One thing that served me incredibly well (finished from BS to Ph.D. in 4.5 years) was to very early on write a proposal that clearly stated exactly what questio

One thing that served me incredibly well (finished from BS to Ph.D. in 4.5 years) was to very early on write a proposal that clearly stated exactly what question I would work on, and stating clearly what the approach and work product would be. This all but eliminated feature creep, and I could then go back and say "you agreed that investigating this was sufficient." Then it was just a matter of doing what I said I would do, and be done. It helped focus my attention, because anything that didn't address the thing in my proposal was more obviously tangential. It focused my entire course of research and kept me from spending too much time on side projects. But most importantly, it established in writing exactly what question I was going to answer and roughly how. I didn't say "I'm going to prove that <foo> does <bar>." I said "I'm going to test to see if <foo> does <bar>." If I do the tests, and show the relationship between <foo> and <bar> I've done science. The goal is to create an experimental plan that tests a theory, not to build the best whatever possible. The job of a Ph.D. student is to demonstrate that they can carry out research, not to build the best widget. If you can, great, but that's not what a Ph.D. means. I saw endless examples of people who waited until they had better data to put together a proposal. The problem with this approach is that it raises the bar by removing risk from the project. If you spend five years demonstrating the <foo> results in <bar> then your adviser will ask for you to spend another two years doing an actual project extending that result. If you clearly elucidate early on that it's not clear that <foo> results in <bar> then simply seeing if that's the case or not is often sufficient. Formal proposals are like a contract between yourself and your committee establishing the scope of project that is sufficient to graduate. Doing this as early as possible can only help you, and in my experience any work prior to establishing the scope merely results in a larger scope later.