4Sight BI Perspective, November 2014
One of the biggest selling points of Business Intelligence is that it helps organizations replace "HiPPO" (Highest Paid Person's Opinion) decision-making with data-driven decision-making. Why is this so important? Because it gets you into the habit of using evidence instead of hunches and preconceptions that can lead to conclusions that are intuitive but wrong. How easy is it to make a wrong decision? Here's a famous example.
The "Let's Make A Deal" game show debuted on TV in the early 1970s and featured funny people wearing funny costumes, engaged in various types of trading with host Monty Hall. Near the end of each show, a contestant was selected to play "The Big Deal," in which he or she would be asked to choose a prize from behind one of three doors. Two of the doors concealed "booby prizes," such as a goat, while the third held an expensive grand prize, usually a car.
"The Big Deal" was an illustration of a logic problem that has come to be known as the Monty Hall Paradox. In its archetypal form, it works like this: Initially, Monty has the contestant pick one of the three doors. But before opening the door to reveal the prize behind it, he opens one of the other doors -- always one with a booby prize -- and then offers the contestant a choice: stick with the door he or she originally picked, or swap that prize for the one hidden behind the third door.
So the question is: should the contestant stick with his/her original choice or swap?
The intuitive answer is that it doesn't matter. After all, we've eliminated one door (the one Monty just showed us, with the goat), and we know that of the two remaining doors, one's got a car and the other's got another goat. So it doesn't matter whether you swap. Or does it?
Well, actually, it does. It turns out that swapping your door for the third one will get you a car two-thirds of the time. How is this possible?
To understand the paradox, it helps to look at it from the beginning -- before you've picked a door in the first place. At the start of the game, you know that there are two doors with goats and one with a car. So when you pick a door at random, two-thirds of the time, your initial door will be one with a goat. When that happens, Monty shows you the other goat -- and the car is behind the third door. So if you swap, you win. The other one-third of the time, your initial pick is the car. When that happens, Monty (again) shows you one of the goats -- and the other goat is behind the third door. So if you swap, you lose.
This is not at all obvious, if you do what Monty wants you to do -- namely, wait until you've already chosen a door before deciding whether or not to swap. Once you've done that, and he's shown you a goat, it is true that one of the two closed doors conceals a car, and the other, another goat -- so the contestant thinks it doesn't matter whether he keeps his door or swaps it. What he fails to realize is that at that point, there's still a two-out-of-three chance that his initial selection was a goat; that's what matters! The logic looks like this:
Premise: After I pick a door, Monty is always going to show me a goat.
Premise: Two times out of three, I will already have picked a goat, so the one Monty's showing me is the second one.
Conclusion: Two times out of three, if I swap my door for the third one, I'll get the car.
Don't Be a Goat!
How does this relate to Business Intelligence and data analytics? Well, the lesson is that when our data show us something we didn't expect, we've got to be committed to believing what the data tell us, even if it contradicts our intuitive grasp of the situation. The Monty Hall Paradox is confusing and counterintuitive, but the contestant really does win two-thirds of the time by swapping doors.
When your company implements a BI system, you're giving yourself the power to identify points where, in the past, decisions were made that didn't reflect the organization's true success drivers. By taking full advantage of your enterprise data, you can avoid the psychology of the game-show contestant, and instead, make management decisions based on the evidence. You might not win the grand prize every time, but the odds will be in your favor.
By John Kafalas
This monthly column covers Business Intelligence and data analytics issues. If you have questions, comments, or topic suggestions, please contact the author.
Copyright © 2014 4Sight Technologies, Inc. All rights reserved.