26 Jun The Art of Data Modelling
Data modelling is an essential tool when managing data. Data models allow businesses to understand, document, capture and manage their data and transform it from raw data into digestible insights. It helps to bridge the gap between business users and technical teams.
A well-designed data model is a blueprint for a data-driven future, so it’s worth putting in the effort to get all teams across it and unpack what you need your data models to deliver. And in the age of unstructured data, this is becoming increasingly difficult to achieve using traditional modelling techniques.
In this article, we’ll share the basics of data modelling and why it is such an art.
What is data modelling?
Data modelling is the process of creating a conceptual representation of data and its relationships. It’s an essential element within the Information Management disciplines helping an organisation understand how data is captured, stored and retrieved for (increasingly) complex data sets.
Data modelling can be done using diagrams, charts, or other visual tools that help to illustrate how data is organised and connected. The goal is to create a clear and concise representation of complex data structures, which can then be used to inform decision-making and improve business processes. By understanding data modelling, businesses can better manage their data and use it to gain insights that drive growth and success.
Why should businesses be investing in data modelling?
Data modelling is crucial for businesses because it helps to ensure that data is organised and structured in a way that makes sense (not just to technical team members but to everyone!) And in the age of self service, it’s becoming increasingly relevant for business users to understand that basics of data modelling due to the popularity of PowerBI and other tools.
Without data modelling, businesses may struggle to understand the relationships between different data points, which can make all the effort of extracting, managing and analysing data redundant. When data is effectively modelled for analytics and reporting, it can have incredible results for businesses, improving processes, cutting costs and making people’s jobs easier.
With effective data modelling, businesses can identify areas where data is missing or incomplete, allowing them to take steps to fill in the gaps and improve their overall data quality. This not only improves their compliance and makes data more secure and with better quality, it also leads to more accurate insights from which to draw conclusions.
What are the types of Data Modelling techniques?
Data modelling techniques typically fall into three categories: conceptual, logical, and physical models.
- Conceptual models provide a high-level view of the data and its relationships. This Data Model defines what the system contains and is designed to organise, scope and define concepts and rules.
- Logical models provide more detail and structure around how the system should be implemented. Business analysts and data architects will create these models to bridge the gap between raw data and technology and using the data in a business context.
- Physical models are the most detailed, showing how the data is stored and accessed in a specific system or database. This model is about implementation of the data model and is typically created by developers.
Each type of model serves a different purpose based on the format of data used, the stage of the data modelling process and the level of abstraction required. For data warehousing, there are industry leading standard techniques around modelling – Kimball and DataVault being very common approaches and philosophies. These are well suited to structured data.
For semi structured data (eg. JSON and XML), which was invented largely from web site traffic, it can be more complex trying to incorporate this into data models. Generally, we do not build data models to cater for unstructured data, as it’s impossible to apply a schema; for analytics purposes we need to use special tools to analyse this data.
Where do we start with the data modelling process?
The data modelling process typically involves several steps, including identifying data requirements, creating a conceptual model, refining the model through a technique called normalisation (removing redundant data), creating common naming conventions, and if required creating a logical and physical model.
Each step requires careful consideration and collaboration between stakeholders, including business analysts, data architects, and developers. By following a structured data modelling process, businesses can ensure that their data is accurately represented and can be easily accessed and analysed.
What are the Key benefits of data modelling?
Despite modern data models being complex due to many different types and sources of data, there are some critical benefits of data modelling that can’t be ignored:
- It provides a shared understanding of data and information between business users, stakeholders and technology teams, by creating common structures (eg. product hierarchies) with common names, separated by business function, therefore making it easier to bring this data together later on (e.g for summary reporting);
- If designed well, it can help to cater for future requirements to meet changing business objectives over time by structuring data from different places into a common shape and definition that everyone agrees on and can use (ie remove ambiguity and hidden rules out of the data sources);
- It can promote a unified view of data around specific data types (eg. products, customers, suppliers etc..) enabling organisations to better understand customer behaviours and improve the customer experience.
The effort to do this is a good investment for any data team to enable large groups of people to use data in a standardised way, making the data understandable and accessible, ultimately leading to improved reporting and analytics, and more timely and accessible insights.