> But I’d say the error rate is actually very low, maybe .1-1% and nowhere near 10-20% of data being messed up.
The thing with CSV-related issues is it's usually not a fixed percentage but instead depends on the data.
I work in the VoIP industry so I deal with the FreePBX Asterisk GUI quite often, and it uses CSV as its bulk import/export format. This mostly makes sense as the data is almost entirely (with one notable exception) simple tables that fit nicely in to rows and columns. The issue I run in to most commonly with this is that it doesn't quote numerical fields, and as a result the fields for extension numbers and voicemail PINs can be problematic when they contain one or more leading zeroes. All of the major spreadsheet software I've used defaults to dropping leading zeroes from columns they've decided contain numerical values, and this results in broken data in these cases. It's of course relatively rare for users to choose a voicemail PIN starting with zero and even more rare for extensions to be set up with a leading zero, but both happen regularly enough that I need to remember to manually configure those columns as "Text" when opening an extension export CSV.
Either way, how often the problem occurs depends entirely on the data being sent through this pipeline. Most sites will never see the problem on the extension column, but one of my sites where the company liked a user's extension to be the last four of their DID when they were initially set up 20 years ago has a dozen of them in a row.