Turning Predictive Analytics Into Tangible Action

Predictive analytics is a series of “statistical techniques from modeling, machine learning, and data mining that analyze current and historical facts to make predictions about future, or otherwise unknown, events.” Predictive analytics are used for a number of different purposes but most often to guide decision-making within an organization whether it be for marketing, sales, or other entities. In a digital world increasingly dependent on more and more data sets it’s becoming more important for organizations to use this data to their advantage. It’s also becoming more difficult to turn predictive analysis into tangible or concrete actions.

In a recent study by Hewlett Packard it was found that many organizations do use predictive analytics, but they fail to reap the benefits due to a disconnect between different departments such as business (Executive/sales/customer service), IT, and data sciences. They noted three points that hinder the advancement of predictive analytics within an organization:

• Business leaders don’t understand the potential of data science and request projects that result in traditional reporting rather than business transformation.
• IT leaders don’t understand the iterative and explorative nature of data science and enforce design and control policies that inhibit analytic discovery.
• Data scientists don’t understand the rigorous IT data warehousing and application development processes that are in place to ensure quality, security, performance, and supportability.

In lay terms, this means there is a disconnect between different areas in the organization. Business leaders stick with traditional reporting, which only shows historical trends but does not help predict future customer behavior. IT does not realize the needs of data scientists and business intelligence professionals. BI does not understand the computing power that it takes to gather, process, analyze, and output the data – as well as other measures such as security that are involved in the process.

Here we’ll turn to a sports analogy to help us understand why this disconnect occurs. Imagine a team of all-stars. The players are all individually great, but may not work well within a team environment. This may occur because they have honed their skill individually for a long time, and have not had to work with others in a collective manner. When you bring them all together, they may have a hard time understanding how to work together. So what happens? They revert back to their individual ways and the team is no better than it was in the beginning. How do we solve this problem? We bring in a savvy coach, or in other words, we teach everyone how to work together to produce the results we’re looking for. The same is true within a business organization.

It’s imperative that business leaders, IT and business intelligence collaborate so each can understand the needs and processes of the other. This allows for each department to recognize the needs of the other and to adapt their approach accordingly. HP reported that this is often achieved “through cross-discipline education, centers of excellence, communities of interest, and multi-functional teams.”

Once the members of the organization understand the method to the madness, it’s time to implement our analytics into every day processes. However, the learning doesn’t necessarily stop. For instance, an organization could use its customer analytics to predict certain customer behavior, such as how long a person will stay a customer (in a recurring fee environment) or how often a customer will return. This data can be implemented into business applications used by sales or customer service, however, if the salesperson or customer service rep does not understand where this data is coming from or how to use it in their position this data goes to waste and the organization does not achieve the growth or understanding it seeks.

Of course, we must not forget the technology side of everything. A proper technological infrastructure and architecture is needed to help achieve the desired results. This architecture needs to support the full analytics lifecycle, HP described this cycle as the following:

• A discovery environment to support flexible, fast, and short-term analytic investigations
• An open analytic workbench that enables data scientists to experiment with the best possible tools for the business problem
• Analytics model management that treats algorithms and business rules as company assets and moves them from laptops to secure central repositories and administration
• Decision and rules engines that enable the business to automate deploying predictive models within the appropriate business context
• A service-oriented architecture that enables the integration of predictive models with real-time applications

When we combine business, technology, and business intelligence and understand what it takes from each we can take full advantage of predictive analytics and collected data. The market moves fast, and our decision-making needs to be fast, and more importantly precise and decisive. When all the cogs are working towards the same goal, and understand how this goal is achieved throughout the organization everything is more efficient.