Page 117 - The McKinsey Mind

04 (083-102B) chapter 4 1/29/02 4:50 PM Page 92

92 The McKinsey Mind

lier assertions, famously said, “When the facts change, I change my
mind. What do you do, sir?” Transferring this to the context of the
McKinsey problem-solving process, when the facts contradict your
hypothesis, you should change your hypothesis, not suppress the
facts. We can’t stress this too much. When you’ve spent a lot of
time and effort coming up with what you consider a brilliant
hypothesis, it’s easy to become wedded to it, refusing to believe
that you just might be wrong.
McKinsey offered several lessons on this topic: “Don’t make
the facts fit your solution”; “Be prepared to kill your babies”
(offered in the context of brainstorming, but it holds just as much
for data analysis); and “Just say, ‘I don’t know.’” What was true
at the Firm holds just as true outside of it. There is an iterative loop
that runs from hypothesis to analysis design to research to inter-
pretation and then, if necessary, back to hypothesis. Only after you
have definitively proved your final, modified hypothesis are you
ready to put together the end product—the advice that you will
give to your client.
When we asked our McKinsey alumni what tools they use to
help them make sense of the data, they almost all mentioned the
80/20 rule. As we discussed earlier in this chapter, 80/20 manifests
itself in a variety of ways. To offer a few more examples, 20 per-
cent of the population in the United States pays 80 percent of the
income tax. Of the students in a classroom, 20 percent occupy 80
percent of a teacher’s time. You might choose 80 percent of the
outfits you wear from 20 percent of your wardrobe. We could go
on and on. The 80/20 rule is not always strictly true; in one case,
the true ratio may be 75/25, in another 90/10. Furthermore, it is
not universally applicable, but it occurs so frequently as to make
it a useful predictive tool.
At McKinsey, the 80/20 is primarily about data, and that’s cer-
tainly true as far as it goes. Applying the 80/20 rule to numerical