As a junior researcher with a little experience in writing articles, I have asked a few senior scientists about the part that they think is the best to start writing a manuscript from. Some experimentalists propose to start with the graphs and their captions. For a theory article, some scientists like to start with the the main body of their calculations and then insert the words in between the formulas.

I personally find it conducive to begin a manuscript with the list of references. It helps the author to define the extent of her report. She can assure herself that she has learned enough about the work that has been reported before (copy-paste citing is certainly not helpful for this aspect.) If not already clear, the author can decide about the terminology and the use of symbols, based on the most commonly used notations in the most relevant references. By reading all the abstracts, the author can summarize for herself what are the most admissible key-terms that she must emphasize on, in her introduction.