UX research participants are usually compensated, and so it's usually done to drill down deeper into problems that analytics found and test possible explainations. I don't see what's unscientific about that.