That is true. I do believe they need to be able to tell wether or not an specific employee has participated though, to make sure an employee doesn't answer twice, or even 20 times. If a survey is 100% anonymous, with nothing tying anyone to anything, any employee could simply invalidate the results by spamming the survey. You'd need to be able to somehow tell wether or not an employee has answered more than once.

The correct and ethical way to do it is obviously randomised links, that can't be traced back to the individual person, with no registration of who the links were sent to. Not that any company has ever done anything the correct or ethical way though. I'm just saying, there needs to be some way to make sure the results are actually an accurate repressentation.