two; they do so at their peril...
Welcome to the second installment of this new monthly
column. Last month we looked at the concept of “quality” and asked some
pertinent questions. Before we proceed with attempting to provide the answers,
we will this month explore some common current “quality” practices within the
canmaking industry.
Tour any canmaking line worldwide – whether two or three-piece – and you will see a production facility replete with inspection systems. Usually – but not exclusively – these systems are examining the results of the manufacturing process, in essence assuring the canmaker does not ship defective product to fillers. Today’s systems are quite sophisticated, having the ability to inspect for all types of container anomalies.
When these systems operate as pass/fail sorters – as they often do - the canmaker is provided all forms of statistical analysis, right? Wrong. A lack of familiarity with Statistics has prodded some in the industry – canmakers and suppliers alike – to interchange Arithmetic and Statistics.
Is there really a difference; are we not just “splitting hairs here?” No, let us examine those differences to better understand them.
A very common measurement in a canmaking line is defect rates. This may encapsulate the entire line, any section thereof, or be pegged to some particular process or machine. It usually is a ratio of the defective cans produced versus the total throughput. This however, is simply division – a function of Arithmetic; it is not a statistic.
So in this case, what does constitute a statistic? Good question.
In a statistical analysis of a can defect, we are interested in a measurement of the severity of the defect, not simply whether or not it “exists.” This requires a way to actually measure the defect and store the magnitude of that defect for analysis.
Why is this important? Yet another good question.
By measuring the defect we may be armed with data that allows us to predict when and if the process in question will drift out of specification. Even more importantly – and this is key to controlling a process – that measurement may be used as an input to the next process in the line. This provides information on when this measurement will cause problems for the next process before we even measure that next process.
It gets even better: by applying true statistical methods we may predict the defect rate before ever making bad product – if the rate is low enough – and know when we are in trouble should the predicted rate rise.
So, how does a canmaker know whether Arithmetic or Statistics are actually being used in a plant? Here are some common indicators that Arithmetic rules:
The words “defect rate” or some form thereof are used to describe “quality.”
It is physically necessary to count the defects before knowing the magnitude of that rate.
There is no measurement data available from the inspection systems, only pass/fail Arithmetic.
Numbers that are not described include standard deviation, normality, capability – there is quite a long list.
There has been a defect existent in a line for weeks, months, maybe years that has yet to be resolved.
As an example remember that Statistical Process Control, which has been around a long time and is “used” frequently in our industry, is only possible when employing Statistics. Six Sigma and other more modern methodologies is totally impossible without Statistics.
Next month we will take a look at true Statistics and its applicability within the canmaking industry.
