Monday, July 06, 2009


1) Ellora's Cave
2) Amber Quill Press
3) Loose Id
4) Samhain
5) Liquid Silver Books
6) Cobblestone Press
7) Torquere
8) Changeling
9) Wild Rose Press
10) Freya's Bower


Anonymous said...

Hi there,

I've come late to this discussion, but could you clarify these figures? Are these annual sales ranking, say for 2008, or are they just for a particular month? And what are the sources?

Thanks very much,

Kris Eton said...

I think the white parts of the columns are average first month's sales for all the publishers. The greyed in parts of the columns are 'total' or overall average sales for all publishers.

In other words, the 'sell through' of a book.

So it is interesting to see that Amber Quill and Loose ID have lower first month's sales (avg.) than Samhain, but have better sell through numbers.

Alisha Rai said...

Hmmm, maybe my data graph skills are rusty, but I thought the bars (the gray area) reflected the margin of error.

Kris Eton said...

If you click on the pic to make it larger, you will see on the side the vertical axis is marked as "sales (first month)/total." That is where I got my info from...and I believe the black lines within each bar on the graph show the range that created the 'averages.'

In other words, if you look at the EC example, the avg. first month sales were ~800, sell through sales were ~1200. However, the range of first month sales goes from ~600 to as high as ~950, the range of sell through sales goes from just under 1000 to almost 1500.

Just my guess here...but I think that is close

veinglory said...

Apologies for the delay in responding, I am travelling for work all this week.

The total bar is the average total sales for all books reported, but most of these are still in print. So it is not sell through, just sales to date.

The white section is sales in the first month.

The vertical bars show standard error whicn indicated variability but is a much small value than total range. (It is the mean divided by the square root of n).

The advantage of the standard error is that non-overlapping SE bar indicated a likely significant difference at the .05 level.