Skip to content

Powering digital transformation with data visualisations

The right visualisation helps you leverage the data you need for successful digital transformation.

As digital transformation continues to gather momentum, data could be your organisation’s biggest asset. Some 65% of global GDP is forecast to be digitalised by 2022 , but automating and scaling processes and operations to meet the needs of larger customer bases will require reliable data-driven intelligence. Data visualisations can help companies harness the vast amounts of structured and unstructured data they are collecting and processing from business transactions, smart devices, industrial equipment, social media and other streams.

In this article, we take a look at the value of data visualisation and the qualities that make a good visualisation.

Why visualisation?

Leveraging data in the right way allows enterprises to digitalise effectively by gaining valuable insights that can enhance their operations and allow them to respond faster and more effectively to changing customer demands. Used successfully, data drives more informed decisions, sharper focus, greater agility and the capability to deliver innovative end-to-end services to customers.

However, consuming massive datasets is not easy. That’s why data visualisations are invaluable. Data visualisations communicate quantitative content in a coherent way, representing data in the form of line graphs, charts, scatter plots or maps. They make data accessible, easy to interpret and more valuable for generating fresh insights. By curating data into easily understood forms, they highlight the trends and identify outliers.

An effective visualisation tells a story. It removes noise and points out the information that matters. Your data instantly becomes more useful because more people across the organisation can understand the datasets.

What makes a good visualisation

As with good design, creating good visualisations involves knowing what rules to follow. The following are some best practices that should ensure that your visualisations are appealing, useful and beneficial to your audience:

A good visualisation answers a question

When designing a visualisation, it can be helpful to start with a question. By thinking about what question you need to answer, you can create a specification that makes it easier to work out which data you need to show and which limitations to impose.

For example, on the case numbers graph on , I asked the question "How do today's COVID-19 case numbers compare to the numbers over the last year?" This allowed me to zoom in on specific data (daily case numbers, weekly average), as well as the context for the graph (over the last year).

Not all graphs need to use time as a comparison. For example, the county graph shows a map along with a bar chart and colour coding that seeks to address the question of how Ireland's COVID-19 cases are distributed by county.

Identifying a question that we would like to answer makes it easier to decide on the best way the data should be presented to answer the question.

A good visualisation is honest

When data visualisations are designed to answer questions, they tell the story of the data. However, it's possible to manipulate the data to tell the story we want to tell. It's difficult to avoid introducing our own perspectives when choosing the data to show, and the way we compare it to other data is a powerful way to make a point, but it's still important to try to represent situations fairly. We should treat our readers — and the data — with the respect they deserve. Failing to do so can hurt our credibility as well as introducing the possibility of bad decision-making.

We can avoid personal bias in several ways. Firstly, we should avoid cherry-picking data that suits us. If there's a long trend in one direction, and we select a small part of the data that shows the opposite, we rob the visualisation of important context and devalue the message.

We should also be careful to state our assumptions. For example, in this herd immunity graph the assumptions are spelled out (estimated target number of adults to vaccinate and the current 7-day average vaccination rate). Being clear on assumptions can help establish credibility while adding important context to the story being told.

Lastly, we should ensure the data and the axes on which the data is measured are drawn and represented accurately. A common trick when comparing bar charts, for example, is to adjust the starting point on the Y-axis to some value greater than zero. This can make two bars that are relatively close in value appear to have a larger difference.

A second device is to present areas using 3D without allowing for the extra volume the third dimension gives the graph. This can result in exaggerated differences between bar charts. Apple used a similar approach in 2008, making their 19.5% market share look larger than it really was.

It is also common to show graphs without labels on the Y-axis and to use pie charts to compare different groups rather than parts of a whole.

It may be tempting to manipulate visualisations to make a stronger case, but it is best avoided.

A good visualisation is efficient

Data-ink ratio, a concept introduced by Edward R. Tufte in his book The Visual Display of Quantitative Data, is a useful term that encourages us to always aim to show the data above all else.

Many visualisations are eye-catching and sometimes intricately or beautifully crafted but still fail to convey the data properly. Excessive flourishes such as shadows, 3D boxes or even heavy use of axis lines can make it more difficult for the data to tell the required story.

When designing the graphs on , we took care only to include elements that actively helped convey the message or data being visualised. This meant removing X and Y axis lines and lightly drawing any lines used. We used minimal axis labels — enough to convey scale without being too heavy. The resulting graphs are much easier for people to scan and understand quickly.

As well as ensuring that visualisations answer a question, it is also important to ensure that there is enough data to warrant a visualisation in the first place. For example, if we were to draw a bar chart comparing two numbers, it would be a large graph with a very small amount of data. In this instance, it might be more efficient to write the numbers in a sentence.

A good visualisation is appealing

This might sound contradictory to what we said earlier, but good visualisations should look attractive. While maintaining strong data-ink ratios and being honest to the given data, an appealing graph will make use of intelligently chosen colours and hues that make it easier to ingest and understand the data.

This can mean choosing contrasting colour sets for large groups of line graphs, gradients that show gradual changes over time or using culturally-significant colours such as red for danger and green for positive.

The Information is Beautiful blog features many good examples of attractive visualisations.

Conversely, you can find plenty of bad visualisations if you want to see what not to do.

What to avoid

Data visualisations make it easier to identify and interpret information that matters. That means avoiding anything that might be ambiguous or confusing, such as:

  • including too much data
  • highlighting certain data that skews interpretation
  • using confusing colour schemes

You can find some examples on the Junk Charts blog.

Political leaders and civil servants have been embracing data visualisation to enlighten the public about the potential impact of the Covid-19 pandemic, but they have not avoided the pitfalls of imperfect visualisations . In October 2020, the U.K.’s chief scientific adviser, Sir Patrick Vallance, presented a graph that was misinterpreted in a subsequent government publication. Even before that error was discovered, the visualisations he used were criticised for including too much detail, frightening viewers instead of informing them.

The future of data visualisation

New technologies are opening up different ways of getting a visual overview and understanding of data. For example, virtual reality (VR) can be used to extract richer meaning from data, and augmented reality (AR) can demonstrate directly how the results of data analytics affect the world in real time. More ways of visualising or communicating data are likely to emerge in the near future, making data analytics more accessible.

Advances in technology are unlikely to change the prerequisites of a quality visualisation, however. Irrespective of how it is created, a good visualisation should always answer a question and be honest, efficient and appealing.

Insight, imagination and expertly engineered solutions to accelerate and sustain progress.