If you're doing follow/unfollow to gain followers on Tumblr (I know there are methods other than that) then this might be of interest to you. The technique that I use that seems to be the one advocated by people on BHW, is to scrape user names from the notes (reblogs/likes) of other blogs that are similar to yours.
I did this for a bunch of NSFW blogs and was able to scrape about 1.5M user names from about a dozen NSFW blogs and then I followed between 170-195 of those users per day and got about 8% follow back on average (not including organic growth). I decided to check the 'quality' of these blogs that I was following by writing a script and collecting data from the API. This is what I found.
1. 6.6% of blogs no longer existed. (even though the accounts were no longer there, their notes still remained)
2. Only 14% were set as NSFW (even though they had rebloged or liked NSFW content)
3. 36% had not posted in over 30 days (basically inactive and therefore basically useless)
The most important stat I came up with is I data mined for blogs that had posted in the last 7 days and had more than 500 posts (the API does not let me see how many followers a blog has apart from my own). My assumption here was that blogs that fell into this category were more likely to a. still be around and therefore possibly follow me b. have followers themselves so that if they reblog my content other people would see it and c. might actually reblog my stuff in the first place.
Of the 1.5M blog database I had, only 42% of blogs fell into this 'active' category. I then cleaned the list and brought it down to around 630,000. I then plugged that into my follow/unfollow bot and my followback rate went from 8% to 19%, my organic growth also went up but I don't have exact stats on that.
Hope some of you find this info useful.