Upwork measures proposal success rate, which is helpful. I believe a different metric would be more helpful: Proposal Success Rate for jobs that were actually awarded. That would alert freelancers to their success level in targeting clients. Being able to predict which jobs are ever going to be awarded is a necessary skill. Today, the freelancer has to calculate this herself/himself. A world-class proposal for a job that won't go to anybody is a signifiicant waste of time, and can prevent freelancers from applying to jobs where client satisfaction would be high - a plus for Upwork as well.
Until I dropped out in 2014, I won about 75% of all actually-awarded jobs to which I applied on Upwork's predecessor, Elance, and its competitor, Guru. I've rejoined the fray, but at a signiificantly reduced level, and have a "newbie" record on both boards. I'm winning about half of the actually-awarded jobs to which I apply. I'm still learning the Upwork client base, and slowly improving my ability to predict which jobs are likely never to be awarded, and which clients I wish to avoid.
Upwork tracks, by client, percentage of jobs that are awarded, which is extremely useful. I've been successful in identifying key words, phrases and requirements that indicate a client likely to be unreasonable, or to be promising unachievable results, and several other red flags. I'm more than willing to provide my red flags list to anyone with the time and i nterest to do a global search of postings.