1. Usually, articles are sent to me as word documents, so I open up the document and begin to "track changes." I Correct some obvious typos and put comments directly into the document. The author of the article will never see these comments or corrections, but I find it helpful for myself. I also take "mental notes," sub-verbalizing what I think. What I'm doing is writing a draft the reader's report in my head. At this point I basically know whether it's an acceptance, a review and resubmit, or a rejection.
2. I get a good night's sleep.
3. The next day I draft a reader's report. I use my marginal comments and my mental notes as the basis of that. I outline the strengths of the paper, the exact points that need to be improved. I might make specific comments on style, organization, and argumentation. When I have the reader's report done, I make sure that it corresponds with my judgment of whether it's a resubmit, rejection, or acceptance. For example, I might have thought it's a resubmit, but reading my own report it's obvious that there are too many serious problems. Or, when I outline the strengths of the article, a reject might be promoted to a resubmit. I never change from acceptance to rejection or rejection to acceptance, though.
4. I get another good night's sleep.
5. I look at my reader's report the next day. Is this still what I think? Do I need to rewrite it to soften the criticism?
In some cases I can do 5 on the same day as 3. In other cases I do the entire report in one day, when the flaws of the article are so obvious that I can tell the editor what they are in much less time.
The most frequent kind of article I read are resubmits, followed closely by rejections and outright acceptances with only minor changes This means that I am usually help an author get hir articles into print.