Specifications and Inspection; some history

At the end of the 18th century, an inventor worked on two projects that would have a profound affect on the tumultuous years to follow in America.

Eli Whitney was a blacksmith who surprised his neighbours and family by enrolling in Yale. He was unable to get work in his native north and headed south to Savannah to work for the wife of the Revolutionary War general Nathaniel Greene. She was staying at the plantation of Phineas Miller whom she intended to marry.

The plantation was not doing well because the only crop they could make money on was cotton and the only variety they could grow was a variety laden with seeds. The labor content of picking the cotton lint out from the seeds was high.

Whitney put his talents to the problem and invented the cotton gin. It was a simple machine that held the seeds behind a wire screen and picked out the lint by means of a rotating drum with hooks. It easily did the work of many men, and made the highly seeded variety of cotton profitable.

Yale did not prepare Whitney to deal with naked greed. The cotton growers of the region stole his invention and after years of turmoil trying to collect from the southern cotton growers, he was patentless and penniless and headed home to the North.

Once again he set his skills to work. This time he developed a rifle design that could be mass-produced. He designed the rifle using templates so that each part could be made separately and it would fit together with randomly selected other parts also manufactured separately to form a working rifle. Until that time each rifle was manufactured by hand and its parts could not be used in other rifles. They were custom made.

Whitney accepted an order to produce 10,000 rifles and commenced to make them. He ran badly behind schedule because his initial efforts encountered numerous problems, but he was able to complete a second order of 15,000 rifles on schedule. The birth of standardization had begun.

The idea was to produce separately, parts needed for a larger assembly. The separately produced parts when manufactured according to a ‘template’ would then fit easily and precisely into the larger assembly as either a new part or as a replacement. At least, as I said, that was the idea. As we all know only too well, the reality is frequently considerably different.

Whitney and his partner Simeon North successfully helped to establish the U. S. arsenals at Springfield, Mass (still the home of firearms manufacturers) and Harper’s Ferry, Va., and the era of interchangeable parts was underway.

As you can imagine, this road has been rocky with a myriad of technical problems that are re-visited with each product to which the idea is applied. The technical problems were, for the most part, overcome, but behind the method itself there were inherent problems more intractable than the technical irritants.

The use of interchangeable parts requires an understanding of statistical thinking that, at that time simply didn’t really exist. Pascal, LaPlace, Poincare, Fermet and others had developed certain statistical principles and theories of probability, but most of that effort was used in games of chance. Statistical thinking, Per Se, was unknown.

Why statistical thinking? Because at heart, the use of interchangeable parts rests on prediction. And prediction, since it is future oriented, is uncertain. Statistics is the tool for making decisions in the light of uncertainty. Thus, statistical thinking is what’s needed here.

Since the science or art (your choice) of statistical thinking had not been developed, the concentration of effort was in the areas of measurement and inspection. Obviously rifle barrels (for example) that are too big aren’t very accurate and those that are too small are dangerous. Since methods didn’t exist to make predictions with a high degree of certainty in manufacturing, usage depended on careful inspection.

Criteria were developed on which to base the inspections. These are called specifications today, and meeting specifications is thought by many to be synonymous with high quality. But if you think about it a bit differently, you can readily see that specifications are a compromise. They are a sign of inadequacy.

The best solution is to make each part identical to the one before it and the one after. Of course this is ‘mission impossible’. Some variation is inevitable. Specifications set the boundaries for acceptable variation. Parts with critical dimensions outside the specification cannot be used as is and must be scrapped or re-worked.

Thus, depending on specifications alone, we cannot be sure it will fit precisely and easily (our criteria, remember?) Already produced parts are examined to see if they meet the specification or not (inspection). Used in this way, specifications are a kind of ‘worst case’.

It is critical to understand this. I don’t think it’s controversial or subject to a lot of question, but it is not a widely held understanding.

Definitions of quality that are based on inspection are not adequate. Fitness for use and zero defects are a minimum set. No problem, then, you say, tighten the spec. But as we seen, that is not an answer. It only adds cost and doesn’t really change anything.

Dependence on inspection is the road to disaster. At best it is a compromise. And it’s never at it’s best. It’s too late to affect the quality of the parts produced. It is expensive. And it doesn’t work. It has been shown time and time again that 100% inspection is only about 80% effective.

And it’s even worse than that. Probably the most pernicious aspect of the inadequacy of inspection is that it leads to ‘corrective action’. As we saw with the funnel, such action is very often a costly mistake.

Certainly we want the advantage of interchangeable parts. Even if we accept the shortcomings of specifications and the inspections they give rise to, we are better off than trying to make 1000s of custom made automobiles. But is this a necessary compromise?

For the answer to that we have to go to the Bell Labs, Walter Shewhart and the explosion of the telephone.