Someone (Rookboston?) posted an excellent analysis of B's general trade-up history a few years ago, which basically was that he targets a group of several players, one of whom he wants to get. If, say, thaere are four players, one of whom he wants, and the third is taken at 17, he will immediately try to trade up to get the last one. However, if two or three are still available at 21, he will trade down, figuring he can trade back up later on. This ensures never grabbing a guy sooner than you have to, does a better job a getting a guy you need, and results in extra late round picks.
Also, an article a couple years ago said there BB and SP list the "impact" players in a draft, of which there are usually 15 to 20. They say that palyers do not form a continuum, but have plateaus of "impact." The idea is to get an impact player with the first round pick, and whenever possible pick at the bottom of each plateau group, rather than at the top. That way you pay less for comparable players, allowing you to spend the money on the lower third of the roster.