Thread

Hm. Hard to say in general. Data infrastructure tends to be quite organically grown, so it is very specific to the companies and their history. I think you get the whole range from „data scattered everywhere“ to „we put a lot of work to have one source of truth.“

@rorcde What I was a bit surprised is that companies that have built data infrastructure for analytics is often in a better shape for ML projects than without. I had expected that the data tends to be very analytics focussed, but I think many metrics require quite detailed data.

@rorcde And then there is always the danger that you „overengineer“ access control and processes, especially for bigger companies. But yeah, I think data is as much an organizational challenge as it is a technological challenge. Who owns data, who is responsible for quality, etc.