BeyeBLOGS | BeyeBLOGS Home | Get Your Own Blog

« Pushmi-pullyu | Main | Triadic Continuum (Prologue) »

August 7, 2007

Visualization

Admittedly, I'm a picture-person. By that, I mean that I tend to think about things using spacial relationships; and I've been this way for a long time. In grade school, I can remember thinking, "I don't remember exactly what year Wyoming became a state, but I know that I read the answer about half way down a left-hand page somewhere in the middle of the chapter." I couldn't remember the detail, but I could recall the context in which that information can be found. Perhaps that just a sort of photographic memory... with astigmatism.

I know that other people think in different ways. My wife, for instance, reads a sentence (or more) at a time. It took me a while to understand that... "by a sentence, you mean one word at a time trailing into sentences?" "No, I mean I look at the page and see phrases and sentences, not single words." Wow. That's so different than the way I read. Maybe that says something about my intelligence, or maybe it just says something about how different people consume data in different ways.

I have a colleague that I've worked with for some years who likes reports with lots of numbers on them. Now, I'm a bit of a math geek, too -- I got a shirt from my daughter a couple years back with this on it:

01000100
01000001
01000100

Still, I think there's more value that can be derived from that sheet of numbers than can be delivered in the bare numbers themselves. Choosing the right pictorial representation can be very hard, but it can also be very powerful. Here are a couple of examples of relatively simple (and underused) visualization techniques that I think are very powerful.

Radar Charts:
If you plot two or more series on a single radar chart (or spider chart) you can deliver information about both which of the series "covers the most area" as well as in which dimension each series "out performs" each other series. So, you've got both point-by-point comparisons as well as an overall comparison available in a single view. The numeric equivalent might be detailed rankings and an overall average ranking.

Bubble Charts:
This kind of chart also allows you to deliver more than one kind of information in a single view. Bubble charts provide for both a horizontal and vertical plot as well as a third measure that is conveyed using the relative size of the marker on the chart. So, you could plot "number of infections" versus "number of flu shots" for a set of regional clinics and use the size of the marker on the chart to represent "number of nurses on duty" -- if you had a reason to think some pattern might come out of those three metrics.

Gantt Chart:
Don't think I'm just referring to project plans, dependencies, and resources here. Gantt charts can be used convey more information than that. A gantt chart, after all, isn't much more than an "activity" versus "time" chart anyway. Imagine a chart with patient volume on the Y-axis and time on the X-axis. Great. You know how patient volume changes throughout the day or over the year. Add a set of gantt charts to the bottom showing holidays, the school year, and other special events and your chart now allows you to correlate patient volume with various other pieces of information other than "time" generically.

You can imagine adding more dimensions to each of these kinds of charts using shape and color. You could add animation over time to show a trend. You could add sound as another mechanism for delivering information, reflect an upward or downward trend, for example. Three-dimensional charts are fine, but imagine a 7-dimensional chart that uses X, Y, Z, color, shape, size, and animation to convey so much more information. Clearly there's a practical limit to how much information someone can consume from a single view of some piece of information, though.

Still, reports with lots of number on them probably isn't enough to justify a multi-million dollar reporting tool any more.

Posted by Paul Boal at August 7, 2007 7:45 PM

Comments

Post a comment




Remember Me?