Well, I decided to enter #MakeoverMonday this week. It’s a weekly dataviz challenge set by Andy Kriebel, open to all. The rules of the game are simple. Each week Andy selects an example dataviz published somewhere on the web – usually something topical used on a news site – and the challenge is to produce your own, presumably better, visualisation using the same data. You’re not supposed to spend more than an hour on it. You can use any tool or medium you like, but results must be tweeted as a picture tagged with #MakeoverMonday. Every submission is pinned to a Pinterest board for all to see. There’s no winner, it’s really just a chance to flex your dataviz muscles, spark debate and get inspired (by others’ submissions).
This week the subject was this graphic on the newly-announced UK sugar tax, taken from an article on the BBC News site.
I won’t go critique the original here, other than to say using side-by-side pie (or doughnut) charts to illustrate differences is not good.
I cheated a little for my submission, in that I also incorporated information from a second infographic in the same article. But I think it is well within the spirit of the competition.
In case you’re interested, I used Tableau to produce my entry, my current favourite BI tool not least because it is so quick to put something like this together once you have your data in shape. But this isn’t a how-did-I post, it’s a why-did-I post.
When designing a graphic like this it has to be eye-catching but, more importantly, you have to have a purpose in mind. What point are you making, what question are you answering (or asking)? This is your objective. If you don’t have an objective how will you be able to judge whether it’s a success?
I decided my objective would be to draw attention to the fact that teenagers consume not only the most sugar from soft-drinks (the subject of the news article) but also the most sugar overall. Two messages in one. The stacked bar chart format works well for this. I’m always a little wary of stacked bar charts which show only percentages, as they can be misleading. When viewers look at percentages they will make inferences about underlying absolute values that may not be true. Basically, if the percentage is “large” they’ll assume the underlying absolute value is “large” and noteworthy.
If you’re struggling to think why this might not be the case, imagine I consumed only 15g of sugar a day, half the recommended daily maximum, and that 7.5g of this came from soft-drinks. Wow, 50% of my sugar intake from soft-drinks! But the absolute value, 7.5g, is not a “large” number in context. It is not nearly as noteworthy as 50% implies.
So I stacked my bar chart with absolute values. The viewer cannot compare percentages accurately in this format, but I don’t need them to. They can still get a broad sense of the proportions, and that’s all I wanted. Limited and deliberate choice of colour was designed to draw attention to the soft-drinks component, whilst keeping the image as simple as possible. The little figures poking above the bars are effectively pure eye-candy, i.e. their purpose serves only to draw the eye. But importantly they do not distract or detract from the data or point made. I’ve deliberately omitted other visually-unnecessary clutter: no axes or grid-lines. The title fits in the top-left nicely, minimising wasted white-space whilst still being the natural place your eye starts.
For a viewer dwelling on the graphic for more than a second or two, the reference to the 30g “recommended maximum” gives them some secondary information, which they can juxtapose mentally with the absolute values in the bar chart, providing a small reward for the effort.
Well, red isn’t always the wisest choice of colours. About one in twelve of the male population has colour-blindness affecting their perception of this colour. Red also has different cultural interpretations in different parts of the world. In the West it is considered “bad”, but in parts of the East it symbolises the exact opposite: “good” (or lucky).
Each of the age categories is given equal weighting (they are all the same width). But there are are a lot more people in the UK ‘Adult 19-64’ population than the other age groups. In fact, there are 4-5 times as many people in this Adult age group than are in the ‘Teenager 11-18’ group. On the other hand, children have more years of life ahead of them (in which they could be affected by obesity). So what would be a fair representation?
Oh, and I haven’t referenced my data source. That’s poor form.