Awhile ago, in a marketing periodical, I recall seeing a recommendation that Big Data should report to the finance function. The rationale for this was that Big Data covers so much beyond marketing. Yet, as I see it, the same rationale can also be used to justify not housing Big Data in finance, since Big Data spans so many functional areas. But, perhaps the intent of that recommendation was to stress the benefits of centralizing Big Data.

Nonetheless, I think there is a far more important issue than whether Big Data belongs in finance, in IT, or somewhere else. I say this as someone who has researched various business success patterns for 25+ years and, before that, early in my career, did predictive modeling, which today is a key function of Big Data. As I see it, the really important issue for Big Data is doing what it takes to get meaningful insights, and knowing where and how to apply those insights. These insights can often require going beyond the numbers or the algorithms to understand what the data really means.

This is the case because Big Data is not a magic bullet. Big Data is just one more resource to add to the arsenal of knowledge generating tools. Yet, Big Data is extremely powerful. And, with today's technology, vast quantities of data can be economically processed and more types of data can be handled than in the past. Yet, Big Data is still not all knowing. Yes, in many situations Big Data can outperform less knowledgeable humans, and more of that is likely in the future. But, at this time, there can still be cases where someone with a strong understanding of what the data means can pick up on areas that Big Data alone might miss.

That's why it's important for people who understand the meaning of the data to be involved in determining what to do with the findings that emerge from Big Data. A data scientist whose expertise is limited to the technical side can discover correlations and develop predictive models. Yet, better insights generally emerge when someone who can go beyond the numbers or algorithms is involved with interpreting the data. This kind of broader understanding can be helpful for determining how to apply what emerges from Big Data. Yet, Big Data staff may have a background that is strictly in numbers and technical areas. If these staffers push for a course of action based upon taking the data literally, without a good grasp of what it really means, the action can easily be wrong, or worse yet, even disastrous.

Of course, the value of a broader understanding that goes beyond the numbers or algorithms can vary--sometimes, it helps a great deal; other times little, if at all. And, some numbers/technically oriented data scientists may eventually gain familiarity with certain topics and become able to go beyond the numbers or algorithms. Conversely, sometimes people with functional expertise thought to be helpful for going beyond the numbers or algorithms may be unable to provide any insight at all about the meaning of the data. Nonetheless, combining the tremendous power of Big Data with the insights of someone who has a good grasp of the situation can provide extremely valuable knowledge.

Human insights can be especially critical in the early stages of adopting Big Data, when there still may be many unknowns that are not yet in the massive accumulation of information that comprises Big Data. Especially in the early stages, when key variables may not yet be included, the findings from Big Data may be affected by the kinds of interpretation issues that can easily occur with data from surveys or smaller studies.

The potential for misinterpretation is illustrated by two recent studies of the relationship between tenure and performance. One study looked at CEO tenure; the other examined employee tenure more generally. Both studies found longer tenure associated with poorer performance. And, both studies could easily be misinterpreted if the data were taken literally. I commented on the interpretation challenges that arise from both studies. I did so by blogging about the CEO tenure study and by discussing the other study when I was quoted in a Human Resource Executive Online article, "The Trouble with Tenure" by Tom Starner.

My blog post pointed out that literal interpretation of the CEO study can be misleading because the study misses the great success of notable outliers like Steve Jobs. Jobs' spectacular results at Apple occurred when he had been at the helm far longer than what the CEO tenure study found associated with the best performance. Taking the CEO study literally would lead to the incorrect conclusion that, due to his long tenure, Jobs' spectacular performance as Apple's leader was weak, which was hardly the case.

In the Human Resource Executive Online article, where a group of us discussed problems with the tenure study, I describe how either-or thinking can easily foster misinterpretation if the study findings about tenure are taken too literally. I point out that longer tenure being associated with weaker performance does not mean companies should strive for staffs comprised heavily of less experienced people, since it can also be advantageous to have people with more experience. A team that blends junior staffers with long timers might be beneficial for various reasons, even if some of the more experienced people are a bit less enthusiastic about the job, and possibly not performing quite as efficiently as their less tenured counterparts.

Both tenure examples entail smaller studies, rather than the kinds of predictive models associated with Big Data. If these two tenure examples had entailed the kind of predictive analytics generally associated with Big Data, a question would emerge: would it be possible to build enough knowledge from the data to overcome the kinds of interpretation challenges that occur with smaller studies? Perhaps, it would. Yet, perhaps, it would not.

Perhaps, in a world where Big Data truly is all knowing, the computer could find enough patterns to sort out and ultimately produce meaningful direction on how to apply the relationship between tenure and performance. But, especially in the early stages of Big Data, knowledgeable people may need to assess the findings to help insure that recommendations reasonably reflect what actually happens in the real world. Some patterns in the data can be quite complex. And, it's not clear how Big Data will perform in handling the complexities arising from important outliers that are exceptions to the patterns, such as Steve Jobs' situation with CEO tenure.

Theories and paradigms can help in this regard. And, at least early on, input from those who understand what the data represents can be invaluable. The challenge remains, however, to tap the acumen of human expertise without letting that human element merely justify conventional wisdom where the data rightfully finds that conventional wisdom needs serious rethinking. But, while this kind of challenge is a signal for caution, there is still great value in combining human insight and understanding with the findings of Big Data--although some situations need human insight more than others do. And, it's important to recognize that centralized Big Data staff whose expertise is solely technical may not be able to add the valuable human insight. Yet, that added human insight might be just what's needed to get meaningful insights out of Big Data, so the findings can be successfully applied.