Whether it’s detailed figures in a company report, important insights in a presentation or stunning numbers for a press release, data is a vital part of our communications. But that’s not to say it’s easy.
Data visualisations (charts, graphs, etc) are a key tool in helping people engage with and understand data. Using them effectively, though, means knowing how they work and how they work best. Visuals aren’t just there to make our data look eye-catching. By representing data with simple shapes like squares and circles, we hijack the human visual system to help readers ‘see’ the data without having to think about it. Visualisations magically help people grasp difficult data instinctively.
This magic gives us two things to think about, however. One is that it only works when we don’t get in the way of the visualisations doing their thing, which means we have to be careful about decorating and fancying up our charts. The other is that because they communicate unconsciously, charts can also change the way a reader thinks about the data without them realising it. In fact, charts tell stories.
This isn’t a deliberate thing. Because of the way we interpret visuals instinctively, charts will frame the way we interpret the data those visuals represent. Charts tell stories whether we want them to or not. And different charts tell different stories.
When it comes to choosing charts for our data, then, we have to reverse engineer this process. We have to think about the stories our data is telling and then choose the chart that will make that story evident.
Our data is usually telling one of five core kinds of story:
- Geographical distribution (which, irritatingly, does not begin with ‘c’).
We’re often comparing or contrasting data points: one brand or country against another, one year or region against another, how one trend is different to another, or how one data point stands out from the rest.
The best charts for comparison are, without a doubt, bar charts. Each data point is represented by a separate rectangle, all aligned to the same base, allowing us to easily see which is the tallest, the shortest, and to directly compare the comparative sizes of the values.
Of course, in many ways, bar charts are the default chart for pretty much every use: everyone recognises them, everyone understands them, everyone can read them.
Only when, however, they are used clearly. It can be really tempting to suspect that your bar chart is boring and to try and liven it up. This is almost always a mistake. Their usefulness lies entirely in their simplicity.
There’s simply too much going on in the first version above. Too many fonts, too many colours, too many visual elements. Calming it all down makes it a lot easier to read.
It can be very hard to see change over time when just looking at a column of numbers, but line charts make it immediately evident.
By joining together the data points with lines, line charts show us how each data point leads to the next, or change from the previous one, the steepness of the slope of the line giving the reader an clear idea of just how different one point is from another.
They need to be used carefully, however. You shouldn’t use them for data that isn’t related, and they don’t work even for every dataset that’s about time -ages, for example, might be about comparing generations rather than one person growing older.
The first chart is perfect data for a line chart, but the second, of changes in age groups, makes much more sense as a column chart than as a line.
Very often with survey or demographic data we’re talking about composition – this whole population is composed of these parts – 52% of people believe one thing, 48% another, the majority of sales come from this division, only a few regions are currently served.
Pie charts are perfect for composition. A circle wonderfully captures completeness, and then we divide it up to show the parts, just as we would cut up a pie into slices. Also, we can read divisions of a circle pretty accurately – after all, it’s how we tell the time from a clock face.
But that’s because a clock face only has three hands. Pie charts need to be simple to be able to work and they’re not great for detail – they work best with fewer slices and with visible differences in value.
Way too much going on in that first chart – and you should never use 3D in flat data visualisation. Fewer segments and calmer visuals are really important in pie charts.
Correlation – that one trend or pattern is related to another – is a common story in data, but a very hard one to express. Traditionally correlation is shown with scatter plots, because you can show multiple datasets at one time, but these are complex charts and rarely used outside of academic papers.
For general audiences you’d more likely chart each dataset separately (in a bar or line maybe) and just show the two charts side by side. Make them clear enough and audiences should be able to compare them easily.
You can pack a lot into a scatter plot, but it takes time to figure out what’s going on. The two charts side-by-side tell the same story much more quickly.
5. Geographic distribution
Geography is easy because there’s already a way that everyone knows to show it: a map. Maps are great – people recognise them, know how to interpret them, are almost always interested to see them.
However – maps are also tricky. Countries are often inconvenient and unusual shapes, and not always recognisable out of context. More importantly, geography is not population, or income. Massive countries like Canada or Russia can be sparsely populated, whole continents might not show up in our datasets. Not to mention the small fact of the majority of the Earth’s surface being water.
Sometimes the right answer for geography is, after all, a bar chart or bubble chart, allowing you to compare all the countries, or a table of the regions.
There’s no data here for North America, but it’s still going to take up a segment of the map. This data might make more sense as a bubble table – hopefully people know where these countries are.
Whatever the story you have to tell, knowing how to match chart types to story is key to leveraging the magic of visualisation, helping your audience see, understand and act on your data.
For more information on data visualisation, check out the book Communicating with Data Visualisation here.