Hot on the heels of my last post, comes another related to the same #MakeoverMonday dataset. This week, a number of different submissions caught my eye, and five in particular were variations of a similar theme. I knew how to achieve a high percentage of the final output, but not all of it. Time to try to plug that knowledge gap.

The submissions in question were, in no particular order, by Adam Crahen, Bethany Fox, Curtis Harris, Matt Chambers and Andy Kriebel. See if you can spot the pattern:

What did I like? The filled space between the Measures. Adam’s example is the only Slope Chart (hence the angled lines), whereas the other examples are basically “just” dual-axis charts. I’ll tackle Adam’s viz first, because I know how a Slope Chart works, and it’s always nice to get off to a decent start…..

I’ve compiled a post about Slope Charts before, and I’ve dusted it down often enough to retain the process pretty well. To recreate Adam’s dashboard is pretty straightforward. First up, I’m going to whack Year on Rows. If I then put the Top 0.1% on Columns, and then repeat this for the Bottom 90% (putting them on a shared y-axis), this happens:

screen-shot-2016-11-28-at-19-41-34

A bit to do still. Easiest part first – replicate the instance of Measure Values in Rows (on a Mac it’s a simple case of Cmd+Drag on the first instance). Right-click the second instance of Measure Values, hit Dual Axis and then right-click the rightmost y-axis to Synchronise Axis. Change the Second Measure Values instance on your Marks Card to Circle, and tweak the sizing of the original Line and new Circle Marks until you get it about right:

Screen Shot 2016-11-28 at 19.49.46.png

The issue now is the lack of a Slope. That is because we aren’t telling Tableau how to break the view up on Columns. It’s just one Continuous (Green Pill) date axis at the moment. I need to put Measure Names on Columns, but watch what happens when you mix Green and Blue Pills:

Screen Shot 2016-11-28 at 19.54.26.png

Green Pills always get pushed to the right, so the view becomes two separate columns – one for each Measure Name. To resolve this, simply make the Year Discrete, and then drag it to the left of Measure Names:

Screen Shot 2016-11-28 at 19.58.37.png

Ah. This looks all sorts of knackered. Tableau’s Marks Card seems to recalibrate itself / has a hissy fit when you cock about with Discrete / Continuous. It’s just a case of setting the first instance of Measure Values to Line, and the second back to Circle. That, combined with a bunch of clean-up to sort out Row and Column Dividers, gridlines (please Tableau – can we just default to having them all turned off?!) and some formatting, results in something vaguely resembling the end product:

screen-shot-2016-11-28-at-20-05-37

Cool! One down (barring the not insignificant tooltips and general dashboard construction,  but I’ll gloss over that).

Matt’s viz is similar enough to Adam’s not to warrant a whole documented rebuild. The principal difference is the absence of a Slope in favour of a standard dual-axis combination of Line and Circle. As I see it, the only complication here is the use of the Path shelf to get the required look. Let’s see what that means, once the bones of the chart have been stuck together:

screen-shot-2016-11-28-at-20-14-55

I’d like to draw your attention to that Year again – it’s Continuous. Look what it does to the Lines. They still think of themselves as a Continuous period of time, so you get these short diagonal lines from one year to the next. If I switch to Discrete and sort out the usual Marks Card hissy fit, we arrive here:

screen-shot-2016-11-28-at-20-18-03

Same shit, different shovel. Here’s where the Path Shelf makes itself useful. If I drag Measure Names onto it, the view is chopped up into bitesize, Discrete chunks. The Measure Names are compared down each individual Year, and the fundamentals of Matt’s dashboard are created:

screen-shot-2016-11-28-at-20-21-49

Time to leave the comfort zone, and take on one of the three similar-ish (conceptually, at least) remaining viz’s. Curtis and Andy get plenty of airtime already, so I’m going to focus on Bethany Fox’s submission. In her viz, Bethany uses parameters in a similar way to my own submission, so I’m not going to tread on that ground again. Instead, my attention is devoted to the aesthete – specifically the way the gap between the plotted Measures has been shaded:

screen-shot-2016-11-28-at-19-30-37

It’s probably quite an obvious thing to most, but I really have no idea how this is done (!). Let’s kick off with a rebuild of the obvious bits, then a pause and look at the original when I get stuck:

screen-shot-2016-11-28-at-20-38-03

An easy start, which is nice. The fiddly bit has two component parts:

  1. Fill the white space between the Measures
  2. Colour it appropriately – where Measure X is higher, colour the gap so, and vice versa

I knew that a Gantt would be an option, but wasn’t 100% sure how to get it to click, so here’s how it is done. First of all, given that I’m plotting the Top 0.1% and the Bottom 90% and allowing no flexibility, I know that I can plot one of those Measures as a secondary axis. I chose the Bottom 90% which was plotted as a Dual-Axis and synchronised. I won’t show an image, as it looks the same as above – I’m just redrawing the Bottom 90% on top of itself.

The cunning part is filling the gap, and that’s all about creating the equivalent of a sort of “Duration” field, which in this instance is simply:

screen-shot-2016-11-28-at-20-45-44

The first thought might be to drag that onto Size, but this will happen:

Screen Shot 2016-11-28 at 20.46.40.png

The secondary axis is sized according to the calc. We don’t want to do that though, we want to plot a “duration” from the baseline. If we consider that the Bottom 90% is already plotted as the baseline, then we just need to change the chart to a Gantt, as that will add the “Gap” calc onto the plotted baseline:

screen-shot-2016-11-28-at-20-48-43

It’s a bit of a pixelated mess at this stage, but if I dial down the size of the Gantt, I should be able to refine the width of each plotted point:

screen-shot-2016-11-28-at-20-49-49

Hmm. So what I gain on one hand, I lose on another. The pixelation is gone, but the overall fill of the space is the compromise. I can live with that. The final part of the conundrum is colouring the gap. If the Bottom 90% is higher than the Top 0.1% I want one colour, else I want another. Simple logic calc?

screen-shot-2016-11-28-at-20-52-30

If I shove that in place of the default Measure Names on Colour, magic happens:

screen-shot-2016-11-28-at-20-53-15

Editing the colours to match the primary axis yields a pretty decent final chart:

screen-shot-2016-11-28-at-20-54-45

Another #MakeoverMonday week concludes with me learning a load of new skills. I’m certainly not the only one on this learning journey each week, and I hugely appreciate the efforts of the two Andy’s in delivering us with this experiment each week. I sincerely hope that the fact we have hit week 48 doesn’t mean that we only have 4 to go for eternity. I’m pretty sure the community will pick the baton up if they relinquish it, as it’s simply too good an initiative to allow it to run out of steam.

Advertisements