Join Bill Shander for an in-depth discussion in this video Exploring chart options, part of Data Visualization for Data Analysts.
- This is a big subject, picking the right chart for the right data. How do you pick the right chart for the data you have? You want to keep your audience interested. You might be worried about using, "Boring charts." At the same time, there is a visual language and user experience reality you have to deal with. Your audience knows how to read and understand certain charts. Sometimes you want to give them something so easy even a dog could understand it, right? Not to mention, that if you have complex data, and whose data isn't more complex than you might think at first? Then you often need to find new and unique ways to display that data.
First things first, bar charts work, they work really well. If your data works in a bar chart, use a bar chart. Humans can pre-attentively process the rectangles. Remember we talked about in the Gestalt principles, pre-attentive processing, that immediate ability to recognize certain things quickly. We know how to read bar charts, we can easily judge the relative size of these rectangles, they're a cliche for a reason, they work, so use them. Feel free to make them look nice, get creative with them, but use bar charts.
Same thing goes with a line chart. If you're showing data over time for example, a line chart's great. Let's talk about a few other chart types that don't get enough play, and that I think the world us ready to see more of. One is the slope graph. They're great for overlaying a lot of items to compare change from one point to another. They can get overwhelming and confusing because there are so many things going on here, but you can always use contrast and color and thickness of a line to really draw one or two to the foreground, and make the rest fade back. A similar concept is something called parallel coordinates.
It's just like a slope graph, but it has more data points. Maybe here we're looking at data over four different points in time. It's often used to show four different attributes of data. I've seen this one for example, used to show different nutritional components of an item of food. You might have four different vitamins, and what the measure of vitamins is in that different food. Again, that can be very overwhelming, but if you're judicious about these of color and contrast, you can make them understandable. Another example that I really like is the dot plot. You can overlay multiple points in a much tighter space, so it's really good for comparing a lot of things, where a grouped bar chart as an example, would be completely overwhelming and useless.
Another chart that I really love is the box plot, box-and-whiskers. It doesn't get enough play outside of stock charts, but where else can you put five plus data points into such a compressed space? Not to mention leaving room to display outliers. If you have to show that range of values, quartiles, and top and bottom percentiles, and averages, and outliers all at the same time, it's really an efficient chart type to use. This can be a little bit less accessible to some of your audience, but with a well-labeled key, I wouldn't hesitate to use it, it's powerful enough to justify the challenges.
Stream graphs are one of those things that they went through a period of overuse, it seemed like they were everywhere for a while, but it's died down, it's something that I'd recommend for certain types of data. There's something very compelling and easy to read about this form for showing the relative share of categories of things over time. It's just like a stacked area chart, but maybe because it sort of looks like an EKG, or maybe because it's not sitting in a traditional chart axis, it's a bit more interesting to your audience. Another type of data that can be difficult to work with is hierarchical data, and there are so many ways to show hierarchical data, and different types of network graphs, but those can get really overwhelming, especially when there's a lot of data.
I love this approach, the tree map. One of the reasons I love it is that it's rectangles. Remember I keep saying it's really easy for humans to parse rectangles and to understand how the area of one rectangle relates to the area of another rectangle, how much bigger is it, etc. If you have lots of layers of data, you can add interactivity to a chart like this, so if I were to click into one of these boxes, it can zoom in and reveal more rectangles within it, so that hierarchical information doesn't get lost, even though it can be greatly simplified using this visual approach.
Tree map is a great way to go to show this type of data. Another example that you’re starting to see more of, even in newspapers for a very general audience, are small multiples. For instance, this collection of maps over here, it's easy to see patterns in the data even when I'm looking at a lot of them side-by-side. You see them very frequently in stock charts where you’ll see small multiples of bar charts, or usually line graphs, even scatterplots, you can see patterns in these small multiples, another approach that I think could stand to be used quite a bit more. My point with all of this is that there're plenty of ways to show data, we know that.
Sometimes it can be hard to find, "The right way." In fact, really there is no, "Right way." You want to keep your audience interested, you want to use alternative forms, but you want to use alternative forms that are understandable. So experiment, as I said before be inspired and just find great forms to show the data and to tell the stories you're trying to tell.
- Why visual communications matter, and how they work
- Communicating via story
- Communicating with color
- Using legends and sources
- Sketching and wireframing
- Rethinking slides, charts, and diagrams
- Rethinking your templates and brand guidelines