Data Mesh: power to the business

Data Mesh: an abstract paradigm or a concrete approach to exploiting data? While data ‘as a product’ may seem complex to define, the approach relies on good business organisation, solid governance and the use of innovative complementary tools. The result is that business units can regain control of their data, for much greater operational efficiency.

The Data Mesh concept

Presented for the first time by Zhamak Dehghani, mother of the concept, the Data Mesh is a decentralised socio-technical approach to managing and accessing large-scale enterprise analytical data.  This approach considers both changing business needs and technological developments. This concept is said to be ‘domain-oriented’, i.e. by business line: sales, accounting, customer service, marketing, etc. 

The great advantage of this approach is that it gives employees the means to develop data services quickly, for use both internally and externally, without having to go through a central team in charge of the company's data. Business units therefore have greater autonomy and, because they are the producers of the data they use to develop new services, they are more efficient.

Promises and cultural pivot of the Data Mesh

Increased innovation, greater internal efficiency, better allocation of resources and skills, the Data Mesh eliminates the traditional bottlenecks that companies generally blame when embarking on a datacentric strategy. The understanding of data attributed to its producers (the businesses) enables them to define their own governance policies, based on the documentation, quality and access specified upstream. It is this autonomy that enables self-service use of data.  

As a result, the Data Mesh approach implies a cultural shift in the way data is perceived within the enterprise. Data is no longer simply an isolated raw material involved in a process. It becomes a (digital) product, a service, accessed by its users via interfaces. They are then responsible for this product, with the freedom to analyze and exploit it, manage its lifecycle, share and promote it. 

Conditions for implementing Data Mesh

The notion of meshing is extremely important. The aim of the Data Mesh is to make data available in sub-domains dedicated to each department, whatever technological choices the company may have made previously: datalake, datawarehouse, etc.  

To enable local, decentralized data management, the company must first have a clear vision of its governance. It will then need to identify its specific areas of activity and define data management responsibilities for each.  

It will then be up to them to set up the appropriate data infrastructure to activate the Data Mesh approach in practice, without forgetting to train their teams in the new working methods and tools available.  

How to implement Data Mesh?

From these reflections, it emerges that the implementation of Data Mesh includes the following 3 pillars in particular:  

• Federated governance, providing a centralized framework and global visibility on responsibilities by domain, and security and compliance policies.
• Data virtualization, in a Logical Data Fabric, to publish, define, understand and discover data across all domains, regardless of how data is stored and managed. 
• A physical, self-service data infrastructure platform, with its own data catalog, data store and connectors.