Well, I started preparing Reds about two months before I first posted. I didn't begin with a PoD, because I was less concerned about counterfactual implications, and more concerned about how I could construct a plausible narrative to achieve some sort of socialist revolution. So I selected a plausible point in history for it to occur (the Great Depression), and began working backwards from there, familiarizing myself in brief with the issues of the era.
That did work out really nice for me, actually, because I was writing a term paper concurrently in a parties and elections class, and the literature review I did for the paper gave me a good background on the issues of labor politics and economics in the late 19th and early 20th century. So I compiled some major sources, some from books (btw, anyone writing a TL that requires work on turn of the century political economy, Richard Franklin Bensel's Sectionalism and American Political Development 1880-1980 is excellent), and the rest from journal papers from political science, history and sociology journals.
So I picked my PoD, sketched out the major divergences and followed them through. When it came time to deal with major topics, like the First World War, for example, I'd track down new sources, devour them and use what I learned to plan out the next section. I'd fill in the gaps and little trivial bits of information and chronologies by Wiki-diving. Writing yearly reviews is mostly just information collected from wikis, but sometimes some of the big sources crop up.
It's gotten to the point that it is complicated enough I keep notes on my computer of future events, and complete copies of everything I've already written so I can keep things consistent. Obviously, it doesn't always work. Usually, though, when I read books or journal papers, I annotate them and write in the margins if I own them, or keep a seperate notebook with detailed notes if they are borrowed.