It’s the early days of a data revolution — but even so, not every decision should be data-driven.
“Big Data in Manufacturing” was the theme of a daylong conference held in Cambridge, Massachusetts, in November 2013 and sponsored by the MIT Forum for Supply Chain Innovation and the Accenture and MIT Alliance in Business Analytics. But the speakers’ insights weren’t restricted to manufacturing. Among them:
A Management Revolution
“We are in the early days of a management revolution. It’s not just the volume of data that’s enormous; it’s also the velocity and variety of data — the fact that you have this fine, very focused data, that it’s real-time. It’s as if the earth for the first time in history is having a skin connected to a nervous system that can detect what’s happening on the planet, anywhere on the planet.”
Four Questions to Answer
“When I think about big data analytics, I think about answering four different questions.
The first is descriptive: Tell me what happened. Sometimes that’s easy; other times it’s harder.
The second is diagnostic: Don’t just tell me what happened, tell me why it happened. That’s a bit more difficult.
If I can do that, I can move to the next level, which is predictive, and ask: What will happen?
And finally, can I use these types of concepts to be prescriptive: How can I make it happen?”
The Limitations of Data
“One risk in analyzing data is making false inferences. I gave my students a big dataset about elementary school kids; we were trying to understand reading scores. One pattern they found was that a child’s shoe size was a good predictor of their score. Why would that be? Well, it turned out that all the data — everything from the first graders to the sixth graders — was pooled together. The older kids could read better — and they also had bigger shoes! So you have to be careful about correlations. In the old days, you’d get excited when you found a pattern.