TDWI | Training & Research | Business Intelligence, Analytics, Big Data, Data Warehousing

Reimagining the Analytics CoE

New technologies, increased user skills, and new competitive demands are driving a new focus on the role of the analytics center of excellence.

• By Brian J. Dooley
• November 2, 2018

Analytics centers of excellence (CoEs) have been around for a decade or more, enabling organizations to increase analytics use and help with the complexities of moving into big data.

For Further Reading: Q&A with Jill Dyché: "Our COE Is About to Kick the Bucket!" Balancing Self-Service, Governed BI, and Analytics Enhance BI Delivery Using a New Framework

As with other CoEs and competency centers, the analytics CoE is a shared services organization established to coordinate complex activities and newly introduced technologies. Under such a mandate, technology change inevitably reaches a point where organizational change is also required.

During the last 10 years enterprises have experienced large increases in analytics use and data availability as well as significant developments in machine learning, deep learning, and predictive analytics. Due to serious data security issues, growth in citizen data science and other self-service practices, and the continued fragmentation of enterprise data, analytics teams are struggling to keep up with the pressure toward complete digital transformation.

Although the initial mandate of the analytics CoE was to establish and support analytics functions across the enterprise, recent changes argue for greater formalization of the CoE, greater differentiation between approaches, and an expanded role.

What Are the Goals?

The analytics CoE must provide a range of support functions for analytics users at all expertise levels. It must also provide training and promote greater understanding of data across the organization. At the same time, it must support everyday analytics processes and overall organizational goals (such as increasing analytics usage and supporting innovation).

The advanced capabilities of predictive and machine learning tools now make it possible to focus greater attention on new initiatives, through both new analytics techniques and discovering benefits and efficiencies for existing processes using advanced analytics.

One way an analytics CoE can accomplish these goals is by empowering citizen data scientists -- individuals skilled in business processes or functional areas with sufficient data science expertise to create their own solutions. Although CoEs can solve the age-old problem of aligning IT with the business, they also require special tools and training, oversight, and control of data resources to ensure that shadow IT -- replete with data and process siloes -- does not take over.

Providing user communities with access to training is critical to the role of the analytics CoE. All employees should have some familiarity with data to ensure processes are understood and secured. Users involved with self-service analytics and analytics-based decision making require extra support. Citizen data scientists need to be trained in data access and data management as well as data security. Training programs also need to include the latest developments in analytics, ML, and AI. Everyone also benefits from sponsoring an internal analytics community with access to external resources and help centers.

Advances in machine learning and predictive analytics are creating further concerns for analytics CoEs to address. Incorporating analytics and AI within business processes demands greater understanding of both processes and consequences, as well as ensuring transparency to avoid the hazards of "black box" operation. Process models themselves need to be managed, and this is likely to become a more critical issue for the CoE.

Organizing the Analytics CoE

CoE organizations differ widely depending on analytics maturity, organization size, outsourcing, IT capabilities, analytics capabilities, industry, centralization, and available analytics resources and personnel. There is no single organizational structure that will satisfy all needs, and the analytics CoE must invariably work in different ways with the IT department, analysts, human resources, and others. It needs executive oversight and support to ensure that its mandates are carried out. An analytics CoE also needs mechanisms (such as ROI and chargebacks) for determining effectiveness.

The analytics CoE needs to fill in the gaps or address the weaknesses in the current analytics environment while planning for the future. This demands a review of data management and data access, personnel, maturity level, analytics use, and particularly the medium- and long-term goals and direction for analytics in the organization.

The CoE needs a mix of full-time and part-time staff, combining data scientists and business analysts. Full-time staff is essential for developing consultancy, creating a broad understanding of analytics, and ensuring continuity. The CoE must also operate flexibly and draw staff from other departments if needed, often on a part-time basis.

Although primarily consultative, development services may also be needed to oversee creation of solutions that are beyond business-staff capabilities. The degree of integration with existing analytics and data management departments also must be considered. In some cases, the analytics department may need to become the CoE. In other cases, analytics may be largely outsourced with the CoE taking an intermediary role, handling the interface between the outsourcer and user departments.

The Bottom Line

Analytics is becoming more pervasive, and organization needs will always change. Current analytics CoEs must be re-evaluated in light of new technologies and growing user skills and experience. This demands the analytics CoE take a flexible approach with a greater focus on the repercussions of incorporating technologies and tools, from AI and ML to predictive analytics.

The greater role of analytics also requires a transformation into an analytics and data organization capable of supporting embedded processes that will demand new levels of understanding throughout the organization.