Analytics is the rage right now. Not just because it’s hyped in the trade press, but because scores of business intelligence (BI) leaders are looking for ways to deliver more value from their data investments. These BI leaders have built data warehouses and delivered reports and dashboards, but now want to take their BI programs to the next level. The question is how.

Unfortunately, the pathway from reporting to analytics is not a straight line. It requires BI leaders to rethink their organizational, data, and analytics architectures. It may also require BI leaders to jettison their BI programs and evangelize new ways of capturing and using data for business gain. 

Step One: From Reactive to Proactive

The first challenge on the analytics pathway is to move from reactive to proactive views of the business. Reports and dashboards present historical views of business activity and are essential for running a business. But BI teams can also mine data to deliver proactive—or predictive—views of the business. These views help companies anticipate customer needs, optimize business processes, and reduce risk and costs by forecasting activities and events.

BI leaders who evangelize the need for analytics often face an uphill battle. Business executives who funded the data warehouse may question why they should spend additional money on data-related technologies that the data warehouse was purported to address. And since the benefits of analytics are not always obvious at the start, some executives may decide that the business case doesn’t justify the investment.

To convince skeptical executives about the value of analytics, new BI leaders cite success stories from other organizations, and in particular, direct competitors, if available. They also identify business executives with analytic vision who are willing to risk money and reputation to back an analytic initiative. Delivering a quick win is a surefire way to get the analytic flywheel moving.

Finally, BI leaders show business executives that analytics is the natural endgame of BI. They use the BI maturity model depicted in figure 1 below to show that additional investment enables BI programs to cross the proverbial Chasm from Teenager to Adult and migrate from reactive to proactive views of the business. (See figure 1.)  

Figure 1. Natural Evolution of BI

Step Two: From BI to Analytics

The second challenge in adopting analytics is organizational. Mining data and creating predictive models requires a skillset that doesn’t exist on most BI teams. To succeed, most organizations create a new analytics team that is unaffiliated with the BI group. The analytics team is comprised of statisticians and data scientists who have knowledge and skill to build predictive models.

If the BI leader plays his cards right, he can make a bid to run the new group, although in practice this is rarely done. Most companies hire a director of advanced analytics who has a background in statistics and machine learning, not BI. Ideally, the directors of BI and analytics work closely together: the BI team prepares data for the analytics team, which in turns, delivers predictive scores and values that the BI team incorporates into production reports.

Making the Leap. One BI director I know made the leap to director of analytics through a series of bold and innovative moves. He sold the CIO on the need for analytics, secured a small team and budget, and launched a grants program in which he invited business units to apply for analytic grants. These grants kickstarted analytic development in the business units, and demonstrated tangible benefits. Subsequently, business unit heads usually chipped in money to complete projects, reaping the full value of the analytic initiatives.

Step Three: From Data Warehousing to Hadoop

The final obstacle to migrating from reporting to analytics is expanding the BI team’s data architecture to support more open-ended data access and analysis. The BI architecture revolves around a data warehouse, which is a repository of clean, integrated data that supplies reports and dashboards with accurate, certified data for decision making. Since a data warehouse often consists of summary data conformed along predefined dimensions, it’s not the ideal data source for creating analytical models.

Statisticians and data scientists prefer to work with the raw data, combining it into wide flat tables to support various types of algorithms. Consequently, most advanced analytics teams create their own data repositories to supply their analytic models. Today, most use Hadoop since it’s free, easy to use, supports any type of data and doesn’t require upfront modeling to load.

Dueling Data Architectures. The result is dueling architectures: the BI team creates and manages a data warehouse while the advanced analytics team creates and manages a Hadoop repository. Both groups pull data from the same sources, and in so doing, trip over each other. The BI team accuses the advanced analytics team of duplicating their work unnecessarily, while the advanced analytics team says the data warehouse is full of errors and unusable.

If the teams sign a truce, they’ll recognize they are building different sides of the same analytical ecosystem. The analytics team creates the data hub that supports advanced analytics, while the BI team creates the data warehouse that supports reports and dashboards. The BI team lends the analytics team its data architects and ETL developers to source and structure appropriate data into Hadoop, freeing the analytics team free to focus on analytics not data management.

Summary

BI leaders face a choice: they can either keep on doing BI, refining the delivery of reports and dashboards to a business audience, or they can leap into the brave new world of analytics.

There is no bad choice here. There is certainly enough challenging BI work to keep an effective BI leader busy for years. Yet every company needs an analytic competency to benefit from proactive BI. And if business executives don’t take the initiative here, then BI leaders are well positioned to fill the gap.