My phone thought I’d be interested in this article on coffee consumption by country. It wound up being a good call, largely because I do my part to bring global coffee consumption numbers upward but also because the first chart caught my eye:
Now, to be clear, this isn’t terrible. You can see where it’s the most expensive, and it’s not particularly hard to find a country. That doesn’t mean there aren’t critiques:
Radial bars make comparisons really hard. If I want to compare Hungary and Italy, they’re on opposites sides the bar height doesn’t help. I can understand why a radial chart was used here - there are a lot of countries and this gets them all on one chart - but it’s typically not my first choice.
Using different types of coffee for the legend is an interesting choice. Data viz purists would say it’s distracting and/or meaningless, though given the audience for this is “people scrolling their phones” I don’t think it’s a bad way to grab attention.
There are some non-data distractions - the grid lines aren’t necessary, the coffee in the middle is probably unnecessary, and the background doesn’t add anything. These are minor things, but nonetheless distract from the data*.
Let’s end on a high note - they didn’t try to make a geographical map out of this. That’s an easy temptation for any analyst, but a map isn’t always the right choice for this sort of thing. You’d never see Luxembourg’s high consumption on a traditional map, for instance.
*I want to be careful here - some data viz practioneers can come off as rather anti-fun. I’m not against flair in a chart per se, and perhaps I’m just more used to building charts in a business context. Still, let’s make it easy to focus on the data, eh?
The article itself does a good job going through the data*, and there’s a table that’s a lot more useful than the chart. It’s paginated, so it’s not overwhelming to look at. I want to stress that it’s really hard to display data when your dimensions have so many members - anything over 10 and the brain begins to break.
*They got their data from this article, which uses a similar technique as mine to show data, something I didn’t see until I’d made my chart. Always nice to get confirmation that the approach is valid, though.
What we can do is panel the data, that is, show ~20 per column and move to the next. This makes it hard to compare individual countries, at least if they’re in separate panels, but we get to keep things in one view. Thomas Sowell’s quip that “There are no solutions, there are only trade-offs; and you try to get the best trade-off you can get, that’s all you can hope for” applies to data viz as much as it does to economics.
This is cleaner, we’ve removed a lot of non-data ink (backgrounds, pictures, gridlines, etc.) so the focus is on the data. Again, nothing against an infographic approach, but in most contexts you want something simple. Especially in a business context, you don’t want to have to spend time explaining things that aren’t related to the data story you’re trying to tell.
You shouldn’t do this for more than one metric. If you need multiple metrics, you’re better off either putting all the dimensions in the same column and letting the user scroll or creating multiple charts. Another option is to show the top 10 (or 15, or 20, or whatever) members of your dimension and grouping the rest into an “other” bucket. You’ll just need to be clear on what defines top n, especially if you do this in an interactive format like Tableau and you’re letting users decide what metric is used for this.
That leads into my final point - you chart choice will depend on how it’s being presented. Interactive provides a lot more options, since it’s generally fine for your users to scroll. If it’s a static chart going into a deck somewhere, you’ll need to be creative. Knowing your options and watching what others do is your best bet, the more tools you have on your belt the more likely you are to have something appropriate to your data and audience.
