Lightning Talk: You Can’t Do Data Mesh Without Governance

Lauren Mafeo, service designer at Steampunk and author of 'Designing Data Governance from the Ground Up,' argues that data mesh architecture cannot succeed without strong governance rooted in organizational culture. She identifies a core problem: despite increasing data volumes, many companies become less data-driven because governance is a cultural challenge, not a technical one.

Data mesh architecture promotes a 'data as products' mindset, spreading ownership across specific domains managed autonomously but within a unified architecture. Mafeo illustrates this with JPMorgan's AWS implementation, where a dynamic catalog maintained by domain-specific processes allows teams to discover and request data more easily. However, this architecture fails without investment in data-driven culture.

Mafeo outlines three essential steps for success:

1. Define data domains by answering four key questions: What are the organization's key areas of interest? What problem spaces does it address? How is it currently architected? Who leads each problem space?

2. Appoint domain-specific data stewards—leaders of each domain who may not hold technical roles. A VP of marketing, for example, defines what data quality looks like for marketing data (fitness for use) without necessarily building ETL pipelines.

3. Define data quality by establishing success and failure conditions for each domain, enabling automation within processes.

Mafeo emphasizes that data governance cannot be siloed; given the scale of modern data, experts from each domain must be involved in co-creating governance with cross-functional teams. Her key takeaway: data mesh allows teams to access, develop, and manage data autonomously, but only after the hard work of building a data-driven culture and automating standards into technical processes is complete.