Tom Fisburne – whom you really should read; in a marketing world full of crap he really cuts through it, much as Scott Adams once did for management – has a great analogy comparing Big Data to teenage sex. But for those people who actually doing Big Data, there’s a Twainian peril: data phrenology.
Phrenology is the study of bumps on the head, used to assess the character of the head’s owner. Unlike its sister study of chiromancy / palmistry, the -ology suffix makes phrenology sound like a science. It isn’t. It’s about coincidence and interpreting those coincidences so that they appear meaningful. See my point about Big Data yet?
Let me elucidate.
The more data you have, the more chance you’ll find coincidences. And the more you invest in Big Data, the greater the pressure for data insight. In other words, not only do you have a lot of patterns, you’re also under pressure to interpret them. That’s a massive potential trap for your business.
This is particularly true when analysing social media data. A couple of years ago a statistic went round that people who liked Burger King on Facebook would spend a few dollars more on each visit than people who hadn’t liked the Facebook page. The implication was that if you could only get people to engage with you on social media, they’d buy more of your product. But this was a syllogistic fallacy. The truth was that these social media types weren’t driving through Burger King and saying: “I’ve liked your Facebook page, so you’d better supersize me!” These were people who liked to eat burgers and wanted their friends to know about whole beef patties, but hold the gherkins. N.B. We neither endorse nor censure any food products on this site. There was also the possibility that they liked Burger King because it ran some campaigns to get people to promote its Facebook page…
Similarly, if you sample the psychometrics of people who follow you on Twitter and find they also discuss Breaking Bad, that doesn’t mean that you should necessarily go an buy advertising space on HBO. Lots of people tweeted about Breaking Bad, just as lots of people like to watch cat videos on YouTube.
Insight = meaning + hypothesis
Before you even start looking at data, think about about what you expect it to show. For example, is data about your product appearing in markets where it’s not sold, geographically or vertically. If so, there may be some data crosstalk. Michael Jackson didn’t just perform in Thriller, he wrote about beer and commanded the British armed forces.
Why have bothered to acquire the data in the first place? It’s not just going to turn up some results that no one ever realised. There’s nothing inherently mystical about it. It’s a test bed for your business assumptions. A way to test hypotheses.
If you’re seeing spikes and trends in data that match your hypotheses, they’re correct. If you’re seeing those hypotheses fail, they’re probably incorrect. And if you’re seeing something else you need to question that trend’s validity: what happened to create a spike? Is it significant or just coincidence?
If you’ve enough relevant data, it will almost certainly beat any gut feeling about business performance. But you can’t expect it to reveal some kind of hidden truth by itself. If you really want to If you want to get meaningful insight from your data, you need to feel your way past the bumps and recognise that your data is only as useful as the .questions you ask it.