In my experience of leading Enterprise Architecture practices, I have often witnessed the debate – centralised vs federated enterprise architecture. The topic has been a standing agenda item for CXOs across organisations for too long now. The discussion has gained even more momentum as each organisation aspires of becoming a data-driven organisation. Organisations across geographies and business domains understand the importance of data in driving competitive advantage and customer satisfaction; and want an architecture that is best suited to help them achieve increased value from their data.
Historically, organisations have implemented some flavour of EA frameworks (e.g. TOGAF, Zachman) to inform the ICT related investment and governance decisions. Typically, these frameworks are modelled on architectural domains like business architecture, data architecture, applications architecture, technology architecture and security architecture; and provide an approach for designing , planning, implementing and governing enterprise ICT architecture. While these frameworks provide useful guidance in developing and managing EA, implementing them is often onerous and terribly slow. Moreover, organisations are moving away from a centralised ICT delivery model to a federated ICT delivery model primarily driven by increased adoption of agile development and product development.
In case of data architecture domain, there is an increased interest in data mesh. While it is not just a data architecture and entails other aspects e.g. organisational (re)design, it can be seen as a target state data architecture in an organisation. I am often asked by my connections if data mesh is the right target state data architecture for their organisation. Here are some questions you should answer to make that choice.
1. How is software sourced in your organisation?
Begin with the basics – assess your current digital landscape to identify how software is provisioned/updated in your organisation. In my experience, most of the public sector organisations have a “buy before build” policy. Such a policy forbids embarking on a “data mesh” journey. Data mesh architecture assumes a high maturity in application development in an organisation. If your organisation is dependent on vendors (COTS products) then do not include data mesh in your data strategy.
2. How mature is your Enterprise Architecture (EA)?
Do you have a well-defined business capabilities and associated taxonomies across business, application, data and technology domains? Or you don’t have an established EA? Your EA maturity will impact the feasibility and suitability of data mesh in your organisation. While a well-defined business architecture may ease the move to a domain-driven design mindset it would also bring a bigger challenge of making the organisation change happen.
3. How are ICT projects delivered ?
Is agile delivery a norm or an exception? Data mesh requires agile development and automated infrastructure as a norm in an organisation. If you are still considering Devops, do not jump onto data mesh yet.
4. How effective is your current Data Analytics architecture?
Is your current data analytics architecture (datawarehouse/data lake/lakehouse) serving the business purpose? Are the data insights sufficient to better inform the decision makers? Is your current IT org better suited to deliver analytics centrally?
5. How is your business/organisation structured?
Do you have application teams who manage their own applications(serving business domains) and would know the associated data better? Are you an organisation where all applications are managed centrally? Data mesh is more suited to organisations where different application teams are delivering business applications/services.
6. What is your risk appetite to fail?
Do you have the appetite to take the risk even if the likelihood of you failing is higher? Does your organisation embrace “fail fast” culture? Data mesh is a new paradigm and would mean failures will be part of the implementation journey. There isn’t a single platform that can meets all principles of data mesh. The implementation journey would be iterative with lots of failures and lessons learnt. Are you ready to make the move?
Overall, if you are not sure don’t start on it. However keep yourself informed of the developments in technology and concepts around data mesh.