Why Data Modelling Still Matters - More Than Ever

In today’s fast-paced landscape, where agile development and cloud-native platforms dominate, data modelling might seem like a relic of the past. But for anyone building business-critical applications or modern analytics platforms, it remains a cornerstone of success.

Modelling for OLTP: The Backbone of Business Applications

Online Transaction Processing (OLTP) systems power the day-to-day operations of businesses, from CRM platforms to e-commerce engines. These systems demand high performance, reliability, and scalability. A well-structured data model ensures:

• Efficient queries and indexing
• Reduced redundancy
• Lower cloud costs

Without proper modelling, databases can become inefficient, leading to performance bottlenecks and increased operational expenses. The physical model must reflect business logic, even if that logic is implemented in code. Otherwise, the database becomes a black box hard to maintain and expensive to scale.

Agile Development: Who Owns the Model Now?

Traditionally, data modelling was the domain of architects and DBAs. They crafted conceptual, logical, and physical models in a waterfall sequence. But agile methodologies have, in my view, shifted the responsibility for the physical model to the developer/engineer.

Today, developers often modify schemas directly, bypassing formal modelling processes. While conceptual and logical models may still be created, they’re rarely updated as development progresses. This leads to “model drift” a disconnect between design intent and implementation. The more I think about this the more I see the conceptual and logical model being owned by the architect or DBA and acting as part of the specification or user story. The developer building the application will reference as they iterate over the build out of the application and the physical data store will morph with this.

Some teams argue modelling slows innovation. Others, including myself advocate for a hybrid approach, involving data professionals earlier in the cycle to prevent costly mistakes. The key is collaboration: modelling should be a shared responsibility across roles, it is part and parcel of the DevOps collaborative culture.

Modelling for Analytics: Start Before You Build

When looking at the analytics side of data, whether you're building a Lakehouse, data lake, or data warehouse, in my view, modelling must come first. Understanding the questions we need to answer and where that data comes from, gap analysis and where the data comes from is key to a successful build out. Without modelling we are more likely to end up with our data swamp.

Trying to run analytics directly on transactional databases leads to performance issues and resource contention. Modern platforms like Databricks, Amazon Redshift, Google BigQuery, Snowflake, and Microsoft Fabric offer powerful tools, but they don’t eliminate the need for modelling. They demand it.

Dimensional modelling fact and dimension tables, slowly changing dimensions, surrogate keys are the foundation of scalable, reliable analytics. Without it, data pipelines become brittle and reporting becomes inconsistent.

The Renaissance of Data Modelling

Far from fading, data modelling is experiencing a revival. With tools like Redgate Data Modeler making modelling accessible to analysts and engineers not just DBAs.

Modelling is no longer just about database design. It’s about communication. A good model is a shared language between business stakeholders, developers, and data professionals. It gets everyone on the same page, clarifies requirements, and reduces rework.

This is going to become more important as more businesses look to get value from their data by using it in conjunction with AI. The adage that garbage in means garbage out is very much the case here and anything that we can do to prevent having to clean up data to work with AI means that we can get to value faster.

Final Thought

In a world obsessed with speed, data modelling offers clarity. It’s the blueprint that ensures our digital infrastructure is scalable, secure, and aligned with business goals.

Agile development may have changed who owns aspects of the models, but it hasn’t changed why we need it. And as data continues to grow in volume and value, the organizations that invest in modelling, early and often, will be the ones best positioned to thrive.

I’m really interested to understand what you are thinking on this and what you see as you build and manage database platforms today.

John Martin

Join the debate, and respond to the editorial on the forums

From 13–17 October, Redgate joined the nationwide celebration of Women in Tech Week, a week dedicated to recognizing and amplifying the voices of women shaping the future of technology.

The latest release of Redgate Monitor visualizes Oracle wait classes and events across single-instance, multitenant (CDB/PDB), and Data Guard environments. It gives DBAs a clear, high-level view of where database time is spent and makes diagnosing and resolving performance issues in Oracle much simpler.

It’s ‘Women in Tech‘ week and to celebrate we’re featuring staff stories which highlight the creativity and power that underpin modern technology. What is often thought of as a world of systems and logic, its true foundation lies in people and the stories that drive us to imagine, build and create a better and more connected world.

The OODA loop—for observe, orient, decide, act—is a framework to understand decision-making in adversarial situations.

In the modern enterprise, information is the new capital. While companies pour resources into artificial intelligence, many discover that technology, standing alone, delivers only expense, not transformation.

Many AI projects fail. The reason is often simple. Teams try to rebuild last decade’s applications but add AI on top

If you’re a SQL Server DBA or developer looking to harness AI for your everyday scripting workflows, this article will walk you through building an AI-powered T-SQL assistant using Python and SQL Server.

As organizations weave AI into more of their operations, senior executives are realizing data engineers hold a central role in bringing these initiatives to life. After all, AI only...

Models trained on short, popular, and/or "superficial" tweets perform worse on benchmarks.

Every new SQL Server release comes with shiny features — but SQL Server 2025 brings more than just enhancements.

VIEWs are an undervalued and underused feature in SQL. They basically consist of a query that has been given a name, and a parameter list, so can be used like an inline macro. Technically, you’re supposed to think of it as a virtual table.

TempDB is one of the banes of my existence.

On Monday afternoon, Amazon confirmed that an outage affecting Amazon Web Services’ cloud hosting, which had impacted millions across the Internet, had been resolved.

AWS has expanded its Customer Carbon Footprint Tool (CCFT) to include Scope 3 emissions data alongside updated Scope 1 and 2 emissions, giving customers more insight into their carbon...

I have the greatest respect for those that organize and create user group tech events.

I remember feeling a little nervous when when I first started contributing to stack exchange. It was supposedly unwelcoming and full of unfriendly people. I even planned on going...

You're not going to believe this - the criminals that took the Qantas data ignored the injunction

In this post, I describe how to create an SSIS Catalog in SQL Server 2025 (RC) using SSMS v22 (Preview 3).

We discussed VALUES in previous articles: Choosing between DISTINCT and VALUES in DAX and Using VALUES in iterators.

If you’ve been working with Direct Lake in Microsoft Fabric, you’ll know its magic resides in its ability to quickly load data.

When I started talking about Fabric Real-Time Intelligence, people were surprised.

Load testing is essential when working with Microsoft Fabric capacity.

Copy job is the go-to solution in Microsoft Fabric Data Factory for simplified data movement, whether you’re moving data across clouds, from on-premises systems, or between services.

In my recent Microsoft Fabric training, I’ve been explaining the difference between the Direct Lake on OneLake and Direct Lake on SQL.

We’re introducing a new feature that gives you more granular control over your Spark compute resources in Microsoft Fabric

If you have Fabric capacities it’s important to be able to monitor your Azure costs relating to them

More often than not, we need to parse data that resides outside SQL Server.

SQL Server Management Studio 22 Preview 3 is out, and it brings with it a new Query Hint Recommendation tool.

Huge pages are a Linux kernel feature that allocates larger memory pages (typically 2 MB or 1 GB instead of the normal 4 KB)

In this article, I introduce you to psql and show you how to connect to and work with your ADP instance.

Postgres has earned its reputation as one of the world's most robust and feature-rich open-source databases

Why Postgres survived being a very distant second to MySQL in the early days of the mid to late 90’s.

A reasonable question to ask, if you are managing a system with variable performance is: “what queries on my system are running slowly?”

For this weeks blog, a quick tip about a feature in Power BI desktop which had flow entirely over my head: You can use RegEx for Find & Replace operations in Power BI Desktop TMDL View!

A few decades ago, I worked with a developer who was respected by everyone on our team.

As Steve writes, there are a bunch of ways to parse strings in SQL. And (which Steve doesn’t write) 2025 is going to change a lot of people’s approaches.

Nobody ever said “I like doing string manipulation in T-SQL”.

I was trying out the new Regular Expression (Regex) functions in SQL Server 2025.

did you realize you can create a temporary stored procedure as well?