I like CSV because its simplicity and ubiquity make it an easy Schelling point in the wide world of corporate communication. Even very non-technical people can, with some effort, figure out how to save a CSV from Excel, and figure out how to open a CSV with Notepad if absolutely necessary.

On the technical side libraries like pandas have undergone extreme selection pressure to be able to read in Excel's weird CSV choices without breaking. At that point we have the luxury of writing them out as "proper" CSV, or as a SQLite database, or as whatever else we care about. It's just a reasonable crossing-over point.

CSV is a flexible solution that is as simple as possible. The next step is JSONL.

https://jsonlines.org/

https://medium.com/@ManueleCaddeo/understanding-jsonl-bc8922...