As the 2016 French Open Tennis began a couple of weeks ago, the withdrawal of Roger Federer through injury was major story, both from a sport perspective (Federer remains a leading contender and his absence will be a major boost to his competitors) and from a data perspective (Federer is so consistent, that he has not missed a grand slam tournament since last century). This got me interested in looking at tennis data, resulting in the visualisation below. Click on the static version to interact:
I’ve highlighted Federer’s Grand Slam record in red at the top of the chart, and have mentioned it below the title. But after getting my data from tennis-data.co.uk, I felt my data was just too good to focus on just one story. So many other things stand out to me: the extraordinary win records and grand slam histories of Nadal and Djokovic; the dogged longevity of Hewitt, still playing 13 years after his last Grand Slam win; the remarkable consistency of Murray and rise of Wawrinka, the unfulfilled potential of Del Potro or Roddick, and many more.
I don’t show this as an example of a highly sophisticated visualisation, or one that uses any technical skill whatsoever. But I do think it’s an example of a fascinating data set (if you like that sort of thing) which doesn’t need to tell a story. If anything, this has gone in the reverse direction – I’ve used a story to prompt me into investigating, visualising and presenting data for the reader to delve further by themselves.
In my professional work, we analyse the data using various statistical and analysis tools and look for stories for our clients, which we then detail in written reports backed up by graphical charts. An interactive visualisation is an extra tool for our client to drill into the data in any way they like, so it feels like a chance for the dashboard’s user to find their own additional story without me leading them down just one path. Hence, I think, the desire not to constrain or focus on a particular story over and above others. This also shows in this previous post on global famous people through history. As discussed in that post, there were opportunities to focus on many subsets or stories within the large set of names in the data, yet I wanted to present them all.
Here’s another example. The spiral graph below has been, rightly, widely shared and admired. It’s achieved its goal perfectly – raising awareness about global temperatures rising.
The thing I like most about the original is the way that the very last year and a bit spirals noticeably away from the centre. It allows us to use emotive terms such as “spiralling out of control”. However there are downsides to it too: (what do the colours mean? Can you easily make out trends in any other year pre-2014 when it has finished? Do the red circles mean anything?) So, as a #makeovermonday challenge a few weeks ago, it was an opportunity to produce a still version of the visualisation
My version is below. I must have done something right as it was named first “viz or the day” by Tableau Public, and then “viz of the week”. Again, click the static version to see the interactive version.
Here, I elected to tell no story at all. As far as I’m concerned, there is only one story, and that is that global temperatures are rising (sorry, climate change deniers, but it’s true). To me, the blue and orange in the visualisation tell that story, so there’s no need for me to over-state it.
A further danger in telling a story is in getting facts wrong or making inaccurate conclusions. I saw other otherwise excellent visualisations making conclusions which weren’t correct (rising temperatures are down to the El Niño effect – incorrect, or at least not strictly accurate, because this is an occasional event that only takes place in certain years therefore does not affect every year; temperatures have risen by over a degree since 1990: incorrect, the baseline is averaged over 1961-1990). As a visualiser, we don’t always have the scientific background or access to the original resource/journalism which has generated our data, so we need to be careful not to overstate our conclusion. In my opinion, if we’ve designed well, there’s no need – the reader can infer the conclusion themselves.
So I prefer not to tell a story with my visualisation if:
(a) the data is so good that there are lots of stories – to tell just one would be to miss several other stories or sub-plots
(b) the data makes the story clear enough and stating any story would be stating the obvious
(c) I’d be introducing causality or drawing conclusions that weren’t strictly accurate
I don’t have the experience or authority in the field of data visualisation to say that this is what you should do in these situations – these are just my thoughts on how I approach things and the reasoning behind those thoughts. So instead I’d love to hear feedback, whether you agree or disagree with my thinking.
Interesting thoughts. In general I agree with you – it isn’t always necessary to tell a “story” with data. And by “story” I mean a narrative, or sequence of annotated points, as opposed to just a finding or insight, which some people mean when they use the term.
A few years ago Moritz Stefaner pointed out that there are also other “genres” of data visualization such as tools and portraits which don’t involve a narrative per se: http://well-formed-data.net/archives/868/look-ma-no-story
Here are some thoughts on your three “no story” criteria:
a) “There are too many good stories” – there is huge value in exploratory visualization, but sometimes you don’t care about the other stories, or they are extraneous to your goal. In those cases you either want to leave the other patterns there for context but in the background (grayed out, for example) or take Cole Nussbaumer Knaflic’s advice in Storytelling With Data and eliminate the clutter and focus your audience’s attention. Only you know what will best help you achieve your communication objective.
b) “Why state the obvious?” – Yes, we wouldn’t want to be overbearing, but there is just one thing to watch out for here. When you look at your data for hours and hours and hours, the key insights seem so obvious, but they still might not be for the person first encountering the data through your visualization. I agree that a full fledged narrative isn’t always warranted, but I have yet to see a well worded annotation that I didn’t appreciate, keeping in mind….
c) “You shouldn’t say more that you can actually say…” SO true, and thank you for pointing this out! It’s a sobering reminder that we all need to examine the stories we tell with a critical eye. Are there biases or assumptions we are introducing into the narrative? Does the data itself warrant the claims we are making? Are they statistically valid statements?
Great thoughts, thanks for contributing to the discussion! I feel like we are all experts in some ways and novices in others, so you should feel empowered to share your thoughts. I hope mine have been helpful.
LikeLiked by 1 person
I agree that not every visualization tell a story, but I think, as you and Ben point out, each visualization should encourage a story to emerge. Whether that story is your own (i.e. you select a single story to tell among the many possibilities) or one that your viewers discover themselves, the choices you make when putting together the visualization should always be in the service of those narratives.
In my professional life I am often tasked with presenting data to a myriad of audiences who need to discover the stories for themselves, so my default approach to visualizations tends to be to amplify whatever signals exist and minimize (or eliminate entirely) any noise. This is why I love opportunities like Makeover Monday where I get the chance to put a narrative stamp on a data set and not worry that I haven’t made it exploratory enough. it’s a good balance for me, and a style of visualization I find that I enjoy more.
Stories come in many shapes and sizes, and so too will the visualizations that birth them.