Data finds its own path. Design the landscape, not the channel.
I come from coastal engineering. Water is stubborn—it finds the path of least resistance. The harder you force it into rigid channels, the more likely it breaks out or silts up. Data behaves the same way.
Organizations build rigid data channels: pipelines, handovers, governance walls. Then people work around them because they just want to get their job done. You can't force data into structures that don't fit reality.
In coastal engineering, we learned to work with nature instead of against it. Build with natural processes, not concrete walls. Data mesh is the same idea: let data meander through domains that understand it, rather than channeling everything through a central dam.
Design the landscape, not the channel. Give teams space and tools. Set standards so streams connect properly. But don't try to control every drop of water.
Rivers silt up without maintenance. Data quality degrades without ownership. The key is putting maintenance where the knowledge is—with the teams who understand what the data means.
You built a central data lake and hired a data team. It worked well at first. Then the team became a bottleneck.
Let the teams who actually understand the data manage it. They know their domain, they know what the data means, and they can answer questions faster.
Data mesh builds on four ideas
Teams manage the data they understand. Split data by business domain, like you split software into services.
Your data serves other teams. Make it good: reliable, documented, and usable.
Build a platform that lets domain teams create and manage their data products without constant central team involvement.
Set standards so data products work together. Define quality rules, security policies, and compliance requirements.
Exploring data mesh patterns, implementations, and architectural decisions
I'm Kees Pruis, a civil engineer turned data architect. I work with organizations in manufacturing, maritime, and logistics to design data architectures that actually work.