So what is Master Data Management and how can it help with the problems that we've now seen?
Note
Master Data Management is a set of tools and processes that aim to deliver a single clean and consistent view of each master data entity that exists within the organization.
In other words, we want each system across the organization to use exactly the same set of master data, in terms of entities, attributes, and members, as long as of course the system actually needs them.
As you might expect, this is easier said than done. Each system will have its own local database, possibly different meanings for different entities or attributes, and its own set of users who are highly accustomed to seeing their data in a certain way.
In order to achieve the above, there is a high-level plan that can be followed for implementing an MDM program:
Getting executive sponsorship—You need to do this with any large IT project, but it's especially necessary with MDM. Someone will need to sell the benefits of MDM to the various group company heads or department heads who may prove to be an initial barrier to implementing MDM.
Defining the scope—Your MDM project needs to have a clearly defined scope so that you know how many systems currently use a given entity, what involvement you will have with these systems, and therefore defining the overall objectives of your MDM program.
Designing a solution—No IT project should ever just launch into the development stage, and the same applies for MDM. It is necessary to thoroughly analyze each source system, using data profiling amongst other techniques, in order to understand the behavior of each entity and attribute that is used.
Develop a model—A standardized model must be created per entity that is capable of housing the entity in the best possible way. This includes choosing the correct names and data types for attributes, all of which is driven by the analysis of the various data sources.
Extract data—Master data from the various legitimate sources must be extracted and loaded into the model. Whether or not this process also occurs on a continuous basis depends on the chosen MDM architecture, which is covered later in this chapter.
Publish data—Once the model has been populated, a method must be devised to allow systems to use data from the newly defined master data model. As with extracting data, how this actually happens depends on the architectural choices.
By going through these steps, each system that falls under the scope of the new MDM program will require some changes in one way or another. These changes may be at the data level, or they may even require architectural/code changes to the system. Given the potential scale of getting each system on board, a sensible approach is to tackle each system in turn, on a piecemeal basis, instead of attempting a big bang approach.