Evaluating Master Data Management Technology
Choosing a vendor for master data management? Use these seven criteria to match available tools to your enterprise's needs.
- By David Loshin
- May 18, 2016
The core components of master data management (MDM), namely record linkage and entity data integration (usually focused on customer data), have been around for many years. It is only since the mid-2000s, however, that vendors have provided MDM tool suites by packaging these functions with entity data models and corresponding data services.
At this point the market is beginning to see a second generation of MDM tool vendors that support customer or product data management applications. What differentiates the newer products from each other and from the earlier crop of products?
Almost every vendor offering satisfies a modern approach to managing master data, but not every approach is right for every enterprise. Each MDM tool provides a composite view of a uniquely identifiable entity's information, but there are often differences in method, function, or data storage.
Here are seven MDM capabilities that you can use to compare different tools or specify requirements to meet your organization's business needs.
#1: Identity resolution
Resolving ambiguity and linking similar records is one of the primary objectives of master data management, so it's not surprising that it tops the list of evaluation criteria. One factor to consider is whether the product only employs deterministic matching or whether it also supports probabilistic matching (that uses similarity scoring to link records that do not share the exact same values).
You should also evaluate the precision and accuracy of the matching (the tool should prescribe predictable limits of false positive matches and avoid false negative non-matches) and how easily business users can adjust settings to tweak the match scoring.
#2: Physical versus virtual
The presumption that a system could create a single "golden record" for each customer or product dominated the early design of MDM products. These early tools created a standalone repository to capture a consolidated record that selected "survived" attribute values from among the linked records.
Today there are alternatives to this physical repository, including virtual repositories that use a master index to point to the original records in their source data sets. The virtual approach no longer forces the creation of a single master record, and it allows for different consumer applications to view master records using different presentation rules.
#3: Master entity modeling options
Early MDM offerings included comprehensive data models to represent commonly used entities such as customer, supplier, or part. This approach may still be suitable for reporting and analytics applications that would use a standalone representation of entity data (such as a customer profiling and analysis application).
That provided data model may not be appropriate for each organization's operational business processes, however. Some vendors now provide a process-oriented approach for users to develop their own representative data models -- integrated modeling is now a differentiating factor for MDM tools.
In order to satisfy the needs of both operational and analytical applications, the data must be synchronized in the master view. When using a single repository, data from the source data sets must be forwarded to the MDM environment for identity resolution, linkage, and consolidation on a periodic basis (such as every night).
Allowing applications to accumulate entity data between those periods means that data subsystems will be inconsistent compared with the data in the master repository. Some of the newer environments provide more frequent synchronization points, synchronize solely through a common physical representation, or avoid the entire issue by using a virtual approach that always provides the most recent source records.
#5: On-premises versus cloud based
A recent innovation in the MDM space is managing master data using a hosted or cloud-based environment instead of a traditional on-premises implementation. Consider whether your organization's existing cloud footprint is suited to integration with a cloud-based MDM implementation, particularly if entity data is stored in SaaS applications such as Salesforce.com.
#6: Data management approach
The typical foundation for storing master data has been the relational database management system (RDBMS). However, because the connections among and between entity records carry information (including identifying information), some vendors are replacing the RDBMS with a graph database to capture master data. Graph databases treat the links between entities as first-class objects, with their own attributes, and provide an alternative means for searching for, linking, and analyzing master data.
#7: Data governance
Finally, there is one feature that has been consistently improving in MDM offerings, and that is operational data governance, such as the ability for data stewards to override automated decisions for linking (or not linking) records or adjustments to the data models.
Data governance capabilities provide users with an increased level of confidence in the quality of the composite master record, and are a must-have for any modern MDM deployment.
Choosing an MDM Tool
Use these seven core characteristics to consider your enterprise's MDM needs and you will surely choose the best MDM tool for your organization.
David Loshin is a recognized thought leader in the areas of data quality and governance, master data management, and business intelligence. David is a prolific author regarding BI best practices via the expert channel at BeyeNETWORK and numerous books on BI and data quality. His valuable MDM insights can be found in his book, Master Data Management, which has been endorsed by data management industry leaders.