Beware Data Visualisation Snake Oil
Martin Squires
Senior leader with extensive experience in customer insight, marketing analytics & data science ? 6 time member of Data IQ Data 100
I haven't had one of my "damn, this is a bad chart" rants for a while but this one set me off twitching this week. It also has an added benefit of letting me show that I don't just pick on right wing examples when I spot crimes against data visualisation, as this one shows the Labour Party having a bad day at the office.
So what set me off when I saw this?
Well, I can't say I'm personally fond of using flags and coins vs just using using "proper" bars (Tufte data:ink ratios and all that) but that's a personal design choice. I actually think there are moments where making charts look attractive means you can bend that guideline a little. HOWEVER, if you are going to bend it then make sure your visual images, in this case the number of coins per bar, actually make sense.
If you've got this far you probably have already spotted what horrified me.
1 coin per 0.1% change in forecast GDP would seem make sense as a design choice but only France and Canada seem to be based on this. Showing Japan at 1.8% with a bar twice the size of Canada 1.5% is just horrible. The bar for Germany is way out of proportion as well.
The German bar does at least make me wonder if this is a deliberate attempt to paint the UK/Conservative Government in a worse light than the forecast suggests or if it's just a very poor chart. For example, if I was going for maximum impact in that way I'd have used 6 coins for the UK stack and 1 for Germany.
What would look better? Here's another chart showing the same data.
While I might critique some of the choices on this chart by Statista, for example, I don't think the logo bottom right adds anything, colour choices on red/green can have colour blindness issues and I might have labelled United Kingdom as UK, the chart itself is far better at doing the essential job of representing the forecast data. Changes here would be more a case of tweaking a fundamentally sound chart rather than rescuing a wreck.
Anyway, rant over for now.
Am I being too want to harsh on the Labour Party chart and does anyone want to defend it? Or maybe someone wants to suggest other ways the Statista one could be tweaked?
Would love to read other peoples views.
Chief Data Officer at Tempcover
2 年Could have been worse… if they used a 3d donut, after all they are only good for eating! Do you think the “analysts” output was altered by PR & Marketing for dramatic, if incorrect, impact?
Chief Customer Officer at Bulbshare
2 年This needs to become a therapeutic group for us people that are kept awake at night by these monstrosities! Viva la chart revolution!
At such times I find rereading Edward Tufte's The Visual Display of Quantitative Data very reassuring.
INSIGHT SHERPAS LIMITED
2 年I'm all for a great visual, particularly when they make data more accessible, but they also need to be accurate ?? and help the reader take away the real (true) point....
Forecasting and Insight Manager at Severn Trent
2 年I don’t mind the idea of using funky coins in the original but the inconsistency in scale is a bit of a shocker!