Visualisation is having a moment. Attractive presentations of data have become more common, not least in the dashboards many executives now rely on. But as David Reed finds out, the way a story is told in pictures can sometimes conceal more than it reveals.
If you wanted to convince somebody of the terrible toll of warfare, how would you do it? You might hand them a copy of the highly-praised “Stalingrad” by Antony Beevor. Powerful, but a lengthy read. You could get them to sit through “Band of Brothers” , but time would still be a factor.
A quicker way to prove the cost of going to war might be to show them Charles Minard’s 1869 chart of Napoleon’s disastrous advance into Russia. Mapping geographical positions against temperatures, it reveals in a simple, stark visual the huge losses suffered by the French army in 1812. It has been hailed as one of the greatest statistical graphics ever created.
For anybody analysing data, one of the major considerations is how to present the outcome. Rows of numbers in a spreadsheet are not compelling to most business executives, which is why visualisation has become ever more important.
The problem is that design and data have grown up in very different worlds. For vivid proof of that fact, take a look at the house ad run by DDB London in the Financial Times during 2011 (http://twitpic.com/105y8s) “There is an expectation that visualisations have got to be beautiful, but are they meaningful?” asks analytics expert Peter Furness. “These graphs break all the principles which have been laid down.”
You might not realise that best practice in visualisation has been codified, but it has. Furness points to a couple of flaws in the agency’s example - using three-dimensional charts that distort the meaning and having scales that do not start at zero. “How could one of the world’s leading advertising agencies - whose business is communication - produce a full-page ad that exposes such a lack of understanding?” he wonders.
One answer is likely to be the low status which analysts have tended to have in marketing. “The data geek” often gets the last five minutes in a presentation, just as everybody is thinking about lunch or going home. Little wonder that few lessons have been learned. And while things are beginning to change since data got sexy - the very reason why DDB London was using a data visualisation ad was to grab a slice of that action - this growing interest presents its own problems.
“The traditional tools used for analytics, like SAS and SPSS, have a visualisation piece in which it is not necessarily straightforward to do something that shows the breadth of the analytical work. What you need to do is tell a real story,” says Furness. That reveals the other side of the visualisation problem - that analysts have not been schooled in creative presentation, but have a deep interest in the underlying data. Creating a presentation in PowerPoint or Excel using basic charts is often as good as it gets.
Not that this is necessarily a bad thing. One of the critical practices in analytics is “plotting, before you chart” . As Anscombe’s Quartet (see box) famously demonstrated, you can sometimes see a problem in the underlying data through sim ple charting. Furness says he visualises data throughout the analytical and modelling process for this very reason. “It is a good way of tracking data quality because you can see patterns in the data which may be data quality problems, for example a line of values along one axis which may have been coded as zeroes,” he says.
Approaching the task in this way uses one of the greatest strengths of the human brain - pattern recognition. Simple visual processing of this sort evolved very early, whereas interpreting symbols (such as written language or visual symbols) is pulling on a much more recently developed area of the mind.
Furness describes it as “an uphill struggle” . He also points out that data can be presented in tables in an engaging way by using elements of animation, such as using colour highlights and underlining to pull out different elements to make the meaning clear without losing track of the underlying data.
With the explosion of big data and digital data sources, this approach is undoubtedly coming under pressure. New techniques, such as social network analysis, require their own visualisation tools. Some of these are genuinely striking (visit informationisbeautiful.net for examples). They also reflect that codification done by Edward Tufte in “The Visual Display of Quantitative Information” (self-published because commercial publishers could not do it justice) and David McCandless in “Information is Beautiful”.
Presenting to senior executives and decision makers undoubtedly demands something with impact. After all, they are time-pressured and want to know what the key take-outs are. Anna Foster, data director at marketing agency TMW, says there is a risk in taking visualisation too far: “You wonder why they have done that. It is usually when people think the data is boring and needs ‘sexing up’.”
She emphasises that “the data is the most important thing and it is really important to get it right. If you do that, it may be enough so it doesn’t need tarting up.” After all, it is usually in the data that the business-critical findings are discovered which will make or save the client money.
Foster has noticed a trend towards more sophisticated visualisation, with practitioners referencing the way Guardian Data presents information or looking at gamification. In some respects this is not new - data trade shows twenty years ago were full of artificial reality vendors demonstrating how analysts would be able to “fly through” data.
“We are usually trying to tell a story with data, for example that something is good and something else correlates to it. That can be difficult to show on one chart. But we do enhance the graphics if the client likes them. A lot prefer three-dimensional charts - one client even likes smiley faces for rising trends,” says Foster.
Whether analysis is done by an in-house team or an external partner, its output has to be “sold” to its audience. The best model in the world will not get looked at if it is not in some way great to look at, not least because senior executives have heard so much about how maths is beautiful or have seen advanced visualisations and want some for themselves. As data adjusts to having so many new friends, it is important it does not feel the need to dress up just to please them.