Excel is fine for opening up a table and doing some quick numberwang.
But, as soon as you have to take your piece of work and start making little variations and tacking bits on, or running it on different bits of data, or God forbid you want to actually test your code (and make no mistake, code is what you are making), well, it all involves rather a lot of clicking and opportunity for fuck-ups.
Of course, R itself hits limitations pretty quickly. While technically it's a general purpose programming language, trying to use it on non-tabular data or to assemble even a medium-sized program leads to pain.
My rough rule of thumb is:
* Checking if it's got vaguely sensible data in? Column https://linux.die.net/man/1/column.
* Just looking, maybe aggregating a column or doing a simple summarise? Excel (or Libreoffice, why not?).
* Giving outputs to someone else or might need to do it more than once? R.
* Need sensible data structures and useful abstractions? Python.
* Need more speed? C or Java or whatever.