Wednesday, May 24, 2006

Amazon as Marketing Data

Amazon.com publishes rankings of books; they appear to be based on sales over a fairly short time span, perhaps a day. It occured to me some time ago that it ought to be possible to use that information to measure the effect of reviews coming out, advertising, talk radio interviews, and the like, information that should be useful to publishers trying to sell books.

I am now doing the experiment. Yesterday Forbes.com published a column of mine and included links to both my web site and the book's web site. My ISP provides information on how many hits each of my pages gets, Amazon on how my sales are doing. By looking at those two measures, one showing attention and one sales, and seeing how they were affected by the Forbes column, I can get some idea of to what degree publicity for my non-fiction work helps sell my novel.

So far it looks as though it does. Daily hits on the book web page for the week before the column appeared ranged from 22 to 39. Yesterday—the column appeared at 3 P.M. EST—the page got 83 hits. The effect is small relative to the number of people reading Forbes—they didn't all go to the web page and then order the book, unfortunately—but substantial relative to the level of attention the page usually gets. So far today it has gotten 58 hits.

The results from Amazon are also positive, although my ranking there bounces around so much that it's hard to interpret them. Today's is at least down to five digits, which isn't impressive but is an improvement over yesterday and many past days.

All of which suggests a business opportunity for Amazon, selling daily or hourly data on book sales to publishers interested in improving their marketing. Of course, for all I know, they may already be doing it.

And if any talk radio hosts happen to be reading this, and looking for a guest ... .

7 comments:

Anonymous said...

I know someone who wrote and published a special interest book -- basically a monograph of use to at most a few thousand academics and students around the world. After an initial spate of sales, she was pretty sure she knew when people bought even a single copy via Amazon. Based on watching what an individual sale would do to the book's Amazon ranking, it seemed that the effect of a single sale would "decay" over about three or four days.

Based on this experience, I concluded that the Amazon sales ranking figures are either some kind of weighted moving average over a window of about four days, or something more complicated that ends up resembling such a thing.

David Friedman said...

Update:

The effect of the Forbes.com piece turned out to be clear on both Amazon and the book web page. The Amazon ranking, which was (unfortunately) more than 100,000 for most of the past several weeks, was down (up) to 27,466 by 5:17 P.M. today, roughly two days after the column appeared, and now seems to be going back up (down). Hits on the book web page were 83 on the 23rd and 90 on the 24th.

Next experiment is Baycon, where I am giving a reading, on panels, and doing various other things to get attention for the book. My guess is that the effect, if any, will be smaller, given the much smaller, if better targetted, audience.

Anonymous said...

David,

You may be in for increased sales for a little longer. Some statistical physics people have studied the return to baseline in the sales of a book after an exposure boost like you describe. It turns out to be a power law, with power about 0.7 (ie extra sales = 1/time^(0.7).)

The power depends though on how many people a reader tells about the book. So if you track this power, you'll have a measure of how many friends your readers have.

(Reference: Physical review letters 93 228701 (2004). There's actually lots of interesting things about book sales in this paper.)

David Friedman said...

"The power depends though on how many people a reader tells about the book. So if you track this power, you'll have a measure of how many friends your readers have."

Interesting.

But it surely also depends on how many of the people who read the book like it. Which is something I am more interested in having a measure of.

Anonymous said...

"But it surely also depends on how many of the people who read the book like it."

Indeed. Now that you mention it, I would guess that this would dominate the how-many-friends aspect.

David Friedman said...

I wrote:

"Next experiment is Baycon, where I am giving a reading, on panels, and doing various other things to get attention for the book. My guess is that the effect, if any, will be smaller, given the much smaller, if better targetted, audience."

I was wrong. By the evening of the final day of Baycon, the book's rating was better than 11,000, probably the highest it has ever been. My guess is that that reflected general exposure, since there weren't very many people at my reading.

Anonymous said...

"My guess is that that reflected general exposure, since there weren't very many people at my reading."

Well, you do tend to talk to just about anyone you run across. :)