If a C programmer asks "do you want to see something cool?", run away.
--John Van Enk

Monday, April 12, 2010

“Find me a rock”

Благодаря @newsycombinator набрел на статью Opinion: The story BCG offered me $16,000 not to tell. Мало того, что я узнал о существовании The Tech, так еще и статья хорошая попалась.

Особенно понравилось

This leads to what I like to call, “Find me a rock” problems. The classic “find me a rock” story is as follows: A manager goes to his engineer one day and asks for a rock. “A rock?” asks the engineer. “Yes, a rock. That isn’t going to be a problem, is it?” replies the manager. The engineer laughs and tells the manager he’ll go pick one up during his lunch break and it will be no problem. After lunch, the manager visits the engineer again and the engineer shows him the rock. The manager looks at it for a moment before telling the engineer, “No, that one won’t work at all. I need a rock.”


There is an interesting kabuki dance to be done when crafting figures to fit a conclusion. The conclusion may be wrong, but you still need to make it believable. You still need numbers to fill out your PowerPoint slides, and the numbers need to have enough internal consistency not to throw up red flags at a casual glance. Honest analysis, even when it has weak areas, is easy to defend. If the numbers look fishy, there’s an explanation — you didn’t have direct data on such and such and had to use estimates from another report, or made a reasonable assumption somewhere. But when the numbers actually are fishy, and there’s no underlying logic to defend, you can’t have any rough areas for others to poke at. And when you know everything is fishy, you can’t tell what will look fishy to someone who hasn’t seen any numbers before.


At one point my manager said to me, “Change the numbers, but don’t change the conclusion.”

Почему мне кажется что такое везде сплошь и рядом?