Article Preview
TopBeing Precise About Analytics
The analytics term is used imprecisely. Sometimes it is employed interchangeably with business intelligence. Another interpretation is that if you view BI as “getting data in” (to the warehouse) and “getting data out” (data access and analysis), analytics is the analysis part of BI. Or finally, analytics is the “rocket science” algorithms (e.g., neural networks) and methods used to find patterns in data (e.g., customer segmentation analysis) or to optimize performance (e.g., revenue management).
It is useful to think of descriptive, predictive, and prescriptive analytics. With descriptive analytics, the objective is to describe what has occurred. With this view, reporting, OLAP, dashboards/scorecards, and data visualization are all examples of descriptive analytics. These are the core and most common BI applications.
Predictive analytics focuses on what will occur in the future. The algorithms and methods for prescriptive analytics include regression analysis, machine learning, and neural networks. These techniques have been around for some time and have traditionally been called data mining. While these methods continue to evolve, the most significant development is their inclusion in analytical workbenches and applications that make them much easier to use.
Prescriptive analytics is intended to show what should occur. It is used to optimize system performance. Revenue management, which strives to optimize the revenue from perishable goods, such as hotel rooms and airline seats, is a good example. Through a combination of forecasting and mathematical programming, prices are dynamically set for the good over time to optimize revenues.
Another perspective is that the progression from descriptive to predictive to prescriptive analytics is a movement from hindsight to insight to foresight (Barnes et al., 2012). First companies want to understand the past, then they want to predict the future, and then they want to optimize what they do.
In most cases, imprecise use of the analytics term does not cause difficulties. There is a problem, however, when discussing the requirements for success with analytics. The requirements for descriptive analytics are different in important ways to predictive and prescriptive analytics. We will refer to predictive and prescriptive analytics as advanced analytics.
Returning to the issue of the different interpretations of the analytics term, this article uses analytics to describe the analysis of data and advanced analytics as the “rocket science” algorithms and methods of predictive and prescriptive analytics. With this interpretation, analytics is a subset of BI rather than an alternative term.
TopDifferent Targets For Analytics
Companies can have different “targets” or approaches to analytics. No one target is better for all firms, and each target can be best for a particular company depending on its situation. All of the targets can potentially deliver significant business value. These are the same targets as for BI (Wixom & Watson, 2010).