Database Design : Data Model Quality

When we walked through the steps for creating a conceptual data model, we validated the model in the final step. We indicated a few of the tasks for validating the model. These tasks ensure that the model is of high quality. But ensuring data quality involves a lot more than indicated in that final step. The importance of high quality in a data model cannot be overemphasized. Further phases follow the data modeling phase for implementing a data system for an organization. If the model is inadequate and of poor quality, then the inadequacy will be propagated to all the phases that follow data modeling. We will examine the reasons for the need for high quality. We will explore quality dimensions as they relate to data models. We will study how to recognize a high-quality data model. Also, we will review the methods for ensuring high quality in a data model. In this section, we just want to introduce the concept of quality in a data model and catch a glimpse of the relevant topics.

Significance of Data Model Quality

Two basic concepts of quality are completeness and correctness. For a data model to be of high quality, it must be both complete and correct. Let us briefly examine these two concepts and see how they relate to the significance of data model quality.

Data Model Completeness.

When you scrutinize a data model for completeness, let us suppose you find that representations of some of the business objects are missing in the model. Consequently, you will also find that any direct relationships among these objects will also be missing. What is the result of lack of completeness ? To that extent, the data model will not truly represent the information requirements of the organization. Therefore, the final data system implemented based on the defective data model will not be able support the business of the company. Business processes that depend on the data about the missing objects and relationships cannot be performed.

Data Model Correctness.

Similarly, let us suppose that the attributes of an object shown in the data model are wrong. Also, assume that two of the relationships are shown in the data model with erroneous cardinality indicators. To that extent, the data model represents the information requirements incorrectly. These errors will filter through to the final data system and will affect the corresponding business processes.

Data Model Characteristics

What makes a data model to be of high quality ? When can we say that a data model is good and adequate ? Can we specify any general characteristics for a high-quality data model ? Let us explore some of these features.

Involves Users.

Unless the relevant users are completely involved during the process of data modeling, the resulting model cannot be good and valuable. The domain experts need to provide continuous input. While reviewing business operations for the purpose of identifying the right business objects, the involvement of the users with appropriate expertise in the particular business domain is absolutely necessary. Also, the right stakeholders must participate in the process. At every iteration in the modeling process, the data model will be used as a means of communication with the domain experts and stakeholders. The input from these users will enable the data modeler to refine the model as it is being created. With this kind of close participation, the data model is expected to be of high data quality.

Covers the Proper Enterprise Segments.

If the goal is to represent the information requirements of the entire enterprise, then your data model must be comprehensive to include all the business processes of the whole enterprise. In this case, the final data system built based on the comprehensive model will be of use for all the users. In practice, however, unless the enterprise is of small to medium size, all information requirements will not come within the scope of the data model. The data model will be created to cover only those enterprise segments of immediate interest. In a large company, it is possible to start with a data system to support the functions of only a few divisions such as marketing and finance. Then the data model will represent the information requirements to support only the business processes of marketing and finance. Here, the emphasis is on knowing what to include and what not to include so that the data model will be correct as well as complete.

Uses Accepted Standard Rules and Conventions.

In the previous section when we reviewed the components of a data model and walked through the steps for creating a conceptual data model, we improvised and used our own simple set of symbols. For the purpose of introducing the data modeling process, these symbols and conventions were sufficient. However, if you showed the data model diagram to someone else, that person may not understand the representations. This is because the symbols and conventions are not an accepted standard. To this extent, our data model is not of high quality. A good data model must be governed by standard rules and diagramming conventions. Only if you use industry-accepted standards can your data model be good and universal. We will introduce some modeling techniques toward the end of this chapter.

Produces High-Quality Design.

One of the primary goals of data modeling is to produce a good blueprint for the final database system of the organization. The completeness and correctness of the blueprint are essential for a successful implementation. A poor data model cannot serve as an effective blueprint. For a data model to be considered a high-quality model, you must be able to complete the design phase effectively and produce an excellent end product. Otherwise, the model lacks quality.

Ensuring Data Model Quality.

The importance of data model quality necessitates measures to ensure the quality. A poorquality data model results in a poor-quality data system. Quality control must be given a high priority in the whole modeling process. Let us just mention how quality considerations must be approached and also a few quality control methods.

Approach to Data Model Quality.

At every step of the data modeling process, you must review and ensure that the completed data model will truly serve each of its two major purposes. Is the data model clear, complete, and accurate to serve as an effective communication tool ? Can the data model be used as a good working blueprint for the data system ? The data model must be reviewed for clarity, completeness, and accuracy at every stage of its creation. Quality control comprises three distinct tasks : review, detection, and fixing. Every step of the way, the model must be reviewed for quality control. There must be techniques and tools for detecting problems with quality. Once quality problems are detected, they must be fixed forthwith. The data modeling team must develop and use proper methods to fix the quality problems.

Quality Control Methods.

Quality control methods include the three tasks of review, detection, and repair. Usually, these tasks are performed in two ways. First, the tasks are performed continuously at every modeling step. In this way, less problems are likely to surface at the end of the modeling process. Second, when the modeling is complete, the overall model is again reviewed for any residual quality problems. When any residual problems are detected at this stage, they are fixed to assure high quality for the complete data model. Who performs the quality control functions ? A good approach is to share these functions. In the review and detection tasks, the data modeling team and the users must work cooperatively. Fixing of quality problems is generally the responsibility of the data modelers.

Source : Ponniah, Paulraj (2007) Data Modeling Fundamentals : A Practical Guide for IT Professionals, Wiley