https://dataproducts.substack.com/p/the-rise-of-data-contracts Thought this was an interesting concept

Our team does this, and it does have its downsides. A few that the post glosses over: • When engineering/product are used to you just pulling the data from their db, it's a lot harder to all of a sudden get them to dedicate time for developing and emitting events. • Generating history is a lot tougher. Compared to CDC/federation, you always live in a little constant fear that some service will go down and it will be a big pain to recover those events. It's also harder to validate/test if you are truly in sync. • You definitely will end up with less data than if you go the CDC route. You will end up making some concessions when prioritizing what data is actually important. Meanwhile Analysts/ML folk are used to just getting everything under the sun. For event sourcing to be successful, you really need your engineering/product to already be event driven. Otherwise you face an uphill battle getting developer time to learn about, create, and maintain events.

This is great context Dan! As always there are a lot of tradeoffs. And not everyone has the luxury of starting greenfield across the board to do all event based stuff.