Navigating the path to data mesh success

In today's data-centric digital economy, the proliferation of critical business data is typically accompanied by the use of multiple data siloes scattered across the enterprise. Data is being stored organization-wide among dozens, even hundreds, of sites and platforms – in legacy on-premise solutions, in the cloud, across departments, business units, lines of business, and more.

The unfortunate result? 'Data scientists are spending an inordinate amount of time' – up to 60 percent by some estimates – searching for, cleansing and preparing compiled data before using it to generate insights. Data lakes were intended to solve this problem by bringing all data into a central location. But the lack of well-defined ownership and data governance has turned data lakes into data swamps.

Enter the data mesh as the next-generation solution – a modern approach to building a data ecosystem that takes advantage of domain autonomy while providing the benefits of domain ownership and federated governance.

A previous article – 'Data mesh: a modern approach to building your data ecosystem' - examined the data mesh concept and its ability to provide easy, secure, reliable access to data across the entire enterprise. And a subsequent data mesh article titled 'Planning the data mesh journey' –explored in detail the key steps your business should take to effectively plan a data mesh initiative and avoid challenges and inconsistencies during the implementation process.

This article emphasizes the necessity of a strategic approach and details the initial steps of a successful data mesh implementation.

There are a number of factors that influence the outcomes when an organization tries to bring their data mesh vision to life. None is more important than having a C-suite or senior leader's sponsorship. This leader, called a 'data mesh champion', can be instrumental in bringing the various elements and functions of your organization together.

Consider the champion's role as the 'cheerleader' for the data mesh initiative – helping to provide the business and its leaders with a solid understanding of what the data mesh concept is and the advantages and value it can deliver enterprise-wide.

Firm-wide clarity on your data mesh initiative is crucial, including acceptance of the required investment. Your champion must lead the way to create and confirm your data mesh vision.

• Domain owner control – Each independent data creator owns and maintains their data. But instead of creating a silo, a data mesh enables each data owner to easily 'publish' their data via standardized APIs, with each owner managing a comprehensive metadata catalog for all teams to access.

• Self-service consumption – Anyone requiring access to data can request it directly from the data owner and use it as they see fit. The metadata catalog should provide data users with the information needed to understand the data's meaning and potential value.

• Federated security and access – Data owners are responsible for the security of their data. They are responsible for handling access requests from data consumers and for provisioning that access — another key element of the self-service model. IT is no longer burdened with such requests.

• Federated governance – Your data mesh requires strong, federated governance. Each data product must be fully interoperable with the others. A governing body — usually internal stakeholders — defines which tools or technologies can be used and sets the standard for the metadata catalog.

Beyond bringing the organization on board for the data mesh journey, your data mesh champion can help bring the vision to life in terms of establishing your target state architecture and tool selection. This is key – and a common problem that KPMG professionals are helping organizations solve today amid the overwhelming choice of vendors offering an array of technology and toolsets.

From there, you need to select pilot domains within the business – starting small, demonstrating value quickly, and scaling as appropriate. KPMG professionals recommend taking a structured approach on the data mesh journey, starting with the minimum viable product (MVP) and following the key steps outlined below.

The journey begins with the MVP, which provides a crucial opportunity to explore possibilities, identify key learnings and set a path that can drive data mesh success. The MVP should serve as a 'showcase' to rest of the organization on the expected benefits of a data mesh. It can also position you to create a playbook for future initiatives. Follow a structured and strategic approach as outlined below:

• Identify the scope of MVP. The scope you choose should deliver meaningful business value that enables you to use the MVP as a 'showcase.' It should be complex enough that you can apply all the principles of a data mesh, yet simple enough that it can be delivered within a few sprints.

• Catalog existing data sources pertaining to the identified scope. Create a comprehensive list of all known data sources that need to be included in the target data ecosystem.

• Define data products and data domain. Identify the key data domain that the MVP will deliver and identify and define all the data products in that domain. Determine which data product or set of data products you will deliver as part of the MVP.

• Highlight the value that the data product delivers. When defining each data product, it's essential to include the benefits or value that the product can deliver. This includes potential use cases that the product can support, analytics that it can support, etc.

• Prioritize which data product to develop based on value. If you have identified more than one data product to deliver as part of the MVP, it’s important to prioritize the products to be built based on the value that each can deliver.

• Design the data product based on your toolset. When complete, review your design with the architecture team. Be sure to leverage your established architecture, tools and framework to effectively design and develop the data product.

• Develop each data product iteratively. Demonstrate capabilities to stakeholders to gain support and engagement. Building a data product iteratively involves developing and refining it through a cyclical process of building, testing and gathering feedback. This allows for adjustments and improvements as the product is developed, demonstrating capabilities to stakeholders and gaining support throughout.

• Test each data product and release it for use. Continuous testing of the data product will help ensure that it is optimized for performance, accuracy and usability. Testing also allows the data product development team to identify areas for improvement and iterate towards a better product.

• Document each data product into a catalog. It's critical that you release the data catalog, metadata and lineage along with the release of the data product or any changes to the data product. In doing so, you will likely gain more trust and confidence in the data.

• Document all lessons learned through the process. It's essential to have a feedback loop that provides critical insights, continuously improving the data mesh journey and results.

Based on key learnings and best practices that the MVP provides, it's important to develop a comprehensive playbook. The playbook can be crucial for the success of your data mesh across the various domains and functions of your business, from HR to finance and beyond.

Each organization is unique – and so is your data mesh playbook. A comprehensive playbook is essential to successfully building and using a data mesh across your business. Each data domain will be responsible for the data products it produces for other functions and domains to consume. The intent of the playbook is to provide all domains with best practices and lessons learned from the MVP, enabling them to effectively create and host their data products. Your playbook should cover in detail:

• A Target Operating Model with well-defined roles – infrastructure team, data product team, architecture team – and data operations under a federated data governance framework. Establishing clearly defined roles within the ecosystem is essential to success.

• A brief 'How To' guide to identify and establish effective personas, use cases, harmonization rules, key metrics, data mapping and standard metadata. This will add consistency across domains.

• How to identify and define data products, along with a template to document the data product and its value. A central location to host this information that the rest of the organization can easily access can prove beneficial in the long-run.

• How the data product should be developed in terms of toolset and standards, how it will be maintained and, most importantly, how it can be consumed. This should also include how to identify a data product 'steward' or owner who understands every aspect of the data, where it is coming from, what it contains and how best to use it.

• Processes, KPIs and OKRs to track and report data mesh progress and value. It's crucial that the quality of each data product being consumed across the business is easily accessible and of consistently high quality. Consistency across teams can be indispensable to success as diverse domains deliver data products for enterprise-wide use.

Now that a comprehensive playbook is available, it's time to scale the implementation. Identify the data products you want to build in each data domain, prioritize them based on business criticality and value, and iteratively develop multiple products in parallel.

Multiple domains can also execute in parallel and each domain can build multiple products simultaneously based on need. HR plans to create seven data products this year? Great news – but you should strategically prioritize which products are built first.

Collaboration across domains is essential, with key members meeting regularly to share feedback and examine lessons learned, as well as to celebrate success and the value delivered to the business. In doing so, you position your business to continually learn and adopt best practices.

Your business is unique and so is every data mesh journey – what works for one organization might not work for others. KPMG professionals can provide guidance and insights based on experience – helping your organization establish an effective playbook that can help bring your data mesh to life.

A data mesh is the modern digital solution to the shortcomings of data lakes – offering autonomy and flexibility for data owners, facilitating greater data experimentation, and fostering innovation while lessening the burden on data teams to field the needs of every data consumer through a single pipeline. A data mesh can deliver a competitive edge compared to traditional data architectures that are often impeded by a lack of data standardization between both data providers and data users within the business.

Given the magnitude of the investment and the effort involved, you'll need a solid strategy before you begin the journey. KPMG professionals are well equipped to help you plan that journey. Whether you're exploring the concept, ready to start, or already navigating the data mesh journey, get in touch with a KPMG data mesh specialist. They're here to talk and answer your questions.