Quote:
Originally Posted by darryl
That is, of course, why authors are upset that the latest report identifies them with their information. I assume you are referring to one author, Dean Wesley-Smith, who writes in a comment on the Passive Voice Blog:
He is one author who has sales through channels DG doesn't currently reach, so his sales on the report would be understated. The Bookstat website claims date coverage of 96% of US ebook purchases and 94% of US Ebook dollar sales.
The plain truth is that the fact that the data coverage is not 100% does not mean it is comprised of "made up numbers" nor that it is not fit for its purpose. 100% accurate figures are indeed not available. But 90 percent plus accurate figures are. By your reasoning a country could never rely on census data, since it is not 100% accurate. Nor could business rely on just about any industry sales report.
|
Actually no, John Scalzi is the author that I was referring to. He's one of the few authors who will publicly break out his sales figures. He mentioned that the Data Guy was off by orders of magnitude in one of his blog post. He compared his actual sales figures with what was estimated and he sold almost twice as much.
The key word in your post is "claims".
Computer models, which is basically what this is, are an interesting tool, but accuracy is a key. Being off that much doesn't imply 90% accuracy. There is a saying in programming, Garbage in, Garbage out. Just because something is driven by a fancy set of calculations doesn't mean it's accurate if the assumptions driving the calculations are faulty. We see that all the time in political polling. Political polls tend to be off by far more than the listed margin of error a large part of the time, at least compared to the actual voting numbers.
It's a bit like the NYT best seller list. It's a great marketing tool and certain can help drive sales for individual books, but most recognize that being on the NYT best seller list doesn't really mean you are at the top in actual book sales.