×

Discussing data mesh with a digital transformation leader

Last Updated: January 9, 2024

Before you begin your data mesh journey, seek advice from a digital transformation leader like Maria José Lopez. That is exactly what we did in Episode 13 of Data Mesh TV.

Data mesh transformations may feel especially daunting early in the journey, when you are just starting to define scope, speed and sponsorship. Part of that challenge is due to data mesh being a relatively new concept, but the transformational change management that you need to succeed is a proven and trusted science. Maria and Onepoint are experienced leaders in enterprise transformations and data mesh, she offered the following tips:

  • Focus on the user experience
  • Abstract the technology complexity
  • Communicate, communicate, communicate
  • Link data value to tangible business outcomes

If you are planning a program to build a Red de Datos (Spanish for data mesh), continue reading below and watch the video. 

4 tips for managing a data mesh transformation

For most of us, the transformation program’s primary objective is not to simply build a data mesh. You are building a ‘mesh’ to enable new capabilities that will enable your business to access and exploit their data assets with speed. Your focus should be on solving the business problem and delivering an immediate and sustainable solution. Data mesh is a framework for building a set of capabilities that will deliver against defined data-driven business outcomes.  Consider the following tips:

1. Focus on the user experience

A great way to build early momentum is to create notable wins for the user. Think about helping the data engineer build, share and reuse data products. Think about helping the data consumer find, reuse and rapidly exploit data products to create insight. Some of those consumers may even want to build their own data products, alleviating the pressure and back-log for the engineers. 

2. Abstract the technology complexity

Data mesh will transform the way all teams manage data, and there will be unique challenges for the data producers and the consumers. One of the key principles in data mesh design is to abstract the operational plane (e.g. back-end)  from the analytical plane. The trick is to keep this simple for your consumers, they should not care or see how data is being moved on the back end. Deploying a high-performing federated query engine like Starburst, will help you avoid or abstract the migration headaches. On the business side, their focus should be learning how to use the data products with their self-service tools. The business consumer should spend 100% of their time on driving insight, decisions and actions. Architects should have the ability to make the best decisions for where data should reside (e.g. cloud, on-prem, centralized, decentralized) without negatively impacting the consumers. 

3. Communicate, communicate, communicate

Every successful transformation program starts with a message to align on purpose, scope and accountability. Focus on the business outcomes and describe how the data mesh capabilities will help, provide specific market case studies. Speak in ‘business’ language, use examples and terms that are relevant to your audience. Throughout the transformation, a steady stream of messages should celebrate wins and articulate challenges to establish transparency and foster trust. Instead of communicating how architecture is changing, focus on communicating why changes are necessary and how data access will improve. Effective internal marketing is a proven way to build and maintain stakeholder support.

4. Link data value to tangible business outcomes

If you are presenting to the board or senior executives, don’t talk about the data mesh transformation, talk about the delivery of business outcomes. Reference metrics that demonstrate greater access, faster insights and adoption. “We have created autonomous domain teams that have embraced a data driven culture. These teams are proficient in using self-service tools to accelerate their use of data. We are building new dashboards in a week vs months, we are integrating new data sets in a day vs weeks and we are making business decisions based on insight that our competitors cannot match.” 

Practical advice for successful data mesh implementations

We have seen quite a few data mesh implementations, and had just enough time to discuss two ideas that we have seen work very well.

Start with data products

We have seen a new trend in data mesh implementations, some organizations are kicking off their journey by focusing only on data products. Maria noted, “The data product is the most important topic because it will help us make sure that we are not centralizing data again, that we are enabling access and truly transforming access and insight.”

Why this is a great idea…

  • This allows your data engineers to get really good at building data products before you create the additional complexity associated with domain management, federated governance and the full self-service capabilities.
  • You can build a central repository of great (e.g. consistent, tested, secure) data products that can be published in a marketplace for consumers to use. As the consumers become proficient in using them, some may begin to form natural domain teams that you can leverage in your next phase.
  • You can evolve organically, allowing consumers to develop skills at their own pace. Teams that are ready to run, can learn how to build their own data products within a domain, following the guidelines provided by the data engineers.

Be careful…

  • Data products need the broader data mesh framework to ensure sustainability and to deliver against the bigger strategic objectives. If you build 500 data products without the underlying processes, you are going to create a mess. Your TCO, risks (compliance, performance) and project timelines will increase. If you don’t enable teams to work autonomously as part of a proper data mesh design, IT will continue to be a bottleneck. To address this challenge you need to agree quickly enough on data domains with company stakeholders and ease federated governance.

Leverage governance as a facilitation mechanism, not a control mechanism

In a data mesh, the federated governance model should be designed to facilitate greater access by applying the right level of governance for each domain. This isn’t a one-size fits all control model, you want to apply fit for purpose controls that allow teams to run as fast as their risk profile permits. If you start your data mesh exercise focused on building unnecessary controls and limitations, you will likely stifle the innovation and reduce the effectiveness of your mesh.

Why is this a great idea…

  • Create immediate value by opening access for teams to fully exploit data products. A mature mesh will have diverse data products, some will require strong controls and others may be more open.
  • Focus on areas where facilitating access will create immediate value without increasing risk. Facilitating the right level of governance will create some early wins to build momentum for your journey.

Be careful…

  • It may be difficult to build stricter controls when they are needed, as teams may expect all data products to have the same level of control. Teams may not understand why “they don’t all work the same.” It is important to capitalize on early success, and use that credibility to explain how data product risk profiles drive access policies.

Reflect, refine, research

  • Reflect: How will you adapt your current data mesh program, are you engaged with a transformation professional?
  • Refine: Consider starting with Data Products if your organization is not ready to commit to the broader organizational changes that come with domain design. Assess your governance approach, is it truly fit for purpose, will it facilitate value for your consumer?
  • Research: Schedule one-on-one time with a digital transformation expert with expertise in data mesh. Reach out to Maria and OnePoint.

Data products:

Enable data producers and consumers to create, publish, discover, and manage curated datasets

Start today

Start Free with
Starburst Galaxy

Up to $500 in usage credits included

  • Query your data lake fast with Starburst's best-in-class MPP SQL query engine
  • Get up and running in less than 5 minutes
  • Easily deploy clusters in AWS, Azure and Google Cloud
For more deployment options:
Download Starburst Enterprise

Please fill in all required fields and ensure you are using a valid email address.

s