Obviously, generating a lexicon is very time-consuming, and there's so much that needs to be done, I often find myself suffering with analysis paralysis, because I just can't decide where to begin.
Word frequency lists don't help all that much, because there's an inverse relationship between how useful a word is and how common it is. The most commonly used word in English is 'the', for instance. If you just made words for the 100 most commonly used words, you wouldn't really be able to make that many sentences, which would make it difficult to work on the grammar.
I've seen some people use the Swadesh list, but I fail to see how that is useful in any way since the list was arbitarily chosen rather than thought through. And besides, it would require you make words for things that are highly unlikely to ever be mentioned, such as a word for 'louse'.
The Conlanger's Lexicon suggests you work by categories, so that that way you can at least form complete sentences, even if this does limit you to writing about one topic. Of course, there's still the problem of which list to begin with. He just lists the categories in the book alphabetically.
Are there other methods to use? And yes, I know about word-generating programs, but those spit out stuff that's much lower quality than stuff generated by hand. And besides, my conlang will have a highly productive word-derivation system, so I don't need that many roots (when I'm writing down lists of words, often times I just write what I want the compound word to literally translate as, though I rarely go back to fill it in because I never get to the words that would comprise the compound).
I've mostly been looking at frequency lists and the Conlanger's Lexipedia, but I would like to know if there was a better way.