|Home About Kurtosis Courses Course Calendar Booking Information Ideas Technique Clients Contact|
Telling tales with data
Practical tips for embedding your numbers into a story
Let's begin with an article that appeared in the Journal of the American Medical Association (JAMA) at the end of last year. Narrative vs Evidence-Based Medicine—And, Not Or was written by Zachary Meisel and in it he said: "Scientific reports are genuinely dispassionate, characterless, and ahistorical. But their translation and dissemination should not be. Stories are an essential part of how individuals understand and use evidence."
Data is supposed to be cold and objective; but the dissemination of your data can be warm and subjective. So go ahead, tell a story with your data. Because if you don't, you run the risk of falling behind. As Meisel continued: "Those who espouse only evidence—without narratives about real people—struggle to control the debate. Typically, they lose."
It’s become pretty much axiomatic these days that if you're really serious about getting your data across to your audience, you need to tell a story with it. Stories are more engaging and convincing than mere data. If you want to influence someone’s behaviour you need to touch their heartstrings and move them to tears. And you won't do that if you only engage their logical left brains. No, you also need to impose yourself on their creative and emotional right brains.
Which all sounds promising and exciting, but we need to remember that it's data we’re talking about here. Data is logical and soul-less and is usually a collection of seemingly disconnected facts. How are we going to fit that into a story?
I decided to do a bit of research on story-telling techniques to see if we could work out ways of telling stories with data
I started with Kurt Vonnegut, if only because his Graph A Story video talk from 2005 is a ridiculously entertaining one. But he’s mainly talking about classic story shapes like Cinderella, which left me wondering what this has got to do with data.
So then I moved on to Nancy Duarte’s book Resonate, where she tackles head-on the idea that there might be a formula for persuading and influencing audiences using presentations. And yes, there is a formula (based on Freytag's pyramid) and Nancy Duarte explains it very well and very engagingly. But because she concentrates on how this dramatic structure works for the likes of Martin Luther King, Abraham Lincoln and Steve Jobs, I'm still left wondering how we can get this to help us with a presentation that's asking whether we can re-classify a six-bedded area in Ward 6 as Breast Surgery instead of Gynaecology.
Vonnegut and Duarte were all very well, but I felt like I was reading War and Peace to get tips on how to write a haiku. There were some great ideas in there; they were just too big to apply to a short form. So I abandoned the high ground and instead went to YouTube and reminded myself of this Honda advert from a few years ago. And then things started to get more fruitful.
Here is practical idea number one. A deductive argument (the Honda ad is a deductive argument, winning a game of chess is a deductive argument, the proof of Fermat's Last Theorem is a deductive argument) can make for a good story. And as data analysts we often find ourselves trying to set out a deductive argument. Here’s an example. If you want to get a clinic's waiting times down from 12 weeks to nine weeks, you’ll need to (a) check if demand and capacity are in balance (they are: approx 24 referrals per week; 24 new patient slots per week); then (b) check that capacity is being used optimally (it is: no new patient slots are being wasted); then (c) check that routine patients are being brought in in turn (they are); then (d) calculate the number of patients you'll need to reduce the backlog by in order to reduce the wait from 12 weeks to nine weeks (answer: 120); then (e) work out how many weeks you've got to bring about this decrease (answer: 12 weeks); then (f) you'll need to increase capacity by 10 new patients per week for the next twelve weeks in order to meet the target.
As long as your reasoning is clear, as long as you don’t miss a step (that would be fatal), and as long as you arrive at the right answer, a deductive argument like this can be compelling for your audience. They're being told the answer to a problem that they previously couldn’t solve. And as long as they understand the answer, they'll like to hear the deductive story of how to solve it. Deductive is seductive.
Here is practical idea number two. It's the whodunit approach. As an example of this, one book that’s always fascinated me when it comes to pondering the whole data-as-narrative thing is Freakonomics by Steven Levitt and Stephen Dubner. Freakonomics is all about numbers but there's hardly a table or chart to be found in any of the book's 242 pages. The most famous chapter in Freakonomics is the one called Where Have All the Criminals Gone? in which they explain why the drop in serious crime in the USA in the 1990s was mainly due to changes in abortion legislation in the 1970s. But the way they piece their argument together is interesting because they actually examine (and reject) seven other explanations for the falling crime rate before they get to the abortion explanation. It’s like they’re teasing us with the other, shakier, explanations. Was it due to innovative policing strategies? Nope. Was it due to the increased use of capital punishment? Nope. Was it due to the booming economy? Nope. And so on, until we get to the right explanation.
I’m reminded of a book that’ll be familiar to those of you (parents or not) who’ve read books to small children at some point in the last couple of decades. Dear Zoo by Rod Campbell. "I wrote to the zoo to send me a pet. They sent me an... elephant. He was too big! I sent him back." And on the book goes. Lion (too fierce), camel (too grumpy), monkey (too jumpy) until we get to the answer (spoiler alert: it’s a puppy).
Here is practical idea number three. Which I stole from a great blog published yesterday by Robert Kosara. It's to do with maps. There’s something about a map that makes for a good story. As he says: "We are so hard-wired for causality and stories that it’s difficult to look at something like a map and not tell a story. We are incredibly creative at filling in the blanks."
And of course a lot of our NHS data analysis stuff is map-based. When a patient moves from A&E to an Assessment Unit they are moving from A to B on a map. If they then get transferred to another ward, that's Point C. And so on. Heck, we even use "the patient journey" as a commonplace expression (and Kosara points out that the word “journey" is itself of course a nod to one of the oldest stories in existence: Homer's Odyssey.
When the poet Philip Larkin was judging the Booker Prize back in the 1970s he complained that too many of the books he'd read had relied on "the classic formula of a beginning, a muddle and an end." Well, try these ideas out with your data stories, and see if it helps the muddle look less muddled.
[15 August 2012]
|© Kurtosis 2012. All Rights Reserved.|