Once an enterprise gets beyond the hype and sets out to implement predictive analytics, success will depend on how they approach it. As with anything new, our approach is generally shaped by past experiences. For predictive analytics, our past experiences in BI and data warehousing shape our approach, or at least part of it.

It is critical to understand which previous BI and DW experiences apply and which do not. There is a fine line between leveraging your applicable past experiences and embracing new techniques. What approaches for predictive analytics are different than past BI projects and should be avoided? What ones apply and should be followed?

The business group’s role is one of the first things you need to assess.

In the typical BI project, the IT team develops dashboards and reports based on business requirements it gathers from discussions with business people. Although some of these business people may be data-savvy and technically adept (often called power users), BI projects rely on IT getting, integrating and presenting the data. The business people are involved in the beginning (requirements) and end of the project (user testing) while IT exclusively handles the work in the middle.

The typical predictive analytics project, however, is different. Although the same business people are involved in requirements and   user testing, a different set of highly skilled business people are involved in  the middle of these phases, in what would otherwise be the exclusive domain of the IT group.  These non-IT people have far more technical and data expertise than your average business person.

It can be an eye-opening experience for IT to realize that the people who build these models are more data-savvy and technically oriented than they are. In fact, predictive model builders often view the IT staff merely as data gatherers whose purpose is to feed their data-hungry models and then develop applications to present their models to other business users. In essence, IT’s role is reversed with predictive modeling — their involvement is at the beginning and end of the projects.

For successful predictive analytic projects, it is critical to understand the respective skills sets and responsibilities of IT, business end-users and predictive model builders. IT needs to understand and respect the roles, skills and responsibilities of predictive modelers. Some projects get derailed or productivity suffers because IT still thinks they are doing the middle part of the project and sometimes have a tough time giving up power to the predictive modelers. IT needs to understand that they are giving up nothing, but rather roles have shifted.