Jean-Francois Puget from IBM (ILOG) published an interesting article that in particular says:

“The purpose of an analytics solution is to enable business decisions based on data.  It does not necessarily mean that such solutions must compute decisions.  It means that whatever their output, they must support decision making.   In short, analytics is a journey from data to decisions and actions, via the use of many different techniques and tools:“

“We can distinguish several analytics levels.  In some cases we just provide views on the data as is. This is called descriptive analytics.  In some other cases we provide insights (eg patterns, or relationships) found in the data.  This is called diagnostic analytics.  In yet other cases we can uses these insights to make some predictions, for instance forecasting future sales levels, or identifying which employee is most likely to leave.  This is called predictive analytics.  Last, in some cases we can compute decisions, eg schedule production to meet forecast sales.  This is called prescriptive analytics.  These different levels depend mostly on what is automated and what is left to humans before decisions can be made.“

Read the entire article “How Does Cognitive Computing Relate To Analytics”

See also:
Predictive Analytics is Becoming Mainstream
The Future of Analytics Is Prescriptive, Not Predictive