Strategy, The Future and Data: Taking Exception with CIO
I have to take exception to the title of cover story of in the July 1, 2010, CIO Magazine. The cover story is “Analyzing the Future (how leading companies are using cutting-edge analytics to change business processes on the fly).” The sub-title is more meaningful than the headline. Journalists need to be careful. The headline is incorrect, and not even possible. One cannot analyze the future because the future has no data. We can only analyze the past, and in this case, what the story focuses on is not anticipation but adaptation. As companies analyze data they make process adjustments. The article talks about BI, but what it is really pointing to a new phenomenon, which uses business data as a sort of sensor (much as our nerves sense our environment, databases sense the environment of a business). I think that is an interesting story, and it sort of the story being told. What isn’t being told is a story of analyzing the future.
Again, the future has no data. This sensing works because there is sufficient data about what is going on to make an adjustment, but that adjustment cannot be permanent, because the environment continues to change. The system has to keep adjusting. It isn’t really anticipating something out more than minutes perhaps, days at the most, because as changes occur to the environment, responses need to be correlated to the stimulus input.
When we talk about the future, proponents of random matrix theory, as discussed recently in New Scientist (Enter the matrix: the deep law that shapes our reality) have demonstrated that data from the past is not a predictor of the future. Using economic data from 1983 to 2005 they explored correlations that looked predictive. They found that some correlations weren’t any better than chance. Their conclusion, all this great historical data only helps make interest rate predictions about a month out. After that the predictions fail. Jean-Philippe Bouchaud of the École Polytechnique in Paris, France is quoted as saying: “Adding more data just doesn’t lead to more predictability as some economists would hope.”
Adjusting to Reality
So first, modelers need to be sophisticated and understand the shortfalls of their models, and second, they need to recognize they can’t analyze the future, just the past, and the past only tells them so much about what the future may hold. And they also need to understand that a model is data itself, and it needs feedback in order to stay accurate. Thus, as I argue here regularly, organizations need a tool like scenario planning for exploring a range of possible futures and responses (and scenarios fall under this need for constant vigilance as well).
And for budget makers and sale forecasters out there, stop reporting on (and worrying about) your accuracy. Accuracy calculations aren’t very productive if the mathematics demonstrates you probably aren’t that accurate beyond a month anyway. Plan for a future, and then watch, and use the new data as feedback for adjustments. As the narrative of the CIO article suggests, it is probably more advantageous to adapt that to try to predict.