In my last few columns, I have focused on technical implementations of object representation; therefore, this month, I'd like to look at how an organization's long-term information strategy affects how data integration projects are performed.

Presented with a collection of legacy data sets for the purpose of enterprise data integration, an information engineer may opt to proceed based on two different philosophies, which for convenience sake I will refer to as "pragmatic" versus "holistic." While either approach should have the side effect of integrating the data, the selected philosophy will have a fundamental effect on the planning, execution and long-term strategic value of the integration project.

The performance target of the pragmatic approach is the same as the side effect: integrated data. Because the goal is to take the legacy data sets and effectively merge them together, the practitioner will probably apply a relatively well-worn process:

• Understand business needs of integration project,
• Evolve "to-be" data model,
• Profile the legacy data,
• Resolve and consolidate frequently used reference data,
• Determine relational structure embedded within legacy data,
• Document source and target meta data,
• Develop transformation routines,
• Populate target data model.

Of course, I have simplified this process by boiling it down to a handful of bullet points. Anyone that has worked on these kinds of projects knows that there are significant challenges in this process, and one may encounter a number of failures before getting it right. And even after the data has been integrated, there remain lingering questions about quality, consistency with previous applications using the data and usefulness of any new system depending on the migrated information.

Alternatively, the holistic approach seeks to understand the business use and value of the information asset as embodied by the data, the structural meta data and an additional layer of meta data that consolidates meaning and intent embedded within the structure and use of the data. One might refer to the latter as "meta knowledge" - information about the knowledge embodied in the data. The holistic approach might apply this kind of process:

• Understand the fundamental entities that interact within the business,
• Evolve a strategy for representation of those entities that can be applied in both operational and analytical/intelligent applications,
• Profile the legacy data,
• Identify key reference data objects and document their application uses and business needs,
• Determine connectivity model between legacy entities and capture business meaning inherent in application use,
• Document source and target meta data,
• Document business rules governing the use of the legacy data and the needs of the "to-be" system,
• Develop transformation routines,
• Facilitate new information use through information consolidation.

I have presented two high-level approaches, which basically do the same thing. The difference lies in practitioner intent for exploiting what you have learned during the integration process. In the pragmatic approach, because the goal is to bring the legacy data sets together, when that goal is reached, the process is finished. In the holistic approach, the goal is to understand how using information adds value across the enterprise, and integrating the data sets is achieved as a by-product.

Is one approach better than the other? It depends on the strategy that your organization has adopted around data. In businesses that rely on data as part of the day-to-day operations, failures in operational processing erode the business, and if there is no significant need for business intelligence, then the pragmatic approach provides the optimal way to achieve business objectives.

On the other hand, an organization that expects to derive synergistic value through information integration may prefer the holistic approach in that it presents opportunities for improving the business. However, the key individuals in that organization must be prepared to justify the effort and resources required to effectively capture and internalize the discovered meta knowledge. This means being able to articulate concepts of information value to senior managers and gain their active sponsorship (as opposed to passive consent). In addition, one must also justify the acquisition of the components to support a meta knowledge infrastructure. That includes, but is not restricted to data profiling, meta data repositories, semantic and ontological meta data organization and business rules management systems, to name a few.

Lastly, and probably most important, is the existence of an information strategy that looks at past use, current needs and future utility of information as a fundamental corporate asset. In this context, a forward-thinking senior manager can establish an information value proposition that persists through and thereby facilitates evolution in information support of the business environment.

...............................................................................