My 10 rules for effective Data Visualization

Data is, they say, the new oil. Gathering data, cleansing data, enriching data… But — what about visualizing data? I’m surprised at how often the value of great sets of data is burdened by inefficient visualization, even by experienced “Business Intelligence” or Analytics consultants.

Yuck!

Yes, art & aesthetics mean a lot in Data, but most of it is Science. You can learn, you can get better. Here are 10 very simple Guidelines I always use that will definitely help you too.

1. Tell a story… or at least, answer a question

Think about your audience. What are they expectations? What do they need? What actions should your visualization help them trigger?

Think about your data. What’s relevant? What message, what story do you want to share?

How are my sales reps performing? That nice! What could we do to help underperforming reps sell more? That’s nicer!

An example of data Storytelling.

Sometimes, title-ing your charts with a question helps you focus on the business question, rather than on lines or bars.

This is a really good article about Data Questions. I also really like this blog for inspiration.

2. Use the right visualizations

Trust me. Chart types are rarely a choice. There are guidelines.

Bars. Use them to compare magnitudes. Yes, even for percentages… pie charts are generally not a good idea. Actually, if you are unsure on what visualization to use, bars are probably the least risky.

Lines. Want to show changes over time? Use a line. For anything else, it’s probably not a good idea.

Never use lines for discrete axis.

Dots. Dots work well instead of bars when you want to highlight the difference (using a non-zero axis)

Scatter. When trying to show the correlation or relationship between to variables. The clustering of the data points (the ‘shape of the cloud’, if you want it) is what you want to focus on (in the following case, life expectancy is clearly related to GDP per capita), as well as outliers (a few countries have an above-average GPD but very low life expectancy… why?)

Notice the use of colour to add context on the country’s continent.

Scatter plots are also used to categorize into quadrants or more “segments”, such as the following example:

There’s plenty of material online to help you choose the right chart. Qlik, for example, has this guide that you’ll probably find very useful.

Choosing the right visualization

3. Use colors wisely… and only if REALLY needed

Colors are probably the most common mistake I see in charts. Colors should be used for a reason.

Useless colors. Probably, absurd sorting of the bars, too.

You can use colors for many useful reasons, such as, helping differentiate metrics visually.

Another example, here color is cleverly used to relate the gaps on the left with the details on the right:

4. Follow the rules, break the rules

Yes, there are rules.

• Did I say that already? Pie charts are rarely a good idea.
• If you reaaaaally need a Pie, always sort slices by size to easy data interpretation. And don’t use more than around 5 or 7 slices.
• Non-zero baselines should be used with care. And never, in a bar chart. They lead to misunderstanding the data.

Non-zero axis are often used to manipulate the audience

• For non-zero baselines, use point charts instead.
• Bullet charts are useful to compare data with previous years and / or with goals.

Sometimes though, you need to break the rules. Sometimes you’ll want to use color differently. Or change the chart type to send a different message. Try to and, if it helps tell the story you want… stick with it.

6. Add context as needed

Is there any text you should add to help the reader understand your story?

Can you visually highlight relevant information? Add a business description to each of the scatter quadrants? Circle or change the color to the top performing country? Add a reference line to showcase what bars are above or below average?

7. Keep it simple, Stupid (KISS)

This is probably my golden rule, and probably a summary of all previous ones.

I continuously keep asking myself: is this really necessary? Why am I using this technique? Would the message I want to convey loose anything without that?

That rules out things like:

• 3D charts (ouch!)
• background grids on charts
• (most) data tables.
• Too many data points (bars, slices, or points)
• Most imagery (some icons can be useful, though!)
• In fact, any “too many objects” in a dashboard.

If it can be removed, it should be removed.

8. Add some spice

At times, you can use less commons visuals, if you know them, such as bullet charts or slope charts. You can even create your own (and, maybe, get famous).

You’ll need to read a lot to discover this hidden gems. And be careful, but daring with them. Finding an opportunity to use one the them, often feels like winning the lottery.

Slope chart

9. Iterate

Your first visualization is, well… wrong.

And the second one is … barely better.

Iterate. Think. Try different stuff. Ask for feedback. Change it again.

These makeovers are good examples on how to do that.

10. And keep learning!

Data visualization is a whole world. There are gurus whom I really recommend following. Blogs with amazing content. There are many examples and videos that you can get inspiration from.

It’s also all around you, unexpectedly: from the news to blogs, websites, politics, economy, COVID… Use every opportunity to think: is that a good idea? how could I make it better?

I hope these guidelines are helpful for you!