In the last couple of years, the popularity of predictive analytics offerings has increased significantly, as many companies looking to improve their existing business intelligence (BI) and analytics faculties are looking to the field of predictive analysis to enhance these capabilities, and in some cases even replace them. But, is predictive analysis really the new BI? Is it a replacement for the good old business intelligence offerings? Or is it really just the evolution of BI?

What Is Predictive Analytics?

Predictive analysis is a branch of a broader discipline called data mining, which comprises important data analysis activities such as exploratory data analysis, descriptive modeling, pattern and rule discovery, data retrieval, and predictive modeling.

For our purposes, we will focus strictly on the predictive modeling scenario that the discipline of predictive analysis is based on. Predictive analytics consists of those tools that make it possible to perform the tasks of a predictive analysis process, which involves the use of various analytical and statistical techniques to build a mathematical model in an attempt to predict the future outcome of a certain scenario of study. From statistics to game theory, predictive analysis techniques make use of historical data to create predictions, usually by capturing relationships between explanatory (independent) variables and “predictors” (predicted variables) from past events. Two major modeling types can be applied: classification (for predicting categorical/discrete variables such as yes/no, likelihood of an event to occur or not occur, risk levels, etc.) and regression (for continuous variables, such as blood pressure, liquid levels, etc.).

To perform its task, the predictive model uses historical data to generate its outcome. So, for example, based on the analysis of historical data it is possible to address specific business cases and attempt the prediction of business scenarios such as customer behavior (shopping and consumption trends), product supply and demand prediction, fraud detection, risk assessment, customer churn rates, and sales and margin predictions.

Predictive analytics applications are those software utilities that enable users to perform such tasks, and because of this nature, predictive analysis and overall data mining techniques are often the basis for many more specific types of business analytics tools, such as specialized industry and line-of-business analysis tools.

Predictive Analysis Is Not BI

Business intelligence applications comprise another set of tools for the analysis of historical information as well, and nowadays increasingly use real-time data to perform a different type of data analysis, such as enabling reporting and dashboarding operations or enabling data analysis by the use of an online analytical processing (OLAP) engine to undergo “slice and dice” processing of data.

So what’s the difference between BI and predictive analytics if both attempt to answer the “What happened?” question and analyze that knowledge to gain insight? In short, both BI and predictive analytics applications address historical data, but whereas predictive analysis uses historical data to gain insights into probable future events, BI uses it to explore the present.

If we consider that both BI and predictive analytics applications are different in terms of the types of problems they address, it makes sense to think that it is not a matter of replacing one with the other, but of starting to think in terms of how can we put both to work in a way that is beneficial for an organization.

Closing the Gap

Because of the different approaches and origins of the fields of BI and predictive analytics, it is common to see a disconnection or gap between these two types of applications; in fact, many organizations seem to view them as two isolated offerings acting on their own. Fortunately though, some vendors are taking action to reshape this trend.
On one hand, many software vendors have come to realize the importance of having BI and predictive analytics acting jointly, and have begun providing users with accessibility to both capabilities through the same product offering. For example, products in the space by MicroStrategy, IBM Cognos, and SAP bring the possibility to have both capabilities with a single software offering.

On the other hand, the presence of alternatives with strong integration capabilities and an important presence in the market, such as FICO, KXEN, and Revolution Analytics, are helping organizations to realize how important the predictive analysis discipline can be for an organization.

Still, one of the challenges that users face when considering the incorporation of a predictive analytics application within their data analysis portfolio is its complexity and the requirement for specialized personnel to perform the task of developing a set of reliable predictive models. Software vendors then face the challenge of evolving their predictive analytics offerings as their BI counterparts provide users with predictive analysis applications that are easier to use and more suitable for business-oriented users who are not necessarily well-versed in the management of these types of applications.

Predictive Analytics, Beyond BI

So, ideally, we should think of BI and predictive analytics applications as partners, where predictive analytics can:

• expand the reach of a BI application, enabling organizations to have reliable tools for both the extraction of data on their current situation and the analysis of potential opportunities and risks;
• enable decision support in all strategic and operational levels of an organizations—predictive analysis has a strong potential use at an operational level, where it can be deployed to detect risk levels;
• reinforce the productivity of BI applications by enabling side by side comparisons between predicted scenarios and real conditions, which can also serve as feedback for improving predictive models; and
• trigger other exploratory analysis, such as the data discovery process, to try to find new and useful questions to ask based on discovered trends or predicted values.

When growing an organization’s BI/analytics set of tools with predictive analysis capabilities, it may be helpful to consider these tips for improved integration:

1. Embed predictive functionality within pre-existing BI capabilities, such as rich data visualization, or integrate it along with BI applications to promote faster adoption.
2. Take advantage of the existing data management infrastructure by incorporating the predictive analytics application into the organization’s data infrastructure, from the data integration process to the data preparation process.
3. As with BI, make predictive analysis capabilities a true integrated part of the complete enterprise software platform, so as to be easily accessible for those users requiring it at all levels of the organization.

Having a predictive analytics solution can definitely provide enormous benefits for an organization, but in order to start exploiting the potential of this type of solution it needs to be applied in a clearly identified business case where it is possible for all stakeholders to appreciate the potential of predictive analytics tools. It is also important to make predictive analytics an integral part of the data management platform and promote its pervasive use where needed, and thus try to overcome some of the difficulties around creating awareness, training, and encouraging adoption of predictive analytics solutions.