Organizational design is like building design: You don’t get a high-quality structure without paying for it.
In recent years, many companies have felt the need for greater coordination of their analytical activities. Some (42 percent of companies in the 2013 Deloitte Analytics Advantage survey) established central analytics groups, but 58 percent had more decentralized approaches, with limited or no central coordination. One relatively decentralized approach is to create a “center of excellence” (CoE), an organizational model with largely decentralized analysts that still incorporates some enterprise-level coordination. It’s been a popular approach to organizing analytics, but I think it’s suboptimal for most organizations. In fact, it often suggests a failure to commit to analytics as a key business tool.
In the CoE structure, analysts are based primarily in business functions and units, but their activities are coordinated by a small central group. The CoEs are typically responsible for issues such as training, adoption of analytical tools, and facilitating communication among analysts. The CoE builds—at least in principle—a community of analysts and can organize or influence their development and their sharing across units. Since the CoE is small, the supposed coordination benefits don’t cost organizations very much, but they often don’t get much value.
Granted, there are some companies that simply don’t need a lot of central coordination and collaboration. CoEs have the greatest appeal in large, diverse businesses with a variety of analytical needs and issues, but that still would benefit from a degree of central coordination. An analytics CoE is a more analytically-focused version of the Gartner-promoted “business intelligence competency center” (BICC), which was typically focused primarily on reporting or descriptive analytics. There isn’t a lot of evidence that BICCs changed the world for the better.
Let’s face it: The key question is whether the CoE has the power and resources to make substantial collaboration and synergies happen. Of course, there are many variations on this model, depending on the powers of the CoE. Do analysts report to it on a dotted-line basis? Or are the business units solidly in charge of their analysts? Does the CoE control the staff development agenda and resources? Does it double as a Program Management Office (PMO), with powers to coordinate priorities and resources across business units? What coordination mechanisms is it empowered to employ? A CoE with a lot of power and analysts reporting to it begins to look like a centralized function, which is probably what it should be.
While almost any organizational model can be made to work with enough tinkering, there are some serious shortcomings to the CoE model for analytics. In general, I have observed that most fully centralized analytics groups are more effective—at least when they employ ways to integrate the work of central analysts with business units and functions (sometimes with so-called “embedded” analysts who are deployed throughout the business).
Many CoEs just don’t have much power to influence what happens with analytics. Because they don’t generally own the people who do analytical work, they can’t manage their careers and projects in any substantial way. Many can’t even influence the direction of analytical projects, which are controlled by business units. In many cases the CoE’s primary coordinating activity is to have occasional information-sharing events. Some have also set up sharing portals. These are fine activities, but they often don’t move the needle on analytics-based performance. A fully decentralized set of analytics people and groups would be worse, but not much more so.
Far preferable to a CoE would be a central analytics and data science organization with analysts assigned to and rotated among business units and functions. That way you’d still get the benefits of centralization, while line organizations get their needs met as well. The central function allows for a critical mass of quants and for central coordination of their skill and career development. It should be a shared service, where anyone with the money and the high-priority requirements can get the help they need. The assignment and rotation allows for close relationships between analysts and decision-makers, and avoids the unresponsive bureaucracy that can result from centralized functions.
Organizational design is like building design: you don’t get a high-quality structure without paying for it. In other business functions like marketing or finance, we don’t generally create a CoE—we create a group that is centralized to some degree. Imagine a CFO or a CMO who tried to coordinate finance or marketing activity across a business, but didn’t have any decision rights or direct reports to carry out their strategies. If companies are serious about analytics, they should treat it like these other established functions.
The politically correct answer to the question, “How should we organize analytics” is, “Any structure has its strengths and weaknesses, but can be made to work.” That doesn’t mean, however, that all structures are equally effective. Centers of excellence are appealing, in part, because they don’t involve a lot of commitment. A fully centralized group suggests that the organization is serious about analytics, and is willing to give it a central, permanent home in the organization chart.