Insights on Excellence | "Insights on Excellence" Archive

Six Sigma quality: Just what is it, anyway?

by Stephen Hawley Martin
for Virginia Business
April 4, 2006

Most people in business today have heard about Six Sigma quality. You might say it's a current buzz word. But exactly what is it?

Six Sigma is a reference to the level of quality produced in a manufacturing process. Most traditional companies believe that 99.9 percent good quality is a terrific achievement. Perhaps by historical standards it is. However, consider what 99.9 percent good quality would mean in everyday life in the United States:

* Unsafe drinking water once a week.
* No electricity for nearly one hour per month.
* 500 wrong surgical procedures per week.
* Two short or long landings at most airports each week.
* 20,000 wrong drug prescriptions per year,
* 2,000 lost articles of mail per hour.

Perhaps 99.9 percent is not so good, after all. World-class companies ship products to their customers with 99.99966 percent good quality. From a statistical point of view, this means that they are shipping Six Sigma quality, no more than 3.4 parts per million defects. That's nearly zero.

The term "six sigma" refers to the number of standard deviations away from the mean (or average) point in a bell curve (also known as a "normal distribution").

If you're not a statistician, the bell curve is a natural phenomenon experienced in large populations of almost anything. Imagine, for example that you are harvesting corn. The size of most of the ears will be centered around the mean (average) of the population. A few ears will be moderately large and fewer still will be very large.

The same relationship appears in the smaller sizes, a moderate number are smaller than the mean, and a still smaller number are very small. If a million ears are harvested, only three to four ears will fall in the very largest category (six sigma from the mean) and only three to four ears will fall into the smallest category (six sigma from the mean in the other direction). The size of all the other ears will fall into a "normal distribution" as defined by the bell curve. This same relationship tends to hold for populations of people. If the height or weight of a large population is measured and plotted on a graph, the statistics will fall into the classical normal distribution.

In manufacturing, the naturally occurring variations in processes will also tend to fall into a normal distribution, for example, the dimensions of stamped or injected-molded parts, the thickness of plating or the amount of solder on a printed circuit board.

Achieving six sigma delivered quality to the customer is not an easy feat, especially considering the "rolled throughput yield" where the yields of each sequential processes are multiplied together to compute the final yield (the percentage of good parts produced by a given process). For example, if there are four processes, each with a 99 percent yield, the rolled throughput yield is (0.99) x (0.99) x (0.99) x (0.99) = 96%.

So how do world-class companies achieve such a small amount of defects delivered to the customer? They use a combination of methods to ensure that a defect is rarely passed on to the next stage of production. A book published by my company, "Lean Transformation: How to Change Your Business into a Lean Enterprise," covers most of them in detail.

-----------------------------------------------------

Stephen Hawley Martin is a former principal of The Martin Agency in Richmond and the author of more than half a dozen books including his newest, Lean Enterprise Leader: How to Get Things Done Without Doing It All Yourself. He is editor and publisher of The Oaklea Press, a book publishing business dedicated primarily to helping business executives increase productivity.