Data visualisation can be described as simply displaying information to enhance understanding of a given topic . A large data table is therefore a visualisation, however not a very effective one as it requires a great amount of effort from a user to interpret any trends. This is the crux of data visualisation best practice: presenting information in the most efficient manner possible to ensure users can reach insights with minimal effort. There is a great range of different charts to choose from, and each has their place depending on the nature of the data and the aim of the Analyst.
Therefore, there is not always a right or a wrong visualisation to use, though some will be more effective than others. There are some reliable visuals such as a bar or line chart which can generally be easily interpreted. However, relying solely on a few chart types can become repetitive and tedious. As a result, Analysts must make thoughtful decisions about how best to present data.
It is therefore important to think critically about the design throughout the development process. This post discusses why it is important to evaluate data visualisations and outlines 5 simple questions to help ensure the visual is fit for purpose.
Data visualisations are designed to display a story or answer a business question. They also represent a series of design decisions made by an Analyst, often with contributions from business users.
During development, the Analyst is repeatedly presented with the same information and become engrossed in the data. The visualisation and story being conveyed become so familiar that it is hard to imagine receiving it as an end-user for the first time. This is often taken for granted, and it can be forgotten that it may not be quite so clear for end-users.
A successful visualisation should be self-explanatory, both in terms of its purpose and the story being told. When a chart is difficult to interpret it has not fulfilled its purpose of enhancing understanding or added any insightful value. This can lead to confusion and low user engagement levels.
It is therefore important to take a step back and critically evaluate the design of your data visualisations. Listed below are examples of questions to ask yourself to ensure your visual is fit for purpose. Asking a colleague is also a great opportunity to see how it will be received for the first time; whether it is hard to interpret or if there is a need for further explanation.
Every visualisation should be there for a reason, which in general depends on the purpose and requirements. This is true for both static, one-off pieces of analysis as well as general reporting. However, for either type of requirement, data visualisations should have a purpose that directly relates to the brief. This does not have to be complex. It could simply be, how sales have changed by month over the last year. Additionally, the answer to the question/story needs to be clear.
A visualisation that correctly answers a business question or shows a story, but is difficult to interpret and draw conclusions from, is not effective. For example, showing product sales over time by month, but with over 20 different products, would not show a great deal on a line chart. Alternatively, you could add a filter to display the Top 5 / Top 10 etc to ensure the patterns are visible.
Data visualisations should be self-explanatory. This boils down to an explicit and often simple design allowing for easy interpretation. Examples of this, depending on the visualisation, are as follows: -A clear and succinct title describing the metrics on display -Dynamic titles that reflect the filter selections -Data labels -Colour and size legends -Number formatting such as percentage (%) and currency (£) -Descriptive axis labels.
Moreover, being explicit also comes from visualisation techniques such as sorting from highest to lowest, filtering on the top/bottom 10, and using colour, size, and orientation. This way, you are clearly presenting the story to the users the story as opposed to asking them to dissect it themselves. Being direct leaves little room for ambiguity and ensures different users will interpret the data in the same way.
The effect of a visualisation strongly relates to the intended audience. Different audiences receive and act on information differently. The audience is important because it dictates the level of granularity and complexity. For example, users with different job roles and levels of seniority will use data in contrasting ways.
Taking customer satisfaction as an example, those working directly with customers could require data visualizations to improve efficiency in responding to complaints. They would require a high level of detail about the complaint such as when it was raised and what department. The visualisation would, therefore, be built to highlight issues that could directly be actioned, such as a hierarchy data table sorted in order of urgency. Conversely, Managers and Directors would need reporting and insights on overall customer satisfaction and benchmarking trends. This would place a greater emphasis on high-level aggregations and Key Performance Indicators (KPIs).
With this in mind, it is important to remember that a single visualisation will not be useful to everyone. However, it should be effective and designed for the intended audience.
There are a multitude of different ways to present the same information. With sophisticated data visualisation tools such as Tableau or Power BI, this can be done relatively quickly. In the context of a dashboard or presentation, it is often easy to duplicate the same information, just presented with a different type of visualisation. This creates clutter and repetition. It is the responsibility of the Analyst to choose the most appropriate chart.
If a visual is effective and self-explanatory on its own, repetition would not add any value. Checking for duplication can avoid potentially undermining the decision-making process and ensure the most appropriate visual has been chosen.
Data-ink refers to the elements of a visualisation that are displaying the data. For example, the height of the bar or the position of a data point in a scatter chart both represent a numeric value. Conversely, non-data-ink are the elements of a visualisation that do not represent the data. Examples include grid lines, shapes, and coloured backgrounds.
Excessive levels of non-data-ink can be distracting and draw attention away from the story and purpose of the visual. Non-data-ink can be minimised by replacing axis with data labels, removing gridlines and borders, and avoiding coloured backgrounds and 3D effects.
However, this is not to say all non-data-ink should be removed. For example, some charts require an axis and using shapes such as up and down arrows to indicate growth. These can enhance a visual and contribute to the story it is telling. It is therefore about balancing the use of non-data-ink, using it to enhance instead of distracting away from the rest of the visual.
The purpose of this post has been to highlight the importance of taking the time to evaluate data visualisations. Assess the story. Imagine how users will interpret it for the first time. Ask who it is for. Check for any repetition. Examine the use of non-data-ink. These are all ways to determine if the visual is fit for purpose and effective. Asking a colleague for their feedback is a great way to assess how people will perceive the visual for the first time. Thank you for reading!