I had this problem when i was a PhD student. Towards the end of my time, i started doing something which fixed it, and which i could have been doing all along: make a list of the figures for my paper, and do whatever bit of work moves the next most important one towards being finished.

I should mention that i was working in cell biology; in cell biology, you earn career points by publishing papers telling a new story, where papers comprise a series of figures, each one showing results which demonstrate some point of your story. Papers also have an introduction, description of the results, and discussion, but those are pretty much filler. The smart scientists i know read the title, the figures, and the materials and methods section. Perhaps not every field is like this.

So, to make a good paper, you have to come up with a series of good figures, which means (a) finding a story you want to tell, (b) working out what points you want to make, and (c) getting results which make those points. Having done that, there's the technical work of making up the figures (learn to love Illustrator), and some writing, and you're done.

Finding the story is usually the easy bit. Your supervisor usually has some untold ideas for stories floating about, or there will be some variation on an existing story that people will enjoy hearing. Or you might just need to spend a year or two trying stuff, which can be fun until it suddenly isn't.

Working out what points you need to make takes time, but once you know the story, is usually not too hard. Crucially, you don't need to know up front that these points are all true. As you work towards them, if you find any that aren't true, you just throw them away, and either replace them, do without them, or change the story a bit.

It's getting the results which is hard. That is the actual work in scientific research. Designing, performing, interpreting, and refining experiments. Finding, preparing, and characterising materials. Learning or inventing techniques. Spotting weaknesses and plugging them with control experiments. Endless, seemingly incessant, tea breaks.

Having a concrete list of the figures you need helps you keep going at getting results, and helps you do the right ones. Before i had a list of figures, i did all sorts of experiments because they seemed interesting, and connected to what i was doing, but which could never have contributed to my paper.