Finding a good dissertation involves some random luck; many extremely smart people don’t stumble on one. That said, there are certain ways to maximize your chances of zeroing in on one. A good dissertation 1) is about something that matters, 2) addresses significant (potential) variation in the real world, 3) shows that you know and understand how other people have explained this (or similar) variation, and 4) makes extremely clear how your explanation of the variation is demonstrably different from other, existing explanations. So:

Choose something that matters to lots of people—and optimally to you.

Think about potential variation in that thing (past, present, or future): how could it be different? How could it have been different? How might it be different in the near future?

a. The best kind of variation is real variation (within a case or across time you have clearly different outcomes of some sort). But counterfactual variation can also be a foundation (and sometimes is all you can get if you’re looking at fairly unique historical events): something very important could have turned out differently.

Try to categorize arguments about why your thing was/could have been/is becoming/will be different. Ask as many people as possible for suggestions on what to read. Look for theoretical logics that might imply something about your thing, even if they aren’t written on it directly.

b. The key here is to track logical arguments, not just read stuff immediately written on your subject. An article written about city councils in Kansas might have an organizational logic that applies to the evolution of the IMF. The only way to discover such sources is to boil down arguments over your subject to abstract logics, and then ask a wide range of people with different specialities if they’ve heard of arguments that take a certain form.

Write a piece about these existing arguments, and why you suspect they are wrong or incomplete. If you can do this clearly, you have a dissertation topic. If you add in some concrete justification of a particular case or cases and some justification of a particular methodological approach, then you have a dissertation proposal.

Effective writing habits vary a great deal across people, but I think these pieces of advice apply in almost all cases:

Go back and forth between research and writing. Whenever you aren’t totally busy doing actual research, write something. Especially in qualitative narratives, only in writing do you often discover what you really need to find out in your research.

Break it into tiny pieces. Figure out when you want to have an overall draft; when you need to finish a given chapter; what you need to accomplish this week; what you need to accomplish today; what you need to accomplish this morning vs. this afternoon. If you meet your goals (which will ALWAYS be overambitious, but so what?), allow yourself to stop working.

ALWAYS write down anything that comes into your head the minute you think it. This is true when you’re actually writing—if you think of a phrasing for an upcoming paragraph, jump ahead and jot it down—or at any other time (carry paper with you everywhere). ALL dissertations turn on wording. Research doesn’t matter much without the best possible phrases to present it. A weak dissertation can become a great one overnight if you find the turn of phrase that illuminates it best.

Get others to read your work constantly. Not necessarily your committee (whom you may not want to bug)—just anyone.