Federated vs Centralised Data Modelling: Which Approach Wins?

In large, complex organisations, data management often feels like a tug-of-war between centralised control and decentralised autonomy. When it comes to data modelling, this can be especially true. Do you consolidate all data and create a single source of truth that everyone draws from? Or do you allow individual departments to model their own data, tailored to their specific needs and leverage domain knowledge? The reality is, there’s no one-size-fits-all answer – both approaches have advantages, challenges, and wider implications to how data models can add value to wider data initiatives.

The Need for Data Modelling in Complex Organisations

In a large organisation, data flows in from countless sources. Sales teams, marketing departments, finance, HR – each function collects, stores, and processes data differently. Without a clear data model to define structure and relationships, this data becomes a scattered patchwork, leading to inconsistencies and inefficiencies. At its core, data modelling provides the framework that turns raw data into usable information. It defines how data is stored, organised, and accessed, enabling departments to make sense of the information they collect and allowing the organisation to make better decisions. But here’s the catch: different departments have different data needs. A marketing team, for example, might be interested in analysing customer behaviour across channels, while finance may focus on transactional data. Creating a single data model that satisfies both might seem ideal, but it can also lead to bottlenecks and compromise the flexibility each team needs. This is where the tension between federated and centralised approaches starts to come in.

The Centralised Approach: One Model to Rule Them All?

A centralised approach to data modelling typically involves creating a unified data platform or repository that serves as the single source of truth for the entire organisation. Data from various departments is standardised and consolidated into a single overarching model, providing consistency across the board. Think of it as building a skyscraper – one blueprint, multiple floors, each department residing in their own section, but all part of the same structure. Some further considerations might be:

  • ✔ Consistency and control: A centralised data model ensures that everyone is working from the same set of definitions and rules. This reduces ambiguity and makes it easier to run cross-departmental reports, audits, and analytics.
  • ✔ Data governance: Centralisation allows for simpler governance, security, and compliance controls, which is critical in industries that deal with sensitive information like healthcare or finance. Data access and privacy policies can be enforced more uniformly across the organisation.
  • ✔ Efficiency: From an infrastructure standpoint, centralising data can reduce redundancy, saving storage costs and making it easier to maintain data quality.
  • ✖ Bottlenecks: When a central team controls the entire data model, it can slow down innovation. Departments may need to wait for approval or changes to the central model before they can proceed with new projects or initiatives.
  • ✖ Loss of flexibility: A one-size-fits-all model rarely satisfies the specific needs of individual teams. Departments with unique data requirements may struggle to fit their data into a rigid, centralised structure.

Centralisation works well for organisations where uniformity and compliance are non-negotiable, but it can feel cumbersome in dynamic environments where teams need to move fast and innovate.

The Federated Approach: Autonomy and Agility

In contrast, a federated approach allows individual departments or business units to maintain control over their own data models. Instead of consolidating everything into a single source of truth, each department builds and manages their own models, tailored to their specific needs. Picture a network of small villages, each with its own layout and architecture, but loosely connected through highways and trade routes. Again, there are more considerations to balance out:

  • ✔ Agility and autonomy: Teams can move quickly, creating and adapting their own models to suit their needs without waiting for approval or intervention from a central authority. This is particularly useful in fast-moving industries where innovation and experimentation are key to staying competitive.
  • ✔ Tailored solutions: Different departments have different data needs, and a federated approach allows them to model their data in a way that reflects their unique processes, KPIs, and objectives.
  • ✖ Inconsistency and silos: Without a centralised structure, different departments may define and model data in conflicting ways, leading to a lack of alignment across the organisation. For instance, "customer" might mean one thing to the sales team and something entirely different to finance, making it difficult to integrate insights at an organisational level.
  • ✖ Data governance risks: With each department managing their own data models, governance, security, and compliance can become fragmented. Ensuring that sensitive data is protected consistently across all units becomes a bigger challenge.

Federation thrives in organisations that value speed, autonomy, and innovation but need to watch for silos and inefficiencies creeping in.

Best of Both Worlds?

For many organisations, the answer lies somewhere between full centralisation and full federation – a hybrid approach that balances control with flexibility. In a hybrid model, core data elements (such as customer IDs, product codes, or financial transactions) are centrally defined and managed, ensuring consistency across the business. Meanwhile, individual departments are given the freedom to create their own models for more specialised or domain-specific data. Imagine a shared foundation that allows each department to build its own custom structures on top. This way, the marketing team can model customer behaviour however they see fit, while still aligning with the central customer database managed by IT. The key is coordination: departments work collaboratively with the central data team to ensure that core data definitions remain consistent across the organisation.

How to Apply This Hybrid Approach:
  1. Define core data elements: Identify the data that must be centrally managed and ensure that everyone uses the same definitions and standards.
  2. Enable autonomy for specific use cases: Allow teams to model non-core data in ways that suit their unique needs, but establish clear communication protocols to ensure compatibility.
  3. Leverage technology: Use modern data management platforms that enable both centralised governance and decentralised flexibility. Cloud platforms, data warehouses, and data lakes can offer the infrastructure for such an approach.
  4. Governance with collaboration: Create a governance framework that provides oversight without stifling innovation. This might involve data stewards or cross-functional teams working together to ensure data models are aligned where necessary but don’t micromanage.

Applying data modelling in a complex organisation requires a careful balancing act. Centralised models offer consistency and control but can be slow and inflexible. Federated models empower teams to move quickly but risk inconsistency and fragmentation. A hybrid approach, which combines the best of both worlds, allows organisations to be both agile and aligned, giving them the flexibility to innovate while maintaining a unified structure where it matters most.
Ultimately, the decision between centralisation and federation should be driven by the organisation’s goals, culture, and industry. But one thing is certain: a strategic approach to data modelling is essential for transforming raw data into meaningful insights that drive the business forward.
  • strip-0

    AFFINITY REPLY

    Affinity Reply are Architecture, Design & Data advisory specialists who accelerate clients to realise new digital capabilities, drive business change and unlock Next Gen Architecture.

CONTACT US