“I'm not sure [psychometric tests] stand up as people tend to answer questions differently on paper compared to how they act / behave in reality.
”
If test-takers try to cheat the psychometric tests (eg because they want to fool the recruiter they're the next Richard Branson

), there's quite often a nifty little formula hidden in the test itself that'll identify just how much cheating is going on and correct for it.
It's quite possible to tell the difference between deliberate cheating and test results that are "off key" for some other reason.
It's part of the research done before tests are published to examine how well each question on the test predicts behaviour. As regards a sales test, for example, you might compare what answers a test-taker gives with his / her relative success rates in achieving various sales performance targets.