Some interesting BookScan observations...
BookScan is one of -- perhaps THE -- industry's best ways of tracking actual sales of books, aside from the actual receipts that finally trickle into the publisher from all quarters. It doesn't get ALL books sold -- estimates I've seen vary, depending on type of book, from getting about half the books sold to getting a quarter of them -- but it does get a large number and geographically widespread.
A few months back, Amazon began to allow authors access to their BookScan data. This has been only partially useful for me in the immediate term because the data didn't reach the really key period, namely the months of release and immediately thereafter for Threshold and GCA.
However, I have had enough time to accumulate some interesting observations. One of the most interesting from the past few months is that aside from a drop immediately following Christmas (unsurprising) there is NO clear downward trend at all; my books seem to be selling at a pretty much constant rate, or even slightly rising over the last 8 weeks, which I'm somewhat surprised by. I was hoping for a long tail, but a tail still decreases, and if this is decent data, the tail is shrinking much more slowly than I would have imagined.
The same area in Author Central also gives you access to long-term ranking data on Amazon, which showed in spectacular fashion the spike in sales for Boundary that followed the release of GCA and Threshold. It's especially striking compared to the sales for my only other in-print book, Mountain Magic, which did go up slightly but not terribly high.
Mountain Magic is not a completely dead property, though; it seems to average 1-2 copies a week through BookScan. Not much, but every sale helps.