Sigma, σ, is a symbol in statistics to denote variation Most phenomena in the world are believed to follow, sooner or later, a pattern of behavior called Normal Distribution. I don’t subscribe to it but that is not the issue for this post. The mean or average value is denoted by the symbol μ (pronounced “mu”). If your process is at 1 σ level that means 32 out of every 100 items produced will be defective (their measured value will lie to the left of μ -1 σ or to the right of μ +1 σ). The curve curves down sharply and not gradually. It is called an exponentially decaying curve. Thus 2 σ will mean 5 out of 100, 3 σ will mean only 0.2 out of 100 defects.
The companies in question claimed that they operate at 6 σ level and they gave a byline that only 3.4 defects will be made in a million opportunities. That is impressive on the face of it. But set against the nature of the normal curve it strikes a discord. 6 σ can not be so lenient as to allow one to make so many mistakes. I painstakingly did all the maths stuff to calculate what 6 σ actually demands. I came up with a figure of 2 over a billion. What an ocean of difference between being allowed to make 3.4 defects per million to 2 defects per billion. Then why are these people making a false claim?
When you say deviation, it is implied that the deviation is with reference to a fixed specification. You are able to say that the train arrived 35 minutes late because there is a scheduled arrival time which is fixed. Suppose the Railways says that the scheduled arrival time can be anywhere between 10.15 AM and 10.55 AM. How do you compute the late arrival of the train when it arrives at 11.00 AM? Railways will claim that the train is only five minutes late because they will conveniently choose the 10.55 option. The great corporations have done exactly the same. They can not hold the process under control beyond a certain capability. So they said that our mean is allowed to make excursions on either side of μ to the extent of 1.5 σ. Where technological improvements show no returns, the strategy is to resort to managerial gimmicks. The corporations would have passed the test of ethics if they had underlined the above assumption along with their 3.4 defects per million tagline.
I can think of only one other instance of similar conjob. That happened 5000 years ago. Yudhistira was made to say that “Aswathama is dead ……………. “ and whisper to the tune of drumbeats “the elephant”.
:) I am so happy that you chose to post this. How about corporates managing people performance through normal distribution? It is the most ridiculous thing I've ever come across! And employees are convinced that this is the "scientific" way of doing it.
ReplyDeleteSuvarna: I am so happy that you chose this example to amplify my theory. A normal distribution is normally symmetric. So corporates do want a certain percentage of their people to remain on the left side! Marching toward excellence should mean that people move to the right on the X-axis. It is ridiculous to fit employee performance on a normal distribution. As if employees are circuits that can be mounted on a table top and their voltage levels / transfer functions can be measured. The force fitting of performance to a normal distribution has given rise to cowardly and unethical practices during appraisal. When I am forced to have 5 people rated as 2, who are my sacrificial lambs?
ReplyDelete(a) Those who have had a 4 or 5 in the last 3 years because, a 4 or 5 three times out of a 4-year cycle entitles them for a promotion; so giving a 2 will not affect their promotion prospects. You remain a goody goody boss! The e-enterprise is not intelligent enough to ask how can a "5" person suddenly degenerate to a "2" person?
(b) those who are likely to move out of the project during April-Jun. By the time the final appraisal ratings are published you are at a safe invisible distance from them - "I gave you a good rating; I don't know what really happened afterwards. You take it up with your present boss". Hands washed off in a cowardly manner.
The reward distribution into 3 components - company performance, project performance and individual performance is not in place in most cases. The middle component is missing. Best option would be to place an amount of kitty at the disposal of the PL and ask him / her to distribute among the team members as he / she deems fit.
On another note I have spoken to executives at Motorola and GE. They all accept my objection in private but complain that they are forced to tow the corporate line. At GE Connecticut HQ, one person said "I wish you say this to our bosses". At Motorola one person said "Our internal confidential notes admit what you are saying". For some time I successfully stalled TCS from going public about 6 sigma initiatives. Venguswamy Ramaswamy was one person who understood my point of view.