Choosing the right data visualisations

When it comes time to present a data story, the importance of choosing the right visualisations to amplify the impact of your data stories can often be overlooked. Whether you're an analyst, a data scientist, someone tasked with designing reports and dashboards, or someone who wants to present their data story in the most visually appropriate form, this article should hopefully equip you with practical insights for creating compelling visual narratives.

I often work in a really interesting space in between people who are analysts and data scientists, designing dashboards and reports and the end users in an organisation for whom those reports are being designed. And often, frustrations arise between these different groups due to misalignment in expectations, where analysts may feel their efforts are under utilised, while end users express dissatisfaction with visuals that fail to address their needs. To overcome these frustrations, I'm a big believer in fostering dialogue and understanding between these groups, and emphasising to both groups the importance of tailoring visualisations to the intended message.

Ultimately, effective data visualisation should align design choices with the intended message - what is the trend or insight that you want to highlight within the data? Different visualisations work in different ways and choosing the right one can be a powerful way to effectively engage the audience. Take for example tools like Excel, which offer suggested charts when you input data. I caution against using these sorts of default options because Excel actually doesn't know what you're trying to share or what message you're trying to get across, and the default may not serve your desired narrative.

Instead, you should be selecting visualisations that best convey the insights you want your audience to take from your data story, because different visualisation types serve distinct purposes. For instance, a waterfall chart efficiently highlights changes between values, streamlining interpretation for users, whereas a lollipop chart can be useful in illustrating gaps or changes in data. Line graphs are ideal for showcasing trends over time, while I recommend box plots to represent spread, because they succinctly convey data distribution.

It's also important, where appropriate, to include context in your visualisation. For example, I recently presented at a data visualisation webinar for the Pacific Data Viz Challenge, where participants from New Caledonia, Fiji, and beyond convened to explore the theme of gender equality. In my presentation I highlighted that when graphing the pay gap over time between genders, it's essential to not only consider national differences but also incorporate global benchmarks. By including a global perspective alongside longitudinal data, we gain a clearer understanding of the significance of observed trends in our own data sets, enabling more informed decision-making.

Alongside using the wrong visualisation type for the intended message, another common pitfall I see is cluttered visualisations. Think carefully about the elements you need in your visualisations and do not include any unnecessary data, such as excessive bars in bar charts or pie wedges in a pie chart, which hinder interpretation. Simplicity and clarity really take the mental load off your audience and allow them to focus on the relevant insights you are trying to convey.

The final point to consider is the design both of the visualisation itself and the overall design of how they are presented. Pie charts can be very much maligned, but I believe are useful for datasets of four or fewer variables. However, they should always be designed in such a way that the first 'wedge' of the pie begins in the 12 o'clock position. This helps orientate the audience and makes it easier for them to determine the size of the data set quickly.

Effective use of colour is also paramount, and I always advocate for consistency and accessibility when using colour. When designing visualisations, ensure that the same shades and colours mean the same things throughout your presentation or report so as not to confuse your audience. Also, it's important to consider how colours work together for people who may be colour blind or have low vision. There are a number of websites available to check the contrast and accessibility of the colours you are using. Furthermore, it's important to remove extraneous elements such as keys, axis labels and grid lines if they don't impact on the way the visualisation is interpreted. This works to reduce cognitive load and enhance clarity for your audience and lets them focus on the task of digesting the insights in the data.

One final point about dashboard design is the arrangement of your visualisations, which should reflect the way we read a book. As we read from top to bottom, left to right, prioritising key visuals in the top left corner ensures immediate impact, with subsequent visuals arranged in descending order of importance. Similarly, reports should front-load impactful visuals to engage readers and streamline comprehension.

So, whether you're a seasoned analyst or a budding data enthusiast, I hope these tips will help the next time you need to choose a visualisation for your data. Remember to choose your visualisations wisely, which will help unleash the power of your data stories. Your audiences will thank you for it.