The five indispensable rules of data visualization

Lesley Lathrop
K2 Data Science & Engineering
5 min readAug 24, 2016

--

You may be thinking, “Rules? But I make art and art knows nothing of rules!”

To which we would reply that the presentation of data is more than just an art. It’s a science, too. And science does have rules.

So data display is both an art and a science, and it has rules. What are those rules? We’ve come up with five. And the best part? Once you learn the rules, you can start breaking them!

RULE 1: BE HONEST

Take a look at these two data graphics. Can you spot what’s wrong with them?

If you said something related to ‘scale’, you would be right.

The graph on the left fails by not setting the y-axis with a zero baseline, thus distorting what’s really going on. The bar on the left shows that six million had signed up for Obamacare. The bar on the right shows the goal. But the bar on the right, which is more than twice the size of the six million bar on the left, shows that there is a difference of just over one million between the actual and the goal.

The chart on the right uses only 15 years of data (and the data selection itself may be problematic, but we won’t go into that here). Actual climate change graphics, such as the one below, use hundreds of years of data from multiple sources.

Source: International Panel on Climate Change (IPCC)

The two subpar graphics above break the first rule in that they present actual data in a misleading way. This is a typical problem in politically or agenda-driven graphics. Don’t do it!

RULE 2: MAKE A POINT

In order to make a point, you first have to have one. Those of us who love working with data find beauty in numbers. But no matter how beautiful your numbers are, if you don’t deliver useful conclusions and insights from the data, you will have violated the second rule.

RULE 3: USE APPROPRIATE CHARTS

The four most commonly used chart types are bar, line, scatter and pie. But these four types are far from the only types available, and they are frequently not the most effective at making your point.

Consider the pie chart. Ask most statisticians what they think of pie charts and they will likely laugh at you. This is because pie charts are notoriously ineffective at showing relationships in data, especially if you have more than two or three categories.

Look at the pie chart on the above left. In that chart, there are so many categories, it’s difficult to discern the point being made. What’s more, research has demonstrated that pie charts are simply too hard for the human eye to understand, because it asks viewers to ascribe quantitative value to angles and areas on sight. A more effective graph of the same data is the horizontal bar chart on the right.

To determine the best type of chart for your data, take a look at this graphic from information designer Anna Vital. And if you need some inspiration to think outside of the conventional visualization box, check out the index of 1000 examples at Visual Complexity.

RULE 4: EMPHASIZE THE MESSAGE

The significant conclusions and insights you’ve drawn from the data are your most important messages. Highlight them visually!

Source: slide:ology by Nancy Duarte

The above chart comes from Nancy Duarte’s book, slide:ology and it demonstrates how best to emphasize your message. Even if you aren’t using one of these chart types, you can still take inspiration to make bolder the thing you want viewers to notice. You need to think a bit like a graphic designer here, using elements such as size and contrast to shape your message.

RULE 5: LOSE THE CLUTTER

The 3D chart below violates this rule on so many levels, it’s hard to know where to start.

First, a cardinal rule of data visualization is this: DON’T USE 3D! Charts in 3D are next-to-impossible for the eye to comprehend, even if you use labels, as this one does. Indeed, the labels in this case add to the sheer confusion this graph causes. So again, don’t use 3D.

Second, there are so many things happening in this chart, it’s not clear what specific message is supposed to be conveyed. The positions of the bars convey one aspect of the data, the colors another, and the sizes still another. If the goal is to display proportions, use a bar chart. If the goal is to display a trend, use a line chart. And if the goal is to display relationships between variables, use a scatter plot.

We have discussed five rules for data visualizations. Know them. Use them. Break them. But don’t forget, sometimes a chart isn’t even the best way to convey your message. The image below powerfully conveys its message because it makes use of the power of simplicity. And, as Leonardo da Vinci is supposed to have said, “Simplicity is the ultimate sophistication.”

Source: London Media

This article was contributed by Lesley Lathrop, staff writer at K2 Data Science, a part-time, remote data science bootcamp. For more information, inquire by email to hello(at)k2datascience.com.

--

--

Freelance writer. Data geek, data wrangler. #DataScience #DataViz #rstats. Crazy lifelong fan of Thoroughbred racing! Find her on Twitter @lesleylathrop