May 2007

If you’ve decided that a Master Data Management implementation would provide significant benefits for your organization, the next step is determining the appropriate strategy to use…

Determining Your MDM Implementation Strategy

To determine the best implementation strategy for your Master Data Management (MDM) project, first ask and answer various questions regarding the problem, use, location, and commitment to the data and the solution. Once you have a confident and firm understanding of the pain points, pitfalls, issues, benefits, risks, and rewards you can decide on a properly suited strategy. There are different strategies to weigh when undertaking an MDM effort. Each strategy has its own advantages and disadvantages offering a variety of approaches to best serve the organization’s needs. Here we will provide an overview of the various strategies, and some questions that can serve as a guide for deciding upon the best strategy for your project.

The styles and uses of MDM vary on the basis of what the business needs and resources are as it relates to the master data in the organization. These strategies range in spectrum from a more central “System of Record” approach to the less central “System of Reference” style. Additionally, the concept of an additional “External Reference” method also exists.

The Transaction Hub Approach

The transaction hub strategy or repository approach falls in the spectrum more as a “System of Reference”. A central repository is created that will contain all of the master data physically stored in a single database, allowing the data to be accessed more readily. The data must be able to support the multiple systems for which it will provide master data and thus should contain all of the necessary attributes as it relates to these systems. Due to the breadth of the data contained in the repository, it is able to act as a transaction hub in support of transactional systems. The data within this repository will be the most up-to-date within the organization.

The Registry Hub Approach

The registry hub strategy for implementation resides on the other side of the spectrum from the transaction hub style. Where the transaction hub style physically locates all master data in a single location, keeping associated attributes for all systems, the registry style keeps the master data in the various transactional systems it is stored. A central registry is created that will map the data in these sources together, creating a common key for retrieval of this data. No master data is physically stored in a registry hub style of implementation. Read-only access is provided for the data as the master data is not distributed from the registry hub for transactions elsewhere.

The Coexistence Hub Approach

The coexistence hub strategy, or hybrid approach, falls in the center of the spectrum of MDM implementations. In this implementation strategy, there will be physical storage of the data in the MDM solution as well as in the transaction systems themselves. For the most important or most queried attributes of the master data, the data may be maintained in the MDM solution to provide faster retrieval of this data. For the less commonplace attributes that are not as frequently used, a registry entry will be used to retrieve this data where it is stored in the transactional system. It is generally accepted that data physically stored in a coexistence hub strategy of implementation may not be the most up-to-date version of the master data.

The External Reference Approach

The external reference strategy of implementation does not reside on the spectrum of MDM implementation from system of reference to system of record. However, depending on the attributes contained in your MDM solution, external reference may need to be integrated. In many industries, there may be an external reference needed to refer to a common industry key. In this approach, these attributes will be linked to a trusted external source for this data.

Planning Questions to Guide Implementation

Is consistent master data a pain point in your organization? Are multiple portions of the organization in disagreement over the true version of the data at hand?
Is there an external source of industry data that is needed for reporting? Is that data currently available and easy to relate to internal sources of master data?
Is there a need to cross reference master data in a variety of transaction systems? Does there need to be a central location from which to update this data or will it only be used in a read-only fashion?
What level of data latency is acceptable in your MDM solution? Will the users of your MDM solution expect to be able to query the MDM solution and retrieve the most up-to-date master data or is a lag of a certain timeframe acceptable? If data latency is not acceptable, is long query time or retrieval acceptable?
Will the data need to be accessed frequently? Does the data need to be highly available with the ability to handle a high rate of concurrent requests? Is this data critical for running operations? Will the data need to be maintained historically over time to support historical transactions? Will multiple transactional systems utilize this data?
Does the master data currently reside in a transactional system? Is the data model supporting this data built to serve in an enterprise level capacity or is the model functioning to serve transactions? Are a large number of requests inhibiting performance of the transactional system, as well as the system dependent upon the data, because the model does not support this type of activity?
Are the organization’s expectations realistic in terms of time, money, personnel, and infrastructure changes? Is the goal of the effort a short term fix or a long term solution? Has enough planning and investigation been done to adequately represent the level of effort?

Putting It All Together

The spectrum over which an organization can implement a Master Data Management strategy varies widely. Your answers to the questions above will influence the selection of a solution that best meets your needs.

Consistent master data is critical to any organization. Without it, organizational divisions are unable to merge their data and speak a common language. System integration and interaction become more difficult, and gains that an organization could ascertain by analyzing data across multiple operational systems become impossible. Keeping master data centrally located as in the repository approach will better allow for management of consistency across your organization. However, with more analysis work and continued monitoring, the registry hub system can facilitate consistency as well. As the registry style entails read-only access to master data, the monitoring and qualification of data would be required across systems to ensure consistency. The registry system will provide a map for your organization to coordinate its data.

For many critical master data entities, a significant amount of attributes for that entity can exist outside of the organization. For example, a given household entity may be a customer of your company, but knowing their purchasing patterns may be outside of the realm of information available to your system analysts. Using an external reference style would allow you to link your internal data to other realms of information beyond your internal data resources. For this reason, aspects of the external reference MDM style of implementation may be right for you.

Often master data is used in a lookup capacity in which the subscribing system needs only to retrieve data specific to a given entity. At other times, it is critical that a central place contain information specific to the entity, but its attributes are updated by multiple systems. According to the mix of entities in your organizations, you may choose to have a system of record style (for those with more updatable attributes) or a system of reference system (when updates of the “centralized entities” are not required). A hybrid approach can also be taken, weighted more heavily of the side of the spectrum your organization falls.

Data latency can prove to be an important consideration in your choice of an MDM solution. Decisions on data latency within an organization often come down to a matter of resources. Whether using the data in an operational fashion or using data for decision support, if the data is expected to be the most up-to-date, the master data should be updated in the system from which the person or system is retrieving the data. If the data does not necessarily need to be the most current, the subscribing system can retrieve it from a read-only system in which the master data is updated periodically. Again, the nature of your master data management system can be moved along the spectrum given the importance of the particular entity or attribute of interest.

Data relevant to support mission critical operations needs to be made readily available to the systems supporting these operations. If attributes of a given data entity are being referenced in a critical fashion by only a single operational system, it may be sufficient to implement a registry hub style of master data management. However, if these attributes are referenced by other transactional systems with a critical need, a dedicated system with resources to ensure availability as would most likely be the case with a repository approach.

In many organizations, the need for master data can overwhelm the transactional system handling these. The demands on these transactional systems may impact your choice in MDM implementation style. When traffic to a transactional system from external systems is impacting the efficiency of your operations, the organization may consider offloading this additional work with a repository approach to MDM.

When reflecting on your organization’s commitment to the implementation of a master data management solution, consider that repository MDM implementations often require more resources to implement up front. They require the integration of this new MDM solution with new and existing transactional and decision support systems. Registry styles are often less resource intensive. Additionally the implementation of registry master data management solutions can facilitate the future implementation of repository style implementations.

We hope this summary of MDM strategies has allowed you to examine how your organization is implementing (or may consider implementing) an MDM solution. Stayed tuned for our next quarterly communication which is set to address implementation challenges that you may face along the way…