http://www.bigdatarepublic.com/author.asp?section_id=3250&doc_id=264987&

Analyzing data “in the stream” is crucial for making effective real-time decisions, which, in turn, is crucial for all manner of business functions, particularly decisions related to fraud that must be made in a matter of seconds.

However, these data streams are useful only when we can systematically extract the most valuable analytic insights from big data. Companies can’t allow themselves to get buried by meaningless noise. The meaningful insights enable an understanding of individual customer behavior and sensitivities to help anticipate needs, predict likely responses to offers or events, and understand how individuals will react to specific treatments. In many situations, this insight must occur in real-time to impact the customer at the right time.

For example, an Ohio-based customer attempts to make a purchase at a department store in Manhattan. Is this a suspicious transaction? A bank has only seconds to decide. A false positive could mean an annoyed and embarrassed customer who may eventually take her business elsewhere. At the very least she’ll use another card for the purchase causing the banks to miss out on fees. At first glance, the transaction may seem suspicious based on geography alone.

Another major impact of big data is that analytics must reduce reliance on persistent data, and allow analytic models to adjust on the fly in the stream. To meet the need of an increasingly dynamic stream, there is an intense research focus on self-learning techniques such as adaptive analytics.

It will be interesting to see how streaming analytics evolve in the next couple years. While the technology is extremely valuable and paves the way for great innovation, it can also be overwhelming or unnecessary in many situations.