W.E. Deming was a physicist but his fame and contribution came mostly from his work on process improvement. He was employed in the U.S. Census Bureau but he assisted in the post WWII raise of Japan with his thinking on quality manufacturing. I have found his writings somewhat difficult and obscure and I felt the same way during my only chance to see him actually perform a presentation.
But he did have some memorable ideas. One of the famous demonstrations he invented was the bowl of red and white beads. He had two wooden paddles, somewhat like small cutting boards with handles. Each had 100 small, shallow indentations in it. A participant was asked to insert a paddle into the bowl and allow the indentations to fill with beads. The white beads were "good" and the red beads were "bad". The participant was admonished and exhorted in a firm way before drawing the sample to avoid the red beads since they were errors, mistakes, boo-boos, costly, embarrassing, etc. Since there were red beads scattered throughout the bowl and the draw was random, there was no way to avoid the baddies. Some baddies were drawn and the culprit was frowned upon. Often, a 2nd participant would also make a draw and whoever had the least red beads was praised as a better or excellent worker while the person with more reddies was asked, "Why aren't you more like so-and-so?So-and-so only had xxx beads but you went and allowed yyyy errors? Why did you do that?"
It can sound funny, ridiculous but it is a model of what happens everyday all over the world.
Another memorable exercise is the clothespin and the bottle. You try to drop a clothespin into a bottle or jar. You stand and drop the clothespin or other object into the bottle or jar with it sitting on the floor at your feet. Some people try to notice which way the drop was off and correct the position of the bottle. But when the variations are simply random, re-positioning the bottle simply adds error to the process. The more corrections, the worse the alignment. If you have a good position, it is best to simply stay with it and weather the failures.
Any process with random components, and all processes have some random but maybe very small parts, is best not tampered with to try and improve it. Doing so can worsen performance.
