Imagine a machine producing beverage cans by drawing aluminum sheets in a punch and die machine from circular cross-sections. If all assignable causes are removed from the machine, if all clearances are within tolerance bands and if the design characteristics of the machine matches the actual capability the process demands and if the material used is within the specification bands, it is reasonable to expect that the process will follow a normal distribution. If the tolerances are made tighter and the process capability is improved, the distribution could be improved to even a 4 sigma to 5 sigma level of variation.
Suppose there are assignable causes in the process, or the material properties like the earing characteristics are not within the specified bands, the distribution would tend to become either right tailed (left skewed) or left tailed (right skewed) or fat tailed. When assignable causes are removed from the process, the process is said to be under statistical control. It is such a condition that allows experiments to be conducted or improvements to be done. Minus the condition, it is difficult to pinpoint a result to a specific action as it is difficult to establish causality.
For human performance to be normally distributed, it would mean no Bill Gates or Einstein in the population engaged to do Math problems in a class of 20 or no Michael Jordon or Magic Johnson engaged in a game of basketball with other minors. It would also mean that the team itself is not made of like the “team of rivals” as Abraham Lincoln constructed his A-team. These examples induce skewedness in the result and the performance can hardly be called normally distributed.
Normal distributions involving human beings have many examples, height or weight in a class could be few of those, but these have hardly anything to do with the performance. Typing errors by expert typists could follow a normal distribution, or dictation mistakes while writing in short hand or dialing a wrong number and the list could go on but these are hardly the ones which could be tabled for serious discussions on performance distribution and normality.
Performance is hardly done by any individual working in isolation; he is part of a group and has to collaborate with so many others. There are external factors at play; some are driven by macro-economic factors. There are training and development inputs that influence results, withdrawal of inputs could actually have a debilitating influence as well. We cannot ignore motivational issues; presence of a mentor in a group could make a tremendous impact.
The job itself can hardly be ignored and some jobs require special skills and preferences, someone on a wrong job with exceptional effort may not make an impression, while someone with the right fit with no effort could drive exceptional results, if the results are themselves crafted ordinarily or with limited challenge. So the target matters significantly.
Tails, whether left or right or fat, abound almost in every performance distribution; given the complexity, may be it is easier to assume that normality is the desirable state and easier to comprehend. So be it.
Leave a Reply
Be the First to Comment!