In that last post, we introduced the concept of The Visualization Frontier. It’s a simple rubric to measure the effectiveness of your data visualization.
We defined effectiveness as the combination of clarity and engagement.
The closer you are to the red line, the more effective you are.
(We hinted at the possibility that the red line is not the same for everyone - that it’s tied to your capability as a data visualizer. A super-skilled information designer might have a red line that stretches almost to the top right corner. A newbie might be a straight line from top left to bottom right.)
In any case, you want to improve. You want to have visualizations that approach that top right corner. How do you do so?
Ask the Right Questions
Let’s assume you have some data to visualize. Are you aiming more towards clarity or engagement? Here are some questions you can ask yourself:
Who is your target audience?
What is the main thing you’re trying to say?
What do you want your audience to do about it?
Try to be as specific as possible as you explore these questions. It might even help to build them into a use case or user story. The process is almost identical to the steps you’d take if you were designing software.
Things that are more decision-oriented will lean toward clarity. If you’re trying to delight people, then it’s more about engagement.
For example, if you’re building a dashboard for an air traffic control center, then you want to be five-star clarity--your audience is already super-engaged. If you want to get a million new Facebook friends, then you need five-star engagement--clarity is secondary.
Visualize the Signal
Separating the signal from the noise is the essence of clarity. And clarity is where you start. Don’t be tempted to imagine the dancing bologna until you’ve locked onto the signal. The signal will tell you what chart type to use, what layout will work, and even how the user can best interact.
Many projects go awry right here. You jump ahead to a cool concept and only later discover that the data doesn’t support it. You do this because design is the fun creative stuff and analyzing data isn’t. Unfortunately, great data visualizers are first and foremost great signal finders. They’re data analysts at heart. This is why there are so few of them.
I would go so far as to say data visualization is a misnomer. It should be called signal visualization. Exploring the data to find the signal isn’t something you should leave to others. It’s fundamental to your job.
Think back to the suicide graphic in the previous article. Do you think the creators of the visual stumbled into that particular format by accident? I highly doubt it.
I imagine they explored a much larger data set that probably included race, region, socio-economic status, and a host of other variables. Through careful exploration, they discovered that age cohort was a strong signal - perhaps the strongest. They then crafted a visual that would bring that particular signal to the forefront. This effort is what makes the visual so effective.
I can’t emphasize this enough: signal must come first, even if your aim is primarily engagement.
Improve the Engagement
Engagement is a creative process, but it can still be systematic. Here are four things you can do:
1. Colour to set the tone
Colour’s impact is instant. You can use it to evoke different moods, to spark an emotional response, or to draw attention to a specific element. Think of it like a soundtrack in a movie. It sets the tone.
Start by choosing a palette that fits your purpose. Red is bold, dangerous, or angry. Orange is playful. Etc. Just don’t overdo it. Each colour you use attracts attention, so the more colours you use, the more your audience’s attention will be split. Don’t forget about grey. It might be the most useful of them all because it conveys information without fighting for attention.
If you don’t have a good feel for colour, simply choose one to highlight the signal, and a second (usually grey) to use for context and background. Two colours are often enough. If you want to dig deeper, this post from Datawrapper is a great starting point.
2. Animate with purpose
Animation is the ghost pepper in your visual toolbox. Too much is deadly. If you overuse it, it’ll catch your eye like those little hooks on old bathroom stall doors. Nobody wants that. Instead, animate to ease your user into the visual or lead them through a narrative.
Go ahead and introduce elements of your visualization sequentially. Users will be able to understand a richer, denser visual that way. Simulate the passing of time by animating your bars or lines on a chart. Peel back contextual items one by one to reveal the signal. Don’t animate willy-nilly. Use it to deepen engagement by fostering understanding.
In keeping with our rather morbid theme, I present a simple, yet very effective animation of Selected Causes of Death. By applying animation to a simple bar chart, the authors have turned a boring column chart into a compelling story:
3. Narrate to connect
Speaking of stories, humans spent millennia spinning them around campfires. Narratives are so hard-wired into our brains that we’ll create them even when they’re not appropriate. Use this to your advantage.
Introduce some of the insights in your data through a story - whether text or video or animation. Lead your user through your visual to both introduce functionality and to demonstrate the possibilities. These narratives become on-ramps to complex visualizations. They help users cross that great chasm between passive consumption and active exploration.
Don’t underestimate how difficult this crossing can be. Only a fraction of your users will do anything more than scroll.
4. Interact to discover
If you’ve done well on colour, animation, and narrative, then your users may actually entrust you with their full attention. They’re no longer passive consumers, but active participants.
Don’t abuse this trust with frivolous controls.
Have a laser-like focus on the purpose you defined above. You’ll be tempted to let your users change the fonts and the colour palette and the chart type (“because some users might want to!”). No you don’t - you’re not building Tableau. Giving users “the option” confuses them more times than not.
Instead, channel your inner Steve Jobs: what’s the fewest number of choices you can offer (while still fostering a sense of ownership)? What’s the least amount of interaction that still allows discovery?
You want a sandbox with walls that make it difficult to get lost or self-deceived, but big enough to own the results. Once again, think back to that suicide visual. It had very little interaction, but it was enough to discover the signal.
This is an incredibly difficult balance. Your design choices here are what will set you apart. How much context, how much comparison, and how much drill-down is needed for the decision at hand? Mock it up and then pare it down.
Have new users play with it and see where they get lost. Keep revisiting your purpose and the signal you’re trying to reveal. You’ll know you’ve arrived when your testers figure it out without a tutorial.
That last five per cent
When your house is built and painted, you’ll feel like you’re almost there. Unfortunately, half the work is still to come. Finishing carpentry is a huge time burner.
Similarly, your “feature-complete prototype” is a long way from prime time. The polish can take forever. Push your project toward greater clarity and engagement. Get others to rate your work on these dimensions. Then, bare down for that last five per cent to the frontier.
In the end, great visualizations are far more difficult to create than they appear. But the extra time and attention is what separates the memorable from the filler.