Extending data modeling to cover the whole enterprise

~ ata modeling has been valuable for database design. Much research has focused on improving tlhe expressive capabilities of data models [3], on clustering [17] and schema integration [2]. Data modeling techniques [9] have become an integral part of software engineering methodologies [8] and have strongly influenced methods for object-oriented analysis and design [11, 15]. Corporate reality is different, however. In many enterprises , data modeling is applied only on a project-to-project basis and data management struggles to keep track of more than 50,000 data elements accumulated over years. Redundancy factors often exceed 10. In an increasingly complex EDP environment this obstructs information access and s.eriously impedes software maintenance. These problems have led several companies and the authors to design enterprise-wide data models (EDMs). Besides the goal to improve the requirements specification for the design of application software and to define a suitable database structure , there are also organizational reasons for the design of an EDM which are of equal importance: EDMs integrate the business views of different departments within an enterprise. Business terms utilized are standardized, facilitating communication among departments. Work flow can be improved if functions that relate to the same class of information are integrated or better coordinated. Links between different departments can be identified and processes can be restructured accordingly. An EDM, which is documented in a repository , is a "better version" of an organization handbook and can be used to increase availability and distribution of knowledge about the enterprise itself, a factor which has been neglected for long. This article is organized as follows: first it is shown that EDMs can differ in level of detail, and the purpose of EDMs is discussed on each level. Then a method fbr the design of EDMs is proposed and finally implications for data model-ing methods are outlined. Effort for EDM design and the benefits of EDMs vary greatly with the size of an EDM. Three different model types can be distinguished: A macro-level enterprise-wide data model represents the concepts which are strategically important to the enterprise and which influence all major business activities. It facilitates communication with high-level managers to pinpoint integration needs, to identify organizational weaknesses and can also be used for navigation and orientation in larger data models. A macro-level model is very stable over time and should be part of every organization handbook. An example of a macro-level data model which has …