14 July 2017
On Wednesday, 21 June 2017, Cmotions was in attendance at the MOA afternoon ‘Data Visualisation with Impact’. The professional association MOA, which was originally focused on market research, is increasingly moving into digital analytics, as became abundantly clear over the course of this inspiring afternoon with speakers from Chart.guide, Antichaos and Philips. What are the most important lessons for effective data visualisation?
Johan de Groot (Freelance Data Visualist/Tableau Specialist at Antichaos) made us face the facts. The historical facts. Data visualisation is nothing new. The engineer, Charles Minard, was already able to chart Napoleon’s march towards Russia back in 1812: how many troops were there heading to Russia and how many – shockingly few – came back? Also a long time ago (1854), there was the physicist, John Snow, who visually depicted the outbreak of cholera in London. It suddenly became clear that the water pump was the source of the spread of the disease.
And if there is one thing that data visualisation is fundamentally about, it is and always has been simply telling stories. Jithesh Rajendran, Senior Manager of Digital Analytics at Philips Global Team, tells strikingly of how “story telling” at Philips is no different from any other story. Start with a context (what is the background to sharing this dashboard?), in the middle come up with an interesting twist (how did you get there?), end with a climax (what is the punchline of this report?).
It doesn’t even need to be digital. What can we make of visually illustrating traffic at a junction in Berlin (Rosenthaler Platz) with real paint?
Data visualisation isn’t necessarily “the next big thing”. But what knowledge have we picked up over all that time to tell a good story about?
Rajendran shows us a Philips advert to illustrate the importance of knowing your audience. The Spiderman actor in the clip knows his audience. He absolutely knows for whom (young children) he is working. He adapts to his audience with his appearance (costume), his message (hope and faith) and his performance (the unexpected stunt when cleaning the hospital windows). He implicitly knows how to attract attention in a targeted way. How different is it actually to effective reporting? Imagine the audience isn’t a child, but a member of the management team. What if the apparition wasn’t a Spiderman costume, but a simple visual representation of your data? Maybe then your presentation will inspire faith in your intentions, like the actor does for the children.
Johan de Groot understood the aforementioned perfectly. The more specific you are to an audience, the better your message is conveyed. Try not to make a graph “for anybody who might be interested”. Know your precise audience. Whilst conveying your message, build a relationship by identifying your own role and expertise. In the fake news era, verifying your sources is no superfluous luxury.
Knowing your audience also helps when determining the degree of abstraction for your data visualisation. Options for clicking on and zooming in on results (“drill down”) should decrease as the management level increases. Having too much data – rather than a compact conclusion – can instead produce too many “small” discussions in your meeting. Your task is not to prove that you have done your job properly. You analyse and bring out the heart of the matter. Your audience don’t want to do the same job twice. They don’t want to analyse it a second time, but to pick up on what you’ve found.
Michiel Dullaert from DePerfecteGrafiek.nl and Chart.Guide highlights that what you want to do with your visualisations is to get people moving by themselves. But what movement can vary widely. A journalist is looking for the burden of proof, an analyst is trying to uncover insights for business operations. One user is concentrating on understanding (the more detail the better!), the other is concentrating on taking action (concise information please!). And just like with a “real” dashboard in your car: you don’t necessarily need to understand everything before you take action. A red light on the dashboard means “take action”, even if you don’t know how it works under the bonnet. De Groot groups the objectives together as Providing information, Inspiring confidence, Allocating budget, or Project support. All of which are different objectives that each call for a tailored approach. Rajendran tells how he used “user personas” to establish the needs of various different data users within Philips worldwide. For example, one persona would be a “Power User”. A factsheet provides an instant overview of this type of user’s role, location and personality traits. This makes it the ultimate starting point for making an effective report.
Dullaert talks about how important it is to know what your addressee is going to do with your information. Make sure you find the “switch” in someone’s mind. Make sure you “flick” that switch with your visualisation. Like how he structured his own CV (or perhaps more of a personal dashboard) in an entirely unusual way and got asked exactly the same questions in job interviews time and again. You can draw the comparison with a “magnet” that you insert into a visualisation, which attracts the attention of the viewer. Attract the person towards you – rather than pushing your data towards them.
Preattentive attributes – patterns that addressees subconsciously recognise more quickly than the mind can keep up with – are hugely helpful in this. Good examples of this are icons and the importance of symbols on road signs, rather than textual instructions. How quickly can you see a trend in a visualisation? You can deduce a message within 250 milliseconds. You see it before you think. You “see and feel” the answer without really thinking about it. You know where to look because it just “is”. The adjacent video shows you how.
As Dullaert demonstrates, the message can jump out in all kinds of different ways. Imagine you have three lines and one of them needs to stand out. Length, orientation, colour, thickness, position, outline or highlight… You have a whole range of tools to get your focal point to stand out.
The key to this is differentiation: when we pick one colour as a contrast, you immediately see what stands out. When there are two colours and the main one is the colour of the addressee’s house style, it still works fairly well. However, lots of different colours at the same time makes it really difficult to allocate your attention. It’s easy to immediately spot a red circle when it’s surrounded by a number of blue circles. However, the more shapes and colours there are, the more difficult it is for us to process the message. So be economical with your preattentive attributes.
You don’t have to invent the wheel every time. As we at Cmotions have done ourselves for lots of different clients: standardise your process. If you’re making fifteen dashboards, you need a template to build on every time you start. This set of standard components and definitions saves time, as well as makes comparability across a large concern much easier too.
Jitesh Rajendran demonstrates how data visualisations are not a local matter. In a global concern like Philips, you want to be able to share those good insights. However, Philips is a company with tens of thousands of employees, hundreds of departments and a large degree of individual freedom. “Reporting” was in silos – there was no overall picture of the digital performance of Philips Global. Working in a structured and standardised way, based on personas, enhanced the organisation. Now Philips knows whether a “5% increase” is actually a lot or a little on a global level, because the reports are now comparable. The outcome: analysis and decision-making are faster, reports can be made at a quicker pace, and transparency has increased across the global business.
There are plenty of tools to quickly “make something from your data”. For example, RAWGraphs. They don’t necessarily do the right thing. De Groot argues that “Bad visualisations” can unintentionally go wrong by choosing the wrong graph (e.g. hundreds of different categories in a pie chart), even when the data is right.
Evil visualisations are the manipulative versions of “bad” ones, which is a serious offence. 3D effects in pie charts which inflate a category are examples of such unwanted “lie charts“. Graphs like this try to unjustly show correlations that aren’t actually there. For example, you can include two Y axes, but only label one of them as a Y axis. For example, you can “show” the “relationship” between abortion and cancer:
Or in the climate debate, we could for our own convenience make the Y axis so long that the effect of a few degrees either way appears insignificant:
Just like the example from Fox News with regards to “Obamacare enrolment”, where the Y axis was distorted even further.
With a funny twist, the example of Tyler Vigen demonstrates the fragility of reporting on correlations. As you can see, the consumption of mozzarella is related to the number of doctorates awarded to engineers…
There’s no getting away from it. Data visualisation is communication. And as old as the phenomenon is, so too are the rules: don’t mislead people. It doesn’t even need to look nice, De Groot argues. First and foremost you need to convey a message. You need to share it in an easy-to-understand way. Or rather:
“The main goal of data visualisation is to communicate information, clearly and effectively through graphical means.”
Visualising things helps our primitive brains to think quickly. And if it also offers just a fraction of the enjoyment of the late Hans Rosling (The Joy of Stats), then it’s a 100% success!
7 September 2022
We don’t know if you’ve heard already, but there is yet another crisis on our horizon:... read more
21 February 2022
Plan your calendar free and make sure you have plenty of party clothes in your closet,... read more
29 April 2020
About the Award The field of data driven marketing is evolving fast. To honor these developments,... read more