When enterprises first started to use mainframe computers, it was fairly easy for an organization running one centralized computer system to maintain a single set of data about the basics of the business, such as product information and customer characteristics. Since those days, however, this information -- which is vital to understanding the performance of the business -- has begun to be collected by and to live within a wide variety of different systems, both operational and transactional, both internal and external to the enterprise.

The processes and software encompassed by the term "master data management" (MDM) have arisen to overcome the data disparities that invariably arise in complex technology environments. The standardization that MDM offers is absolutely critical to the success of business performance management (BPM) initiatives, as only faulty conclusions about performance can be drawn from uneven and inconsistent data.

Master data can be defined as persistent reference data about business entities, such as customers, products, or accounts, about which there is an agreed-upon view across the organization. This differs from metadata, which is a term used in data warehousing and data integration initiatives to refer to the data schema -- that is, the structure and definitions of the types of data in a database. Master data refers, instead, to the data itself, information such as the ZIP code of a customer's location or the color of a product. Master data management, then, is a set of policies and procedures for accessing, managing, and maintaining a single version or view of master data, and for coordinating the data with subscribing systems across the enterprise for the purposes of maintenance, analysis, and reporting. For example, MDM software can help complex companies maintain a standardized chart of accounts.

A year ago RSM McGladrey Employer Services (RSMES), an HR and benefits administration company, set out to improve its master data management. The company had a master database, a Great Plains General Ledger system, that stored information about accounts, cost centers, geographic locations, and similar items, but employees were limited by the inflexible account structure inherent to the original design more than five years ago. All account structure streamlining and new business requirement dimensionality had to be manually changed in the hierarchies of the company's analytics system, and accounts had to be created in the G/L. This meant having to maintain data in multiple systems, which was time-consuming and inefficient. Like most G/Ls, RSMES's system used a long account number within which each dimension of the account was expressed as a specific sequence of digits. Adding one logical member (i.e., a new cost center) required 50 or more new G/L accounts to be manually entered.

Implementing a master data management system enabled RSMES's financial analysts to begin entering the information in one system and having it flow automatically down into all of its chart-of-accounts systems. The result has been improved consistency, accuracy, and up-front usability of its chart-of-accounts data across all systems companywide.

MDM or Point Solutions?

Customer data integration (CDI) and product information management (PIM) applications can be viewed as functionally focused master data management solutions. There are a number of highly targeted solutions available for managing these specific types of master data. If maintaining consistency in the company's customer information is the only thing keeping you up at night, then a CDI solution may be all you need in the way of MDM. The same can be said of PIM if management of product data or optimization of supply chain efficiency is your exclusive focus.

Organizations that need to manage a broader range of master data to solve an overall business problem that involves customers, suppliers, and products may not find a CDI or PIM solution to be the best approach. If your goal is to integrate with a BPM initiative at the enterprise level, these types of functional MDM solutions on their own are definitely too limited to serve your purpose.

Another type of software that is sometimes considered to be an adequate substitute for MDM is extract, transform, and load (ETL) software. It's a common misconception that if you have ETL, you don't need a master data management application. While ETL enables companies to import data from multiple sources; reformat and cleanse the information; and then load it into another database, datamart, data warehouse, or other operational system, it is just that: a mechanism -- used primarily by IT -- for taking data from "here" and moving it to "there." In a perfect world, business users would maintain all the information that the company needs in an enterprisewide software system, and simply moving data around and running cleansing processes would ensure that it was consistent across the company. But in the real world, the data that is needed for BPM may not reside in the company's operational systems. A lot of it is in the heads of the business users, in their spreadsheets, and in departmental business intelligence or BPM initiatives. ETL doesn't provide a way to capture new information.

Further, transformations performed in ETL need mappings that are owned and managed by the business. An accountant can't be expected to use an ETL tool to map a G/L account transformation. And under Sarbanes-Oxley rules, IT developers should not be responsible for that kind of data. Ultimately, the master data for mapping transformations should be managed outside of the ETL tools, then utilized by those tools as a trusted source of valuable mapping information.

The bottom line is, if you need to proactively manage, generate, and synchronize data in one place, you will need MDM in addition to ETL. However, for optimum performance management data, there is one more step in data management -- the management of dimensional data.

Added Value of Enterprise Dimension Management

Companies that are evaluating MDM to support BPM initiatives are finding that generic MDM imposes limitations on the way they leverage BPM. Enterprise dimension management (EDM), on the other hand, is a type of MDM that helps users work better with the unique dimensions of enterprise master data in the context of their specific roles within the organization. For a marketing manager, a meaningful piece of master data associated with the product dimension might be color or size, whereas a sales manager might need to view the data in context of price and territory. This creates the need to support a variety of data hierarchies, some extremely complex.

Unlike generic MDM, EDM is specifically designed to support large numbers of rapidly changing dimensional models and hierarchies. Data is viewed in the context of tree structures that can be easily created, moved, and edited in the context of business functions, by business users, to support real-time data analysis by BPM applications. EDM facilitates the rapid creation of collections of hierarchical data to support BPM software's ad-hoc reporting capabilities. The creation of such hierarchical data collections is not an inherent strength of basic MDM, and thus constitutes a limitation in its support of BPM.

Enterprise dimension management manages data dimensions and hierarchies without the overhead of building and maintaining redundant instances of data in the enterprise model. Data hierarchies can be developed, managed, and revised quickly and flexibly, without increasing data maintenance resource requirements or reducing IT system performance.

In addition, EDM technologies are designed to integrate directly with subscribing BPM systems, providing export views that present the appropriate data in the context supported by the BPM system, and including all the integration necessary to seamlessly support BPM applications. Since general MDM isn't designed for this kind of integration, it requires extensive custom programming and maintenance to support BPM applications in this way, so downstream changes in MDM models and BPM applications require significant maintenance overhead to preserve their integration.

Organizations that are just doing enterprise resource planning (ERP) or that are concerned only about managing a unique list of customers may not care much about having multilevel data hierarchies. But analytics applications need the ability to deal with multiple hierarchies and many pieces of master data for a given subject area in order to get an accurate analysis. They need to be able to visualize the complex trees or hierarchies that result from multilevel data relationships and to test the effects on these hierarchies as changes are made. For this, analytics applications -- especially BPM systems -- need the across-the-board approach, the consistent view, of MDM. This view is further improved upon by the dimensional flexibility and agility provided by EDM.

Operational Data Made Manageable

The benefits of EDM in supporting enterprise BPM are clearly illustrated by the data management experience of Tiger Brands, a South African company that manufactures and distributes food products to retail centers through a consortium of companies. In 2004, with 23 business units using 23 different customer tracking systems, Tiger ended up with an unmanageable 130,000-member data hierarchy, in which each customer was duplicated numerous times. This made it impossible to get a clear picture of customers' cross-product buying profiles. In fact, when the organization did attempt to produce customer buying profiles, it found its data being corrected by customers.

The company needed a way to evaluate performance across a number of data hierarchies around each customer -- trade market, channel, region, etc. -- as well as externally procured customer sales data. So in January of 2005, Tiger implemented a dimensional management approach to MDM that enabled the company to establish a greatly streamlined customer hierarchy with only 11,000 discrete customer accounts mapped to those customers' 130,000 identities in the 23 systems. This allows Tiger to accurately measure performance by criteria such as "In this region, report on these customers by trade marketing code" -- regardless of how the individual business units track such information in their internal systems.

Tiger also uses EDM to unify its supplier list enterprisewide. The company needs to accurately track and report on how much business it does with Black Economic Empowered (BEE) vendors, in support of South Africa's Black Economic Empowerment Act of 2003. Each of Tiger's individual business units is encouraged to identify opportunities for relationships with BEE vendors, but before the EDM implementation, the divisions had trouble identifying and reporting on how much of their business was done with BEE vendors. The EDM approach enables the company to establish a consistent vendor hierarchy in which corporate managers can accurately measure the amount of business Tiger does with each vendor across the enterprise, regardless of how each of the company's individual business units tracks vendor information in its internal systems.

Tapping Into Human BI

A business's primary source of operational intelligence resides in the human knowledge repository of the company's workers and their spreadsheets. The real-time integration of this human intelligence into the company's data repositories is required for any meaningful BPM initiative. In order to integrate its human knowledge base with its other master data, a business must open up MDM to users across the organization. However, technology alone can't make this happen.

In terms of "human systems," data exchange and analysis have as much to do with processes as with data structures. For example, an ERP order-entry system may require a bill-to customer, along with that customer's address, to be created before an employee can book an order. Since customer and address are things that the order-entry system cares about, the software may allow users to build a geographic hierarchy only from that information. A user who wants to build a report based on market segment, such as consumer-product goods vs. health-care offerings, cannot do so through this ERP system. It is not practical to customize an ERP system every time a user needs a new information hierarchy; doing so would be both expensive and ineffective, since ownership of the resulting hierarchies in the operational system couldn't be assigned to the appropriate business functions. Therefore, the optimal master data management solution must be able to leverage the human systems' analytic and dynamic relationships with process, as well as accommodating the context of these relationships within operational roles.

In order to capture the information stored in people throughout the company, an MDM system that is going to fuel corporate BPM must include workflow management, chains of command, and change management or version control. McKesson Provider Technologies, a technology leader in the health-care industry, gives its business managers direct access to the dimensions in its data management system, putting responsibility for changing hierarchies and cost-center structures in the hands of the appropriate members of the organization. Role-based security and business rules governing changes to the dimensions ensure consistency in the management of the data. This uniformity in data management makes the hierarchies easier to follow and understand across organizational units.

Through version control and change management features, MDM systems maintain central control of data management, even though they allow business users to assume more role-appropriate responsibility for maintaining their portions of the data hierarchies. McKesson Provider Technologies can automatically track each dimensional change in its MDM system, including who performed the transaction, what changed, and when it changed. The software also provides the ability to search and report on individual changes to dimensions -- even reverse them, if necessary.

Version control capabilities become especially critical in the face of business changes and reorganizations of business units or products. These kinds of operational modifications require dimension changes, but a BPM system needs access to older versions of the hierarchy for its historical reporting duties. Meaningful performance measurement cannot be accomplished in a vacuum of historical information. An ERP system may be happy knowing only about the customer as it exists today, but analytics applications such as BPM need access to multiple versions of how a customer looked across history. The ability of an MDM solution to support parallel versions is especially helpful during planning and reorganization cycles. Furthermore, because dimension structures drive the financial and nonfinancial reporting results of the organization, a dimension management system must, as a part of the business's overall MDM strategy, provide the tools to ensure compliance with corporate governance requirements including Sarbanes-Oxley, GAAP, and International Accounting Standards (IAS) reporting standards.

While the goal of MDM is to standardize data and data management processes across the enterprise, on its own it doesn't go far enough to ensure the success of analytics initiatives such as BPM. Businesses implementing BPM need a data-management solution that addresses the master data hierarchy requirements of analytics applications, ensuring that measurement data can be rolled up accurately across organizational and operational boundaries, and that BPM initiatives capture all available dimension data to obtain their greatest value. This is how EDM improves the octane of master data management to boost BPM.

Ian Ahern is the CEO of Stratature, an enterprise dimension management sofware vendor. Ahern has 20 years' experience delivering BPM and business intelligence to organizations worldwide.