Visualising your data as precisely and accurately as possible prevents you from misleading your audience, and, possibly even more importantly, yourself since you will also base your interpretation on your (exploratory) graphs. As different graph types and graphical elements affect the way we perceive information, they will also affect your and the audience’s interpretation of the data. Therefore, when you are designing a graph, it is important to always keep your data in mind.

Some things to specifically pay attention to are:

• Always consider where your data comes from. What does it tell you and what would using different visualisations on it imply?
• Check your graphs and plots for errors and outliers. They may often still need to be included in your final graph, but it is important you are aware of them.
• Keep your data visible as much as possible. Often, it better to plot all data points instead of their averages for example, as this is more transparent.
• When using colour, use an appropriate colour scheme. See Colour use in data visualisation for more information.
• When you want to compare variables, make sure your axis ranges are similar.
• X-Y graphs can (un)intentionally misrepresent the effect size and therefore mislead your audience. This occurs when zero is used as the origin of the dependent variable’s axis even though your data has a natural zero point. To prevent this, you can add a zero-break, or you can start your y-axis at zero. The latter solution depends on whether this makes sense with your variable¹⁵

Keep your graphs as simple as possible without losing relevant aspects of your message. This means that you should avoid any ‘chart-junk’ in your graphs, i.e. elements that are unnecessary to comprehend the graph. These will make it more difficult for your audience to understand the message you’re trying to convey¹³. For example, this means that you do not add colours or patters to your graph just to make it look nice, but consider carefully whether gridlines and small tick marks are necessary or not.

Additionally, do not use the default settings of your software as these are never adapted to your specific type of graph, data, and message. When you are making graphs using a programming language, specific data visualisation packages will often lead to nicer looking visuals than using the default data plotting settings. For more information on software for data visualisation, see Tools for creating figures.