Matt Blind over at ComicSnob has been compiling lists of manga bestsellers from online sources, now he explains how he does it:

While I enjoy compiling the rankings, and am a little proud of the results, I’m the first to admit they’re flawed. Online sales are just one (already distorted) piece of the overall sales picture, and despite all the math involved, we still don’t have any actual sales numbers: we traffic in comparative descriptions. It’s easy enough to find the top 10 or even top 50 manga selling from most sites, and we can trust that #1 is beating #22 by a fair margin — but 1 vs 2, or 27 vs 28? How many more copies does a manga have to sell to gain even one spot in an online bestseller list?

I try to make up for that by looking at several sites. (and with another trick introduced this week; more on that later.) Also, the rankings from each site are weighted — though to date that has been a simple matter of counting the top sites (Amazon and B&N) twice, nothing fancy. It so easy I’m surprised other folks aren’t doing it… well, other than the aforementioned soul-draining tedium of raw data entry.