The power (and perils) of data

This piece from Germany’s C3 references a couple of great examples of the latter, including dating site OK Cupid, which trawls through its data to produce interesting tidbits on the contemporary dating scene (shock finding: older men are more inclined to message younger women than vice versa) and Expedia’s crunching of data to generate sound travel advice for the jam-packed US Labor Day weekend.

We could add others with which we had the pleasure to be involved, including this groundbreaking report from Philips, which combined the results of an ambitious international survey with third-party data to develop a roadmap for the future of healthcare.

So far, so good. But as C3 rightly points out, whether you’re a journalist or marketer, in approaching and using data it’s important to be aware of its limitations. Data is no more inherently conclusive or free of bias as any other source of information, and should be subject to the same levels of scrutiny.

This isn’t a new story, of course: the phrase, “There are three kinds of lies: lies, damned lies and statistics” was popularised by Mark Twain more than a century ago. Which means that if you’re not questioning your own data, someone else very likely will; a recent survey by KPMG and Forrester Consulting found that most decision-makers don’t even trust the data insights their companies generate internally.

Beyond the issue of trust, there’s the question of whether data really connects on an emotional level. As one of the most powerful quotes in this excellent Vanity Fair piece on how data has transformed decision-making puts it:

“No one ever made a decision because of a number. They need a story.”

Having seen firsthand what data can (and can’t) do, we’re staunch advocates of putting it to good use. But as our recent reading has underlined, it’s important that data is used with principles in mind. Here are those that we see as the bedrock for any solid data-driven storytelling:

*Strive for transparency: Being as open and specific as possible about where the data comes (without sacrificing privacy standards) will add to its credibility; avoiding the matter will do the opposite. In publishing the results of a survey, this would include details such as the methods used and the number and composition of respondents.

*Practice acceptance: Maybe you’ve commissioned a poll and the data doesn’t quite tell the story or support the thesis you had envisioned. That’s okay, and no reason to discard the results — surely they contain other information worth sharing, and if they’ve confounded your expectations chances are other people would find them interesting as well. Also avoid cherry-picking findings to fit a pre-generated thesis, as it’s almost always obvious when this tactic has been adopted and it risks discrediting the whole exercise.

*Be selective: At the risk of appearing to contradict the above point it’s also important to be at least somewhat selective about the data you use and share. The ‘big data’ term exists for a reason; any data-gathering exercise inevitably produces a staggering amount of statistics. Rather than attempting to ‘go broad’, pick one theme or issue to target through research or a survey and ‘go deep’; the results will inevitably be more interesting. And when you do have findings, don’t plan to publish them all. Instead, look for consistent patterns or data points that seem to challenge conventional wisdom, and concentrate on examining and sharing those if they stand up.

*Remember data is a starting point: Regardless of the topic (yes, even the wild and wonderful world of online dating) audiences aren’t engaged by data alone, and a page chock-full of statistics or charts, no matter how tastefully designed, will cause a lot of eyes to glaze over. Proprietary data should be seen as a starting point for stories and campaigns that are fleshed out with anecdotes from internal and external experts, case studies and research from other sources, to build credibility and bring the numbers to life.