So, you wake up one day, you look in the mirror and you ask yourself…Does this shirt make me look fat?
(I mean, who really does that right?)
Well the truth is.. it might! But who knows? So maybe you’ll just log into Facebook, send a quick pic to your bestie, and illumination will come from the heavens to help unravel this mystery.
A few minutes later your phone buzzes a few times and the holy grail is yours for the taking.
This shirt DOES make you look fat! Corrections are in order, changes are to be made. This is a moment for the ages. So off it goes. Thrown into the closet and everything is right with the world again.
But…something doesn’t quite sit right, there has to be more to this story.
I mean c’mon, you only asked for one friend’s opinion. Maybe you should ask 3 friends. That’s right, you are having thoughts, you need a second opinion so you pop back over to messenger and send the same message to 3 other people.
Boom. Two no’s and a yes.
Wow, this is getting very confusing.
What was a simple decision just became a complex one. Now you have 4 pieces of advice (points of data), and they are all telling you conflicting things.. Even broken clocks are right twice a day right?
Now let’s ask a different question. This may seem like its totally off point but just stay with me.
If we roll a six sided dice, it is obvious that it has a 1 in 6 chance to land on any given number right?
So, theoretically if I roll it 6 times I should get 1, 2, 3, 4, 5, 6..just not in that order, right?
well…maybe not.
It might be be a 2, 6 times.
It might be two 3’s, and 4, 2’s
THERE IS NO WAY TO KNOW!
But..if you roll the same dice a million times? Guess what.. You will get very close to the factual mathematical truth that each number is hit very close to 16% of the time.
This, is the power of having enough data.
Now, how does this apply to our little problem?
MY FRIENDS ARENT LIKE DICE, THEY KNOW ME you say. they KNOW WHAT I LIKE
Well, they might, but what you like isn’t important here. They know what they like. They have biases, life experiences, and ideas on how things should be.
Besides, wouldn’t a good friend just lie to you to make you feel good? This advice is the probably faulty in the first place.
Let’s play another little game.
Let’s ask 1 million people if they think this shirt makes you look fat.
The results depend on a lot of factors. They depend on how pretty you are, how much these strangers like your appearance and identify with your struggle as an individual. We could probably include some backstory that your parents just died and make them feel sorry for you so they say no.
People just aren’t reliable when asked a direct question right? Not when asking a bunch of strangers, probably when asking a few friends, but a few friends just isn’t enough data.
So…lets try a new way of asking.
Instead of asking if the shirt makes you look fat or bad or whatever. Let’s just give a million people a choice.
Shirt A
Shirt B
Shirt A is the shirt in question, and shirt B is some obviously ugly shirt that you hate.
This is a simple A, B test.
Now we can ask a million people which one they like more, by asking them to click the one they want.
The results will be far more accurate to the truth than your one friend, or even 4 friends you asked.
Sometimes you ask for a second opinion, but in the world of data science you would need a million opinions to find the actual truth, and how exhausting would that be?
Data scientists use this all the time these days to help them determine if a product is good or bad. They create facebook or google advertisements. If you click it, you like the shirt, if you don’t click it, you don’t.
If they want to compare two shirts they make two ads and see which one gets more clicks. Easy right?
Moral of the story, In God We Trust, All Others Bring Data.