Big Data is a bit of a problem for businesses. The fact is that data is growing enormously, both in its volume and importance. Also, we’ll soon see a big push on usable open data and its value. So, many organizations must move on Big Data.
Yet, I haven’t found a use case that will deliver for any and every company. McKinsey recently asked eight executives from companies with leading data analytics programs about their experiences. According to the McKinsey report, “[t]he reality of where and how data analytics can improve performance varies dramatically by company and industry.”
One problem may be that Big Data requires a paradigm shift in how businesses approach data. Typically, business is goal-oriented with data: You run a report because you need a specific set of data on a specific topic.
In fact, with Big Data, what you find might not even qualify as an answer or a fact. But if that’s the case, then what good is it?
The answer lies in determining how to best understand Big Data, and so far, that’s still a point of uncertainty, even among the experts.
The leading theory is that Big Data requires an explorer mentality, but two recent pieces by data experts argue for a more scientific, or at least methodical, approach.
The first is a Forbes review of the book Big Data at Work: Dispelling the Myths, Uncovering the Opportunities, written by industry observer and Babson College President’s Distinguished Professor in Management and Information Technology, Tom Davenport.
The review highlights Davenport’s brief discussion about the ideology of Big Data, “especially the misguided belief that the collection of data is a goal in itself and that the data can speak to us and answer questions we never knew we should have asked.” Davenport warns against sifting through the data without purpose, which can cost you too much in time and money. Instead, he recommends approaching the data with a hypothesis formulated even before you’ve analyzed the data.
Apparently, he didn’t explain how you do such a thing, because the reviewer calls that out as a topic that needs more discussion.
So, for Davenport, you need a hypothesis to learn from data. Meanwhile, Jim Harris suggests yet another approach. I’ll share more about that in my next blog post.