« A tale of four charts | Main | The hard work of entertaining »



I'm amazed that you found that many rational comments. If nothing else it shows that Americans (I'm not) have got to the point where they can't debate anything rationally once it has a political spin.

John Munoz

Great find Omegatron and great post Kaiser.

As you know, there is no shortage of bar charts in print and online that violate the primary rule of bar charts. The primary rule is, all bar charts MUST have a zero-based axis. Bar charts are great because they allow us to compare the absolute difference between groups and you can't do an accurate comparison unless the baseline starts at zero.

If anyone can serve up an example of where this is not true, please show it here.

John C. Munoz

Peter Jung

Big fail. (BTW Kaiser, this is my first comment, but your blog is one of my favorites.) I shared this amusing fail on an internal company work blog Thursday night.

Nitpicky point, but although their current version is improved, I wouldn't go quite so far as to call it "fixed," as I think there's still room for improvement.

John, I agree under most circumstances. Two possible exceptions I'd like to throw out:
(1) To show differences from a performance goal or standard, I think it's acceptable to set the goal as a new baseline and have bars going up and down from that new baseline. I think this can be especially useful if the differences between values are small compared to the absolute values, whereas a standard zero-baseline bar chart would be ineffective. This is almost equivalent to subtracting the goal from all values and plotting the differences, except it retains a sense of scale.
(2) For data that's a bit more ordinal than quantitative, zero may not be meaningful. An example I have in mind is a survey where the response choices are [1-strongly disagree, 2-somewhat disagree...5-strongly agree]. It may make sense to average results for each question, then plot these averages in a series of horizontal bars where the baseline is 3, and the bars go left and right...although I guess this is the same as assigning values -2 through 2 instead of 1 through 5.

Bill Towne

It seems to me that there was a problem with some script making the chart and not some conscious decision to play with the scales.


Bill: you are absolutely right. Despite what some of the readers implied, this is unlikely to be conscious manipulation. A while ago, I had an example from Google Finance that also must have resulted from a bug in the computer script. And that's exactly why I cited the comments because an innocent mistake often is not interpreted as such.

Peter & John: good discussion. To synthesize what you're both saying, the important thing is to have a reference level. Oftentimes, this is the zero level. In the cases cited by Peter, it is not zero. In Peter's case(1), we should probably plot the differences from standard, thus still using zero as the baseline but having it in the middle of the chart, with positive values going right and negative going left.

The comments to this entry are closed.

Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR.

See my Youtube and Flickr.

Book Blog

Link to junkcharts

Graphics design by Amanda Lee

The Read

Keep in Touch

follow me on Twitter