Consider a surgeon: how many lives does she save in her career?  Perhaps she does 6-9 surgeries per week of which 2 are life-saving, working 45 weeks per year from age 25 to 70. Multiplying up we get 4000 life-saving surgeries. This is pretty good: to get similar results by giving to GiveWell's current top charity you'd need to donate $200,000 every year for forty years.
But let's imagine she gets sick of the job and retires early, at 47. What happens to the people that she would have been operating on? The other doctors work somewhat more to keep the operating room staffed while the hospital hires a replacement. The 2,000 lives she would have saved in the remainder of her career end up being saved by other surgeons.
What is the positive effect of becoming a surgeon, then, if anything you do someone else would probably do instead? Perhaps you're an unusually good one, managing a success rate much better than the person who'd be in your place. Perhaps you save 4,400 lives over your career while your replacement would save only 4,000. This is 400 people who get to keep living because you chose surgery instead of some non-lifesaving job.
We tend to be optimistic about our abilities, however, so we should acknowledge that you could just as well have a success rate much worse than your replacement. If you save 3,600 lives instead of 4,000 you are effectively killing 400 people over your career.
We need to make a distinction between the good that happens because of our choices and the good that happens through us but because of other people's choices. Would the people employing you just hire someone else to do the same thing if you weren't there? If you're in a position to target your work, however, perhaps as an academic choosing to work on existential risk, a doctor raising money and building institutions to help especially poor people, or a research analyst starting a charity evaluator, then you're doing more. This is a big reason that the make money to give away approach is sensible: the people you displace in choosing a relatively high paying job wouldn't be donating very much.
If you're choosing a career in part for altruistic reasons you need to consider replaceability: how much more good would you be doing than the person you'd replace?
 This is all written in terms of "lives saved", but I realize that this isn't the best metric. It's just an easy one to think about, and a good approximation for what we do care about.