Day 2 of a client hackathon. We're building a data product using Data Mesh principles. Rather than thinking about data as a left-to-right, source-to-consumer model, we design data products outside-in, by starting with the output ports, which are closest to the user. Then we discuss input ports, which allows us to explore what data we have. Only when we know whether we can meet user needs with the data we have, do we worry about the processing and architecture.

This approach lets us split up and parallelize work, using test- and behavior-driven design. And that's how we build a full end-to-end data product in two days.