Same reasons webdevs crush pages with 112 js and css files when you could, you know, just render the page like you want it in the backend and save 90% of that overhead. It's human nature to "stick with the working solution" even after it outgrows itself many times over.
You'd be very surprised at how many smaller businesses are basically run using Excel spreadsheets. Accountants that are only slightly technical absolutely adore them.
I know of a fair few banks that settle trades using Excel spreadsheets.
It's a familiar interface with well-understood operational primitives (copy, apply, combine, etc.). SQL/Python/writing declarative code introduces an interface which has a substantially steeper learning curve for anyone who isn't a power user, and is a step removed from the ease of use of drag-and-drop.
How is that manageable or efficient and not error prone compared to a database or just csv processed with python or other code?