Caveat Emptor: Data Oriented Marketing Has Limits

The first note here is that I'm a huge fan of analytics platforms and what they enable businesses to do. It's awesome to know that you can break down the behavior of all the folks visiting your site, signing up for an email list, etc and tailor your attempts to engage each group.

I've had a lot of experience riding the wave of a data engine's ability to predict things. numberFire is generally judged, and maybe misjudged, by the algorithm's ability to make a call on a future outcome. That call would help someone win or lose a fantasy football game or a bet. We would always get questions like "Did your system see Colin Kaepernick coming?" or "When Linsanity happened, did you guys know it was going to happen before it did?" Nope, and nope. Sure, after 5 games or so, we'd be able to get a pretty good beat on a guy. But, even then, we would have been totally wrong on J-Lin again - he flamed out when Mello came back and hasn't done a ton since.

Greg Satell contributed a great piece to Forbes that sums up what's going on here. People expect that because you have so much data, you're going to be able to predict everything, perfectly. The problem with that view is that it overlooks creative happenings or a certain moment of genius. Data tries to capture the happenings of the past as precisely as it can in order to give you a probability of certain events occurring in the future. Two things to keep in mind there:

1) At least at numberFire, we were dependent on imprecise mechanisms of capturing past behavior - box scores don't tell you everything you might want to know, but increased camera views coupled with unstructured data mining technologies are certainly helping with that. Google Analytics, Omniture, or whatever other technology you might use will have some bugs in it. Granted, the amount of information those technologies are able to capture about behavior on a website puts box scores to shame, but it's still a point to keep in mind.

2) If you're batting .600 (or for those of you less acquainted with sports, if you're right 60% of the time), you're making money, perhaps even a lot of money depending on how much you threw down and how directly correlated that is to ROI. Real talk, if you're winning 53% of the time in Vegas, you're making money. 53% leaves out 47% of circumstances in which you're totally wrong, in which something crazy happens that nobody, given current technologies, can call.

The point is this: data is great in most circumstances. Most circumstances, however, are not all circumstances. As a marketer, it's incumbent upon you/me to keep that in mind and not stop trying to think about what's missing, keeping an ear to the ground, keeping on thinking and not just computing.