If you want people to look at your paper you need to spend significant time coming up with an engaging title and abstract.
If you want people to keep reading and engage with the scientific content you need to produce clear figures and write effective figure captions. This is not easy. Here are a few suggestions that bear in mind the following reality.
Most people will look at the figures to decide whether or not they think the paper is worth reading. Some will do just that. Thus, the main messages need to be contained in the figures and they need to be quickly ascertained.
1. Try to begin the caption with a short title sentence that summarises the main point of the figure.
Why are you including the figure in the paper? Hopefully, not because "I took lots of data" or "I did lots of calculations." What do we really learn from the figure?
For example "Resistivity exceeds Mott-Ioffe-Regel limit at high temperatures", rather than "Resistivity versus temperature."
2. Make sure the caption is self-contained. The reader should be able to understand the figure without having to refer to the text to find definitions of notation, parameter values, or what the significance of the figure is. For example, don't include acronyms [DMFT, MP2, ARPES, ....] unless they are defined. If the reader has to work really hard to understand the figure they will quickly lose interest.
3. Re-write and re-write. Get feedback. It is hard work.
Generally, the above means that captions can be quite long, e.g. a solid paragraph. Occasionally referees may not like this and say some of the details should be relegated to the main text. I disagree.
I learnt the above from John Wilkins, my postdoctoral advisor. He said the first step in writing a PRL is to work on the figures and captions.