There are many of us out there who are driven by statistics. Bloggers especially. We compare the number of subscribers to our feed, the traffic we see, sometimes even the money we make. This post is about how Twitter SHOULD NOT be something we use as a measuring stick.

The Experiment

I have a website that I run every March for the NCAA March Madness tournament. This will be the 8th year I've run it, and until now, it has grown completely organically. Friends and colleagues told their friends and colleagues, and before I knew it, I had over 500 people filling out brackets on my website.

With the rise of Twitter, Facebook, and other social networks, I thought I'd see what I could do with a little promotion.

So I created a new Twitter handle: cutdownthenets. I started by following about 25 some of my close friends on Twitter. But that seemed silly, because they were already going to know about my site. I am trying to reach people that I've never met before.

How do I find a huge number of people that are interested in sports, brackets, or the NCAA tournament? Well, ESPN has a Twitter presence, what if I just follow everyone that is following ESPN? They've got over 40,000 followers. Surely some of them will be interested in my site. Heck, if I got 1% of them, that would be 400 new participants in my little office pool.

So I opened ESPN's "followers" page, and started following every single person I could. Fatigue started setting in around 1,000 though. So I left it at that. Based on my expectations, if I only got 1% of those people, I'd have 10 new followers the next day.

When I woke up this morning, cutdownthenets had over 200 followers.



Are you kidding me? Surely this can't be right. As it turns out, it is. About 20% of a random sample of Twitter users followed me back, without knowing anything about me, and certainly without reading any of my posts. (I'd only made 5 posts TOTAL at that point).

In fact, my 1% number did come into play in all of this. About 1% of the users that I followed had an automatic follow and DM action set up somewhere. I got lots of messages like these:

What This All Means

What this little experiment has made me realize is that the number of followers someone has is almost completely artificial. There's TONS of services that try to compare people based on the number of Twitter followers they have, and that's just wrong. If I wanted more followers, all I need to do is start following a ton of people. Heck, I could even "unfollow" those people the next day to really inflate my Twitter Ratio.

The quality of a Twitter user should be determined by how relevant their content is to their followers. There are plenty of popular users that have a signal/noise ration approaching ZERO. They don't participate in the conversation, it's just a constant stream of noise in the form of links, messages, and more links. (I'm looking at you, Guy Kawasaki.)

So the next time you get followed by someone, take the time to look at their posts. Read their bio. Truly determine whether listening to that person is valuable to you. I'm re-evaluating the people I follow because of this. There's a bunch of noise out there, and I'm in search of value.

How This Can Be Harnessed For Good

Twitter could be a very valuable tool for companies. Some, like Dell, really seem to get that. They can provide valuable information and resources directly to their customers. But this can be taken a step further.

What if every time someone mentioned your product on Twitter, you knew about it? What if every time someone had a negative experience with your company, you heard about it, and could help to remedy the situation? Using the Search features at http://search.twitter.com, or a desktop application like TweetDeck, I am able to pay attention to the specific terms that I am interested in, and interact with people that use those terms. The terms I am currently following are "Zune", "March Madness", and "Microsoft Tag." If ANY Twitter user mentions one of those things, I immediately get a notification in TweetDeck, and can reply to them, follow them, etc.

It allows me to meet and interact with other people that have similar interests to me. THIS is where the value of Twitter is. Finding common interests, and having conversations about them. Just getting a HUGE number of followers is not.

I no longer care about how many followers you have. It doesn't matter.

Labels: statistics, twitter

It always surprises me when I talk about the Monty Hall problem, how many people argue with me about what the "right" answer is. It goes against most everyone's ideas of what is right. Here's the problem:

The Three Doors

You are presented with three doors, and are told that behind ONE of them, there is a shiny, new car. Behind the other two doors, however, there are goats. The object of this game, obviously, is to choose the door that has the car.

So make your choice. Do you choose Door #1, Door #2, or Door #3?

The Reveal

Once you have chosen a door, your gracious host, Monty Hall, chooses to open one of the doors you did not select. He will ALWAYS open a door that contains a goat. For the purposes of illustration, let's assume you chose door #2 initially. In this case, then, Monty knows that there is a goat behind Door #1. So he opens it.

The Choice

This is when the Monty Hall problem rears its ugly head. Once that door is revealed, he offers you a choice. You can keep the door you initially selected, or you can switch to the other unopened door. It's entirely your choice. What would YOU do? Would you stay with your original door? Would you choose the other one? (Leave a comment on this post with your choice right now...before you continue reading...) In our example, again, we are going to stay with our initial choice.

The Result

After you've made your choice, Monty will open the door you DIDN'T select. Here's a picture of how our scenario played out:



So the point of this whole post comes down to probability:

Why You Should Always Switch Doors

This is where the controversy comes in. Many people (probably including you) disagree with this. The common thought is that no matter what you do, your initial choice of the three doors can't actually be improved. I'm writing this to tell you that you're wrong.

In fact, you can DOUBLE your odds of winning by choosing the other unopened door. Let me illustrate the flow of choices you get to make, and how they improve with a "switch." In this example, we assume that the "contestant" always chooses Door #1.



Recall back to how the problem works: You pick one of three doors. So your odds are 33% (1/3) that you're going to pick the car. Said another way, there's a 66% (2/3) chance that you're going to pick a goat.

Here's what that means, as you look at the image above: 2 out of 3 choices result in you choosing a door with a goat. Switching, when you have selected a goat initially, results in getting the car 100% of the time. So if you ALWAYS switch, you've got a 66% chance of winning the car. If you NEVER switch, which means you stick with your initial choice, you are forcing yourself to take a 33% chance of the car being behind your door.

Yes, that does still mean that 1/3 of the time you will still lose by switching, but I'll take that over losing 2/3 of the time by NOT switching.

In the end, you DOUBLE your odds of winning the car by switching every time.

Let's Discuss This Further

I'd like to think that this topic will generate some conversation. You probably have stared at the diagram, and yet, inside, you still disagree. You might say that switching is a 50/50 shot. That's where I disagree. If you're simply playing the percentages, switching will always give you the best results.

Leave a comment. Leave several. Let's have a lively discussion on this one.

UPDATED: Marvin S. Schwartz, PhD. has written a simple C# application that lets you play this game. Here's a link to his Monty Hall source code.

Labels: let's make a deal, monty hall, probability, statistics