You've got a budget, you need to decide where to allocate the resources. That's it, really.
Much of it comes with experience, learning what is essential versus what is optional. Good designers know both. Certain elements are key to the gameplay, certain other elements are key to the feel. Often the key emotional components are not key mechanical components, so the designer needs to recognize both. Then the designer needs to recognize what are nice, things that contribute but aren't essential.
Start with a big list, sort it. Keep it updated and sorted as time goes on.
One effective approach is to start with that list, get estimates for what they all cost, and go through it until you run out of budget. Anything below that mark gets cut. It's often good to do that in a range, these top x things are green and low risk, this range is yellow and might get worked on, the rest is red, wish list that won't be accomplished. It works well for everything from bi-weekly sprints to overall feature wish-lists.
While working on the Sims Store, with each person creating hundreds of items we got pretty accurate at estimating the amount of work as time went on. Designers could come up with their wish lists and each item would get an effort score from each groups, Effects, Animation, Modeling, Engineering, Sound, (so FAMES) score. Designers knew how much of each resource they had to spend, so could allocate them accordingly. For example, the designer could put together a set of features that were animation heavy for one game object, but would need to compensate with another object that is light on animation. Similarly one object could be engineering heavy with a lot of code complexity, but the next would need to be engineering light. Plenty of good ideas were cut because there wasn't time in the budget from a given discipline. There were also lots of little suggestions of easy or nearly-free features that could be thrown in when someone knew there was an easy gain in the system that could be leveraged.