The Design Process: The Good, the Bad and the Ugly

September 16, 2020

This summer, I was busy writing a chapter on the data design process for the book Big Data Visualisation: Recent Developments and Challenges. It will be published by CRC Press in 2021—stay tuned! In the meantime, I’m making some of the chapter content available here on my website. Last week I shared the process visuals and cheat-sheets. This week, I’ll discuss the design decisions that led me to the visualisation which is the centrepiece of the chapter.

Below is what the final visual looks like. It shows the Goodreads ratings for the authors of the best data design books. The authors are grouped by their backgrounds.

The visualisation I created for the book Big Data Visualisation: Recent Developments and Challenges.

When I showed this final version to a close friend of mine, his reaction was “Wow, this is beautiful, you’ve improved your design skills so much!”. But when I shared that it took me a full month to give birth to this design, he was shocked.

So why did it take me so long?! I’ll walk you through the good and bad choices that led me to the final design. I won’t talk about the data itself or the editorial choices in this article (more on that in the book!).

Take One: The Big Circle

I knew quite early on that I wanted to create a radial chart, and I had a clear vision of it in my head. I had fallen in love with the visualisation below by Sara Piccolomini and planned to borrow three brilliant ideas from it:

  • the subtle colours for categories and text,
  • the perfect circle formed by the bar labels,
  • and the thick circle-like lines in the middle that indicate different categories.

The Black Data of Piemonte by Sara Piccolomini

In addition to the three ideas above, I wanted to add little circles in the middle of the chart with another encoding—the average star-rating on Goodreads for each author. This was inspired by Sonja Kuijpers:

The Art of the High Line by Sonja Kuijpers

I was ready to roll! To begin with, I created the base for the radial chart with Flourish. Once I was done stealing ideas from Sara and Sonja, this is how the result looked like in my Figma workspace:

Take one: the big awkward circle

Well, this just didn’t work. As much as I wanted the bar labels to form a perfect circle while far away from their bars, the data didn’t care. And the data always has its way. I needed to make the radius of the circle narrower to avoid those awkward gaps between the author names. And let’s not even mention those cut-off pieces of circles in the middle that look like clunky pancakes.

I created another version of the radial chart with RawGraphs, and worked to make the circle smaller. It now looked a little better.

Take one: a less awkward smaller circle

Take Two: The Little Circles

For some time, I entertained the idea that the positioning of the labels as above could work. So I moved on to creating an encoding with the little circles. I tried different variations which you can see below. Those bubbles, tiny or big, coloured or not, didn’t look good.

Let’s recap: by now, I had created a big radial chart for total ratings and small circles for the average number of stars. I didn’t like either.

Take Three: The Contrast

At this stage, I was feeling frustrated and didn’t know how to fix the visualisation. Inspiration came, as it often does, from looking at other people’s work. I was reading the book Visual Journalism and stumbled upon this visualisation by Valerio Pellegrini and Michele Mauri.

Wikiflows - One Year of Wikipedia, by Valerio Pellegrini and Michele Mauri

It was a breakthrough for me. Turns out, the bar labels can look good even when they’re close to the bars—there’s no need for a perfect circle. In addition, bright colours and black labels provide strong contrast that is nice to look at. I decided to adjust my design based on these revelations.

Much better, isn’t it?!


Take Four: The Layout and More Circles

My initial plan was to create an A4-sized visual like the one from Sara Piccolomini I referred to earlier. For that to work, I thought I’d add a description and a legend at the top and a couple more graphics at the bottom.

So I did, and this what it looked like in Figma:

Take four: the layout. Trying out the vertical layout for the first time

I’m sure you’ll agree that the above is hideous. Too much colour! So I attempted multiple other versions with less rainbow, hoping that would help.

Take four: the layout. Checking if the vertical layout could work with less colour

You’ll notice that I also got rid of the little circles in the middle and tried to reintroduce the original idea of thick circle-like lines. But I still wasn’t convinced. So I thought, ok, what if I used the connecting lines from a dendrogram to group categories? I already had them ready from RawGraphs.

Take four: the layout. Bringing back some of the elements from the initial dendrogram

Ah, finally some light at the end of the tunnel! After that, I tried multiple layout options until I landed on a horizontal design, with the title and the descriptions placed on the right side of the chart.

The last image you see is what I shared with my peers to collect feedback. After some more iterations on colour and wording…

…I finally had a design that I was happy with! 🎉

The final design!

Final Thoughts

In this process of trial and error, there are two lessons for all of us.

First, don’t get attached to your initial ideas. You may get lucky and they’ll work out. Quite possibly though, none of them will look good and you’ll have to throw them out and figure out something new.

Second, good design is as little design as possible. If your design doesn’t work, chances are you should remove some elements from it. And then remove some more.

✨But remember, even when you’re in the dark phase of your project in which nothing works, keep pushing, and it will turn out great.