First answer – as a pedantic Englishman, the letter “U”, of course.
As I keep saying in this blog, I’m not really in a position to give great expert advice on data visualisation principles. Many experts have written and spoken about colour, not least Matt Francis earlier this week, who gave a reprise of this talk here: http://ontour.tableau.com/london/content/850. Many more have written great advice based on experience, sound colour theory, and up-to-date knowledge of best practices.But I do come across a lot of interesting issues when I visualise data, many of which I am encountering for the first time. One thing that has come up again and again, just in the last day or two, is the use of colour, so some of my thoughts are below.
So which two colours should you use in a two-colour visualisation? In my “day job” we have often presented charts using shades of red and green in the past. Its traffic light connotations of “stop” and “go” or “bad” and “good” are pretty well known and well used. Should we be doing that? Not really, these days you have to be careful to avoid combinations of colours associated with colour-blindness, with red/green being the most common of all. How about red and blue? Red and blue have political connotations – identify areas or categories as red or blue in the USA and the reader will automatically associate with republican or democrat – we designers in the UK need to be aware of and remember that. Red and blue, or even pink and blue, is OK for male/female, but is it too predictable/stereotypical, and lacking in imagination?
Anything geographical in the UK gets quite complicated. Red and blue could also be construed along political lines (Conservative/Labour), but there are several smaller parties in UK politics. With former third party (Liberal Democrats) a kind of orange/yellow, current third party (Scottish National Party) also a prominent yellow, and minority parties UKIP (purple) and Greens (er, green) then the non-political colours are disappearing fast. So which two colours to use for Brexit mapping? (also discussed in my previous page here)
The above map is otherwise excellent but it uses blue/red. Since the Brexit choice was not a straight decision along political lines, it looks a little unintuitive to politically-aware UK readers. Leave was generally mapped in blue, though the blue Conservative party was very divided over the issue. Remain was usually mapped in yellow – a shade not seen much in England/Wales but with strong political connotations in Scotland. At least the yellow correspondence of Remain voting areas and SNP voting constituencies made sense, but there was a feeling that this was a compromise colour pair with no obvious better candidate.
After all, if you’re doing any visualisation at all in Northern Ireland, ironically red and blue is probably as good as anything, given the deep-seated colour connotations of green and orange, in relation to Catholicism and Protestantism respectively.
There’s a groundswell of opinion for blue and orange as colourblind-safe default colours when two colours are needed – certainly these are used by Tableau, my usual software of choice. By way of shameful promotion, here’s an example I prepared for today’s Wimbledon men’s singles final. I think I was OK with blue and orange here.
Blue and orange aren’t emotive or controversial, right? Wrong! They are deeply significant in Dundee – they identify as football teams Dundee or Dundee United. Famously, the iconic character Zippy from UK children’s programme Rainbow was supposed to be blue, but presenter Geoffrey Hayes refused a blue puppet, being a Dundee United supporter he didn’t want anything Dundee coloured. So, he looks like this:
More lesser known stories. I once had a guy working in my team who refused to take a diary from me. Why? It was red. Liverpool colours. He supports Everton, their city rivals. I thought he was joking, and though he was smiling, he was absolutely serious. I used to know a magazine editor who refused to publish any image or photo which contained orange. Why? She supported Preston North End whose biggest rivals, Blackpool, play in orange. The examples might be getting progressively more daft (and explain a lot about inbred football-related tribal mentality in the UK!) but they are all true. The key thing is to know your audience and not use anything that could be misconstrued or cause offence. I never gave my colleague anything red again and we got on fine after that!
So onto today, where I have had a look at a visualisation as part of the #MakeoverMonday community visualisation I like to take part in. This week, the original chart was this one (below). It’s a nice small multiple chart which shows some of the dramatic differences between Republican and Democrat well, though the half-circle shapes are somewhat unorthodox.
So this one is easy. Simple divergent bar charts. Red and blue. Make it clean/simple and mobile-optimised, sortable on republic or democratic. Job done.
However, I wanted to do something different. This is an emotive dataset which emphasises the differences between pro- and anti- gun control opinions in the light of a dreadful tragic event. It doesn’t offer a political opinion, but it carefully chooses subjects to make the point of the differing tone of replies. Because this was an option to experiment and practice, I wanted to add a “rainbow” element to the chart in order to express that there was more going on than just political “red” and “blue”. My revised chart (still just as simple) was below:
I know this breaks rules. The rainbow colours don’t add to the story. There’s nothing to associate criticising Obama with the colour violet for example, and the “busy” rainbow colours aren’t needed. They make the brain, and the eyes, do extra work. So why put them in? They indicate that the creator of the visualisation (me) wanted to say something extra. A futile gesture of LGBT support and something to catch the eye away from the normal and predictable red, white and blue, though still including the red/blue elements to distinguish the number of mentions per topic per party.
I’ve been told the rainbow colours shouldn’t be there. They don’t add to the visualisation. I know they break the rules! If I was a journalist, I’d leave them out. If I was an editor, I’d demand they were removed. If I was a report writer, or a data analyst (which, in real life, I am), there’s no way I’d include them in the first place.
But this is an exercise to get us thinking and being more creative. “Know your audience” was the advice I gave earlier. This isn’t a professional creation, where my audience, whomever they might be, should be able to enjoy the visualisation while taking insight from it as quickly, simply and efficiently. This time, my audience is me and a few blog readers/data visualisation enthusiasts, so the colours are staying, with the caveat that they form part of an interesting albeit rambling discussion!