Worse than a 3D pie chart
I have seen my share of good charts and I have seen my share of bad charts, but I never expected what I saw today.
As you may know, Hjalli and I are writing a book about chart design. We will guide you through choosing the best chart for your story, and to create beautiful and effective charts. The book will be aimed at those who want, or need, to get a chart out there but aren’t that interested in the why’s. We start by looking at the charts the big boys do by default and go from there, examining the parts and how to improve them.
The first chapter about chart design is about tables, which was fun to write. There was more to say than we expected.
The next chapter focuses on line charts, where I used Numbers, Excel and DataGraph to create default versions of a line chart. As I knew, there were things that could be better designed in the default versions of all applications. None of the defaults is useable in our opinion. Numbers the least.
Today, I dove into the details of my bar chart design. Called up the author of DataGraph to discuss moving axis labels by a pixel. Stared at my screen for half an hour, wondering if I want to keep the x axis on the bar chart or not. Then I opened up Numbers and Excel to create the defaults. DataGraph was already open, since I can almost do perfect charts in it already. So I started with the DataGraph default. It disappointed me a bit. The y axis didn’t automatically label my bars. Other than that, it was not pretty but useable. Numbers, to be fair, does automatically label the bars.
Next up was Numbers. At first glance it looked fine, the color of the bars was okay and the bars were labeled correctly. As I checked off items in the designing-a-bar-chart list in my head, everything seemed fine. Until it didn’t. At all.
In disbelief, I went straight to Excel to see if this alarm went off there as well. It did. And there was much wailing and gnashing of teeth. I felt like George Taylor in the Planet of the apes: “You Maniacs! You blew it up! Ah, damn you! God damn you all to hell!”
The default bar charts from Apple Numbers and Microsoft Excel.
They didn’t include the zero on the x axis! This is no small omission. Their default bar chart is a lie! When comparing the bars, you must compare the full length of the bars.
You can tell both applications to include the zero, but that should not be needed. Creating a bar or column chart without the zero on the axis shouldn’t even be possible. This is worse than a 3D pie chart.
There, I’ve said it.
It’s nice to hear that you think it shouldn’t be possible to create a bar chart without zero on the axis but your FORGOT TO TELL US WHY!! I know why but this article has zero educational value without some explanation.
jaakkoh
September 14, 2012 at 4:13 am
Although I wasn’t aiming for educational value, you are correct that I could have been more helpful for those who want to know why a bar chart should always include zero on the quantitative axis. I have added a link to the article that leads to Jon Peltier’s blog where he explains this quite well.
Thorri
September 14, 2012 at 9:32 am
There’s a pretty good article on why you should start at zero here http://www.dashboardinsight.com/news/news-articles/using_dashboards_good_evil.aspx
Jon Hazell
September 19, 2012 at 3:30 pm
You should try this with Tableau! perfect defaults are baked into the product from the start – a subtle difference that creeps up on you as you become a Tableau user.
Tom Brown
September 21, 2012 at 11:03 am
I have heard many good things about Tableau, but I don’t use Windows.
Thorri
September 21, 2012 at 11:14 am
Tableau are currently working on a Mac version – until then, you can use parallels or some other windows emulator. Tableau is good enough that Apple are now a major customer.
Tom Brown
September 21, 2012 at 11:15 am
What version of Excel? I barely use the software but when I booted up VM and ran Excel 2010 on Windows it started my axis at 0. Unrelated, it’d be great to consider R’s defaults and see how these stack up.
Scott
September 26, 2012 at 9:46 am
I am using Excel:mac 2011. R defaults is a good idea. Thanks.
Thorri
September 26, 2012 at 10:32 am
This was pointed out by Darrell Huff in his small but valuable book “How to Lie with Statistics”, first published in 1954.
Statistics Guru
May 8, 2013 at 4:22 am