Prediction

Lessons for Marketers: Why Big Data got the US Election Wrong

How could so many very, very smart people armed with the best technology in the world have been so wrong? The answer is rooted in a simple, basic fact. People are not good at predicting their future behavior. We can tell you about our past actions and why we took them. We can tell you how we feel about things that are happening now. Somehow, however, when it comes to telling you what we are going to do even if the action is just a few hours away, we are utterly unreliable.

The Clinton team (and the rest of the world) just re-learned this old lesson. Implications for marketers are clear:

1) Past data is a gold mine, but it is not magical. We should amass as much information about past behaviors and attitudes as possible. Cohort data (how people like us behave and feel) enriches this backwards-looking data significantly. Past data can also be predictive, but the reliability of these predictions depends up consistency of ALL conditions. As soon as something changes in the environment of the subjects, all bets are off...