Software Modeling vs Data Modeling

Part of the chapter on Application Data Modeling

This is a chapter in Part 3 that dives into the technical approaches to creating and maintaining data models. This chapter - which I hope to have completed and published the draft this week - starts with data modeling for software applications. After that, the following chapters will look at data modeling for analytics, followed by machine learning/AI. The following chapters should be published throughout December.

This section¹ discusses the differences between modeling for software applications and data modeling. People often get confused about these two, and there are differences to be aware of.

As always, leave comments and feedback. Your thoughts are much appreciated.

Thanks,

Joe

---

Our first use case is data modeling for software applications. Good data models make applications more performant and reduce bugs. Getting the data model right at the application level also means that downstream consumers, depending on your data, have a viable foundation for their needs. This is an often-overlooked point in data modeling, and it has serious consequences for downstream users, whether the data is consumed for analytics or ML/AI.

It's critical to understand that modern applications no longer operate in isolation. In the past, and still common today, data would be lobbed over the wall from developers to business analysts and data scientists. Things are changing. We don’t live in a one-dimensional world anymore. Today and tomorrow's new workflows and systems feature complex feedback loops between applications, analytics, and machine learning use cases. For example, an e-commerce application might feed customer behavior data to a data warehouse used by business analysts or ML models that learn and improve the application's recommendation engine. This same data might flow back into the e-commerce application. Rinse and repeat.

It’s also becoming increasingly common for applications to incorporate different forms of data, whether structured, semi-structured, or unstructured. So, software engineers and application developers play a central role in ensuring the data they create is used downstream as productively as possible. We need to think of data as operating in a feedback loop in a symbiotic way, not just throwing it over the wall like we used to.

Before we discuss the details of modeling application data, let’s examine why modeling for software applications and data is intertwined but different.