Enterprise-wide data model

New technologies are not a universal remedy: ways to achieve an enterprise-wide data model

Today almost all enterprises are fighting against a profusion of data, simultaneously suffering from a lack of useful information. Applications have grown isolated and exist in their own more or less well-documented data and file world. An important task of information management is to convert the multitude of data into a manageable amount of significant information. "Information as a resource" has integrated itself in the series of terms that have become common knowledge for data users. This keyword is commonly used and everyone now considers information to be of equal importance to the classical production factors capital, human resources and plant. Information management is an old hat which has finally been recognized and allocated its own organizational unit.

The persons appointed the responsibility for this information management are those who have so far been responsible for information systems: the DP or Organizational Manager. As an additional admonition, these managers are then required by general management to also consider old data as a new resource, and treat it with the corresponding diligence.

This viewpoint may be exaggerated, however the impression is given in many enterprises that by appointing an Information Manager, enough has been done to keep up with the new trend, and it is now possible to return to day-to-day business with responsibilities for:

Hardware and software selection and implementation,
Design of a hardware and software architecture for centralized and decentralized applications,
Provision of the infrastructure for information users in the various enterprise sectors,
Maintenance of standards and procedures.

But is that really all that information management needs to do? It is indisputable that the strategic direction of Information Technology is a considerable complex task of information management, the tasks mentioned above having become considerably more complex than they ever were.

Information management has lost its way in the data-processing jungle. The technical range, with its overabundance of possibilities, has not just become more extensive and complex, but has also brought with it compatibility and integration problems due to the lack of standardization; just consider the range of different network types, communication technologies, CIM products.

It’s no wonder that information management can these days easily err in the data-processing jungle. But let’s assume that the IT-technical world was different: strategically concise, tidier, clearly structured and without any technical problems.

What would then stop the enterprise from finally being able to fully utilize the longed-for possibilities to exchange all information as desired?

Everyone could then:

within the realms of his authentication, independently
use and alter others’ information, create new information and make it available to others?

What is stopping them? This picture might be enticing, but unfortunately extremely deceptive. Because even the most perfect technology cannot hide the fact that, although bits and bytes can be distributed as wished, their information content could still continue to be unknown, or at least be misinterpretable.

By now it should be clear, that today’s information management insufficiently fulfils the fundamental tasks of tomorrow:

Information planning and information strategy
Design of an application architecture
Planning software applications

These three fields of responsibility are closely linked together, as an expedient planning strategy of individual software applications needs to be based on a previously compiled applications architecture, designed for the future.

The application architecture itself will need to be based on the results of the information plan and strategy, so that this task can be regarded as, in the long-term, the central logical basis.

The following remarks will therefore be confined to this basic function. There are two aspects to information planning. It demands firstly that you deal with the information itself - specifically and in detail. And it needs the managerial functions that create and process the information. However the lynchpin remains the information itself.

We are still confused – but on a global scale

So, initially the information is in the foreground. Information cannot be classified as such, until the data has been complemented by its semantic content, i.e. its meaning, thus becoming interpretable. However the current situation in most enterprises still predominantly mirrors the conventional picture of data processing and not that of targeted information processing. Applications systems that have grown isolated exist in their own world, where no one system is aware of the other, and which, at best, are only able to communicate via elaborate interfaces.

Data communication demands a common data appreciation though. However homonyms (terms with the same name but a different meaning) and synonyms (terms with different names but the same meaning) have become the order of the day in both application systems as well as in individual departments.

Applications, whose job it is to compile summaries and analyses, composed from base data from different operative systems for planning purposes or even as a tool to support enterprise decision making, find it extremely difficult to deliver reliable results. Reliability can only be achieved, when it can be assured that the base data do not just have the same name, but also the same meaning.

As clear definitions and descriptions for the data meaning are still missing in many enterprises, it is right to doubt the informational value of many an analysis or report. This situation cannot however be improved by implementation of new technology, which serves no other purpose than to distribute the dubious data more quickly.

New technologies alone may even make this problem worse, by ingeniously helping to expand localized chaotic situations into global ones, based on the principle, "We are still confused, but on a global scale".

Structuring data comprehensively and usefully

One of the most important tasks of information management is therefore to transform the multitude of existing data into a manageable quantity of meaningful information, in a structure that is both comprehensible and therefore usable for all information users.

This structure is the well-known data model. A data model is an illustration of the enterprise’s information (or parts thereof) and their interrelations from a purely managerial point of view, independent of how they might be realized in the data-processing world. These days the importance of such an enterprise-wide data model is almost indisputable and its design and maintenance should be a task for data management, which is an integral constituent of information management.

Unfortunately in reality, surprisingly few enterprises dare to venture the construction of such a model. One the reasons for this appears to be fear of the word "enterprise-wide", as it gives the impression of an impossibly huge and insurmountable task.

But there are in fact realistic and viable ways by which "enterprise-wide" can be approached step by step, without having the rug pulled out from under your feet. One of these methods leads to what should here be called "enterprise-wide data model", the other leads to the resulting "enterprise data model".

The construction of both models is based on the same theoretically established and empirically tested method, that of the data model, which however will not be gone into detail here. Both models differ in their aim and, more than anything else, in their level of detail. Both models should enable information planning and information utilization globally across all projects, nevertheless each with a somewhat different specificity.

The enterprise-wide data model

The enterprise-wide data model corresponds to today’s current established data model, and has the certainly extremely ambitious aim to achieve the following:

A complete base of all information that the enterprise has to offer (including a professional data catalog), which is able to serve both as a detailed fundament of information and communication between departments, and aid with data processing.
To provide a specification from which database structures can then be derived.
To keep project interfaces small.

How is it possible to meet these high demands? Such a detailed data model cannot realistically be achieved in one simple step, but needs to be constructed from many small sub- data models. Each single partial model results from a project, which applies methodical data analysis. Each project creates a project-related data model, confined to its own informational area. The terms and concepts used in this data model however need to be clearly defined and be valid for the total enterprise.

The enterprise-wide data model evolves from the bottom up, arising from the union of the single project results into one consolidated structure.

Practice shows that this method has the following advantages:

Each project recognizes the benefits of data analysis itself as an aid.
The resulting "project data model" can be utilized immediately.
The project result has been achieved to the great level of detail required, yet with a manageable amount of effort.

Problems arise however with this method when consolidating the partial models. It often becomes apparent at the interface of two projects, that the supposed enterprise-wide denomination and definition of the data is only actually fully valid from the limited project viewpoint, and now needs to be synchronized with the other projects. Information streaming increases project effectiveness.

This fine-tuning can be an elaborate process, which also in addition needs to take into account the human factor, namely the danger of those involved mistaking their own contributions and efforts as their property.

The process is also elaborate, because alteration to names and structures could have an effect on the results of other projects (e.g. functional flow descriptions), and other projects may need to adapt their results accordingly.

It is only possible to minimize this project-related annoyance if:

Each sub-project is adequately informed of the enterprise’s strategy with regard to mutual information, and feels sufficiently obliged to comply.
Each project is kept informed right from the beginning of the results of previous or progress of current projects, and is able to use these actively, thereby saving expenditure, and even more importantly, effort.

This method produces immediate results, as even the initial results of the first project are a step towards information organization, without which information management is powerless in the long-term.

However the enterprise-wide data model cannot be used as a basis for information planning until at least two years later, as it takes this long for the results of the individual projects to be delivered, quality-controlled and synchronized with each other.

The enterprise data model however demonstrates its benefits rapidly, because it is constructed as an independent assignment, detached from other projects and with a different target: that of the enterprise data model.

The enterprise data model primarily follows a different target direction to the enterprise-wide data model. It does not aim to achieve a detailed data catalog and will never represent a complete base of all data in the enterprise.

In contrast it should:

Provide an summary of the enterprise’s information at top level,
Recognize information supply areas ("theme" databases), according to which this information can grouped and summarized,
Be a decision aid, enabling projects to be defined more precisely at their initiation, and last but not least,
Produce a result with which enterprise management can demonstrate to all information users that they take the term "information" seriously.

These goals of a collective consolidated structural summary of the enterprise cannot be attained by joining the individual project results, bottom-up, and then integrating them upwards. An enterprise data model can only be developed as an independent and self-contained project, with participation of all management levels, as summary and not detail information is required here. By management we mean the specialist departments and sectors, as it is only here that data can be defined from a managerial point-of-view.

Data collation is achieved through interviews relevant for the level concerned and related professionally and technically.

Its goes without saying that the definition and description of this data will have a different quality to that of the enterprise-wide data model. In the enterprise data model relationships are identified clearly and precisely, always with the aim of simplification and rough abstraction. (Keep in mind that the objective here should not be a constant information refinement down to the last detail (i.e. an endless top-down process), but the construction of an approximate summary model that can continue to be maintained at this crude level.)

Project model with clear task definition

As the enterprise model is a self-contained investment, without direct implementation into a data application, its initial construction should be completed within a maximum 6 month time frame. The model can then be used immediately at project initiation.

The enterprise data model can be made immediately available and passed on to projects, along with a clear definition of which of the subject areas described belong to the project objectives. Project boundaries can be defined and established in relationship to each other right from the start, and subproject intersections can at least be fixed at a rough level.

This anticipatory description of intersections considerably reduces later investment for project adjustments, as each project knows its own "data limits", and is also aware from the start which other information management project partners he will need to confer with for the fine-tuning phase.

The results of this project refinement are not included in the enterprise data model, but grow together to form their own independent enterprise-wide data model.

The enterprise-wide data model and enterprise data model are therefore not "either-or" models, but are, in the true sense of the word, "as-well-as" complements. But what do both these models have to do with future-oriented information planning and information strategy, are they not managed by data administration?

The problem today lies in the term administration, which has little to do with management and consequently with strategy and planning. Many enterprises have started constructing an enterprise-wide data model, but its use is still mainly limited to the unification of terms and information correlations. Compared to the alternative of prevailing data chaos this limitation is however still a huge step forward, which itself is worth a certain amount of effort.

However data and information planning require more, and open up in the long-term other perspectives. Information planning should achieve the goal of comparing the enterprise’s current information supply situation with future visions and to assess them, in order to recognize supply deficits and to be able to simulate and optimize future supply situations. This however assumes that not only the enterprise’s information is known, but also the functions that are connected to the use of this information. Principally only the comparison of a data model with the functions or functional areas that are necessary for the improvement or extensions of strategically important business areas, allows informational gaps and superfluous informational ballast to be detected. But what does this mean, in view of both the above-mentioned types of data model?

The enterprise-wide data model is only suitable for detail planning aspects, because of its level of detail; that is, when dealing with the concrete definition for contained single subject areas. In contrast, this model is unsuitable for planning or strategic aspects. And yet it is the only model, in many cases, that is (at least in part) used by information management.

Foundation for theme databases

But who begins with strategic management tasks in other enterprise areas at the operative level, such as the Internal Revenue? Information management is often degraded to the level of dealing with nothing other than the daily business (enterprise-wide data model), instead of setting the basis for management tasks: the enterprise data model.

The enterprise data model offers management (and not just information management), for the first time and within a short period of time, a defined basis for their own information and informational areas (theme databases) which is comprehensible to all.

This model allows a meaningful and clear classification of information for the enterprise’s functional areas and organizational units. It supports the construction of a more comprehensive model, with which a conscious design and simulation of future information management is made possible.

And there is one more advantage: the enterprise data model includes business management, in its target-oriented way of thinking, in information and informational correlations right from the beginning. Data model and information management become accessible to all concerned, becoming today’s obligation to go on the "search for tomorrow’s information".

back to top of page
<< Database design and database normalization | Database technology articles | Firebird databases as the back-end to enterprise software systems >>