Comments on: Bad Graphics – Stacked Pyramid Chart

By: mermaldad

mermaldad — Tue, 08 Dec 2009 16:34:26 +0000

To me, the purpose of a stacked pyramid is to express a relationship between the layers. The lowest layer is the foundation upon which higher layers depend. As you noted, the eye naturally sees the volume of each layer as indicating the magnitude of whatever is being measured. A reasonably good example of this is the USDA food pyramid, where the volume (or area) of the layer indicates how many servings of each group one should eat. Moreover, if the because the eye is only so-so at comparing the volumes of the layers, the pyramid is best used when the numbers aren’t the main point.

The chart above ignores all of this. It’s not that the stacked pyramid is bad, just that it is easily mis-used.

By: Jon Peltier

Jon Peltier — Sat, 26 Apr 2008 14:32:55 +0000

I suspect they use thickness because (a) the algorithms are easier to implement, and (b) they haven’t even thought of using a different apparent ’size’ property of the sections. probably a good thing, because these can be improved on only by using a uniform section, viewing the section as a 2D shape, and unstacking the sections.

By: Damir

Damir — Sat, 26 Apr 2008 12:58:03 +0000

One could simply rotate the bar-chat around the category axis to get truncated-conical shape which can then be used as a funnel or pyramid chart. In this case the width of a disc-element would be proportional to the category value. Why do they do thickness is beyond me.

By: Jon Peltier

Jon Peltier — Tue, 22 Apr 2008 21:52:47 +0000

Oh, that’s nasty. I guess it plots by thickness, too, like the pyramid.

I’ve seen funnel charts which were supposed to indicate a process of filtering out items, so each stage of the funnel is supposed to contain fewer items. This would be good, for example, for tracking the success of development projects, where you might start with 100 projects, 90 pass through the initial feasibility stage, 75 pass preliminary design, 50 go on to detail design, 25 pass technical review, and 10 are released to manufacturing.

However, I think that a column or bar chart best illustrates a filtering process at work. The funnel analogy is lacking, because it seems to me that everything squeezes through the entire ever-diminishing cross section of a funnel. Pressure and friction increase, and nothing is really filtered out. And if you get an artistic graphic designer drawing the funnel from a 3D perspective, I don’t know what it looks like.

By: Damir

Damir — Tue, 22 Apr 2008 11:54:40 +0000

I guess the only worse one– or equally bad–is the funnel chart. Here are the two showing the same data-set: http://www.damirsystems.com/?p=99

By: mr tom

mr tom — Fri, 18 Apr 2008 08:35:16 +0000

Hi Matthew.

I’m a recent convert to microcharts.

It has some bugs and some annoyances. It takes a bit of getting used to.

But I love it.

It lets me do stuff that excel simply doesn’t otherwise do.

Well worth the investment, and there’s a limited time demo you can use to convince yourself.

By: Jon Peltier

Jon Peltier — Thu, 17 Apr 2008 17:10:44 +0000

To some extent I agree with Rod. To the extent that the dial is moving rapidly enough that you can sense its rate of change, then the dial provides that feedback. As Rod points out, this is useful in a motor vehicle, since you can watch the RPM ramp up over the course of half a second or even a few seconds.

But dials in dashboards are not updated at the frequency of the tachometer in your car. I think Rod and Matthew are both saying this. Thus you need a different display approach that shows current and historical values. A line chart satisfies this approach.

If you only need to show an instantaneous value, Stephen Few’s bullet chart is much more efficient in terms of space. You could fit at least three or four bullet graphs in the space of a single gauge. The bullet graph is described here: http://www.perceptualedge.com/articles/misc/Bullet_Graph_Design_Spec.pdf, and Charley Kyd’s Excel implementation is here: http://www.exceluser.com/explore/bullet.htm.

By: Matthew Pfluger

Matthew Pfluger — Thu, 17 Apr 2008 14:41:16 +0000

Rod, I respectfully disagree. In my opinion, dials only show an instantaneous reading of a value. Dials do a poor job of showing rate of change unless you either constantly refresh data or watch for a long time, something people are unlikely to do with Excel dashboards.

For example, the auto shop guys that I know install dial gauges not to see rate of change but quickly see the current value; engine RPM, how close to Red Line, road speed, etc. Other graphs (lines) are better suited for rate of change. However, when I only need an instantaneous value in relation to a min and max, I have no issue with dial graphs.

By: Rod McInnis

Rod McInnis — Thu, 17 Apr 2008 14:18:53 +0000

Speedometers in automobiles tell you a lot more than how fast you are going. The reason that all cars use dials is that they allow the driver to see the rate of change in addition to its current value. Speedometers don’t just tell you how fast you are going but also tell you how rapidly that speed it changing. By looking at a tachometer you can see how quickly you are approaching the redline so you can time your shift appropriately. Back in the 80s a lot of sports cars showed up with digital displays that looked really cool but were a big step backwards for data visualization in that application.

Like everyone agrees, dials don’t do a good job at presenting business information because it does not change rapidly while you are look at the gauge. For rapidly changing data that you need immediate information on both the value and the direction/rate of change, dials make a very useful and elegant chart.

I have been coming to this site and directing others to peltiertech.com for years now. The bolg is a great addition to give me a reason to stop in more frequently. Good work!

By: Matthew Pfluger

Matthew Pfluger — Thu, 17 Apr 2008 13:21:10 +0000

mr tom,

What has been your experience been with MicroCharts? I just discovered them while reading posts on this topic, and I am thoroughly impressed. So are the managers and bean counters at work, and they are interested in learning more. Could you offer some perspecitve?

Thanks,
Matthew Pfluger