Data Mesh Strategy Framework

The Data Mesh Strategy Framework is designed to help you formulate and implement a data mesh strategy for your organization. It outlines the typical phases observed during the implementation of the data mesh strategy. Consider the relevant phases for your organization based on where you are in your data strategy and cloud maturity journey. Sometimes, customers focus first on tools and technologies for their data mesh strategy. Instead, we recommend aligning your strategy with the business value delivered by your organization.

The Data Mesh Strategy Framework comprises five phases:

• Discover

• Align

• Launch

• Scale

• Evolve

Discover phase

In the discover phase, dive deep into the business and data landscape of your organization. The objective of this phase is to gather information that helps you design the data mesh. In this phase, qualify and gain clarity on the following topics:

• The current structure of the business and whether any reorganization is planned

• The amount of data generated by each line of business

• The data sources of the organization and the type of data each line of business generates—for example, comma-separated values (CSV) data, image data, video data, and IoT data

• The velocity of data generation (batch data or streaming data)

• The current process to manage data access

• The location of data storage: cloud or on-premises

• Whether the data solution must support a hybrid scenario

• If data is on-premises, whether any cloud migration is planned

• The security and compliance guardrails of the data

• Current data-driven use cases: their maturity and tenancy (cloud or on-premises)

Align phase

After you gather the necessary data points during the discover phase, define the boundary of your data mesh solution, based on your organizational structure. Ideally, you want to have one data mesh solution that encompasses your entire organization. However, large organizations sometimes adopt multiple implementations of their data mesh solution. If this scenario applies to you, consider building a data mesh solution for each commercial brand or each geographic region. While defining the boundaries, consider whether the solution structure is a one-way door or two-way door decision. At Amazon, a one-way door decision is considered to be nearly irreversible. On the other hand, a two-way door decision can be reverted without any significant consequence.

Align with your stakeholders on the scope of the minimum viable product (MVP):

• The technical features of the MVP.

• The lighthouse, or pilot, use cases (business user requirements) for implementing the data mesh–based data solution. The experience gathered from implementing the lighthouse use cases helps create the blueprint to implement future use cases.

• Metrics to measure the success of the MVP.

• The desired scope of the data solution beyond the MVP phase (solution growth).

To identify the technical features of the solution, work backwards from the data user experience. For the MVP, select the minimum required features to fulfill the user experience. While choosing the lighthouse use cases, consider the following:

• Use cases with high cloud maturity

• Use cases for advanced data users

• Use cases that deliver feasible business value

• Use cases whose requirements can be fulfilled starting with the baseline data solution features

Launch phase

After all stakeholders are aligned on the scope and supported use cases, build the MVP of the data mesh–based data solution. Adopt agile practices, such as Scrum or Kanban, for an iterative-build approach to realizing value. Define a roadmap and milestones for the MVP, and establish the data governance mechanisms. The launch phase includes the following key activities:

• Identify the data domains of the data mesh.

• Define the tenancies of the domains.

• Add the lighthouse use cases to the data solution.

• Add the data products to support the lighthouse use cases in the data solution.

• Define the business and technical metadata of the data products.

• Build the data-access management workflow.

• Build data-access patterns for consumer teams.

• Build security and compliance guardrails.

• Build tools to measure data quality and data lineage.

• Build observability tools to notify users, monitor resource use, and track success metrics.

• Roll out the MVP to production.

• Conduct educational and promotional activities. 

At the end of the MVP phase, evaluate the outcomes to measure the success of the launch phase.

Scale phase

In this phase, expand the MVP solution, iterating the scope of the full solution, based on the outcome of the MVP phase. Introduce the features that were planned for implementation after the MVP phase, and add support for the early-adopter use cases. Continue educating the stakeholders on feature enhancements, additions, and operating and maintaining the solution.

Evolve phase

When building a data solution, you're never done. Manage the lifecycle of the solution by revisiting what you have built. Introduce optimizations and new or enhanced capabilities that meet the needs of business users. For example, add generative artificial intelligence (generative AI) capabilities to enrich the business metadata of the data products. Add the late-adopter use-cases to the data solution.

The following figure displays a summary of the activities and the change in the number of supported business use cases in each phase.

The users associated with the lighthouse use cases are the first to adopt the data mesh–based data solution. In the scale phase, more early adopters begin to use the data solution. In the evolve phase, the late adopters follow.