Hans Rosling has been justly acclaimed for a couple of TED talks on global health in which he makes mesmerizing use of his (and now Google’s) GapMinder software, which he uses to tell compelling stories with data. The software is very cool, but what really makes the stories come to life is Rosling’s narrative. Data analysis, for him, is a performance art.
I’ve been thinking about this because I’ve been trying to investigate a perceived crime wave in my home town. You’d think it would be straightforward to get hold of the data but, after four months, I’m still trying. Meanwhile, however, I found some historical data at the Bureau of Justice, and I decided to see what I could make of that.
The visualizations shown in today’s screencast were done with Many Eyes, which is another very cool piece of software. But what I realized while making them is that narrated animation is really the secret sauce. Analytical software, whether it’s Excel or GapMinder or Many Eyes or something else, is necessary but not sufficient. The stories that people will understand, and remember, are the ones that have been performed well.
Now I’m no Hans Rosling, and you certainly won’t see me swallow a sword at the end of this screencast — as he amazingly does at the end of this video. But I will be trying to emulate his example when I tell stories with data. And I’m struck, once again, by the way in which screencasting can bring software interaction to life.
The charts used in my screencast could have been made in Excel or in any other charting package. By making them in Many Eyes, I added the important new dimension of social analysis. So you can visit the data sets there, comment on the visualizations, and add your own visualizations. But data analysis as performance art goes beyond the snapshots produced by analytical tools. It lives in the interstitial spaces between the snapshots, traces a narrative arc, shows as it tells.
14 thoughts on “Data analysis as performance art”
Very interesting post. I wonder what your thoughts are on narration as a narrow path through a fairly rich space of possible interpretations. Traditional visualizations, such as those in Time or National Geographic, reflected the biases of the editors and ‘narrators’ of those magazines. For example, it would be interesting to see your data mapped onto population statistics like average household income. In your narration, I’m prevented from seeing that.
Clearly, the big win in your scenario comes from the social analysis provided by Many Eyes. That way, viewers — possibly those worried citizens in Keene — can perform their own analysis as well.
“Clearly, the big win in your scenario comes from the social analysis provided by Many Eyes.”
Agreed. I am really hoping to encourage the idea that regular folks can not only access data, but incorporate it into their routine civic discourse.
And I am particularly looking for ways in which that civic discourse can be grounded in data, in a way that’s much more fluid.
Many Eyes is moving in the direction of more annotation tools, and I’ll bet in time it (and similar kinds of social analysis services) will become platforms for doing presentations.
It’s interesting because I recently saw the founders give presentations based on some of the stuff in Many Eyes, and of course their tour through the data made it come alive in away that it doesn’t when you simply visit the site. Of course that excitement then translates into making you want to visit the site and engage with the data, so it’s a virtuous circle.
“regular folks can not only *access* data, but *incorporate* it into their routine civic *discourse*”
Absolutely! I still think Many Eyes, for example, is missing a trick with the *access* part. You can’t get any more accessible than an HTML table at a public URL, using standard elements for headers, footers, rows, columns, cells, a caption, etc. But Many Eyes hides the data as a partial “preview” at a randomised URI in a table with no id.
You need that URI fragment to do proper *incorporation* – transformation, visualization, filters, pivots, etc. You shouldn’t have to upload any data, you should just provide the URIs, and a set of mashup services can get going!
As for *discourse* – explaining, annotating, and discussing the data collaboratively – this is another area where HTML beats spreadsheets hands down – they can’t even layout text properly!
See my blog post for an idea of how this should all work – whenever you add a table to your blog, spreadsheet-like functionality should be available within the blog editor.
I enjoyed watching this John. I luv a good data story. I also find your theory interesting particularly since i work in a lab where we believe that people learn by seeing and by hearing. But i think there is more to it. When i look at data being presented to me, i always look for data sources, methodology, the type of data being presented, how the data are being shown, what is on the axis and then i wonder if that is the logical and accurate way to show that data statistically, about who is telling me the story, what they know about social data analysis, who they work for, who paid for the research and why they are telling me their stories. This helps me build trust or not in what i am being shown and told. When I see Hans do his schtick, I have trust in what he is saying because of his credentials, i know his previous work, i can read reports associated with his presentation, these reports contain the methodology by which he derived these results and the data are there to support his and his associates conclusions. Bref, I can go back and validate and verify or attest or contest the results and assumptions. I also have faith that he will uphold the reputation of the institution he is representing and he will back up his statements. At least I hope so! Therefore, it is more than the fancy show and narrative, there is a kind of implicit trust. Because of that trust i can just sit back watch the show, then go and really see what was being said in the associated documentation.
In the case of your show and tell, there were many charts being shown about a part of the US i know little or nothing about, and i could not easily pause to go back slowly to assess if your statements made sense with the data shown on the screen. It was also hard to follow along and i could not go back to review any texts associated with what you were saying and showing nor were there any metadata nor methodological statements i could read. You also went through tons of geography that i have no reference to in my mental map. You did tell a good story, and it was good to hear it and see it, an excellent experiment, but i think there is more to data representation than good narrative and there is no one magic bullet but a number of factors that make the data speak.
I look forward to more experiments.
See Tufte (http://www.edwardtufte.com/tufte/) – especially his space shuttle booklet. He does not deal with sound but does deal with data representation and visualization quite well and has some excellent examples.
hi john – i, myself, have been interested in crime stats as well ever since i moved to buffalo a month ago.
There’s this church downtown, near wear my wife works, with maybe 30 small signs that read “peace starts with ME” and behind those are this larger sign that shows the number of days since buffalo’s last homicide (7). To say the least, it was a bit unsettling that my wife had to work there (especially since she has to work late at night).
anyways, my point being, that buffalo even has a program called citistat, which began in 2005; however, they offer very little data. I think I’ll start bugging them as well.
I enjoyed the screencast, btw :)
Nice info on your blog keep going and provide more info like that and improve your website traffic more and more by the help of going good keywords and content
Loved the way you presented the analysis. Very classy.