Data visualisation best practices

For my week 2 challenge in the Data School Down Under, I was tasked with making improvements to one of my data visualisations to fit best practices. In this blog, I want to walk you through how I approached this challenge. When visualising data I find that there are two important questions you want to ask yourself: What information do you want to communicate, and how do you want to present this information? To demonstrate this, I will be making edits to my first data visualisation (the one I used as an application for the Data School) and explaining the reasoning behind them.

My first visualisation used movie grossing data in North America from 1995 to 2021. The data is easy to grasp and was quite great for a first visualisation. Below you can see my visualisation. There are some good things and some bad things that can be improved upon.

Example 1. Sorting the values of the graph.

It is important to sort the graphs because it makes the graph easier to read. The reader can easily locate the highest and lowest values and everything in between.

Example 1 – without sort

Example 1 – with sorting

Example 2. Position any important text or interactive features on the left.

Reading is often from left to right. It would be most natural to format the most important things on the left which you want the user to see first. In the case of this dashboard, I want the user to see the filters, parameters, and legends to most left, the graphs second in the middle, and the analysis last to the right.

Example 2 – Poor formatting

Example 2 – Improved formatting

Example 3. Double encoding colour

Colour legends can be reinforced without excessively using legends throughout the dashboard. One way to do this is by colouring words in the title as well. Here I colour ‘average’ and ‘total’ in the title using the same colours I used in the graph. This creates the association between the colour purple with ‘average gross’ and blue with ‘total gross’.

Example 4. Removing unnecessary graph/chart types.

Even though I was excited to learn how to make a radial chart, one of the things that I have learnt during my time at the data school is that certain chart types are just not appropriate. I have come to realise the radial chart did not add anything to the analysis but rather made it a bit confusing. Instead, it may be best to hold back on some fancy charts and stick to bar charts, which the radial chart essentially is but just shaped differently. Even though bar charts can be boring they are a staple for a reason.

Example 5.1 Colour choices.

Using too many colours can make the dashboard chaotic and overwhelming to look at. Sticking to two shades of one or two colours makes the dashboard appear more coherent thematically and much more pleasant to look at overall.

Example 5.2 Colour choices

Another colour change that I would do is to simplify colour legends. The usage of four colours in the legend was quite unnecessary for a waterfall chart. Because the main purpose of a waterfall chart is to represent change and differences from the previous value it would be best represented by two colours: one to show an increase and another to show a decrease.

Example 5.2 – Colour choices

Example 5.2 – Colour choices with the simplified legend

These are some of the changes that have made my first dashboard in line with some of the best practices of data visualisation. This is the end product that you can compare to my original one here.


Leo Huynh
Author: Leo Huynh

Leo did a Bachelor of Science degree in psychology and neuroscience at the University of Sydney. While studying at university, he developed an interest in data science and visualisation. He specifically appreciates how data can be used to inform decision making. During his free time, he enjoys playing video games, cooking, and listening to music. His favourite food is Bún bò Huế (Vietnamese rice noodle dish with sliced beef).