Data is not the same as information. In this big data era, marketers can collect data in a matter of seconds, but information is not that easy to obtain. We have to transform or combine data to get good, useful information. And information is the thing that we really need to predict the future and to make marketing decisions.
There are mainly two approaches for us to transform data into useful information: traditional experts and statistical machines. The former one tends to make decisions based on their experience and intuition, and the latter one can achieve these tasks really fast based on statistical prediction by using models, randomized testing and regression equation etc.
So who is more reliable and accurate? This is a question that has been asked and debated for many years. For example, John Landgraf, the president and general manager of FX Networks, thought that numbers-crunching would never do better job than a good programmer in predicting the future. However, some facts show that equations beat experts with more accurate predictions. I read an interesting story recently about it from Super Crunchers, wrote by Ian Ayres:
Ted Ruger, a law professor from the University of Pennsylvania, thought that it was impossible that statistical machines can win experts in predicting the outcome of Supreme Court cases, so he decided to run a race, to compare the accuracy of super crunching predictions and 83 legal experts. Both of these two teams were asked to predict in advance the votes of the individual justices for every case that was argued in the Supreme Court’s 2002 term. The experts team was consist of famous law professors, practitioners and pundits while super crunchers utilized six variables and 628 historical cases to accomplish the prediction. The result is the experts lost, with only 59.1 percent correctly, while the statistical model got 75 percent right.
We can find many stories like this in real life that equations get higher correct prediction percentage than experts. So why this happens?
Suffer from biases
Humans are likely to suffer from biases, which prevent our ability making accurate prediction. We tend to overestimate unusual events and underestimate common causes. Mostly this kind of bias is subjective and can lead to wrong judgment without realizing it.
Overconfidence
We are overconfident about our predictions while statistical predictions are not. Equations always come with their own confidence interval and the quality of the prediction, but we humans usually give our predictions with high confidence, especially those experts who have rich experience in their field.
Insist on mistaken beliefs
We tend to cling to our preexisting and mistaken beliefs and refuse to change them when facing new evidence. However, super crunching models have no emotional commitments to their previous prediction. They can change immediately with new data. I have ever had this kind experience before in the statistics class. Professors asked us to design surveys on an interesting issue and analyzed the result based on the data we got. Before analyzing the data we actually had some answers we preferred in our minds, based on past experience and hypothesis, so we tended to ignore the data that conflicted with the conclusion in our mind and tried to explain it in an exceptional way.
Only works on simple things
Humans are bad at predicting complicated things because we are not clear about how much weight should be placed on the individual factors. At the same time, regression equations are better than us at figuring out appropriate weights of variables. We can handle the prediction with maybe two or three variables, but how about 8 or 9 variables at the same time without using statistical models?
It seems that in the big data era, we should give more credibility on statistical machines than traditional experts in making accurate prediction about the future. It seems that experts are gradually becoming the servant of the statistical machine. But, we cannot underestimate the power of human, because it’s human who decide the variables and who design the equations. And because there is a growing trend that marketers should pay more attention on niches and long tail marketing than the mass market, the role of experts are important when dealing with particular issues.