Data Rocks
Jan 15, 20207 min
Updated: May 21, 2020
Makeover Mondays is a Tableau focused project run by Eva Murray and Charlie Hutchenson, challenging the data community to create a more effective version of a given visualisation. People then post their creations on Twitter and can tag them for review (tagging #MMVizReview). It is a fun and engaging way to practice your data viz skills, get the rust off of your Tableau knowledge and learn a lot along the way.
I've been using Tableau since 2016, and although I knew of the Makeover Monday project for quite a while, I've never found the time to join. As I recently made some crucial adjustments to my lifestyle, I committed myself to participating in every week of 2020. This week's submission was my first ever #MakeoverMonday and this post will give a sneak peek of my experience and process.
The original visualisation was extracted from a Vox article, exploring a Gallup survey about American's favourite sports. This is what it looked like:
The chart is straightforward - TOP 3 sports and their change over time.
It works quite well within the context of the article
The title and sub-title are self-explanatory, and the labels well-placed - the reader doesn't need to guess anything.
The lines are very smooth in previous periods and very squiggly in recent years - that might indicate that there isn't data for each year and the line is actually only "connecting the dots" where data is present - but that isn't clear.
I know the measure is a percentage, but from the chart alone I'm not sure percentage of what. Reading the article it seems like it's a share of people who favour one sport or another, but it would be helpful to have that stated somewhere near the Y-axis.
I'm not into sports. Nor do I live in the US. When I first saw the original viz I thought to myself I wouldn't be able to pull it off. The data source itself though is incredibly simple: it includes the sports in a category column and the percentage proportion of people who picked each of them across a few years. There are missing years in the data, but nothing that would impact the analysis too much. It fundamentally only displays the proportional trend over time of each choice. How can it be any interesting?
I then turned to the closest person to me at the time (my partner, he has a vague idea of what I do, but is not a data viz person) and asked him what would interest him in a survey about favourite sports in the US. His answer: "how many people said they just don't like any sports? I'd probably be grouped with them."
EUREKA!
I then remembered this Data Viz Today podcast I listened to not long ago, about how to make Sports visualisations engaging for a diverse audience - including people who just don't care about the topic. While the takeaway advice is not exactly where I landed, it pointed me more or less in the right direction: interactivity would probably do the trick.
I also had in the back of my mind something I saw Alberto Cairo briefly post about on his blog. He called it adding a "me layer" to a data visualisation. It means that adding some sort of feature (or layer) that helps people see themselves (or their choices) in a visualisation, makes it that more relevant at an individual level, helping to engage people.
I then had a general outline of what I wanted my viz to say:
These are the favourite sports according to a survey made with US people;
If you don't like sports, none is the 3rd most popular pick;
If you like sports, you would be interested in seeing who you'd be grouped with;
If you're not in the US and this is different in your country, here's how your favourite sport would compare to theirs.
The next step for me is always a big sketching session. This is where most of my time goes and probably where I need to get more efficient to meet the 1 hours target of Makeover Mondays. I like sketching crazy ideas before I roll up my sleeves and start developing.
There are several reasons why I swear by sketching as a habit before you get to do anything concrete in any software, but I'll get to that in a different post.
Several colourful pens later, here's a snip of where I landed:
Also known as the real fun part!
The data had years spread through columns (being one column for each year), and I don't like very much to work with years spread like that, so I did a Pivot between columns and rows - which repeats the Categories across each year but leaves all results in one neat values column I can easily use to write formulas.
I opted to do this in Tableau itself, although I could have also done it in the spreadsheet. It's a choice. I was concerned about time, so I didn't really give it any long consideration.
This is an excellent tutorial on how to do this.
Since I decided to go with a waffle chart, I added an extra sheet to the data, to allow me to plot the squares easily. The full detailed tutorial where I learned how to build waffle charts in Tableau can be found here.
With the data in the shape I wanted, I first created the waffle chart.
You can see from my sketch that the initial idea was to have the chart with people shapes, to go with the theme of bundling together with people that share your interests.
When I showed it to my partner though, his natural reaction was to try to click in each shape, expecting for something else to happen.
I guess I individualized the chart a bit too much - misleading him to think he would find more details about each person or group of people by clicking or hovering over them. I then decided it would be just as effective and less confusing to just go with the typical waffle squares as they get the message across, cutting the unnecessary fluff.
The line chart and table are pretty standard, except that I couldn't figure out an elegant way to do the index ranking within the same table but before the dimensions column.
I tricked Tableau by placing another table, with just one column with an index formula right beside the original one, in a container - surely there are more elegant ways to achieve this!
Putting it all together, I created a string list parameter (dropdown) based on the Sports category field. For the sake of clarity, I grouped all sports with 0% as "Others", including the "Other" option. There are just 5% of people in the last year with data that fall under this category, so I believe it's a reasonable adjustment.
I then create a (T|F) formula to add to the sheets I want the parameter to hide in case no sport is selected (like in the default view).
Tip: The parameter show/hide relies on containers collapsing when they become empty due to the parameter selection, but for some reason, Tableau doesn't collapse the containers that have titles. The parameters also can't be applied to text boxes. The workaround is to create sheets just for the titles and text fields, and include them as sheets above your charts.
With all that setup, the last visual detail was to create a colour formula to follow the parameter selection, and highlight the user's choice in purple:
Last, but not least, the text you see beside the waffle is formula based too:
How will my "audience" react to it? There's only one way to find out...
After everything was completed, I showed the final product to my partner again.
It was great because he spotted typos, situations where measures had funny values going on, unintended behaviour of containers, etc. Even though it's not necessary to have someone reviewing your Makeover Monday viz before you post it, it is incredibly invaluable to get feedback.
I tagged #MMVizReview on Twitter when posting my viz. This means that the Makeover Monday team got to talk about it and provide feedback live in their weekly webinar.
Because of my time zone (that's New Zealand!), I can't join live unless I wake up at 5 am... So I watched the uploaded session shortly after.
Their feedback:
To add a big % label over the waffle chart to make extra visible what the number of people interested in your choice was;
To set up the highlighted line in the chart to come to the front of the other lines when selected;
By the end of the week, I had woken up to a tweet saying my viz had been chosen among so many great ones, to feature as their favourites of the week!!!
As someone who genuinely believes that practice and iterations make it closer to perfect, I gladly took the feedback in and added a second dashboard tab to my workbook incorporating the suggestions (which you can see and download from Tableau Public).
I have a personal flow I like to follow to build any visualisation, and my main struggle with this first Makeover Monday was to keep it within the suggested 1 hour for development. I had a few interruptions on the way, had to work full time, get back home and still finish the viz - it was quite messy. I'll try to do better next time!
The Tableau community is great! I've felt incredibly welcome and saw so many inspiring views posted!
I don't personally like social media. I use it only for work purposes, and this is my very first post ever on Twitter, so everything is very new to me. It can get overwhelming quickly.
The workbook I created is incredibly messy. There are formulas that I ended up not using, sheet names make little sense, etc. I'll get better on this one too as I know people might use it as reference to learn - don't do this with your workbooks. Be tidy.
The fact that the very first viz I've posted to the broader community was so well received was equal parts flattering and unexpected.
You can follow me on Twitter @DataRocksNZ and check my upcoming Makeover Monday submissions.
And while you're at it, check my Tableau Public Profile