Performance Management and MDM Convergence

Enterprise performance management (performance management for short) and master data management (MDM) are two of the most talked-about topics in business intelligence (BI) today. Although the industry treats them as two distinct disciplines with their own methodologies, tools and implementations, in reality, they are two dimensions of an enterprise information management strategy. Without the key performance indicators (KPIs) of performance management, MDM becomes an exercise in data integration, and without MDM, performance management cannot achieve the promised enterprise-level impact. This convergence will become one of the driving forces of our industry for the next couple of years. The convergence will be both logical, in how to apply various methods and techniques, and technical, in how to make the various technologies work together.

Performance Management Simplified

It goes by many names in the industry: corporate performance management (CPM), enterprise performance management, business performance management, etc. Performance management can take many forms, such as reporting systems, analytic applications, dashboards, scorecards, balanced scorecards and strategy maps. In the end, though, all are concerned with a similar issue: providing the organization with visible, actionable metrics to illustrate and guide its performance.

Let’s examine the two concepts of visibility and actionability.

Visibility is not as simple a concept as the name would imply. Many times, the performance data that really matters is locked away in an obscure system from which it cannot be easily teased out without some technical programming. The data has to become readily available, or visible, to the decision-makers so that they can use the data.

Actionability refers to the ability to take some form of corrective action based on the data presented. Whether the presentation is in a computer system is irrelevant. What is key is that the data being presented is something that can be controlled and manipulated by business users.

By tying together actionable and visible metrics into a cause-and-effect relationship, you get an analysis chain. Figure 1 illustrates a simple example of metrics that have varying degrees of visibility. Strategic accounts attainment has direct impact on change in customers, which has a direct impact on financial health.

An organization will have hundreds of these analysis chains, but there are normally a dozen or so that ultimately drive the business. We will use this model to tie in MDM shortly.

Figure 1: Example of an Analysis Chain

MDM Simplified

The best way to start with master data is with logical understanding. Great technical tools exist to build the physical master data architecture, but without the right logical model, they will fail.

Because master data, whether product, customer, employee, vendor or others, could contain hundreds of elements, segmentation is needed to follow a divide-and-conquer analysis. If not, the analytic effort to understand hundreds of attributes making up the customer or product in one big group will become an endless cycle of analysis paralysis.

Segmentation should first be broken down into four levels.

Level 1: Identification. Identification of the elements that determine uniqueness is much easier said than done. One would think a universal ID like Social Security number would work, but it is almost never the right answer. Uniqueness is almost always determined by a concatenation of keys that are either naturally occurring (Social Security number, tax ID) or system owned (customer number, vendor number).

Level 2: Common elements. Level 2 is reserved for elements that are commonly queried but are not required for uniqueness. Common elements should be used by a large group of users, a large amount of eventual data queries or both. The common level can contain attributive information to complement the identification level, such as when the ZIP code is in level 1 and the town name is in level 2.

Level 3: Extended internal. Level 3 is where the large group of internal profile and attributive data elements will go. If elements are not used by a large group and not queried heavily, they end up in level 3 if they are sourced from internal systems. Internal systems are any systems that your enterprise has complete control over without dependencies or codevelopment by other organizations.

Level 4: Extended external. Level 4 is for the outside data brought in house. The outside data may be purchased data, such as credit ratings, product rankings, address correction files, or shared data, such as co-owned systems with business partners, customers or vendors. Two classic examples of extended external elements are market share data and external shipper (FedEx, UPS, etc.) data.

With this four-level understanding and the analysis chain, we can now combine them to show how performance management and MDM converge on a logical level.