Pyramid charts

Pyramid charts! The best under-used chart in the world! Pyramid charts seem to be pretty much synonymous with "population pyramid charts", like these:



The shapes of some demographic pyramids are so stereotypical they can be referred to in shorthand without even showing the chart.

But pyramids are great for so much more! I understand that you can show the same data with a pair of bars, but I find such charts pretty much impossible to interpret and, well, boring. Pyramids leverage our incredible sensitivity to asymmetry to good result. Here's a simple example from The Washington Post:


So much better than some bars. I mocked up the same data in a bar chart in Domo:pyramid_wapo_1_alt.png













Is it just me, or is this chart just super boring? I'd be tempted to put it in a table and save some space. But with a pyramid or slope chart chart? It would be much more interesting. For our project, we could use these kind of displays everywhere. A lot of our work is in visualizing change over time and difference between facilities. So, our left/right split could be by department within two facilities or departments in one facility for two different periods.


Pyramid charts can be much richer than their classic use in population pyramids, as shown from this second example from The Washington Post:pyramid_wapo_2.png



























Regardless of where you fall on the various lines above personally, this is an outstanding chart. The asymmetry is exploited masterfully and the color shading makes each bar able to carry an extra value/category/degree of data. Impressive. Clearly, I'm a fan of pyramid charts and, again, we could use them constantly to compare and contrast departments/divisions/facilities/teams either against their cohorts or against time. 


I've got one extra idea: Display overlap. A traditional pyramid has a dividing line at zero and then a positive number line heading in either direction. The modal is some attribute like gender, sampling period, or what have you. In our case, we often want to compare overlapping groups. So, for example, imagine that you're comparing teams where some members are part of the same group, or markets where some products are in both markets, and some are not. I'll mock something up to show what I'm talking about. Here, imagine a company that sells breakfast foods in the USA and Australia. They're interested in revenue by product category, which is easy to do in a pyramid chart, like this (with American-sounding product categories):




















So far, so normal. But now they say "well, we've got some products that are sold in both markets, and some that are sold only in each, what does that look like?" Depending on what the data is, there are a few really interesting ways to graph this:


* If you're doing something more like counts say "products in each category", then you can treat the rectangles as small, individual Venn diagrams.


* Given that it might be hard to draw a Venn that's proportional both the the intersection and the individual lines, bars might be a help:


* Best yet might be just what the Wapo did with their chart, a category breakdown within each bar:



None of the calculations in a pyramid, even the more complex one, are particularly hard - it's a kind of weird bar chart. They're very easy to interpret, exploit our natural sensitivity to symmetry/asymmetry and are pretty flexible to boot.


Please consider them for a future release, I'd guess they would prove pretty popular!


Note: I asked about pyramids a few weeks ago and Godiepi share a clever workaround:


It turns out that the negative numbers are a bit hard to explain, but I appreciated the workaround. So, thanks again to Godiepi if he visits this thread.


Thank you.

3 votes

· Last Updated

This discussion has been closed.