The term “data stack” is thrown around a lot. But what does it mean? In reality, a data stack is a practical concept that describes how data moves through an organisation, from its creation to informing decision making.
At its core, a data stack is the combination of tools, technologies, and processes that support the journey from raw data to decision making. It is not just about dashboards or reports; it includes everything that happens behind the scenes to ensure data is accurate, consistent, and usable. Understanding your data stack is essential if you want analytics that scale with your business rather than becoming a source of friction.
Most data stacks, regardless of industry or size, can be understood as a series of connected layers, each with a specific role to play.
Data Sources
Data begins its journey in operational systems. These might include finance platforms, CRMs, ERPs, e-commerce systems, or marketing tools. Each of these systems is designed to support day-to-day operations, not analytics. As a result, data is often fragmented, structured differently from system to system, and optimised for transactions rather than insight.
ETL: Extract, Transform, Load
To make this data useful, it must be moved and prepared. This is where extract, transform, and load (ETL) processes come into play. Data is pulled from source systems and then cleaned, standardised, and reshaped so that it can be analysed. Business rules are applied, errors are corrected, and formats are aligned. This step is crucial: if transformations are rushed or inconsistent, problems will surface later in the form of mistrusted reports and conflicting numbers.
Storage and data warehousing
Once prepared, data is stored in a central location, typically a database or cloud-based data warehouse. This storage layer provides a single place for data to live, independent of the systems that generated it. Centralising data in this way allows organisations to retain historical information, scale as data volumes grow, and run analytics without placing strain on operational systems.
Data modelling and the Semantic Layer
The next layer focuses on modelling the data so it reflects how the organisation thinks and operates. Relationships between tables are defined, metrics and KPIs are calculated, and business definitions are formalised. Often referred to as the semantic layer, this is where concepts such as revenue, margin, or active customers are defined once and reused consistently across all reporting. A strong modelling layer ensures that everyone is working from the same logic, regardless of which report they are looking at.
Reporting and Analytics
Only then does data reach the reporting and analytics layer. This is the front end that users interact with through dashboards, reports, and visualisations. When the underlying stack is well designed, reporting becomes simpler, clearer, and more reliable. Insights are easier to trust because they are built on consistent logic and clean data, rather than on manual workarounds.
Many organisations arrive at their current setup without ever consciously designing a data stack. These “accidental” stacks tend to evolve out of necessity, relying heavily on spreadsheet extracts, manual joins, and one-off reports built to answer immediate questions. Over time, logic is duplicated across files, complexity increases, and small changes become risky. What once felt flexible quickly becomes fragile and difficult to maintain.
Modern data stacks are different because they are intentional. Each layer exists for a reason, and responsibilities are clearly separated. Instead of recreating logic in every report, definitions are centralised. Instead of relying on manual processes, data flows are automated. This intentional design creates the single source of truth that many organisations aspire to but struggle to achieve.
Ultimately, a data stack is not just a technical architecture. It is the foundation for how an organisation uses data. When designed well, it allows analytics to grow alongside the business, supporting better decisions without adding unnecessary complexity.