I recently had the pleasure of attending a one-day tutorial on data visualisation, given by Andy Kirk (@visualisingdata) on the Genome Campus. I was particularly glad that we were able to organise for Andy to run his tutorial here since, rather like Noah Iliinsky and Miriah Myer, Andy frames his guide to data visualisation in terms of a design process; something close to my [UX design] heart.
Taking a step-by-step approach to exploring one’s data, learning about the audience and their goals, deciding on the purpose of the visualisation, and taking time to experiment with different possible solutions is an essential grounding to set down for people, I think.
Along with around 20 researchers and developers from the Sanger Institute, we spent the day learning about and applying the data visualisation design process that Andy covers in much more depth in his book, Data visualisation: a successful design process.
The data visualisation design process:
An effective and efficient design strategy for making sense of data, finding stories and presenting stories
A five-step plan for designing visualisations
Andy suggests five main steps in the process:
- Establish the visualisation’s purpose and identify key factors
- Acquire, prepare and explore your data
- Establish editorial focus with your subject matter
- Conceive your visualisation design specification
- Construct your data visualisation solution
He was careful to point out, too, that there is (and should be) a degree of iteration and experimentation, so that step 5 can often feedback into step 4, for example.
Sketching to communicate and explore
The tutorial is, as you might expect, a mixture of lecture and practical work, illustrated with examples (good and bad!) from all sorts of data domains. Towards the end of the session, we spent about an hour, in groups, discussing and sketching out ideas for data visualisation solutions that would meet a design challenge that Andy had set us.
I was very happy to see all the groups presenting back their sketches and explaining their design choices. I am always very keen to encourage my colleagues to work on ideas using pen and paper as much as possible, only moving into code when they’re ready to craft more solid solutions. The same goes for your data visualisations!
Eight hats of data visualisation: archetypes and roles
One of the frameworks that I particularly like that Andy describes, both in this tutorial and his book, is the “eight hats” of data visualisation design. These archetypes or roles are meant to give us different, complementary ways of looking at data visualisation challenges. It is based on Edward de Bono’s Six Thinking Hats process for parallel thinking, something that I was introduced to a few years ago (thanks to Jason Mesut), and a process which I think is a great way to tackle complex problems… which many data visualisation challenges in biology are.
The eight hats that Andy describes are
- the initiator
- the data scientist
- the journalist
- the computer scientist
- the designer
- the cognitive scientist
- the communicator
- the project manager
He envisions their activity throughout the data visualisation design process like this:
Running the tutorial again at the Wellcome Trust Genome Campus
Andy travels the world, consulting and running this course. We are hoping to have him back in early June, to run this tutorial again. The participants will be a mix of Sanger Institute and EMBL-EBI staff and based on my experience, I’m sure they will get a lot out of it. Applying a design process to your visualisation work is definitely a powerful way to get the most from your data.