During Innovation Week 2021, our MSc students were lucky to have a keynote talk from Gavin Allan, BI & Data Science at Barclays Corporate Banking about data visualisation. He has put together some of the highlights from his talk for you now and we have included a recording of his keynote talk below.
One thing that has clearly shone through over my last few years working in data analytics & data science is the importance of data visualisation – and consequently, data storytelling – to engage an audience in data insights.
The aim is always to generate value from data, but to do this we first need decision makers to take action from insights. To take action, they must first understand, trust, and interact with the data insights presented to them. This is where good data visualisation comes in. It plays a pivotal role in helping to build this understanding and trust with decision makers.
Data Visualisation: Interesting vs Useful
You might think I would focus first on design, but there is a far more important step before we get to that. We first need to consider the story we want to portray in our visualisation as after all, to show everything would most likely overwhelm our audience.
How do we do this?
Crucially, any insights we choose to visualise should be useful for our audience. To determine what is useful, we need to work with them to understand their needs and wants. It can be all to tempting to show off insights that might be very interesting but are not useful – these types of insights are dangerous as they engage our audience but don’t provide any value to them! Most of the time, the insights we produce confirm what we think, or perhaps know, to be the case. Remember that this is usually still useful as it confirms our current knowledge is still valid, or maybe you might prove a belief that hasn’t been backed up by data previously.
Data Visualisation and Bias
Before we touch on good design, we’ll first look at an example of where things go wrong:
On the left is a visualisation taken directly from the Daily Mail, and on the right, is a visualisation I have built based on official government data. Colour schemes and design principles aside, there is a more fundamental problem with the Daily Mail visualisation. You may have noticed that the X axis is variable – ranging from a 14-day interval in June to a 4-day interval in August – and completely skews the data as a result. It is important to always question the representation of the data you are presented with and not to take it at face value.
Data Visualisation Principles
If that is an example of what not to do, what should we be doing?
Before I begin, I would like to first say that the principles set out below are credit to Andy Kirk’s brilliant book on Data Driven Design. In this, he sets out that good data visualisation should be trustworthy, accessible and elegant.
To be trustworthy, we should always strive for neutrality and minimise bias. In doing so, we should actively call out anomalies in the data and not hide anything because it doesn’t fit our story. Naturally, we want to understand what we see, and it can be all too appealing to build up a compelling narrative to fit our data. Quantitative insights often tell us a lot about the ‘what’ but offer limited insight into the ‘why’. Keep this in mind and avoid making any assumptions beyond your own knowledge when building the story. As an aside, consulting qualitative research is a great way of adding depth and meaning to your quantitative insights!
To be accessible, start with considering your audience and how much time they have to consume your visualisation. A senior stakeholder is unlikely to want to invest time in understanding your Sankey chart no matter how beautiful it is! Consider those who are colour blind and pick colours accordingly – orange and blue is a winning combination.
Finally, and most importantly, your visualisation should be elegant – both beautiful and useful. Consider our interesting vs useful discussion earlier and focus on keeping the design clean. Keep the number of objects on screen to a minimum to focus on the insight itself.
Telling the Story – Data Storytelling
Just because we have made our insights accessible through beautiful and useable data visualisations, does not mean we have earnt the right to an engaged audience. We must also be able to tell the story.
To do this effectively, you must know your audience and speak in their language. They are unlikely to be interested in statistics, they will want to know what it means to them – fewer complaints, £X more revenue, £X reduction in costs. Making this connection is what will drive interest in data insights!
Another eminently powerful technique I have found is to take your audience on the journey. By that, I mean to start a few levels above the end insight you are looking to show and set the scene. This allows them to follow your thought process and builds trust. When done well, your audience will spot the insight for themselves in your final visualisation making it far more memorable and powerful to them.
I see data visualisation, and storytelling, as an essential tool in generating value from data. Done right it engages audiences in data insights, builds trust and ultimately drives action.
Watch Gavin Allan’s keynote speech: Making an impact with Data Storytelling
More about The Data Lab MSc Innovation Week:
Our Principal Data Scientist, Joanna McKenzie gives her thoughts on Design Thinking in data science projects and the need for a Creative Data Science approach after joining students at Innovation Week: Creative data science and design thinking in data science projects.
Find out more about the AI Strategy Session as part of Innovation Week: Reflections on the AI Strategy session at the TDL MSc Innovation Week