Enterprise Data Modeling

Introduction

The formalized practice of data modeling has occurred since the 1960s and has grown steadily in importance ever since. Today it is commonly utilized by IT professionals to identify the requirements necessary for handling data for the purposes of better supporting the business objects of an organization.

Just like how when moving into a new house, individuals tend to map out where to place furniture, electronics, etc., data modeling minimizes the difficulty of adapting to new environments and simplifies decision-making in complex situations.

As such, data modeling has become an integral part of maintaining the IT landscape and ensuring an efficient means of storing and analyzing data.

What is enterprise data modeling?

Data modeling is a process for conceptualizing the relationships between different types of information in an organization, independent of the organization’s structure, processes, people, or domains.

Data models are a representation of data objects and the relationships between those objects which helps users across disciplines store and interact with data more effectively for a variety of use cases.

This visual guide helps when performing data governance and creating data policies. Data modeling is a way to help organizations become more data-driven. 

Source: LeanIX GmbH

IT professionals design data model structures based on the actual ways in which IT entities, personnel, and business capabilities interface with one another. Such objects become the main categories, or boxes, in the model. These objects are all interconnected, and the connections (or relationships) between the items are used to visualize data and guide policies governing this data.

When creating a data model to represent the infrastructure of an information system, it is important to make the models logical and easily understandable for those requiring insights on data objects in relation to their business needs.

Overall, the process of data modeling entails defining the attributes of all data objects and connecting the relationships between the different types of information that need to be stored. This map, or diagram, helps IT professionals understand what key data needs to be stored and easily retrieved.

Why should you use data modeling?

Data models reflect data that are absolutely essential for a business's continuing operation. Its structure helps align databases across the physical, conceptual, and logical levels. The primary goals of using data modeling are:

• Users can take advantage of having this information neatly structured to identify technical and functional overlap, foster business intelligence, and optimize how data are organized.
  
  
• Controlling how and where data interact in either a server or cloud environment is crucial for implementing systems that equally benefit business and IT professionals.
  
  
• A data model can be used to validate the technical and functional benefits of current and future data objects while also revealing if databases are correctly represented.

Thanks to easy-to-understand representations of the underlying data, it is particularly helpful for developers when creating a physical database (e.g., missing or redundant data can be easily spotted to save time for developers).

If data is not accurately represented, there is a greater likelihood of false outputs from analytics reports and miscalculated strategic decisions.

Though it is easy to become overwhelmed by manual documentation efforts when outlining a data model, the efforts are invaluable when upgrading infrastructure.

Benefits of data modeling

There are benefits to using data models of all types.

1. The first of which is being able to reliably ensure that data objects in an IT landscape are correctly represented.
2. This information can then be utilized to define connections between primary and foreign keys, tables, and procedures.
3. A data model can then be used to build a physical database if sufficiently detailed.
4. Data models can also be leveraged to communicate to business stakeholders throughout organizations.
5. Locating accurate sources of data to auto-fill the model.

Challenges of data modeling

Unfortunately, there are also some challenges when using data models. In order to effectively create a data model:

1. The creator should have a firm understanding of the characteristics of the data that is already physically stored.
2. A data model is also a system that can result in complex application development, thereby making these processes difficult to manage as well.
3. All changes to the data model, both large and small, require developers to modify the entire application system.

Types of enterprise data models

Organizations can benefit from three specific types of data models depending on the information needing to be delivered. The three different types of data models are conceptual, logical, and physical.

Conceptual data models

Conceptual models reflect high-level and static business structures. In most cases, they are only generalized representations highlighting which business objects are involved in an information system.

For theorizing new solutions and efficiently organizing rules, a conceptual model should be employed. This model is commonly used by data architects and stakeholders.

Logical data models

Logical data models focus on data attributes, IT entity types, and relationships between the entities. A logical data model is useful for understanding the nature and compositions of data but not its actual implementation.

They are commonly used by business analysts and data architects to help develop a database management system, a technical map of structures, and rules for the model.

Physical data models

Physical data modes cover aspects related to the design and implementation of databases. These cover the structure of databases, including all relational databases and objects.

They are typically employed by developers and database analysts to show the execution of a structure with the use of a database management system.

Choosing the correct type of data model for an organization rests on knowing the specific needs of a business. However, significant attention must be placed on the variety of stakeholder preferences involved in building a working data model.

Data science professionals, for example, are likely to want models offering full visual views — the likes of which provided with physical and logical data models. Conversely, business representatives interested more in outcomes rather than technical details are likely to select a conceptual data model.

Enterprise data modeling techniques

There are three fundamental data modeling techniques: Entity Relationship Diagrams (ERDs), Unified Modeling Language Diagrams (UMLs), and Data Dictionaries.

1. An Entity Relationship Diagram is the default technique for data modeling and works especially well when modeling tabular data. This technique involves making graphical representations of data objects alongside their attributes and relationships.
   ERDs are very useful when designing traditional and Excel-based databases. They are also ideal for securing clear visuals of database schemas, along with top-level data.
   
   
2. Unified Modeling Language encompasses a series of notations for designing and modeling information structures. Used by many as a general-purpose software notation, UMLs reflect either the behavior or structure of data objects and employ different diagram types for doing so. One of these diagrams is a class diagram, which relates to defining the classes, methods, and attributes of databases.
   
   
3. Data dictionaries are based on a tabular definition of data assets. This is a grouping of tables and data sets with an accompanying list of attributes and columns. Other optional sections of a data dictionary are item descriptions, additional constraints, and relationships between columns and tables.

Enterprise data modeling best practices

1. Don’t create redundancies: Good data objects do not overlap; they are mutually exclusive. A good test is to check whether you can assign Level 2 data objects without any ambiguity.
   
   
2. Rely on business capabilities: It is very easy to find which data objects exist once you have mapped your business capabilities. This is why we recommend first creating a business capability map.
   
   
3. Long-term stability: Properly defined data objects are fairly stable over time, persisting throughout any organizational changes. Only major business changes should affect them.
   
   
4. Cross-organizational: Don’t get too specific. Data objects should remain the same, independent of any changes that might happen to the organizational structure.
   
   
5. Use existing data models: Many applications (e.g., SAP) will already have existing data object models. Familiarize yourself with these when creating your own map.
   
   
6. Breadth rather than depth: While more levels can help to get a better structure, it comes at the cost of increased complexity. Go for breadth and build your map with no more than three levels.
   
   
7. Involve relevant parties: Leverage insights from representatives throughout the business. Those responsible for different parts of the business are likely to have the best overviews of data objects. Consider using surveys to collect information.

Data modeling and the LeanIX EAM platform

Using the configurable solution from LeanIX as an example, objects in an enterprise architecture data model can include:

• Provider
• IT Component
• Tech Category
• Application
• Interface
• Project
• Data Object
• User Groups
• Business Capability
• Process

Conclusion

Enterprise data modeling is essential for standardizing organizational assets and optimizing information systems. Though the practice has occurred in various forms for many years, its importance has grown exponentially in the present era of DevOps.

The process of data modeling helps IT professionals define data requirements to support the business objects of an organization. To learn more about data modeling with and at LeanIX, here is information on our flexible data model.