The Viz That Wasn’t (And Then Was Again)
Let me be clear from the outset of this post, it’s not my intention here to criticise Makeover Monday. It’s a ludicrously brilliant project, which has spawned new vizzers, created a community and helped people to focus their careers and get employed (but that’s another post altogether). For a very good reason, it even carries its own section on this blog. What I am perhaps highlighting are the dangers of vizzing without looking both ways. Andy’s review of last week’s contributions already addresses a number of issues, and this is really an offshoot of some of that comment.
On Thursday I got caught up in a discussion online around my reservations regarding the dataset. At the time, little was known about the original work, since it was translated from Russian and the method behind the survey was unknown. Plenty of contributions were happy to pile in with interpretations that I simply felt uncomfortable with, but was more than ready to acknowledge that this was a personal issue for me, rather than everyone else being wrong. That’s not in the spirit of things.
The Vizzer Spectrum
There is clearly a spectrum in the world of vizzing. At one end live the pure artists, the other end home to the analysts. In truth, hardly anyone occupies the extreme ends of the spectrum, but each operates in a small range depending upon how they’re choosing to apply themselves on the occasion, and their relative skills with the technical tool they’re trying to use. I very much place myself towards the analyst end of the spectrum, hence I tend to build visualisations with the message I’m trying to convey in mind analytically, with the visual aid of charts. I admire from afar those whose Tableau projects are visually stunning and carry little text. I’m not there yet, however. Here’s another way to demonstrate the same thing, in pictoral form (and this only serves to underline my point that I’m no artist):
I appreciate that for many out there, Makeover Monday is an opportunity to stretch the legs a little, and to escape the boundaries that their regular work might impose upon them. They want to explore different chart types and more elegant forms of telling a story, and a ripe new data set each week affords them that opportunity. Design comes first, and in some cases the integrity of the final product can be compromised. That’s their prerogative and I’m fine with it, so long as this is suitably flagged up.
For analysis-led vizzers it’s a battle. If the numbers don’t add up, such as in this case, we want to know why that would be. We can speculate all day long, but we recognise that the answer affects how we represent the data. Increasing comprehension of the source data also adds to what we know, which can in turn help to improve the viz if we’re able to supplement it with context. For me, context is everything, but I really do appreciate that, given the confines of an hour and an ambition to turn out something eye-catching, the designers may feel they need to sacrifice a little governance.
The viz that inspires this week’s activities is called, somewhat ironically, “The Secret Of Success”. It features a radar chart that plots three social groups as dimensions and makes no attempt to explain what the numbers on the axes relate to. In the data source, these are presented as a rate, but all we know from the original is that the question was “what are the main reasons for success” and that there are eight responses possible. Are we to expect that each respondent could only choose one, more than one, or as many as they liked? This isn’t clear – we simply don’t know. But because we see that the sum of responses is greater than 100%, it seems easy to infer that they weren’t just given one choice. Even then, however, this is our assumption. We don’t know whether the vizzer in question, Maksim Abrosimov, may have made an innocent error and somehow multiplied each answer through by some weird factor.
It was for this reason that I steered clear. The tiny artist in me failed to speak up over the big analyst. And then, like a fairy godmother, Chris Love came along:
A little digging into the source of this weeks #makeovermonday – here's the original afaict https://t.co/vjX3LwN3ob @VizWizBI @TriMyData
— Chris Love (@ChrisLuv) March 28, 2017
There was an original report available that included some charts. And the viz used as inspiration for Makeover Monday was, in itself, a makeover. And an extremely arty one at that. Which seems to serve my point, a little – Maksim has sacrificed integrity for the benefit of style, and finds himself or herself at the other extreme of the spectrum to me.
As Chris helpfully calls out, and some further use of Google Translate supports, there were so few respondents to the survey who identified themselves as “Rich” that analysing this is a risky approach without plenty of caveats. Of 1,500 responses, 1% identified as rich. These figures are rounded to the nearest whole number, so we can take it as anywhere in the range from 0.5% to 1.49%. That equates to somewhere between 8 and 22 people. The sample is too small to be meaningful, the possible error margin too high. Yes, you can plot the data, but you should also sprinkle on it a massive health warning.
However, now of course I’m in possession of some context. Perhaps I can contribute an effort myself this week after all? I can strip out the ‘Rich’ responses and compare simply the other two main categories. Which, as it turns out, don’t include “Middle Class” but “Average Income”, which in itself carries different connotations of course.
And I can also cite the original source of the data (which isn’t visual.ly, which is the source of a bad viz). With that, I’m happy I can put something together. So, here’s my effort. And, of course, I will be glad to receive any stylistic tips that might help me to shuffle towards the middle of the spectrum!