Leverage Data Quality to Build an Effective Enterprise Architecture

InfoManagement Direct, May 21, 2009

Conventional wisdom among information architects is that solving the data quality problem is really a matter of solving the information architecture problem. It is generally agreed that unifying the business and technology architectures of an organization is the way to provide the necessary contextual environment to solve data issues and ensure the integrity of information. In A Practical Guide to Enterprise Architecture, James McGovern et al write that data integrity suffers from compromises in the enterprise architecture because people are pressured to produce more in less time and the proper cross-checks on the data are not performed. Information architects agree that addressing faulty data only at the physical data level cannot work because the data layer does not capture the requisite semantics to accurately understand data spanning business processes.

More than 25 years ago, Steven Spewak made the case for architecting a stable business model before designing the information systems that support it. In his enterprise architecture planning methodology, Spewak provided guidance for implementing the top two rows of the Zachman Framework: Scope (Planner) and Business Model (Owner). EAP was innovative for its data-driven approach emphasizing data dependencies being defined before system implementation and the order of implementation activities based on the data dependencies. Layer 3 of the EAP framework plans the future architecture, defining the data dependencies by understanding the major kinds of data needed by the business. Application and technology architectures are then designed specifically to manage an appropriate environment for the data supporting business processes. This planned approach defines a consistent method for centrally collecting, migrating and storing data, leading to improved data quality by making enterprise information accessible and timely for any business need, all with commensurate cost reductions and efficiencies.

One of the modern implementations of the EAP framework is the Federal Enterprise Architecture, an initiative that aims to provide a common methodology for IT acquisition in the federal government. The primary purpose of the FEA is to identify opportunities to simplify processes and unify work within similar lines of business by developing a common taxonomy for describing data and IT resources. Federal agencies have adopted the guidance laid out in the five FEA reference models to build out their corresponding architecture layers (performance, business, data, application and services, and technology). This guidance is intended to help establish a well-architected enterprise business model and the data dependencies to guide business and IT modernization efforts.

Yet even by adopting these and similar EAP-like approaches, federal agencies and other large organizations continue to experience data quality problems: overlapping data within functional areas, costly interfaces between incongruent systems, reengineered modes of information sharing not working as planned and new system development contributing more of the same. It has now become clear that despite a quarter century of advice on better business architecture and data-driven approaches, “information politics” and other factors inhibit coordinated approaches to building out the enterprise frameworks required to transition organizations to the Information Age.

Given this reality, it might be time to rethink the notion that effective information architecture development will solve the data quality problem. In recent years, a handful of large organizations and federal agencies have established mature data quality practices, and it is now possible to see that the impact of these initiatives goes far beyond data management and information exchange improvements. Whereas a data quality improvement program’s principal goal is to identify and standardize the quality of performance-related data - reassuring data consumers of the credibility of information upon which they base their decisions - a byproduct of the practice is the establishment of new, enterprise-wide practices that can be leveraged for other organizational initiatives, most notably enterprise architecture. This is, then, a somewhat radical bottom-up approach, viewing data quality improvement as a key enabler of effective information architecture development rather than the other way around.

Data quality principles and initiatives can enable better delivery of integrated services, the cornerstone of the FEA goals. In the sections that follow, the layers of the FEA and value proposition of the DQIP are briefly explained in reference to how data quality provides a foundation to build an organization’s enterprise architecture.

DQIP and the Performance Reference Model

The FEA performance reference model is a standardized framework to measure the performance of major IT investments and their contribution to program performance. By utilizing a number of existing approaches to performance measurement, the PRM identifies performance improvement opportunities that span traditional organizational structures and boundaries. Data quality initiatives can lay the groundwork for PRM development by discovering the systems and data most responsible for high-priority business performance reporting. The DQIP certifies data tied to performance at rigorous data quality standards, giving the organization confidence that the data is fit for use and will contribute to program performance across the organization.

Data quality’s information value cost chain supports the difficult task of estimating the value of IT investments. This process maps performance data’s complete lifecycle to include the logistics of their creation and the steps of their transformation into “finished” information products. The process also includes detailed descriptions of the servicing of the data (their maintenance as well as support to customers using the data). Costs are attached to the data at each stage of its lifecycle. These costs can then be compared against the real and intrinsic value of the data to support the organization’s adherence to a five-year strategic plan or other key objective. Information products that do not yield a profit (i.e., their costs of production and maintenance over their lifecycle exceed their value to the organization’s bottom line) would be prime targets for reprocessing.