In my last post on data visualization, I had a couple tools recommended to me to try out. One had a limited trial period that I didn’t take advantage of in time, and the other, a tool by VisualizeFree, was too buggy to work. I uploaded my data easily enough (still had to clean it up first like some other ones), then I can’t view the actual visualization. Lame. I could email support, but I’m too lazy to do that again.
I think I got more out of reading this article on what is (or should be) the point of data visualization. Its based on a talk given by Manuel Lima of VisualComplexity.com, who curates that collection of data visualization examples and resources. The main point from Lima was this:
“We need to make a transition from tools of curiosity to tools of functionality.”
Which is true, there are many tools that provide interactivity but not much substance. On the other hand tools for fun or aesthetics only can also drive innovation. I’m torn. It seemed like a lot of people were disagreeing over separating “information art” from data visualization, but this article states it more eloquently than I can. Manuel Lima also listed some key principles for data visualization that I did not know and should probably keep in mind:
- form follows function
- start with a question
- interactivity is key
- cite your source
- the power of narrative
I’ve become much more judgmental toward charts and graphs thanks to blogs like Junk Charts, Flowing Data, and Simple Complexity. So when I tried out one of the many Twitter measurement tools, Graph Edge, and received my first report, I was confused by the charts.
Followers and "Legitimate" Followers
This graph shows my Twitter followers and “legitimate” followers (they have a definition for it). Because the lower limit of my y-axis is the number of legitimate followers, it gives a false impression of having a very low number of legitimate followers. Why not start with 0?
Net Twitter follower change
This line graph shows the follows, unfollows, and net follower change over time for my Twitter account. But I thought it was strange to include negative numbers, because it looked like I had negative 1 unfollows. Anyway, there’s room for improvement here.
Misc.
Today I created a Twitter list for #measure and web analytics people (like @ABTests did) on TLists. I think the value is that its curated so you can see recommended people who actually tweet about web analytics on a regular basis, and you can follow many people at once/discover new people. It may not be worthwhile once Twitter implements lists though. And I successfully integrated my GWO data with GA thanks to this post from the GWO Tricks blog!
All posts
sarahd23 on Twitter
Nice post!
Thanks for adding me to the #measure list. There were some folks on there I was not following yet!
-Rudi
Hi Rudi!
No problem, I think you’re definitely worth having on that list! I’ve also found some new and interesting people from the list as well.
Sarah
I’ve been designing custom visualizations in our recent product for several years, and it is often not a clear cut process. There are always tradeoffs. In some of our graphs we use the lowest value (not 0) as the basis in order to preserve more visual detail, since large numbers with little fluctuation can just become straight lines otherwise. On the other hand I see your point, too – since it makes the lowest value seem disproportionately small, and often lets people assume it is 0.
I’m not sure if there is always a good solution to this sort of thing. In tools that are used frequently I would still argue for using non-zero values as a base, since you can get used to it… but on the other hand there are so many ‘casual’, one-time users these days, that it becomes a very difficult question to answer.
Hi William,
Those are interesting points you bring up, that makes sense as to why you would sometimes want to show change through graphs not starting at 0. I guess you either have to assume users will develop workarounds to it in their mind or go with the standard.
Maybe one solution would be to show a graph in the default format with the lowest value at 0, and have an option to see in more detail that would alter the lowest value relative to the numbers? Then again, I have no idea how difficult that would be to program, or how useful. Its an interesting problem, for sure.
Sarah