The third instalment of #MakeoverMonday in 2017 encouraged participants to visualise all the accounts Donald Trump retweeted during his presidential campaign. My own submission was nothing of note, but it did at least reinforce some logo swapping knowledge I picked up last week. In addition, it gave me an opportunity to practise using String functions to tidy up a pretty hectic “date” field:


It might not be the most elegant solution, but it works and that’s good enough for me. Incidentally, I also pulled the source data into the Tableau 10.2 Beta and made use of the magnificent new automatic date parsing functionality. Here’s the source data on import:


Initially it’s regarded as a string because of all the crap at the end. How does the automatic date parsing work? You just change the data type to Date. That’s it. Instant clean up:


Early on in the Beta I threw a few different formats at the functionality and it handled the ones you’d expect it to, and struggled with crazy ones. It won’t resolve all date parsing issues, but it’s going to go down a storm when it goes live. Part of me is a little sad about this as it’ll mean foregoing a few opportunities to practise some of the Date and String functions, but I can probably live with that.

Having said that, I did want to see how a few other people converted the string into a usable format. Here’s the most popular approach:


That’s Miguel Cisneros using the DATEPARSE Function:


So basically you need to mimic the format of the string using a series of letters denoting what period of time each value in the string represents. But what does the “E” mean? Why are some “m’s” capitalised whilst some are in lower case? In this rare instance, I didn’t find the Tableau documentation up to scratch, so I Googled and found this article, which includes a comprehensive table confirming that among other things:

E = day of week

M = month in yer

m = minute in hour

In fact, the four other dashboards I downloaded (by Andy Kriebel, Tom O’Hara, Josh Tapley and Sarah Bartlett) all used the exact same calculated field, so I guess that’s the best way to do it! I’ll tell myself that they all just downloaded a workbook and ripped of another users’ approach to make myself feel better.

But what other visualisations caught my eye this week? I’ll be honest, I didn’t find the variation this week to be huge, and I didn’t really want to revisit line charts, jitter, word clouds or bar charts. What I really didn’t want to do was look at any form of mathematics, and then Miguel Cisneros ruined my plan with a completely unique submission. Last time I tackled radial charts, I had to lie down in a darkened room for a couple of hours.

With words of encouragement ringing in my ears, I dived in:

Screen Shot 2017-01-16 at 20.47.20.png

Deep breath. And go! There are five separate worksheets, and I’m just going to focus on the “main” element with the large radial chart. If I delve into small multiple radials, well that would be a world of pain (but I did really like the calcs used to set out Rows and Columns which differed to all other small multiple examples I’ve seen. Maybe one for the future…..).

Screen Shot 2017-01-16 at 20.53.35.png

[Y] is:


With [Difference (or, Day of Month)]:


And [Angle]:


[Seconds Since Midnight]:


[Seconds in a Day]


Phew! It’s at this point that I begin to question whether I’ll trust Miguel’s definition of “simplicity” in the future, but let’s work this backwards and see if any of it sticks. 60 seconds * 60 minutes * 24 hours gives a uniform 86,400 seconds per day.

That’s required as the basis from which to plot the [Angle] as Miguel then calculates the [Seconds Since Midnight] by multiplying each hour by 3,600 (as 60 minutes in an hour * 60 seconds in a minute is 3,600), each minute by 60 (i.e. the number of seconds in a minute) and each second by itself.

Dividing one by the other, and having that value multiplied by 2 * PI() gives you an [Angle]. I’ll not worry about the precise mechanics of that, other than to show a table of some of these key calculations:

Screen Shot 2017-01-16 at 21.27.23.png



No extra calculations to deconstruct. What does the above look like in a reduced data table?

Screen Shot 2017-01-16 at 21.29.45.png

Next, I shamelessly looked at a kids maths resource, and in particular paid attention to an interactive diagram:

Screen Shot 2017-01-16 at 21.41.25.png

If nothing else, it reaffirmed that my maths days are well behind me! When I start using phrases like “round and round” and “in and out” in my head to describe what [X] and [Y] are doing, I know I’m in trouble. Do I vaguely understand what’s going on? Well, yes, if you have a really, really vague in your spectrum of vagaries. Could I do something like this from scratch? Hell no.

A final observation was the clever use of a background image upon which to plot the data. I was bamboozled for a while by the fragmented concentric circles and hour labels, but I needn’t have been:


Screen Shot 2017-01-16 at 21.54.56.png

In summary: it’s a beautiful viz with a lot of clever components. Aesthetically beautiful and structurally it absolutely fits the data by plotting tweets around the clock. But: it’s too much for me to get my head round so I need to be able to step back, doff my hat at these relatively complex visualisations in future, and then deconstruct a bar chart!