The problem with claiming upwards of 95% accuracy when guessing ebook sales based on relative chart positions is that when the people with the real numbers tell us the true figures, and the guestimates turn out to be only 60%-70% accurate, we have a problem.

Bookstat, the $10 million paywall book sales calculation site run by Paul Abbassi, alias Data Guy, that morphed from the free to all Author Earnings Report, makes just such a claim to accuracy.

But lately, thanks to the UK trade journal The Bookseller recently electing to publish the top ten ebook sales list as provided by Bookstat, we are seeing a very different picture emerge, that serves only to confirm past questions raised by TNPS about the Bookstat numbers, which as Author Earnings numbers has a track record for being less than solid.

Soon after The Bookseller began publishing the UK top ten ebook sales list numbers as provided by Bookstat, some publishers chose to report back to The Bookseller the real numbers, thatwere not quite in line with Bookstat’s guestimates.

And when it became clear that sizeable discrepancies were a regular feature (The Bookseller reports the publisher numbers as a note of clarification) TNPS began tracking the differences.

Having wildly overguessed in March, Bookstat reined in its numbers for Hachette titles in April before Easter, but the first report since Easter shows that once again Bookstat is quite unable to offer us accurate numbers for Hachette titles, which begs the obvious question – are any of the other numbers accurate?

Here’s the latest Bookstat UK ebook top ten chart as presented by The Bookseller.

What’s not to like? A detailed breakdown of the UK’s bestselling ebooks across all platforms. Or so Bookstat would have us believe.

Except, The Bookseller carries a quiet caveat:

Clarification: Hachette has stated that The Flatshare sold 20,176 units, Blood Orange 15,442 e-books, and In Five Years 12,392 digital copies.

Okay, so nobody expects perfection. But we have been told by Bookstat to expect a 5% either way accuracy.

Here’s the problem.

Bookstat tells us that The Flatshare sold 15,335 ebook units in the given time period. This while claiming 95% accuracy.

Hachette reported to The Bookseller that the real sales were 20,176 – an under-estimate of 4,821 units. A 25% error.

Bookstat tells us that Blood Orange sold 10,295 ebook units.

Hachette reported sales of 15.442 for the same period – an underestimate of 5,147 units, amounting to a 33% error. An error of 33%.

Bookstat tells us In Five Years sold 8,602 ebook units.

Hachette tells us the real number was 12,392 – an underestimate of 3,790, amounting to a 30% error.

The other publishers choose not to report if their numbers match the Bookstat numbers, but when 30% of ten guestimates by Bookstat can be so far out, what chance any, let alone all the others are withing the 5% error margin Bookstat stands by?