Time to tackle some of the more important parts of life. I saw this delivery schedule in a cooler at a DC Whole Foods. Bell’s Beer is from Kalamazoo and we take pretty big pride in our hometown beer. So we (I mean, I) get really geeked when Bell’s shows up during travels. But this delivery schedule? It’s not working for this data visualizationist (I just made that term up – how does it sound?).
I couldn’t put my finger on precisely what bothered me so much about it, so I took a snapshot and mulled it over for a few weeks. Here is the revised chart I settled on.
The revision still isn’t sitting right with me, but I’ll tell you what important changes I did make:
1. I reversed the order of the listings. Previously, the beers that are delivered year-round were closest to the headers listing the months. That meant that the ones that were the most narrowly distributed were furthest away from the month listings, making them the hardest to decipher. In the revised version, I also just put the month listings across the bottom of the chart, too. Why not?
2. The beers are in a very slightly different order, now depicted by the length of months they are on the market. Just adding a little more logic.
3. The labels over the bars make for less seek-and-find. Previously, the viewer would have to locate the desired beer label, trace the bar to the right, and then simultaneously locate the months across the top, traveling that information down the graph, to find where the paths crossed, just to determine if Favorite Beer was in stock. Too much. This way, we remove one element of difficulty in decoding the chart. Still, I’m unhappy with how I had to abbreviate the top two beers to make them fit in the label.
4. White background. The textured orange background (intended to be beer) was too busy and conflicted with the colors used to differentiate each bar. I’m not totally in love with my new color scheme, but it’s a step up.
If I had access to the beer label icons, I would have still placed them along the left side of the chart. That visual cue is important for the viewer to quickly identify the beer in question.
Posted by Stephanie Evergreen on May 15, 2012
When giving advice about chart styles, one of my friends likes to say “save 3D for the movies.” She’s right – research shows that 3D charts actually slow reader comprehension. I’ve long advocated for the use of ChartTamer as a friendly way to restrict data visualization to that which will actually support reader cognition. In a recent webinar I heard Cole Nussbaumer at Storytelling with Data give another compelling reason why we shouldn’t graph in 3D – Excel doesn’t use the front line or the back line of the 3D bar when aligning with the y axis – it graphs the midpoint of the bar. What??? Yes, she said Excel uses the middle of the 3D rendering as the determiner of the data point. I didn’t believe it. How could anything so antithetical to cognition actually survive in this sink-or-swim society? So I tried it myself:
Here is a 3D column chart of some fake data about customer satisfaction with prices in various departments of a natural health food store, before and after their physical move to a new location:
Now I know I’m breaking some serious graphing rules here, but stick with me. I super enlarged the y axis and honed in on one range of it so that we could really see where the gridlines line up with the tops of the columns.
See the blue bar in Produce? The actual data point for that blue bar is 85. Neither the front line, the back line, nor the midpoint of the top of the column are at 85. See the blue bar for Beauty? The actual data point there is 83. Again, none of the possible points of measurement on the 3D column are accurately expressing 83. Maybe more like 82.75, but definitely clearly absolutely not reaching the gridline marking 83.
I selected the exact same raw data table and created a similar column chart, just in 2D:
This time the blue bar in Produce is exactly at 85 and the blue bar in Beauty accurately represents the data point of 83.
I wasn’t able to reproduce Cole’s assertion that Excel uses the midpoint of the 3D column. But it is pretty clear that the audience misinterpretation often cited in research that results from the use of 3D charting is due to more than the complexity of analyzing in the third dimension – it is also because the columns simply aren’t accurately visualized.
Posted by Stephanie Evergreen on January 9, 2012