Introducing Banking Customer Segmentation Commons
A community focused on customer segmentation in small and regional banks
Read articleLogic Over Legacy: Why skipping to physical modelling is killing your data strategy and how to fix it with ERD design.
Conceptual, logical, and physical models are not competing artefacts but different lenses on the same reality, each optimised for a different outcome.
Choosing the wrong layer for the problem you’re solving is how teams end up arguing SQL when they should be debating semantics, or debating semantics when the real constraint is latency.
Data modelling isn’t a mandatory art project designed to justify an architect’s salary. If you think an Entity Relationship Diagram (ERD) is just a collection of boxes and lines to be filed away as “documentation,” you’ve already failed. At its core, data modelling is the act of giving data shape and constraints so it becomes reliable reasoning material. It is the literal bridge between a business’s messy reality and the rigid constraints of a high-performance system.
A data product is a reusable, measurable asset engineered with the discipline of product thinking. To build one, you must come to the understanding that you are designing for the truth. A Logical Data Model is that blueprint of your company’s intelligence.
And to build a system that scales, you must navigate three distinct layers of abstraction: The Conceptual, The Logical, and The Physical. Each serves a specific outcome, and confusing them is how you end up with expensive databases that nobody knows how to query.
[state-of-data-products]
Data models exist for one reason: to ensure your data actually means something. They are tools for alignment, scale, and correctness. In the modern landscape, we are moving away from centralised monoliths toward decentralised domains where teams own their data. In this world, models are built around value streams instead of just storage preferences.
The most common failure in modern data teams is choosing the wrong model at the wrong time.
If you’re arguing about data types (physical) before you’ve defined what a “customer” actually is (conceptual), you are building a high-speed engine for a car that doesn’t have a steering wheel. You’ll get somewhere fast, but it won’t be where the business needs to go.

Conceptual data modelling pertains to defining the business language. The Conceptual Model is the “What.” It is the pure, unfiltered language of the business.
At this stage, jargon like “primary keys” or “varchar” has no place. You are defining entities and the high-level relationships between them.
Does an “Order” belong to a “Customer” or a “Household”? This is where the foundation for your Entity Relationship Diagram begins. This stage demands conversations, not code. If you can’t explain your data in plain English at the conceptual level, you have no business trying to code it.
Logical data modelling means to structure business meaning without technology. The logical model is where the heavy lifting of ERD design happens.
In a logical model, we are able to define your attributes, keys, and relationships while staying platform-agnostic. The logical database model doesn’t care if you’re using a cloud database, a relational one or even a graph database. All it cares about are the rules of your business universe.
This is the most critical layer for modern organisations because this is where the Semantic Layer is born. A Logical Data Model further allows you to expand into “Metric Trees.” It’s the source of truth that prevents data silos. If you get the Logical Model right, your technology choices become interchangeable.
A Physical Data Model optimises for storage and performance. The Physical Model is the “How,” and becomes the execution artefact.
The physical data model is where we talk about tables, indexes, partitions, and constraints. It’s tuned for the machine instead of the human. The physical layer succeeds when it remains porous and interoperable, allowing different domains to talk to each other without breaking trust. If the conceptual and logical models are the strategy, the physical model is the tactical execution.
When aligning stakeholders, use conceptual data models: If you’re in a boardroom, do not show them a schema. Show them a conceptual map of their business entities and value streams.
When designing analytics, AI, and Data Products, use logical data models: This is your semantic foundation. If you want AI agents to understand your data without hallucinating, they need the structured context of a Logical Data Model.
When it’s time to implement, use physical data models: This is for the data engineers. It’s the final step in turning a bulletproof business requirement into a functional, high-performance database.
Everything. When you skip the logical phase, you bake your business rules directly into your hardware. The moment the business changes, your entire system shatters because there is no separation between the “Meaning” of the data and the “Storage” of the data.
AI requires context. If your data is just a pile of physical tables with cryptic names, your AI is flying blind. An Entity Relationship Diagram rooted in a strong Logical Model provides the map that AI needs to navigate your data with accuracy and avoid the “safe bet fallacy” of ad-hoc pipelines.
[playbook]
An Entity Relationship Diagram is not a static picture but is the evolution from language to logic. It starts as high-level bubbles in a conceptual meeting and matures into a detailed Logical Data Model full of formal definitions and rules of engagement.
The industry has confused “drawing ERDs” with ERD design. Anyone can put boxes on a screen. True design is about the cardinalities and the business constraints that make data valid. Your ERDs are most valuable at the logical layer because that is where you finalise the “Laws of the Data” even before the first table is ever created.
Stop letting your database dictate your business rules. Start with intent, solidify the logic, and only then worry about the physical implementation. That is how you build an architecture that stands the test of time.
A no-cost session with a practitioner who's been where you are.
Modern Data 101 is a global movement redefining how the world thinks about data. Built by the team behind The Modern Data Company, the company pioneering DataOS, the data operating system, Modern Data 101 brings together a global community of practitioners, architects, and leaders across 150+ countries at the intersection of data, product thinking, and AI.
While DataOS advances the technology and infrastructure on-ground to help enterprises operationalize data as products, Modern Data 101 advances the ideas, practices, and conversations that shape how the industry approaches data and AI. Together, they connect technological innovation with the broader movement to rethink how organizations build, govern, and activate data.
At its core, Modern Data 101 exists to simplify the journey from raw data to tangible, measurable impact. It champions data product thinking and next-generation architectures that unify and activate insights and automation across analytics, applications, and operational workflows.
In a world shifting from data stacks to AI ecosystems, Modern Data 101 helps the industry rethink what's possible.
A 9-dimension self-assessment used by 100+ data teams to benchmark strategy, ownership, and platform readiness.


Modern Data 101 is where the data community thinks out loud. When you're ready to move from articles to architecture, data products, governed AI pipelines, or a full Data Operating System; the team behind this community can help you build it.