What is Analytical Data Modeling?

Introduction to Part X, Analytical Data Modeling

Now we’re on to analytical data modeling, something that many people have asked about. Here, I provide some thoughts on what makes analytical data modeling unique and how it differs from operational data modeling. This section might change depending on whether I think of anything else that’s missing. If you have suggestions for additions for this section, please add them in the comments.

In this part, I will not provide in-depth coverage of popular techniques like Kimball and Data Vault, as these are thoroughly covered in many excellent books and articles by the creators of these approaches. I urge everyone to read the classics; I’m merely guiding you to them. Instead, the coverage for analytical data modeling will be a mix of introducing the topic to beginners and providing context on where I think various approaches fit today for today’s Mixed Model Arts data modeler.

Thanks,

Joe

---

Now that we’ve seen various approaches to operational data modeling, let’s turn our attention to the next use case - analytics. The goal of analytical data modeling is to structure data for analysis and decision making. As I alluded to very early in this book, most data is consumed by machines - applications and ML/AI models. Humans consume data for analytics.

The data used in analytical data modeling most often, but not always, depends on data sourced from upstream applications. This data can be received in various ways, including being pushed via a stream, pulled from a database or API, or read from a file, among others. In the cases where analytical data does not derive from applications, it might come from other analytical sources. For now, we’ll assume that we’re dealing with data derived from applications.

Figure x-y: Application data (typically) feeds analytics

In this book, we’ll focus on how to model data for analytics, mostly using tables. Even though semi-structured data is more widely used, tables are still the primary format for structured data in analytics. And even when LLMs provide natural language interfaces, when it comes to analytics, what they query underneath is still often a table.

In the context of data modeling, I’ll use the terms operational and application interchangeably. How does analytical data modeling differ from operational data modeling? Let’s find out.