I know of two common methods: by using z-score or by rank, the latter is simpler to explain let's start with that first. Say they interview 200 candidates, each candidate will have three rank numbers between 1 & 200 for the three components. After weightings are applied the three weighted numbers are summed then ranked again against the other 199 applicants.
For example A is ranked 5th ATAR, 10th UMAT, 50th Interview whereas B is ranked 10th, 50th, 5th respectively. If the three components are equally weighted A and B would be equal in the final rank. With 40/20/40 aka 2:1:2 coefficients A= (5*2)+(10*1)+(50*2)= 120, B= (10*2)+(50*1)+(5*2)= 80. In ascending order B's final rank is higher.
In the other method the three components are each standardised to a common scale called z-score. It's a number indicating how many standard devs you are above or below the cohort's mean. Therefore regardless of how ATAR or UMAT or Interview score is presented you end up with a z-score mostly between -3 bottom and +3 top (meaning you are 3 standard devs above the mean for that component).
Weightings are applied to each candidate's three z-scores, then they are summed and ranked in descending order for the final ranking.