Here’s a way to understand the difference between a traditional analytics team and one that’s set up to exploit big data: compare bank tellers to orchestra musicians.
Traditional analytics teams, often centered in IT departments, are the bank tellers. Like bank patrons waiting in line to conduct standard transactions, business users line up to hand business intelligence and data warehouse experts their requests for reports on questions that have been well-defined in advance. Then they wait for their results.
Orchestras, on the other hand, employ a core group of musicians, but the conductor has to assemble different teams to play a variety of pieces. The resulting performance, and its value to the audience, depends on the skills, knowledge and experience that the musicians, working together, bring to the stage.
“Big data is pretty complex music,” says Bill Hostmann, vice president and distinguished analyst at Gartner. And the expertise companies need to use it effectively depends on the questions they are trying to answer. “A one-size-fits-all approach doesn’t work.”
Having a central analytics team may be a sign that your organization is mature in its analytical ways, says Jack Levis, a vice president with the Institute for Operations and Management Science (INFORMS), a professional society, and director of process management with UPS. But when it comes to embedding analytics into daily business decisions (the goal of many current analytics initiatives, even when they don’t involve massive or unstructured data sets) “you can’t separate the brains”—the statisticians and developers—”from the operations.”
In other words, hiring data scientists and training technologists on Hadoop isn’t the only step business leaders need to take in order to build their capacity to use big data. They also need a structure that makes it easy to coordinate expertise across the enterprise and facilitate collaboration.
The Organizational Focal Point
Many business and IT leaders are familiar with the concept of a center for excellence (CoE). Typically, such groups provide a focal point for a particular type of business expertise, allowing an organization to share skills, develop standards and disseminate best practices.
In IT, such groups can help to break down functional silos, says Michael Kopp, technology strategist with Compuware’s application performance management CoE. For example, application performance improves—and customer demands can be managed more easily—when developers, testers, quality assurance and production experts share knowledge and cooperate, as well as communicate with business leaders who know how the application will be used.
Similarly, having an interdisciplinary team focused on analytics—whether you call it a CoE or something else—provides a way to organize people so they can collaborate on identifying relevant questions, collecting and preparing data, building statistical models and delivering results in a way business leaders can use them. IBM, which last month announced a service to help customers build analytics centers of excellence, says such groups often manage analytics strategy, “including centralization of the data and implementation of technology.”
When Sandra Woodley joined the University of Texas System two years ago as vice chancellor for strategic initiatives, the analytics group that reported to her focused mainly on producing descriptive data and statistics. “Your traditional fact book,” Woodley says. Part of her charge: to develop the capacity for predictive analytics, in order to help the 15 institutions within the UT system use data become more cost effective, improve faculty productivity and serve its 200,000 students better.
Woodley invested in a new business intelligence and data warehouse system from SAS. She hired additional people to organize and provision the data—which they aggregate from many terabytes across the UT system and other sources—as well as manage data quality. And she hired researchers “to do deep dive analytical studies. We’ve set up a research agenda a year out,” Woodley says, to support the system-wide strategic plan. Among their projects, the researchers are studying what makes students successful, including how factors influencing graduation rates differ from school to school within the Texas system.
One tenet of the team is to involve experts across the UT system in developing new research. Dashboards allow users to access data for their own analyses. Meanwhile, the team partners with key constituents to craft its studies. For example, for its research on student success, the team consulted with subject-matter experts throughout the system, including administrators and institution-based research directors. “We rely heavily on existing relationships,” Woodley says. Who gets involved in a particular project happens organically, depending on the topic. “We know who the experts are. We do the heavy lifting, they provide ideas and they can react to the drafts.”
No Analyst Is an Island
The ability to involve experts in business areas, is critical, says Hostmann. Whoever leads big data initiatives needs to be able to “reach across the white space in the org charts” to tap employees with relevant skills and get “the commitment and the buy in” from managers.
Levis observes that mathematicians can develop great algorithms to optimize business processes, but the results might be impossible to implement. If you want analytics to be used for front-line decision-making, “you need data—often the analysts don’t think about where they’re going to get the data—you need the mathematics, you need the interface. And when you get a bad answer, you have no idea which one of the three caused the problem, so you need people who can research it.” His analytics teams at UPS include “PhDs who are mathematicians, software engineers who can do data cleanup and displays and can iterate with IT systems [and] business people who understand enough of the math” to know how to make a new data-driven process work.”
Using analytics to improve operations has become critical in healthcare, says Randall Gaboriault, CIO at Christiana Care Health System, which is among the largest healthcare providers in the United States (it ranks 17th in hospital admissions). Regulations issued as a result of the Affordable Care Act penalize hospitals with high readmission rates and reward providers that deliver quality care. The rules aim to cut costs by changing how healthcare providers get paid; instead of revenue from each procedure, they’ll be held accountable for managing each patient’s health. Predictive analytics can help providers uncover factors that put patients at risk and take preventive measures.
So Gaboriault is revamping Christiana Care’s big data analytics team. Now, within the organization’s business intelligence/data warehouse group, one team is responsible for technology, training and delivering projects. But another team, enterprise analytics, is charged with mining data across functions and business processes. Domain experts on this team, who come from the business, collaborate with clinicians and administrators to ask “radical” questions, such as what can be learned about treating cardiac patients by analyzing all the data about them, including information from their primary care doctors.
All the technology and data expertise is a shared resource. Although the big data analytics capabilities live within IT, the point of it is to empower business users with access to data and support. Says Gaboriault: “We’re taking the IT organization out of the middle.”
Elana Varon is contributing editor with Data Informed. Tell her your stories about leading and managing data driven organizations at firstname.lastname@example.org. Follow her on Twitter @elanavaron.