here's the best advice you can follow:

NEVER go "overtime" on questions.

on quant, this means that you should NEVER, EVER spend more than 3 minutes on a problem. ideally, you should try to stay within about 2:30 (2:45 at the very most), but you can use the 3:00 barrier as the absolute event horizon of the proverbial black hole of time management.

after you reach that point, you will pretty much never be doing anything except deliberating. DO NOT EVER DELIBERATE; instead, either
(1) immediately select one of the remaining choices (if you're deliberating between 2 or more equally attractive choices) - don't be buridan's ass
(2) immediately abandon the technique you're trying to use, and try a secondary technique such as number plugging (if you're deliberating on a strategy or approach)

if you follow these recommendations, then you should find that you either (a) solve problems within the allotted time, or (b) just don't have a clue. if (b) is the case, then just "punt" the question - use process of elimination, or, in the worst-case scenario, just guess at random, but do so QUICKLY. remember that any random guess is just as good/bad as any other random guess, so there's no point in thinking too hard about which random guess to make.

also, make sure you solve problems according to this hierarchy:
(1) try to find a "textbook answer key" method for the problem.
if not, IMMEDIATELY
(2) try to plug in numbers (for problems with variables, word problems, or equations) or just list stuff (for probability and counting problems).
if not, IMMEDIATELY
(3) try to estimate the answer
if not, IMMEDIATELY
(4) try to use some sort of process-of-elimination
if not, IMMEDIATELY
(5) guess.

if you can learn to work your way through this hierarchy QUICKLY (i.e., actually respecting the meaning of the word "immediately" where you see it), then you should see most of your time-management problems begin to evaporate.
most time is lost when test takers don't drop to the next level of the hierarchy fast enough, or don't at all.