Data Architecture

A modern data architecture is subject to 6 fundamental principles:

​

1. Data is a common good: A data architecture must dissolve department-specific data silos and provide all stakeholders with a holistic view of the company.

2. Users need appropriate access to data: Modern data architectures must provide interfaces that allow users to easily consume data with tools that are appropriate for them.

3. Security is essential: A modern data architecture is designed for security and supports data policies and access controls at the raw data level.

4. Common vocabulary: Shared data assets such as product catalogues, fiscal calendar dimensions and KPI definitions require a common vocabulary to avoid conflicts in the analysis phase.

5. Curate data: Invest in core data curation capabilities (model important data relationships, cleanse raw data and curate relevant dimensions and metrics).

6. Make data flows more agile: Reduce the number of data moves required to reduce costs, increase data timeliness and optimise business agility.
    

 

Modern data architectures must be designed to unlock the benefits of new technologies such as artificial intelligence (AI), automation, the Internet of Things (IoT) and blockchain. Modern data architectures should thus have the following characteristics:

​

 

Cloud-native: modern data architectures should be designed to support elastic scaling, high availability, end-to-end security for data (in motion and at rest), and cost and performance scalability.
 

 

Scalable data pipelines: To take advantage of new technologies, data architectures should support real-time data streaming and micro-batch data bursts.
 

 

Seamless data integration: A data architecture should integrate with existing applications via standard APIs. Interfaces should also be optimised for sharing data across systems, locations and organisations.
 

 

Real-time data: Modern data architectures should provide the ability for automated and active data validation, classification, management and control.
 

 

Decoupled and extensible: Data architectures should be designed to be loosely coupled. This ensures that services can perform minimal tasks independently of other services.