IQ is a normal distribution with 100 as the mean, i.e. it is a relative measurement. So, say you wanna measure a 10 y/o IQ, you just run that result against peers only. IQ is known to be correlated with 'success', i.e. money, happiness, school performance. It might not be a perfect test, but it is indicative (and there lies it's functionality).

But you meant to say how do we measure intelligence. Perhaps the better question is, what is the goal for measuring? What does anyone want to get out of it?

By no means should it be interpreted as a final verdict. All tests have a sensitivity vs. selectivity trade-off. Which means no test can be fully discerning nor fully registering. But you're gonna have to use 'a' test (dependent on the goal).