Design principles

Preattentive attributes affect users before they start to process information. Design principles help users as they try to make sense of our charts. So now let’s look at some major design principles that apply to creating charts and data.

Simplicity

Simplicity is one of the most important principles in chart design. We can apply simplicity in many areas of chart or visualisation design. First, ask yourself if you really need a chart or visualisation. Information is sometimes better presented in text or a table.

If a chart is the best choice, make it as simple as possible for your audience to understand.

In his book, The visual display of quantitative information, Edward Tufte coined the term ‘chartjunk’ to represent anything in a chart that’s not needed to understand the components of the data.[1] Both the term and the definition imply that chartjunk should be removed and that, as a result, important aspects of your data will stand out. People have different perceptions of what’s important, but keeping this principle in mind can help improve the design of your charts and visualisations.

Examples of chartjunk include:

• distracting formatting (eg, 3-D effects, borders, or background images)
• gridlines
• redundant legends or titles.

We’ll see how to apply this guidance to specific chart types later in the weeks. For now you can check out some examples of how you can go from Excel defaults to ‘clean’ charts by reading the PDF below.

Read: The Shaffer 4 C’s of data visualisation: ‘Clean’ examples [2]

Finally, remember that you don’t have to show all of your data in one plot. It’s often better to keep individual plots simple and clear and to use multiple plots to make comparisons, show trends, and demonstrate relationships between variables.

The example below starts with a stacked bar chart and converts it into three separate bar charts. In this format, it’s easier to compare the three-way split across categories and the different proportions within each category.

Micro versus macro

Some charts are good for seeing the big picture but not good for seeing subtle differences. Choropleth maps use colour, but users might not be able to see small differences between them. You might also find that important regions are too small to be shown on a map.

If we apply this principle to bar charts, you can imagine that having wide ranges in your data can distort your visualisation. If you have a small number of excessively large bars, you could consider using multiple charts to show both the full scale (macro) and a ‘zoomed in’ (micro) view.

Presenting too much information

Avoid adding too much information, or too many layers, to a single chart. Having too many layers eliminates the advantages of processing data visually because the user has to inspect each element individually to make sense of the chart.

Look at the two graphs below to see how confusing multiple layers can be.

Consistent intervals

Don’t skip values when your data is numerical. Trends are misrepresented if the intervals between your dates are inconsistent. The left-hand chart (below) suggests there are values along the entire chart, but when the data is plotted as a bar chart, you can see there are several 0-values.

Baseline or no baseline?

Earlier in the course, we mentioned that bar charts need a baseline of zero to avoid being misleading.

Read: Fox News continues charting excellence [3]

This article analyses an example that deliberately uses this strategy to mislead the reader.

However, you don’t need to include a zero baseline on a line chart. In fact, in some cases, including a zero baseline can obscure data. To provide a clearer view of the data, line charts can be drawn without connecting the axis. You can see an example of this below.

Concise messaging

Messaging can help your audience understand a chart, but it is easily overused. When labelling the axis on your charts, consider:

• rotating the bars if the category names are long
• putting value labels on bars to preserve the clean lines of the bar lengths.

Here’s an example.

As well as the axis labels, you could include legends that explain what a particular colour, dot-style, or line type represents. Legends can be contentious in the chart world. You’ll often find a legend below the chart or items listed arbitrarily, forcing users to waste time by looking up values.

Instead of using legends, consider placing labels next to data (i.e. beside the lines). Another benefit of labelling data elements individually is that you don’t need to rely on colour to identify them. Keep it simple!

References

_{1. Tufte ER. The visual display of quantitative information. 2nd ed. CT: Graphics; 2001.}

_{2. The Shaffer 4 C’s of data visualization: ‘clean’ examples [PDF]. Available from: https://dataplusscience.com/files/Shaffer%204C%20-%20Clean%20Examples.pdf}

_{3. Fox News continues charting excellence [Internet]. Flowing Data. Available from: https://flowingdata.com/2012/08/06/fox-news-continues-charting-excellence/}