Over the past few years, master data management has emerged as a key area of information management. The architectural debate was initially about whether to be data domain-specific or not, with vendors focused on providing customer or product-specific hubs. In the past couple of years, the industry has realized that organizations want a uniform approach to all their master data, and vendors have started to address this requirement. There is still a significant divide about the style of implementation, with some vendors specializing in certain areas, and some confusion has arisen around terms such as “operational MDM,” “registry” and “analytic MDM” used for various approaches to implementing MDM. Implementation approaches focus either on managing master data associated with business intelligence and reporting (termed analytic MDM) or managing master data associated with transactional systems (termed operational MDM). There has also been discussion of the use of federated approaches for MDM.

The Information Difference considers it important for both organizations and vendors to understand how MDM is generally being implemented so as to gain insight into the underlying reasons for the approaches selected and the available experience to date. We have conducted a survey aimed at gaining deeper insight into the views and plans of businesses regarding their current or planned MDM initiatives, focused on the styles and architectures adopted or planned to be implemented.

188 respondents completed the survey from all around the world, the majority from North America (59 percent) and Europe (20 percent). Most of the respondents were from companies having annual revenues greater than $1 billion and represented a wide spectrum of industries. The responses were split between two groups – those that had already adopted MDM and those planning to do so. The key findings from the survey are summarized below.

One-Third of Organizations Have MDM

Fully one-third of the organizations responding have already adopted MDM and a further 32 percent plan to do so within three years, as shown in Figure 1.

Types of Data Managed

There is a high diversity of data domain types in the two groups, with an average of five data types being/planned to be managed by MDM. These mostly include but are not limited to product and customer. Less than 15 percent in both groups were focused on a single data type.

Single or Multiple MDM Database Hub

Around two-thirds of organizations had implemented (or planned to implement) using a single hub/database for MDM. Surprisingly, a significant portion (20 percent for those already having MDM and 25 percent of those planning to implement) had opted for a federated MDM architecture – mostly following their organizational structure (line of business/business unit) rather than geography.

Scope of Master Data Managed

Encouragingly, almost two-thirds reported that their current or planned scope was enterprise-wide. This is illustrated in Figure 2.

Analytic or Operational MDM?

Of those who had already adopted MDM, 23 percent had adopted analytic MDM, 37 percent operational MDM and a further 33 percent both (see Figure 3).

A similar trend was found for those planning to implement MDM. More than half had implemented analytic MDM, based on the need to improve their management reporting.

Success was High

Reported success rates, shown in Figure 4, were high and respondents generally considered their implementations “somewhat successful” (60 percent for analytic MDM and 63 percent for operational MDM). Significantly around a quarter of respondents told us their implementations had been “very successful” (26 percent for analytic MDM and 31 percent for operational MDM).

Less than 6 percent reported that their implementations had had little or no effect.

Size of the MDM Hub

For approximately one-third of organizations, the size of the MDM hub was between 1 and 2 million records. Overall sizes reported ranged from 20,000 to 25 million records.

Architecture for MDM

There are broadly four styles of MDM systems:

  • Registry - the system contains pointers to where master data lives in operational systems,
  • Transaction - the system becomes the one and only source of master data,
  • Co-existence - the system contains master data where practical, with links to other master data sources where impractical, and
  • Consolidation - the system contains master data as copy. This style is suitable for analytical purposes.

The majority of those who already have MDM implementations (see Figure 5) elected to use the co-existence model (28 percent) closely followed by the consolidation model (22 percent). Surprisingly, as many as 17 percent had chosen the potentially more challenging transaction model. Among those planning to implement there was no clear preferred option.

Costs of Implementation and Maintenance

The average cost of implementation was about $7 million with a median of $3.5 million. The corresponding figures for annual maintenance of the MDM systems were a mean of eight full-time equivalents, with median value of 4.5 FTEs.

The full report is available for purchase from The Information Difference.