Madhuri Dass Woudenberg is Head of Communications at the Global Development Network, a public international organisation that supports high quality, policy-oriented, social science research in developing and transition countries, to promote better lives.
She is also a strategy, advocacy and communications specialist, with over 15 years of experience across Asia and Africa.
Besides data visualisation, she is interested in web and new media, writing, designing, films, event management, communications training and emergency response communications. She is also an expert trainer in many of these topics.
Madhuri is on Twitter at @MadhuriDass. The author’s views are personal.
I help social science researchers think about how to plan or commission data visualisations for their results.
Many think that designing a great visualisation will somehow elevate their findings. This is not always true.
The consulting field on data visualisation, unfortunately, is filled with advice on which colors or charting methods to use, or how to adapt them for use on mobile phones. Which is all very well, but it obfuscates.
People forget that a ‘visualisation’ of any kind is just an aid. It needs to say something solid.
Many beautiful, not to mention expensive, data visualisations blur into insignificance because they are unclear – not in their design, but in their recommendation. There’s no magic or secret formula to developing a great data visualisation. Just check that the stuff – rather than the fluff – is the real focus.
To me, a good data visualisation provides an insight or proof of a claim.
Begin by looking the research question, and how your findings measure up against it. Then check how relevant those findings are to the world. This means looking for the newness in the research, or the importance of it, and capturing that in the visualisation.
If you think that your results will break the internet, or if a policymaker simply must see your conclusions before he spends public money on the wrong idea, then do design something beautiful and unforgettable. But if your research does not make a concrete, relevant contribution to the general discourse on the topic, and if you cannot ‘see the headline’ of the piece even before it is designed, then don’t bother trying.
Assuming you do have something solid to say, use your ‘headline’ recommendation to carefully select the data points that support your research in the best way. Be honest with your content: include all the bits that contribute to the storyline, to give the full picture. Exclude anything that is not relevant. Add some context or background, but stick to the main story.
Data visualisations should make things simple to understand. If your story is complex, decide on a series, rather than loading everything into one visual. Or break it up, moving from simple ideas to complex ones. Compare scenarios to make a point. These storytelling decisions are fundamental to the design process of good data visualisation.
Even once the content is decided, it can be quite hard to distil the precise elements of visualisation. This is the time to get creative. If the aim of your visualisation is to show the deterioration of the education system, you could potentially show school busses falling off a bridge to depict the number of children being failed by the system each year. If you wanted to show the impact of drought, you could animate the appearance of brown spots on a heat map – and the disappearance of nearby villages over time.
Remember that your values will show through the objects you choose for the visualisation, so choose carefully. Some may misinterpret school busses headed for the abyss. Create something that your target audience can understand right away: Bright yellow buses may work in Canada, but not in Mongolia. Also, note that showing the impact of your findings on people – rather than objects or places – is more persuasive. Brown patches of drought are not nearly as alarming as disappearing villages, for example.
If you need a high-tech solution, to help users play with some of the variables online for instance, then get a data visualisation professional to help. Other than reviewing their work for accuracy and comprehension, you should also champion minimalist design. Less is definitely more and white space aids restful reading and cognitive grasp. Professionals can also advise on the best typography to create proper visual hierarchies.
I always urge researchers to drop the dense in favor of the lucid. Why can’t something dense also be lucid, you might ask? Sometimes it can, especially with the clever use of colors or style. Color can make the data points pop so people notice it right away. Extra details can be stripped away to convey the message more clearly. Often, a simple pie chart or variation of the bar graph can say it best.
As you finalise your data visualisation, remember that it should be able to stand alone. Include dates, sources or sample sizes in sub-text, if needed. Add the basics, like color keys or units of measurement. You don’t always have to point out exceptions or explain how you dealt with bias, but do point out where people can access the full data set and research paper.
Think of the people you want to reach with your data – what do you want them to do once they’ve understood it? Use the answer to caption your work. Think through and address anything in the data that may be questioned.
You should always test the visualisation. Ask your companion on a bus or a non-researcher friend or family member if they get it before you breathe a sigh of relief about getting it right.
In the end, solid data visualisation can do more than just grab people’s attention. It can change people’s minds.