Most of the reactions seem fall into three categories: 1. "You (the individual) should stop using Excel". Good advice. As the article mentions, though, if yo

Most of the reactions seem fall into three categories: 1. "You (the individual) should stop using Excel". Good advice. As the article mentions, though, if you ever send a tabular data file to someone else to edit, there's a fair chance they will open it in Excel and corrupt the data. Excel's CSV import/export cycle has even more traps than using xlsx consistently, so in some sense individual avoidance of Excel makes the problem worse. You could send SQLite files or somesuch so your colleagues can't work with your data, I suppose. 2. "You (all scientists / whatever non-programmer group) should stop using Excel for data analysis". True, but getting everyone to change is hard. Both top-down (thou shalt not use Excel!) and market-driven (Airtable! LibreOffice!) haven't worked so far. Hopefully there will be more progress on this front. 3. "Scientists should learn to use their tools". Even if a user knows all the scenarios in which Excel corrupts data, it cannot in general be stopped from doing so. If you set column types up front, type every value manually (no copy-paste) & save and check on each edit, and then never re-save the file again or let anyone else re-save it, you're safe. Probably. I sort of expect a bunch of replies saying this isn't safe at all. But if the file is edited, you need to check the whole thing. Eventually an error will slip through. The only way I've found to safely work with Excel files, or files that might be edited in Excel, is to put the data files under version control and always check the diffs. Often someone else's import/export cycle will change the quoting style on the whole file. In this case, the work needs to be repeated so the diffs are clean and the changes can be verified. This is a necessary but very obnoxious part of this process. Version control targeted at non-programmers and non-text files would help a lot. There are also tools to to flag potential data corruption errors [1], but error detection in the absence of the original data won't be perfect. And if it is manually run, it won't be run consistently. Most people don't have continuous integration pipelines for their data. Excel is very powerful and has a lot of potential to improve non-programmer productivity. It is unfortunate that avoiding data corruption is such a minefield. The situation could be significantly improved by anyone with a decent product and a great marketing strategy. [1] http://maplab.imppc.org/truke/ http://maplab.imppc.org/truke/