Pitfall 1: Lack of Executive Buy-In
MDM serves as the language of a business. It represents the crucial reference data that defines the dimensions of an organization and how its associates should report information. It's common for a single business unit to embark on an MDM implementation focusing solely on how they define their data elements and entities. Trouble arises when this activity detracts from a corporate standard or produces information inconsistent with the viewpoint of senior leadership. For an MDM implementation to be successful, senior stakeholders must see the value of the initiative and act in an enforcement role to ensure accountability amongst various stakeholders. This is especially true when process re-engineering and data governance initiatives come into play.
Pitfall 2: Not Focusing on Business Processes
It's common to think that technology automation can act as an acceptable alternative to a defunct operational process. This couldn't be farther from the truth and may in fact be the main contributor to failed MDM implementations. In order to create a single view of a reporting entity, for example, "customer," a project must include ample time for process optimization and re-engineering. At each stage of the data chain, from point of origin and data entry through data consolidation and reporting, clear business processes are necessary to support the flow of data and, ultimately, the integrity of that data. This is where executive buy-in plays an important role, since it is common for business units to resist change and potentially surrender control. Leaders must be prepared for difficult discussions around standardized processes and the role of data stewardship.
Pitfall 3: Lack of Data Governance
Pitfall two serves as a catalyst for the third: for any MDM solution to be successful, there needs to be an overarching data governance discipline. At the core of MDM are the business rules, decision rights and stakeholders that ensure an MDM solution is not just a project with a specific end date, but an ongoing program and core competency for the organization. As part of the data governance component of an MDM implementation, many areas need to be addressed, including data terminology and taxonomies, data stewardship, decision rights, accountability, corporate policies and standards. It's important when planning an MDM implementation that the governance components for predeployment and postdeployment stages are addressed.
Pitfall 4: Starting Big, Ending Small
Many implementers think an MDM initiative must start with a clean slate, break down all of the silos and rebuild Rome. Over the last several years, commoditization has been a trend in business, spanning industries such as banking and telecommunications. This makes it ever more challenging for businesses to centrally manage reference data with the goal of creating a single view of the business. It's tempting to start over when there are so many silos and disjointed processes. But trying to redefine how all of a company's business units, merged organizations and corporate entities fit together is going to be a multiyear initiative with a project scope that becomes a moving target. Any "big bang" program is dangerous on many levels, but mainly because we live in a world of uncertainty. Especially today, with the worldwide financial crisis, there is no telling what will happen to any given organization or industry. An MDM implementation should have a manageable scope focused on specific dimensions of the business, such as the financial chart of accounts structure. This will allow for faster deployment followed by a period for monitoring and improvement. A tighter scope also allows your business to be more agile in dealing with change while fostering learning and improvement during the migration and postdeployment stages.
Pitfall 5: Lack of Data Validation
Implementations that are centered on data can be challenging to test. Unlike traditional software implementations where expected results are driven by use cases and functional specifications, MDM implementations require a significant amount of data validation at various points within the architecture. For an implementation to be truly successful, a solid data validation plan is required both during the implementation and also as part of an ongoing production process. If the scope of the MDM plan only validates the inputs and outputs of the solution, it will become susceptible to downstream issues. Therefore, it's imperative that thorough end-to-end data validation testing is anticipated and completed.
Mike Cochrane is the CIO of Palladium Group, Inc. He has appeared at industry conferences and is a frequent contributor to Information Management as an expert in performance management,enterprise information management and business intelligence. He can be reached at email@example.com.