Three elements influence the effectiveness of an enterprise application: The quality of the code, the user experience (UX) and the data model. This last element is not only the responsibility of the IT department but must be co-designed by relevant stakeholders to achieve an optimal result.
We explain what is behind a data model, what significance it plays for internal processes and which types of models companies should use in which cases.
Definition: What Is a Data Model?
In the literature, there are various definitions of the term data model, some of which are very detailed. In our view, a simpler, pragmatic description that ensures a common basis of understanding is sufficient for everyday business.
A data model, also database schema or data schema, describes current or planned data structures for a use case in the company. Data and their relationships to one another are visualized, whereby the visualization can take place from a technical or business perspective and include IT systems, employees and workflows.
What Is the Benefit of a Data Model?
Improved Decision Quality
Data modeling, i.e. the creation of a data model, is an essential step in software development and a prerequisite for optimal utilization of data in companies. After all, the quality of decisions based on analysis and BI tools is ultimately also based on the quality of the data models used.
Data models provide a static overview of the data and data relationships in an area of use. Particularly in complex use cases, it is helpful to obtain these before the actual coding. If IT and business departments discuss data models in advance, corrections in the development process can be minimized. Data modeling can thus reduce programming costs by 60 to 75 percent.
Data models make it easier to network the data sources in a company in such a way that technical capacities are used efficiently. This reduces the costs for hardware and storage space. At the same time, the IT infrastructure becomes clearer, so that maintenance can be carried out faster and more cost-effectively.
A semantic data model creates a basis for communication for all stakeholders in the project. Creating it, forces employees to structure their ideas clearly and comprehensibly. Stakeholders use a common vocabulary when referring to the data model for requirements and change requests.
High Data Quality
Data models form a reference point for quality assurance, so that an as-is/target comparison between planning and the current implementation status can be made at any time. Since they provide an overview of the data structures in the company, they help to plan effective data integration processes.
Data Model Types
Data model is a generic term. Depending on the project phase and department, different methods are used for data modeling.
Conceptual data model / semantic data model
This type of data model is usually created first and by business stakeholders to develop business processes. Using text and graphics, the model describes real-world objects and their properties and relates these business objects to each other. For example, a product such as a hydraulic pump would be listed as an object, and the product category, weight, and color would be listed as properties.
Logical data model
The logical data model is created from a technical perspective. It extends the conceptual data model with information, for example on data formats, attributes and key properties. Thus, the data model can be transferred to a vendor-independent database management system (DBMS).
Physical data model
For use in operations, the logical data model must be extended and formatted in a vendor-specific manner, depending on the DBMS being used. In the relational data model, for example, the data must be assigned to tables and columns and authorizations, triggers and views must be created. The transfer and adaptation of the logical data model to the syntax of the respective DBMS can be partially automated.
Create a Data Model: How Should Companies Proceed?
Which methods and technologies companies use to create a data model depends on the application purpose and the existing IT infrastructure. However, some basic principles have always proven themselves.
A top-down approach is always recommended. The technology is subordinated to the application scenario. In terms of data modeling, it means that business decision-makers first create a semantic data model, followed by the IT department developing the logical data model. This is coordinated, released and translated into a physical data model.
Ideally, a data model is created incrementally, in successive iterations, whereby technical and business stakeholders should be in close contact with each other. Mistakes made in data modeling result in considerable costs.
Companies should ensure that their data modeling tool can interface with their existing systems as easily as possible. Future company developments and possibly increasing requirements for data modeling should also be taken into account. Ideally, solutions should be easily scalable or modularly expandable.
The use of versioning is also recommended. Many data modeling tools offer a history as standard, so that users can track changes and their effects down to the business object level. Database versioning not only helps with troubleshooting, but also speeds up problem solving many times over and contributes to transparency in data management.
Since data models lay the foundation for databases, the quality of the data model determines the value companies can derive from their data. For this reason, and because the lifespan of data models extends far beyond that of individual applications, companies should devote sufficient time and resources to their design and implementation.
Frequently Asked Questions About Data Model
Simple, semantic and graphically prepared data models can be created with standard software such as Microsoft PowerPoint or Word. However, quality assurance is hardly possible here and IT cannot reuse the model. Therefore, anyone who wants to create data models should use specialized tools in which export and automation functions are integrated.
The relational data model is the most widely used model for structuring databases. Relational databases consist of a large number of tables in which data is stored. Since the tables can access each other, redundancy-free data storage becomes possible and companies can save storage capacity. However, non-relational databases, in which data is not organized in rows and tables, but contextually, are being used more and more frequently, as they are more agile and better suited to new technologies such as artificial intelligence.
The term Semantic comes from the ancient Greek for meaning. A semantic data model thus focuses on the description of data and data relationships at the user level. The term is used synonymously with the conceptual data model. In contrast to the semantic data model, the logical data model is dedicated to the syntax, the structures and rules according to which the data are to be organized at the technical level.
Data success is plannable