Launch, grow, and unlock your career in data

March 11, 2013

We broke up because: Comparing pies and slope graphs as forms of data visualization


Lee Byron, who has previously work with David McCandless, has done an interesting analysis that breaks down the timing and methods for breakups.  I really like his time-series analysis.

The time-series analysis does a great job of pointing out the peaks and valleys of breakups.  It seems to me that people either want more freedom at spring break or break up as a result of spring break or both.  People also break up at a high rate before Christmas.  Is this to avoid giving a gift?

Lee ruined my day with his pie chart analysis.

There are a few problems with these charts:

  1. The first pie chart adds up to 101%.  Pie charts are supposed to be base 100.
  2. The charts go in completely reverse order.  Tilt your head so that “in person” is at 12 o’clock and you’ll see what I mean.  The left pie chart reads counter clockwise and the right pie chart reads clockwise.  This makes looking across the pie charts nearly impossible.
  3. The whole idea for this analysis is to compare different generations.  Pie charts do not do this job (or any job) well.

Lee’s analysis is well done.

Compared to people born before 1975, people born after 1984 are twice as likely to breakup via the digital world, twice as likely to breakup over the phone and far less likely to decide to talk it out over coffee.

It’s a shame he didn’t make a better chart choice to support his message.  One alternative Lee could have considered, and one that I’ve been falling in love with since Alberto Cairo explained why he likes them, is a slope chart.  I made this chart in Excel is less than one minute. 


Going back to Lee’s summary, the slope chart does a much better job of supporting his message.  You can clearly see the shift away from personal interaction towards phone and instant messaging (less personal interaction). 

Having four kids of my own, I can clearly see this shift in communication.  I do my best to encourage my kids to call their friends instead of texting, and it’s a continuous challenge.  Texting is way more convenient and takes much of the emotion out of the conversation.  I can see how this would make breakups easier, however I can’t help but wonder what the long term impact will be on society due to this shift away from in-person contact.

Download the Excel file here.


  1. Great post, Andy. Just one issue: I think the slopegraph should only be used when you have time on your x-axis, which isn't quite the case here. Given the categorical data, I'd use a horizontal bar chart (though, admittedly, that may too frequently be my go-to graph!).

    1. Thanks for the comments Cole. Interesting that you say the axis should only be time. That was always what I thought too. When I talked to Alberto about it, he showed me why it can work with categorical data too. Basically it helps you show relationships between categorical data, much like a scatter plot would.

      Tufte's definition is very strict, with time as the only option, but then again he's not exactly the most flexible person in the world.