Thursday, May 8, 2014

Data's Dirty Little Secret

If you ride New York City subways, you have probably come across Poetry in Motion, a noble attempt by the MTA to deliver us from Dr. Z.  One of these missives really stuck with me:

“Out of the crooked timber of humanity, nothing entirely straight can be built.

This quote came from Immanuel Kant.  I had no idea what  Kant meant when I read "The Critique of Pure Reason" in college, and I understand why I didn't know why then.  Herr Kant was warning me about marketing data.
I know, again with the data.  Stick with me.  You might learn something.  If not, you can at least tell people you read an impassioned exegesis of Kant today without really lying.

I've talked before about how purchase data can mislead marketers, even though marketers generally prize purchase data above all other types of data.  When I visited the topic two years ago, I made the argument that purchase data only record what people did, not necessarily what they wanted to do.  For instance, a buyer may want to buy khakis, but if all she only sees corduroys at the store, she may buy them instead.  In turn, the retailer doesn't learn what the buyer really wants, which may mean mis-targeted communications down the road.

The Kant quote reminded me of another scary thing about data: people are f***ing weird.

And this is a guy who started a rock band without a bass player

OK, more accurately to Kant's point, people often behave inconsistently.  As the old jingle went, "sometimes you feel like a nut, sometimes you don't."  People buy the same soda every day until they don't.  Or they never like grapefruit until one day, they do.  Happens every day.

Well, so what?  

"So what?" is that marketing data are data.  We expect data to have precision.  Data points are things like "water covers 71% of the earth's surface" or "New York City had 2,245 murders in 1990."  Exact.  Unwavering.  True.

Marketing data don't enjoy the same claim to truth.  Sure, Ford sold 763,402 F0-150 trucks in 2013, but those are over three-quarters of a million people who might never buy another Ford product again.  Or they may never buy anything else but Fords.  Who knows?

Someone biased against data might take this exercise as a reason to write off marketing data entirely.  After all, if you can't trust marketing data the way you can trust, say, sports data, then why bother?

On the contrary, I think this ambiguity should encourage marketers to embrace data more strongly.  Murky data give marketers a chance to experiment.  On the chance that the data don't show what you think they show, you have the opportunity to test out new ideas and new approaches knowing that, what the hell, it just might work.

This "damn the torpedoes" approach has limits, of course.  If your audience buys electric guitars, don't use that as an indication to test offers for accordions.  (Seriously, don't.)  However, if data on electric guitar buyers suggest that they won't buy acoustic guitars, it might make sense to test that inference.

See what you can make of those crooked timbers.

No comments:

Post a Comment