There can be few organisations of substance which have not understood, even internalised, the importance of data to their competitiveness. Simply stated, data is the lifeblood of any organisation, be it private or public sector, commercial or not-for-gain.
Given that this is the case, it is incumbent on all organisations to gain the deepest possible understanding of their data, and this should begin with an enterprise data modelling exercise.
Enterprise data modelling (EDM) is the practice of creating a graphical model of the data used by an organisation. Part of its value lies in the ability to visualise all the data in the organisation, in effect gaining a helicopter view of the data, the entities which generate it and consume it, and the relationships between them, and the subject areas, along with unified terms and definitions contained in a centrally held metadata dictionary or repository.
An EDM should be developed in conjunction with decision-makers, as they will be the ultimate customers of the exercise. It depicts the key information requirements of an organisation, such as accounts, products and customers, and yields a common view of the data, along with terms and data-related business rules.
Modelling can be viewed by some, and often is, as an academic exercise, so it needs strong-willed and responsible promotion, sponsorship and support if it is to rise above this charge and deliver long-term value.
Top priority
Modelling needs strong-willed and responsible promotion, sponsorship and support if it is to deliver long-term value.
Mervyn Mooi is director of Knowledge Integration Dynamics.
EDM is one of the first duties of a data steward, once such a person has been appointed in an organisation, and is closely aligned with the broader topic of data governance.
While there are various approaches (some proprietary, some generic) to EDM, all vendors and standards bodies agree that there are three levels of data modelling:
* Conceptual: These are high-level schematics of business entities and their relationships to subject matter. These are easily understood by business people.
* Logical: These are translated into one or more logical models that depict details of the subject matter in more technical terms.
* Physical: These are the actual database and environmental object specifications of code that is executed to create/configure the objects. At this level, the models are highly technical, and only for consumption by technologists such as DBAs and programmers.
Taken together, these levels form a total view of the data structure across an organisation. They have typically served to facilitate client/server applications which depend on a central database.
EDM has evolved to the point that it today represents data at all levels and across silos in organisations. The ultimate goal of the exercise is data integration and collaboration, with enhanced efficiency the ultimate benefit.
Overcoming disconnect
The EDM can help break down the traditional disconnect between business and IT, as the two camps are now conversing from a common set of definitions and reference models. They also make it far easier to accommodate change, including regulatory compliance initiatives such as Sarbanes-Oxley and Basel II, both of which have central data issues.
However, a caveat: the full benefits of data models will only be realised if the target market (developers, analysts and end-users) are not knowledgeable about data modelling concepts and best-practice data design principles. Furthermore, the access and approachability (ease of use/understanding/comprehension) of the models is a key factor in enabling the users. If this does not happen, people will turn to the technologists to explain the models in business terms, which undermines the overall value of the exercise.
Models and accompanying metadata must be well documented, and easy to extract, use and understand, preferably in a tool where dynamic viewing, browsing, searching, matching and analysis can be performed. A round of training on what the data models are and how to interpret them will also be necessary before the models are published.
In part two of this series, I will look at how to begin with EDM, along with its role in process modelling, and how it can ultimately become part of a broader enterprise architecture.
* Mervyn Mooi is director of Knowledge Integration Dynamics.
Share