Eternal optimist
Light entertainment

Scale restoration

The original graph threw us off our sense of scale.  It seemed to be saying all these oil companies are roughly the same size but one grew much faster than the others.  The red color and the setting off of the data above the title of the chart seemed to announce some important find.

The junkart version on the right reversed everything to our normal sense of scale.  It is a version of the bumps chart, one of my favorites.


So we find that Total is the smallest of these oil companies, about half the size of ExxonMobil -- you wouldn't know that from those abysmal bubbles!   Adding to the problem is that the growth data is used to sort the companies while the actual production data is hidden in the data labels.

Total is indeed growing faster but BP is not far behind.  The fall of ExxonMobil and Royal Dutch Shell is equally intriguing.

Reference: "Total, the French Oil Company, Places Its Bets Globally", New York Times, Feb 21, 2009.


Feed You can follow this conversation by subscribing to the comment feed for this post.


Bump chart terminology: Are there different terms for the ordered charts (like the English rowing results) and scaled charts (like this oil company chart)?


I don't know if I can agree with the junkart version. The big story the Times is trying to convey is growth rate. Their graph essentially plots the oil companies' growth rates on a y-axis, which makes it crystal clear that Total is faring much better, growth-wise, than its competitors.

I agree that the bubbles are far from ideal for conveying size, especially a series of sizes that are relatively close in value. But this is really just window dressing to flesh the graph out a little, and doesn't detract from their main point about growth rates.

The junkart chart does a better job of displaying all the data. But the disadvantage is that it doesn't draw your eye toward what is significant. And isn't this what charts are supposed to help us do - to see the patterns in the data?


Dave: I think others have made the same point before. They prefer to call a bumps chart with a continuous scale something else. Would they call a log-scaled line chart something else other than a line chart?

Chris: Given the data, I am not sure if Total's growth is the big story here. That's the point I'm making. And you're right that if the point is to show the significance of Total's growth, then the junkart version is not better.


Dave, I'm one who thinks a chart of value against ordinal category is just a line chart. I reserve the term bumps chart for a chart of *rank* against ordinal category. Rank is itself an ordinal category, which can be occupied by only one data point per year, or stage, etc. When a point moves up into a new rank, it "bumps" another line down a rank, and vice versa.

Here are two examples of what I call "proper" bumps charts: Junk Charts Sep 2007: "Knowledge Transfer" showing the Tour de France, and XLCubed blog article showing the English Premier League. Why are they both sports data like the Oxford Bumps Races? I suppose because sports is much more concerned with rank than other areas are.

I could certainly imagine a bumps chart of records (vinyl music ones, I mean), as that's a another famously ranked field. The data set for that could be gorgeously rich, if you chose a big pop chart that had been going for many decades.

I don't suppose business or science is, or should be, concerned with rank very often.


What do you think about the tables presented here by a statistician?

drilling rigs

i never looked at it like this before. this is some great information.

New Jersey Church Restoration

The junkart version seems kind of off. Thanks for the post though.


While this visualization is a significant improvement from the original, there are still a few issues if you just look at the visualization shown in write-up. Firstly, the axis labels are incorrect, as the x-axis is "year" not "barrels of oil". Secondly, how can the "Total" line (assuming that's total barrels of oil between the two years) be less then four companies? Shouldn't the total be a summation of the four, or am I missing something? Lastly, all the companies, with the exception of BP, have a decline in production from 2002 to 2008, so how can the slope of the total be positive?

It is only by going back to the original visualization that these questions are answered. The "Total" line is actually the "average production amounts of all oil producing companies". It is smaller then the four companies that are shown, because the industry is made up of a larger number of "smaller and more nimble" producers, which have "outpaced [the] other major oil producers". The four companies shown in the visualization are the largest companies, and the point of the original visualization was to show that while the top companies are in decline, the production values of smaller companies (which there are more of) are up.

Because of these omissions, Junk Chart's redesign is clearly not a stand alone visualization and is actually miss leading.


What software are you normally using to create such relatively simple bump charts like the one you created for this article?


Thomas: You can use Excel so long as you pick the right options (e.g. use a line chart; don't center the x-axis labels) and you may need to do some cosmetic changes in addition.

The comments to this entry are closed.