Bandlines Are Already Commercially Available
In the recent Jan/Feb/Mar 2013 edition of the Visual Business Intelligence Newsletter, I introduced an extension of Tufte’s sparklines called bandlines. In less than two weeks since their introduction, without any prior knowledge of them, the software company XL Cubed has incorporated bandlines into their product. Here’s a screen shot of an example:
XL Cubed works as a companion to Excel. It can be used to extend Excel’s data visualization and analysis capabilities, including the addition of bullet graphs. You can read about this implementation in their recent blog post appropriately titled Bandlines in XL Cubed. This is not an endorsement of the product (I don’t give endorsements), but merely an announcement that one software company so far has recognized the value of bandlines and beat all others out of the gate.
Take care,
8 Comments on “Bandlines Are Already Commercially Available”
Hello Stephen,
I’ve recently created this data visualization, where I implemented sort of bandlines in it using Tableau Software. http://healthintelligence.drupalgardens.com/content/mortality-world-1970-2010
Your feedback is appreciated
Best regards,
Ramon
Yes, don’t mean to piss on XL Cubed’s fire but Tableau has already been able to do this for a long time using reference lines…
http://img502.imageshack.us/img502/9161/screenshot016n.png
To me, the big benefit of pulling together a set of graph-making rules like Bandlines is consistency (learn-once, read-many), but the example deviates quite a bit from the rules. The gray levels are inverted and there appears to be a different rule for point highlighting (and point coloring).
Similarly for the previous commenters’ examples: adding bands to a line graph doesn’t make it a Bandline graph.
Here’s another example in Tableau, using a presumably realistic data set of monthly coffee store sales. Except for the outlier markers, Tableau makes this easy. Outliers are tricky only because Tableau (ver. 7) doesn’t provide a quantile function.
http://public.tableausoftware.com/views/bandlines/0-Intro?:embed=y
I also compared bandlines, bandlines with sparkstrips, and sparklines. It of course depends on the dataset, use case, and personal preference, but I kind of like having the text values.
I agree that unless there’s a good reason for not following Few’s formatting example, doing so makes it easier, especially if you’re not including sparkstrips or text labels.
Ramon and Jim,
Thanks for sharing the examples of bandlines that you were able to create using Tableau. In Jim’s example, one thing that caught my eye was the fact that the white reference line for the median appears on top of the sparkline, which causes breaks in it, which is a problem. Can the reference line sit behind the sparkline?
Unfortunately Tableau puts reference lines in the foreground. Usually this works well, but not in this case. You can minimize the effect by making the line semi-transparent (I think I used 70% here).
One alternative is to eliminate the median line, which I didn’t like. Another is to use four shades of grey, one for each quartile—this is, I think, what Ramon and Robin did and what I’ve seen a couple of other Tableau users do. It is the quickest way to add ‘bandlines’ in Tableau. But I find it hard to decipher four shades, especially in miniature graphs with screen resolution.
Jim,
The fact that Tableau’s reference lines can only be displayed in the foreground has been an occasional limitation all along. I remember mentioning this to the folks at Tableau when reference lines were first in beta. I agree that using four shades of a color for the bands rather than three with the middle band subdivided by a line doesn’t work as well, which is why I took the approach that I did in my design specification.
Stephen and Jim,
Just to recognize that my implementation was the simpler and incomplete one. I want to take the opportunity to congratulate Jim for his implementation and to Richard Leeke for the implementation of the percentile algorithm in Tableau. I’ve learned a lot from both and now I’m working in its application to a real public health analysis.
I’ll share my results.