Thursday, September 19, 2013

Iron Viz TCC13 - Why do old movies get better with age?

I was fascinated by the Iron Viz competition at TCC13 (Tableau Customer Conference). Amazing what 3 people could do in 20 minutes. It demonstrated 2 of the biggest benefits of Tableau -

1. Speed to Insight - How quickly one can get insights using the tool. Agreed these are the Iron Viz contestants and know Tableau inside out, but it still shows the power of drag and drop visualization that anyone can leverage.

2. Same data, different insights - 3 people, using the same data identified 3 totally different insights. This is the power of User Generated Analytics. Give people the right tools and the data and you will be surprised at what you discover. You will find out insights about what 'you don't know you don't know'.

This brings me to another benefit that ties in to Nate Silver's talk about biases at TCC13. Sometimes data can be misleading. I started down this path when I started looking at the visualization created by Ryan Sleeper (the winning entry). 

Ryan asked a simple question based on his 20 min analysis - Do movies get better with age? If you play around with the visualization (change the age), you will see that newer movies are always showing up with a lower rating than older movies. Which makes you want to believe that the answer to the question is - Yes. If you answered Yes, then you get to Nate's theme around biases and a topic that is well expounded on by Daniel Kahneman in his book - Thinking Fast and Slow.

The beauty of Tableau is that it lets me ask questions and asking questions is key to making sure you don't misinterpret data / fall prey to biases and believe what you want to believe.

The question I wanted asnswered was - why does the data indicate that old movies get better with age? 

So I downloaded the workbook and started playing around with the data. The power of collaborative analysis. I made a minor modificantion to Ryan's viz - I added  the # of Ratings in addition to Average Rating (a very minor change). You can see the result here

It becomes very clear that there are way way more ratings for newer movies than older movies and since average ratings for movies in the last 5 years is 7.5 and the number of ratings is 326K (compared to a total sample size of 449K), this is totally going to skew the results downwards.

My conclusion - New movies stand no chance.

This then leads to another follow on question - Why are older movies rated well? Not sure the data helps here but a hypothesis is that - if you are watching a movie that is 50 years old, chances are you heard good things about it and if you go to the trouble of rating it, you are going to rate it well. As far as newer movies go, you will watch a lot of good and bad movies and there are quite a few blockbusters that are pretty bad, that you will rate lower and want to make sure no one else wastes money on.

This is the other big benefit of Tableau -

3. It lets you ask questions and takes you down this journey of data exploration. It makes the exploration fun and easy. It also shows you how multiple people can analyze the same dashboard, collaborate on the answers and come up with insights that you might not have thought about. 

Lastly, coming back to biases - One way to avoid biases, is to make sure you have multiple people looking at the data and allow multiple people to play with the data. User Generated Analytics = reduced bias.

No comments:

Post a Comment