In a Data Leaders peer discussion, Data and Analytics leaders compare their interpretations of and approaches to data mesh and discuss the arising obstacles when enabling cross-domain data sharing.
Data Leaders members can read this post and connect with peers via the Data Leaders Hub.
Data Leaders members looked to validate the initial steps of their data democracy programme “inspired” by data mesh. The centralised team’s purpose is to create a standardised and common data architecture that can be deployed and used by all affiliate countries with the objective of increasing data sharing within and across countries.
They began their journey 18 months ago with a move to a public cloud provider and aligning data leaders in-country to adopt the new data-mesh-inspired approach which involves moving to a common language, self-serve platform and a data-product approach. The initial phase is focused on proving the value of the approach with each country data leader engaged in creating MVPs cross-domain with the data team, and aligning as they determine which elements can and cannot be common.
Questions to Peers
- How were your data activities organised before data mesh and how has the way of managing data been impacted since then?
- Should a data governance maturity assessment be a prerequisite before implementing a data mesh?
- What is the most challenging aspect in terms of organisation, mindset transformation, implementation and how can you mitigate them?
Value Outcome
Over the course of the discussion, Data Leaders members were able to:
- Uncover new pitfalls to data mesh implementation,
- Explore different approaches to governance,
- Learn common approaches to adopting data mesh in a variety of industries.
Peer Insights
The majority of the discussions’ participants are taking their first tentative steps towards data mesh, picking and choosing the elements that best suit their data strategy and organisation. From the outset of the discussion, there was an expectation that people would have different interpretations of data mesh and as the conversation progressed, ideas on what does and doesn’t constitute data mesh were raised.
What do we mean by data mesh?
D&A leaders shared their different interpretations and approaches to data mesh. There was a general consensus that data mesh is a philosophy, a challenge requiring both technological and cultural change. There is a common belief taken by CDAOs that data mesh principles can be cherry-picked based on what best works for the given organisation.
The majority begin their approach by creating domains and domain leaders to manage data in logical groupings and then gradually introduce data mesh concepts that bring value to the organisation without fully disrupting it.
While the discussion raised the approach of a centralised data team being the driving force behind its implementation, one D&A leader argued that this approach could not be considered data mesh. Instead, it was proposed that data mesh ownership must be fully domain-driven, that data producers must be data owners and that a centralised team bringing all the data into one place and governing it centrally for self-service could not be considered data mesh but rather a traditional approach with more governance.
According to this view, data mesh therefore requires a much closer collaboration between executives from the product/business organisation, IT and data team to create a much tighter connection between the areas and a much larger effort to ensure product owners are educated and incentivised to be fully accountable for their data.
This will require new roles and capabilities including consultants and enablers to teach people how to use the self-serve platform, subject matter experts in domain-driven design to teach how to build data products and governance teams that are “fun to work with.” This presents a big challenge in terms of re-engineering teams from centralised to decentralised.
Starting a data mesh implementation is easier than sustaining it.
Discussing the member’s current approach to data mesh, a D&A leader expressed his view that the biggest challenge to implementing data mesh is not getting buy-in but rather ensuring sustainability. In his experience, while enthusiastic to participate at the outset, many domain teams responsible for building business functions see data and analytics as a secondary responsibility and struggle to sustain data products beyond the MVP phase.
Additionally, allowing other teams to access the data opens flood gates and raises questions about sustainability. To ensure sustainability, companies must focus on bringing new value propositions and finding intelligent use cases, while also considering the cost and time required for implementation. Overall, it is harder to maintain a data mesh than to start one.
Another CDAO shared how, from a technical stack point of view, the risk of investing in data mesh is too high, so instead his organisation is adopting it from an organisational standpoint, supporting domains to gradually take ownership of managing their data initially by providing the infrastructure and skillsets to create data products. The next step is to tackle the issue of data product access and security and how to move away from the duplication of reports to a more scalable solution.
Governance in the data mesh: is it the chicken or the egg?
The discussion turned to the relationship between governance and data mesh: is it necessary first to have strong data governance in place before embarking on a data mesh rollout or is data mesh a framework and impetus for a data governance programme? The experience was mixed.
One CDAO expressed that his organisation did not have global governance in place before adopting a data mesh approach, but that having a clear data strategy with associated top-level sponsorship, focused on data sharing enabled them to implement their cloud-based solution and defined rules for data access and progress thereafter.
Another said he turned to some of the core tenets of data mesh – specifically data as a product – as a way to instigate a company-wide governance programme to meet regulatory requirements. However, they did not use the term “data mesh” or other technical terms to avoid confusion and instead used it as a concept to roll out a central data governance program without the team even realising it.
Another D&A leader shared that after his first experience of implementing a data mesh approach with weak governance, second time around he is now focused on governance, specifically on data cataloguing and creating a comprehensive understanding of how people use and process the data, emphasising the importance of understanding the story of the data for analytics and machine learning, and “retraining the muscle” to pay attention to data and describe what it is.
D&A leaders are also looking to governance automation to iron out complexity in compliance and security, while still seeing the need for human intervention and collaboration.
This article has originally been published exclusively for the members of Data Leaders service. Our service gives unique access to trusted, timely, relevant insights from Chief Data and Analytics Officers from across the world.
To find out how you and your organisation can benefit from becoming part of our service, contact us here.
If you are a Data Leaders Member login here to access more related content: