Jammos Analytics

Today I want to discuss Facts and Dimensions and, more specifically, whether they’re still relevant in dimensional modelling.

Introduction

Dimensional modelling, a crucial technique in data warehousing, relies on two main parts: facts and dimensions. Facts contain numerical data, while dimensions provide context and categorisation for these facts. Traditionally, this method focused on making slim, denormalised structures for faster query performance. But with better computing power and storage, a question arises: are facts and dimensions going to stick around in dimensional modelling?

Dimensional Modelling – A Brief Overview

Dimensional modelling has long preferred keeping fact tables lean, with few columns, for quick query processing. This was to make data retrieval smooth and keep things efficient. However, as technology improves, the restrictions caused by hardware and storage space are becoming less of an issue. Modern systems handle more data and complex structures without losing speed.

The Emergence of Wider Fact Tables

Some experts suggest that wider fact tables, containing more attributes and metrics, could revolutionise reporting and analysis. They propose including related dimensions directly into these tables, reducing the need for complex joins with dimension tables during queries. Supporters believe this could simplify data models and queries for more efficient reporting.

In my own experience, across multiple roles, I have seen this shift take place. The emergence of technologies such as Delta Live Tables has meant that wide facts can be defined much more easily and, certainly when it comes to building dashboards, they can have much better performance.

The idea of wider fact tables challenges the traditional view of dimensional modelling, which emphasises keeping dimensions separate from facts to maintain data integrity and scalability. Critics argue that wider fact tables might create redundancy, increase storage needs, and possibly make data retrieval less flexible.

Summary

With technology advancing, an important question emerges: should dimensional modelling move towards wider fact tables or stick to the traditional lean and normalised structures?

We invite your thoughts on this ongoing discussion. Do you think wider fact tables are the future for better reporting in dimensional modelling? Or do you prefer keeping structures lean and normalised? How do you see the future of dimensional modelling amidst advancing technology?

Your insights are valuable in shaping this conversation. Please share your thoughts below and be part of the discussion on the future of facts and dimensions in dimensional modelling.