For our first Friday presentation, we were asked take a look back at one of our initial two applications to The Data School and make changes by applying what we had learnt about best visualisation practices and Tableau in our first week. I chose to fix up my first application, and taking a look back at my initial application to The Data School, a few things immediately stand out;

1) The colours are a visual nightmare, past me decided to use a strange mixture of light blue, cream, and a light grey background. While this was due to the fact that one of my main variables I was charting was represented in white (as this was the colour White, quite literally), its still a lot of clutter in a way that doesn’t take advantage of cohesive colour palettes or any sort of stylistic canon. In my revisit to the dashboard, I removed all these unnecessary elements, keeping only a single colour background, which made everything cleaner. My original intent with the bizarre bars separating each section was so I could express which charts were meant to accompany which pieces of insights, but simply emboldening the chart titles and centering them achieves an almost identical intended effect without the sensory overload.

2) There’s just a LOT of text present. I’ve had to stretch out the textboxes to their upper limit before they intrude into my charts just to fit in the volume of information I was trying to convey, and had to rely on information tooltips to condense down some of text. While this is inherently good as I’ve tried to prioritise a large font size, and the mouseover tooltips are useful for giving specific answers to users with those exact questions, it can still be refined further down if anything just to give the dashboard breathing room. Just a bit further down I’ve got ANOTHER mouseover tooltip explaining more in-depth information, and its just better to have these condensed into the one tooltip (especially when they also look identical). I’ve also toned down the over-explanatory language of my initial dashboard overview, but due to the nature of the dashboard being presented towards a hypothetical audience with no prior knowledge of the subject matter, some of it is still retained, just in a more concise way that doesn’t eat up as much real estate.

3) A few of the charts are quite a bit redundant. In particular within the first section of higher level analysis within my dashboard, I’ve got two charts which functionally convey the same information. While its nice to have a specific number tied to average pick order per colour, its equally as tedious having to explain why a higher number is bad for this very statistic. It doesn’t help that the individual statistic of average pick order is then repeated in the chart next to it, just displayed in scatterplot form. Even further down in the dashboard I’ve made this mistake, charting the Top 20 cards in terms of their Power statistic, but then expressed that again in a scatterplot. In my reviz, I’ve removed the first bar chart in each instance, and enlarged out the scatterplots so that the individual details can be seen more easily. This way, the granularity of the individual card data (in particular with regards to the Power charts) is preserved, but there’s also further context, especially with the addition of an average reference line to really highlight the trends of the data.

4) Speaking of scatterplots, one of the most resonant pieces of feedback I received in my initial application was that the different shapes used to represent each colour was too visually distracting and unnecessary, which honestly can be applied to a lot of the small details within my initial dashboard. So I’ve amended this in a series of smaller changes. The shapes are now all filled circles (though having a black border would make them even better), chart lines are gone, and axes rulers were changed to a darker colour to match the new, lighter background. Font sizes were changed to be consistent, especially across the titles of my various charts and analyses, to be more cohesive and presentable as a final dashboard, and tooltips were all also similarly reformatted to be cleaner and more presentable.

There were a lot more changes I could make given more time (a better background colour, changing the chart type for a strange histogram that I didn’t delve too much here due to how much work would be needed to address its issues, and possibly a whole new orientation for the actual dashboard), but with just the things learnt within four days, I can already see so many things that could be changed and revised, so I’m excited to see what happens when I take a look back at this after even more training.

Daniel Yam
Author: Daniel Yam