Friday, 6 June 2014

Data visualisation workshop with Andy Kirk

This blog post was written on the train on the way home from attending a Data Visualisation course by Andy Kirk : 

So today i had the pleasure of finally meeting Andy Kirk in person by attending his data visualisation course in the big city of London. Now i have stalked Andy for sometime on twitter and been enthralled with his tweets on travel and other equally exiting topics. Occasionally he posts about some tweets about data viz. It seems he is somewhat of an expert in this field, he's  written a couple of books and everything. So I decided to pop down and see what his lessons were like. Despite arriving late due to missing my train by a fraction of a second, seriously I was pressing the open door button but nothing, the course started well with an introduction to the topic. 

Its always a tense time at one of these things when you are being shown bad data viz when you hope and pray that one of yours is going to pop up. Well luckily none of mine were featured in the rogues gallery. Andy had a some real choice examples of just bad bad bad stuff, including the worst pie chart i think i could ever imagine. 

The course went though the thought process of designing a good data viz, it was software agnostic so covered the broader topics of what to and not to include. Andy spent a good amount of time explaining the common data visualisations, bars, scatter plots etc and when and were they should be used. 

There's a strand of the data viz world that argues that everything could be a bar chart. That's possibly true but also possibly
a world without joy.
Amanda Cox, New York Times

There is a lot of voices in the dataviz community that hate chart junk and can be quite dogmatic about it. And yes in a purist sense its fair to say that a lot of the time a simple bar chart would convey the data in a more straightforward way. But, as the quotes andy used said, thats quite boring. It really depends on the situation where the viz is going to be viewed and what the purpose is. If the viz is being viewed as part of a dashboard then a minimal colour scheme and bars or bullets is probably the best choice, you want to see at  glance if there are any issues that need looking into. In that case you have a captive audience that are looking at the dashboard as part of a routine task. On the other hand, if the viz is part of a webpage, or news paper, you only have  a second or two to catch the eye of the reader. You need to engage with them quickly, grab the attention and hook them in. Now in that situation a set of bars all the same colour might not do it, maybe a different viz style would be best. The take home message was, it really depends.
There is no magic bullet for these things, in each case its a judgement call on each situation. Often the most critical people are not the target audience. The Twighlight films might not go down well with 50 something male film critics, but as far as the teenage audience is concerned they love em. The film is made for the audience, not the critic, and so should your viz. So what if it breaks a few rules and gets the purists up in arms, if the audience engage with it, understand it and get the message that you were trying to put across then its worked. 

That same question then leads you to decide wether your data viz is going to be exploratory or explanatory. Are you going to give your audience a data set with a collection of filters and interactions and let them loose to see what insight they can find? Do you want them to read the viz? If the purpose is to explain something then do you need more text, more tool tips. Are you trying to elicit an emotional response from the audience? Which ever of those choices you make then influences the direction in which your viz will take. 

"It’s a common mistake to think that charts are just a fancy way of showing numbers. They’re not. They’re tools for understanding."
Robert Kosara ,

This message really hit home with me and made me think about how i have created vizzes in the past. Am i just creating a bar chart for the sake of it because i think i have to create one rather than show the numbers in a table? Does the bar add anything to the understanding. I'd like to think that i do aid the understanding but it will certainly be something that i will consider in the future. 

I thoroughly enjoyed the course, probably one of the best i have ever attended really. It felt good to know that what i have been doing so far is "correct" and that i hadn't fallen into some of the bad habits Andy talked about. However, its gonna take a long time to get the image of this out of my mind. 

Matt Francis

Author & Editor

Tableau Zen Master, Social Ambassador, Wrangler of Data, Vizzer of Data