👆 Click on “Broadview” for more book news

Whenever I visualize data, whether it’s a static graph, a dynamic graph, a report, a part of a blog, or even a Twitter image, I follow these five principles. 

1. Display data. 

2. Reduce clutter out.

3. Combination of graphics and text.

4. Avoid using pasta drawings.

5. Start with gray. 

Presenting data and reducing clutter means reducing excess gridlines, markers, and shadows, all of which can interfere with the actual data.

Strong titles, better labels, and useful comments will make the chart work with the text around it.

When a chart has many data series, you can strategically use color to highlight the series of interest, or split a dense chart into multiple small charts.

Taken together, these five principles remind me to focus on the needs of my audience and how to tell a story with visual data.

Readers can only understand your focus, point of view, or story if they see your data. This doesn’t mean you’re going to show all the data, but you want to highlight the data that supports the point. As creators of charts, we are challenged with how much data to present and how best to present it.

This U.S. dot density map uses U.S. decennial census data since 2010, with each point representing a person, which is the distribution of the nation’s 308 million residents in census tracts (one census tract is equivalent to one block). Note that this map has nothing but data, no state boundaries, no roads, no city signs, and no signs of lakes and rivers. But we can still see that this is the United States, because people tend to live in border and coastal areas, and these data traces outline the shape of the entire country.

The similar principle of Gestalt helps us to see how people are concentrated in the United States. Image source: Copyright, 2013, Weldon Cooper Public Service Center, Rector and Visitors of the University of Virginia1, Dustin M. Produced by Dustin A.Cable

This does not mean that we have to display all the data all the time. Sometimes charts show so much data that it’s hard to see which data is more important. For example, these two line charts show the average number of years of education in 50 countries around the world.

In the chart on the left, each country is represented by a polyline of a different color. This leaves the whole chart so confusing that it is impossible to see the trends in any one country.

In the chart on the right, six countries of interest are highlighted, all of which are grayed out as background information.

In this way, the reader can see at a glance the country we want to emphasize. This is not to say that we want to show the least amount of data, but that we want to show the most important data.

Only a few countries are highlighted (right) so that the chart is easier to read

Using unnecessary visual elements can distract the reader and clutter the page.

There are a lot of pitfalls that cause confusion in the chart to avoid. There are some basic elements, such as tick marks and gridlines that are too thick, which can be removed almost directly.

Some charts use data markers, such as squares, circles, and triangles, to distinguish between series, but when the markers overlap, they can make the chart look messy.

When charts that work well with simple, solid colors, never do textures or gradient fills. When unnecessary 3D (stereoscopic effects) are used, the data is distorted.

There are also charts that contain too many text and labels, making the whole space chaotic and crowded.

Take this three-dimensional histogram of the average years of education in the United States and Germany.

You’ve probably seen this kind of 3D chart before – distracting, difficult to read, and distorted

If you think no one would design such a strange diagram, then you are wrong.

This is a direct copy of the chart, including its gradient style. Three-dimensional bars and flashing stripes, mismatched data and axis labels, indicate the accuracy of the data with a large number of decimal places, but in practice there is no such effect – all of this mixes together to form a chart that is hard to read and, to be honest, looks uncomfortable.

At the same time, 3D graphics can distort data. This distortion occurs because of the use of unnecessary 3D perspective effects. Simplifying the chart by discarding these irrelevant, distracting elements can make your point clearer and easier to understand.

While our understanding of perception, and how the eyes and brain work, is mostly rooted in scientific research, deciding what visual effects to use is often subjective. For example, what kind of chart to use, where to place labels and notes, what colors and fonts to use, etc.

Using a basic histogram eliminates the clutter and distortion caused by 3D effects, so charts are easier to read and understand

In some cases, using a chart is objectively wrong, but in most cases, it is up to you to judge subjectively. As you create and read more and more visualizations, you’ll broaden your horizons, improve your aesthetic abilities, and find a balance between art and science.

Although we mainly focus on the elements that create visual charts, such as bars, points, or polylines, the text description of the chart is equally important. We often think of text and annotations as something to think about after the fact, but these elements can help the reader understand what the chart contains, as well as the chart itself.

Amanda Cox, the data editor for The New York Times, once said, “The notes section is the most important … Otherwise it’s the equivalent of saying, ‘It’s all here, figure it out for yourself.’ ” 

Adding the right annotations to the chart is critical from the point of view of helping the reader understand.

There are three ways to blend charts and visuals: remove the legend, create an attractive title, and add some details.

1. Remove the legend as much as possible and directly label the data 

Placing labels directly on the chart makes it easier for readers to find the corresponding data

2. Write the headline like the headline of a newspaper 

A good title needs to be able to grasp the main points of the chart and tell the reader what conclusions can be drawn from it. I call these “strong headlines” or “newspaper-style headlines.”

The title of this chart from the Pew Research Center tells you exactly what you should learn from it

3. Add a comment 

Once the chart is made and the title is decided, ask yourself if you would add some more text descriptions.

Sometimes there are peaks or valleys, discrete values or fluctuation values in the data that need to be interpreted. Adding details to the chart will help you derive your argument or key points. If you are using a non-standard chart, you also have to explain how to read it.

The short description in the image on the right explains some of the basic characteristics of the data

Spaghetti Chart is a term used in manufacturing to refer to the kind of chart that holds a lot of data.

Obviously, when a chart contains too much information — a line chart looks like a bunch of spaghetti strips, and a map of dozens of colors and icons, or one bar after another that fills the entire page. This is indeed a challenge when a chart contains a lot of data, but we don’t need to put all the data in one chart.

Two examples of small multiples. The image on the left, from Zeit Online, shows the average temperature in Germany over the past 140 years. The image on the right, from the Centers for Disease Control and Prevention, shows how facial hair affects the installation of a respirator. The connection principle of Gestalt helps us keep track of changes in the diagram

We can break down a chart into multiple charts. This is called a grid or panel chart, also known as a grid chart, or a small sequence diagram. These smaller charts use the same scale, axis, and extent, but spread the data across multiple charts. In other words, instead of putting all the data in one chart, create multiple smaller versions on top of the underlying data. 

Small sequence diagrams are not a new or revolutionary way of expressing data. In 1878, photographer Eadweard Muybridge wanted to determine whether a horse was fully vacant as it galloped. Mukbridge developed a technique to photograph a galloping horse, which can take a series of fast-moving photographs (what we now call stop-frame). His photographs prove that the horse did indeed leave the ground completely as it galloped. Image sequences, which also give a sense of dynamism, are an early example of small sequence diagrams.

Photographer Edwid Mubridge used small sequences as early as 1878 to determine whether a horse was fully vacant as it galloped

Small sequence diagrams have at least three advantages.

First, once the reader knows how to read one of the charts, they will read the others.

Second, you can display a lot of information without confusing the reader.

Third, the reader can make comparisons across multiple variables.

The Guardian’s example shows the results of the 2016 Brexit resolution on six different demographic variables. The horizontal axis remains unchanged, and it is easy to see the relational direction of each demographic indicator.

The Guardian’s multiple small scatter charts show the relationship between voting choices and six demographic variables. The similar principle of Gestalt makes it easy to see two types of data in each scatter plot

However, there are also some drawbacks in such series diagrams, which can be confusing if not avoided.

First, the charts should be arranged in a logical order. Instead of letting the reader go around the entire page, use intuitive sorting methods, such as geographic location or alphabetical order. 

Second, the chart should use the same layout, size, font, and color. Keep in mind that we are dividing a chart into multiple charts, so it should look like one chart has been copied multiple times. The vertical and horizontal axes may change, but you cannot use the blue dot to represent “no” in one chart and “yes” in another.

Third, sequence diagrams should be relatively readable. You don’t have to ask the reader to zoom in and interpret all the details in the chart in detail, your aim is to give them a holistic pattern. The size of these charts is small, so including notes and labels, or repeating lengthy axis labels and data markers, can overwhelm the reader. 

I’ll end this section with a practical tip, a simple step in creating a clear, easy-to-understand visualization: start with gray. Whenever you draw a chart, start with an all-gray element. This forces you to be more purposeful and strategic in your use of colors, labels, and other elements. 

Let’s take a simple chart of average years of schooling, which this time shows only 10 countries. With the colors and labels (the chart in the upper left corner), I can put this chart into my report or handout, do a little work, and add an attractive title so that the reader can know which labels correspond to which polylines. However, if all the polylines are grayed out (the chart in the upper right corner), the reader cannot accomplish the same task because they do not know which polyline corresponds to which country.

Setting all the data to gray out first forces you to think about your purpose and where exactly you want to draw your reader’s attention

Now I can adjust this chart purposefully.

I can add colors and change the thickness of the lines to better highlight the information I want to highlight, like two of them.

The chart in the lower left corner puts all the labels on the chart, while the chart in the lower right corner only directly identifies two countries, and it is obvious that the chart in the lower right corner can convey information more effectively.

Starting with gray forces us to purposefully choose which elements to place in the foreground.

The above is an excerpt from the book “A Guide to Better Data Visualization”, so read this book to learn more about how to effectively present data!

Time limit order minus 50, quickly scan the code to buy it!

Publisher: Liu Enhui

Review: Chen Xinyi


 Hot text recommended  

Refuse to be bells and whistles, and zero foundation can also give you machine learning to understand

Book list | There are 10 good data analysis books not to miss this year!

The Ten Deadly Sins of Cognitive Bias in Digital Transformation

Suddenly enlightened, I know what a real mind map is!

▼Click to read the original article to learn more about this book~