Beware of Different Flavors of MDM
Many companies are facing a harsh reality: an improved economy, increasing pressure from more global competitors and the increasing speed of business are driving them to deliver more service-led differentiation and to develop faster and more efficient go-to-market strategies - just to stay in business. One way many enterprises are trying to stay ahead of the competition is to make more effective use of the data they already have. In order to be successful, they require unification and integration of enterprise-wide data from multiple systems, including order management, customer relationship management (CRM), supply chain management, enterprise resource planning (ERP), logistics and provisioning.
How, for example, do you resolve the multiple data issues surrounding the following hypothetical initiatives?
- A freight forwarding company wants to unify operations to deliver support and tracking of a third-party logistics initiative.
- A managed service provider is looking for "triple play" cross-sell opportunities.
- A major logistics supplier wants to integrate sales operations to unify touchpoints with its top 100 corporate customers.
- A pharmaceutical company is trying to convert areas of revenue leakage into new sales opportunities.
- A technology company is laser focused on enhancing customer intelligence and monetizing its lead-generation programs to improve marketing productivity.
While the specific initiatives vary, one thing remains constant: the success of these initiatives is dependent on accurate and relevant master data records for customer, supplier or other critical data elements. Data accuracy, however, remains an elusive goal. As data management professionals know, the same customer or vendor record is often represented differently across multiple corporate systems, databases and applications because different employees have entered variations of a name or an address, each system uses its own naming convention and schema, and assigns its own unique identifiers. Complicating matters further, applications such as billing, service and order management are often replicated across product lines or business units, further fragmenting relevant data.
Is Master Data Management the Answer?
To address the need for more efficient and effective data management, many companies are considering master data management (MDM) solutions, which automate data integration across various systems and applications through the concept of a data hub. Depending on an organization's needs, this may include data matching, corporate or consumer householding, data governance and standardization across systems.
At its core, the concept of master data management is a good one. Using a central data store, or hub, companies can reconcile data across their heterogeneous systems and applications. Because the hub communicates bidirectionally, companies can provide integrated and accurate data to all their important enterprise applications while still maintaining a single version of the truth as transactional or analytical applications make changes or updates to the central data store.
However, aside from the concept, it's important to understand that there are multiple flavors of MDM. In evaluating MDM approaches, information architects should be aware of the different models of MDM deployment and be careful to choose the approach that best meets their data needs, time horizons and budget. Currently, there are two competing approaches to MDM: infrastructure-driven and solutions-driven.
Infrastructure-Driven MDM
Many MDM solutions today are infrastructure-driven, taking a bottom-up, enterprise-wide approach to data management. These solutions suggest an "uber" hub approach, seeking to consolidate all of the enterprise's data into a single, literal data store or federated model that can be accessed through a variety of transactional interfaces. When completed, the infrastructure-driven approach to MDM provides a great potential benefit - a single, atomic data store that eliminates confusion caused by different, duplicate representations of the same record across multiple systems.
The downside of this approach lies in the pragmatic implementation as it often takes millions of dollars and years to re-engineer existing systems. One of the biggest hurdles that IT faces in such an implementation is reconciling what the schema, hierarchy and extent of unified record will be. Today, because each is coded with a unique data schema in mind, each application has its own idea of what the ideal record structure should be and is written to maximize performance, usability, scalability and reliability based on that specific schema. In creating a literal master record, it is often pragmatically impossible to absorb and reconcile all of the metadata, columns/values and formats from all of the existing databases into a single "mega" record because of the size, complexity and performance constraints on the ensuing single database.
Using the infrastructure-driven approach, project teams must rationalize this problem by subsetting and appending the existing records to minimize the size and complexity of data. This dumbing down of the data into a lowest common denominator, however, loses rich information that is required by specific business units or functional organizations to function effectively.
As an example, a third-party logistics provider that is on an acquisition spree must constantly update customer master records and organizational trees to represent major customer contracts, spending and their logistics requirements in order to apply proper crediting, billing and service quality commitments. In order to enable future customer records to best match and align with existing customer data trees, the system must be able to maintain and learn from all of source customer records. The rich, contextual data from each original source record is critical to the continuous improvement of future match accuracy as new data is added from each new acquisition. The dumbing down of this information leads to lost associations that dilute the leverage value of each acquisition; it becomes impossible to create a "single face" to each major account customer across the various acquired companies.
Another issue is that the one-size-fits-all concept extends to how the employee or business unit can view the information. Because the goal is to store data in a single data schema in the infrastructure hub, there is often only one way to filter and view the information and its associated hierarchy. This generic view is often unrealistic and leads to poor use and loss of relevance of the master data. As an example, when examining master customer records, the support organization may want to view the severity of trouble tickets sorted by the affinity level of the customer rolled up in an organizational hierarchy. Finance, on the other hand, may want to look at the customers according to their days of sales outstanding, viewed by the value of their master (corporate) contracts.
Combined with other issues, implementing infrastructure-driven MDM, can take years to accomplish due to the need to re-engineer distributed applications and data to fit a single universal data model - and manage the internal politics required to resolve such a daunting task. This approach sounds good in the abstract, but is really not a practical one for most companies.
Solutions-Driven Master Data Management
A new class of suppliers is starting to deliver a different approach to master data management called solutions-driven MDM. Their goal is to gain the highest levels of data accuracy across applications but, unlike the infrastructure-driven MDM, to do so in a more cost-effective, nondisruptive manner. Solutions-driven MDM seeks to only integrate the data needed to meet the needs of an individual project and incrementally enhance the data hub on an as-needed basis as business usage expands. Typically, this approach does not attempt to physically consolidate the data, but instead creates a reference index and leaves the data where it is.
The solutions-driven approach to MDM brings several benefits. First, this approach, because it does not try to physically consolidate all of the master data across the enterprise, focuses on maximizing match accuracy to drive better and more accurate transactions and decisions. Also, since the hub can maintain the rich metadata of each individual application's records, there is no need to alter the data at source databases or applications. Instead, this "reference" type of model can "point" to each record, maintaining the vast majority of data context as a means for maximizing long- term match accuracy of new data elements. Similarly, the hierarchy of the data is flexible and can be represented in various forms optimizing its use and business ROI by adapting its hierarchy in a flexible way to meet the needs of different business users.
Second, the solutions-driven approach allows rapid initial ROI and proof of concept, making it easier to obtain funding and project buyoff. Because this type of master hub does not require re-engineering of applications and databases, there is less political baggage and fewer headaches than the infrastructure approach. It should not be implied, however, that this approach requires no cross-departmental collaboration, all master data models do. However, the solutions-driven approach significantly eases cross- departmental data politics, as finance and sales do not need to fight over which representation of a customer is the "right one" for all applications to use.
A drawback to this approach is that it is not a universal resource. When another project requires that the master record include another data source that is not currently captive in the solutions-driven hub, it must be incrementally added
Over the past decade, IT management professionals have explored a variety of techniques to manage customer data across the enterprise. However, without clear solutions on how to unify and make sense of this data, The Data Warehousing Institute estimates that companies are losing approximately $611 billion annually because of problems associated with bad customer data. [i] MDM presents the first comprehensive and realistic answer to automating and managing data across multiple systems. When choosing a MDM solution, however, enterprise decision-makers must understand that infrastructure-driven and solutions-driven MDM offer very different approaches to a company's data issues with greatly varying costs and time to value. Each provides opportunities for different business units to re-establish themselves with critical customers and/or suppliers or address distribution-facing business issues.
Reference:
1. Eckerson, Wayne W. "Data Quality and the Bottom Line," TDWI Reports, 2002, http://www.c4dq.com/readingroom/TDWI-DQReport.pdf
Bob Hagenau is the vice president of products and corporate development for Purisma, a D&B company. D&B is a source of commercial information and insight on businesses, and its Purisma solutions provide master data management platforms for enterprises.
