Today I did a short presentation at the University of Queensland where I briefly discussed the differences between exploratory and narrative models of data storytelling.
For a long time I've thought that a blend of the two represents the holy grail of data storytelling. Stories that provide a strong narrative thread to guide a reader through the data, but also provide jumping off points to explore the data further in a less directed way.
Creating any kind of hybrid like this is challenging from both technical and storytelling points of view, but getting the mix between the two right for any given story can be really difficult. There's no silver bullet here, each story will be best served by a different mix depending on a variety of factors.
I love exploratory interfaces to data and they can, unquestionably, tell powerful stories.
I'm a total data nerd and there's not much I like more than exploring a big dataset with some kind of interactive interface to find a story that interests me. But as much as I love that—and I do—when I'm just a regular consumer reading a story, I don't have much time for it. The data has to be super interesting to me personally to hold my attention for long.
An exploratory interface which reveals an interesting dataset—all the genres of music ever, for instance—will fail to hold my attention for long. This is because, even though I'm into music and I like reading about it and exploring it, with this kind of exploration I'll fail to find an interesting story to follow without a lot of effort.
And music is a relatively popular topic compared to, say, government revenue and expenditure.
On the other hand, an equally exploratory interface on something like where to rent or buy will hold my interest without any trouble, should I be looking to move house.
At the other end of the spectrum in data storytelling are stories with a set narrative. They might involve elements which expose data (charts, graphs, visualisations), but they offer the reader no chance to explore beyond the author's chosen views on the dataset.
An example of this style is my own reporting on forecast error in Australian federal budgets. As a reader you're only seeing the view of the data I want you to see. But the great advantage here is that a writer can use narrative elements to keep a reader interested in what is otherwise a pretty boring dataset.
Getting the balance right
Purely exploratory interfaces have their place in storytelling, but for the most part, readers need a narrative to make sense of the data.
So where should we aim for on the spectrum? Well, I think it depends on a number of factors. Here are three that I think are important:
1. The story itself
Sometimes the story already has a strong narrative, so you don't need to put much of one around it. An example which comes to mind here are the various 'all of Trump's tweets' datasets. Examples:
Most readers will bring their own strong narrative to any story about Trump, so there's less need to construct one in order to hold the audience's attention.
2. The audience
How well your audience knows the data has an impact on how much narrative they need. There's a really interesting presentation by Chad Skelton where he talks about exactly this.
A key point he makes is that you should lean toward a narrative driven approach whenever there is a gap between your understanding of the data and that of your audience (which is pretty much always in journalism). The bigger the gap, the more narrative driven your approach should be.
3. The platform
More and more of the news audience (and I think this probably holds true for other audiences too) are moving to mobile.
This has forced us to think carefully about how to use visulaisation and I think it's forced us to add more narrative (often resulting in better stories) where once we would have left it to the audience to explore and draw their own conclusions.