MDM for Performance Management

Philip Russom
Research Director for data management

Topic: MDM

Master data management (MDM) and business performance management (BPM) are two of the most rapidly adopted practices in business intelligence (BI) and data warehousing (DW) in recent years. Yet, few organizations have fully leveraged the synergy between the two. Allow me to explain some of the benefits of MDM specifically applied to BPM, as well as a few tips for making it happen.

First, let's step back for a moment and define our terms:

For the purposes of this article, business performance management is a business methodology (usually supported by BI technology) that depends on well-defined numeric measures, metrics, and key performance indicators (KPIs) as accurate representations of corporate performance. Although it's possible that metrics can exist and be useful out of context, they achieve greater business value when placed within a hierarchy, where measures roll-up into metrics, which roll-up into KPIs. The results of BPM are often presented in dashboard or scorecard-style reports and analyses.

Master data management is a mix of people, process, and technology that drives people in multiple functional roles to agree on standard definitions of common business entities (customer, product, sales, etc.), then represent these definitions in the master and reference data of related IT systems. This, in turn, enables the consistent, apples-to-apples sharing of data across multiple IT systems and the business units that own those systems.

MDM plays two different -- but equally important -- roles in BPM:

• MDM applied to operational systems. This makes data sources for BI and BPM better understood, which leads to more accurate selection of preferred sources for specific BPM purposes.
• MDM applied to BPM data. This data is inherently complex due to its hierarchical and multidimensional data models. MDM simplifies the management of BPM's complex data by supporting its models and providing change management functions.

A number of problems arise from a lack of MDM for BPM:

• Metrics and KPIs are calculated from the wrong data. When the wrong data is presented in management dashboards, bad decisions are made based on bad data.
• Analytic cubes degrade. We think first of dashboards or scorecards as the delivery vehicles for performance data. Yet, online analytic processing (OLAP) is commonly applied to the analysis of multidimensional BPM. Either way, the data content of dashboards and cubes suffers from poor MDM.
• Roll-ups for measures into metrics and metrics into KPIs are easily broken. When the roll-up of corporate performance is misrepresented, top-level perception of performance is skewed -- or just plain wrong. Furthermore, as with any roll-up, this one is a complex data consolidation process. Even a small error can halt the roll-up before it completes, which usually forces a restart after a time-consuming fix.

A KPI, by definition, is calculated from a rich hierarchy of measures and metrics. The levels of hierarchy are analogous to the multiple IT systems across which MDM coordinates master data. Therefore, the hierarchical levels of BPM (and similar multidimensional data structures in other areas of BI) benefit from their own MDM solution. Despite the need, hierarchy management and dimension management are too seldom part of any MDM solution. This can be a lethal failure for data that's inherently hierarchical and dimensional, as is almost always the case with BPM data.

With appalling rapidity, users flatten master and reference data models that should otherwise be hierarchically or multidimensional. This is due to the difficulties of multidimensional data modeling (which users don't fully understand, in general) and the lack of maturity for hierarchy management (in some MDM tools and most tools for data integration and quality). Before BPM can be fully served by MDM, relevant vendor products and user practices need far deeper support for hierarchies and multi-dimensional data models.

As if the complex data models weren't challenging enough, BPM data also involves slowly changing dimensions. Therefore, you need tools that support hierarchies, plus change management, impact analysis, versioning, and roll- back for slowly changing dimensions. Workflow for proposing, reviewing, and approving changes helps, too.

Summary and Recommendations

• MDM has an impact on BPM, whether MDM is practiced for upstream data sources or for the hierarchies of BPM. Therefore, if you care about BPM, don't ignore MDM. Coordinate and align the two.
• MDM within BI and DW tools and platforms -- specifically for BPM -- is still not common enough. More users need to apply their MDM talents to hierarchical metrics.
• Hierarchies are implicit in any metric-driven practice such as BPM. Therefore, BPM is best coordinated with an MDM solution that supports hierarchy management.
• Slowly changing dimensions are an issue with most DW and BI applications, including BPM. With MDM for BPM, look for tools and platforms supporting hierarchy management, plus versioning, roll-back, change management, and other functions that help with slowly changing dimensions.

Philip Russom is the research director for data management at TDWI. Philip can be reached at [email protected] .