The goto place is the Nielsen Norman Group¹. You can get any of their books, it will describe how to do tests. It's basically catch up to 5 users from your target audience, ask them to do the tasks you want, observe without saying a word, and don't ever come out with any variation of "users are stupid" explanation because even if they are, they are the ones you are creating the software for; fix your software, and repeat.
But if you want details about why those large "record the user" sessions aren't worth it, you'll have to dig into their papers.
But if you want details about why those large "record the user" sessions aren't worth it, you'll have to dig into their papers.
1 - https://www.nngroup.com/