Everyone loves a mystery – especially when it involves a secret code or an algorithm that only a select few fully comprehend. Google’s PageRank algorithm is one example. It has been the subject of hundreds of articles, blog posts and even a book. Amazon’s sales rank algorithm is another example. Like PageRank, it takes into account many factors, weights them and then produces a single number that indicates how well a title is selling. There is mystery attached to all three areas – the choice of factors, the weights and the calculation.
Amazon says very little about its sales rank number, providing only this modest overview of its basic function:
As an added service for customers, authors, publishers, artists, labels, and studios, we show how items in our catalog are selling. The lower the number, the higher the sales for that particular item. The calculation is based on Amazon.com sales and is updated each hour to reflect recent and historical sales of every item sold on Amazon.com.
The sales rank is designed so that a lower value represents greater sales. Typically your sales rank will decay (grow larger) when your title is not selling, and jump downward when sales materialize.
There have been many attempts to analyze the sales rank algorithm. Morris Rosenthal of Fonner Books provides a more detailed explanation with accompanying video. His analysis shows how the sales rank correlates to books sold weekly, for books with a rank of 1,000 or greater, and daily for books with a rank of less than 1,000. Both plots show a gently curving line on a log-log scale. In studying the algorithm over several years, he has found that the newer version (released in October, 2006) is less influenced by historical sales, and also that it is more predictable and correlates better with sales events such as promotions or reviews.
Another analysis of the sales rank provided by Dog Ear Publishing, pinpoints some of the characteristics of Amazon’s calculation of sales rank. These include:
Sales rank is based on a comparison of your success relative to about 5.2 million other titles that have sold at least one copy
Ranking are updated hourly
The top 5,000 books seem to follow a different ranking methodology than their higher valued cousins. This seems to include considering the time between sales and averaging sales to keep rankings in this group more consistent over time.
Volatility is sales rank is greatest for books with a sales rank between 50,000 and 250,000. Titles on either side of the sales rank spectrum tend to change more slowly over time.
Brent Sampson of the Web Pro News new analyzed the frequency of update for sales rank and provided an interpretation of what rankings in certain ranges indicate. One interesting finding – the ranking is predictive in nature. That is, trending calculation is applied to arrive at a computerized sales trajectory. This can mean that even if a book’s sales pick up, it’s sales rank may temporarily fall.
Chris Anderson, author of The Long Tail, has published the results of experiments he and others have conducted to study the effects of purchases on sales rank. He has also used this data to determine the percentage of Amazon sales accounted for by titles in the long tail (i.e. – average sales rank over 100,000). He concluded that they made up between a quarter to a third of Amazon’s overall book sales.
As Google does with PageRank, Amazon periodically tweaks the algorithm and this is often the subject of controversy. Freakonomics’ author Steven Dubner wrote in the New York Times recently about what he perceived as a change in the weighting of top ranked books which favored hardcover paperbacks over older, but still best selling paperback titles.
There are a number of handy tools available for monitoring your Amazon sales rank. If you want to quickly check your title’s sales rank without navigating through Amazon, you can use the applet at Sales Rank Express. You can also install a handy Amazon sales rank widget on Steve Weber’s blog. Finally, Charteous has produced a nice tool for tracking Amazon sales rank over time. The site allows you to track your Amazon sales rank over time as a line chart. The charting begins when you sign up. You can also compare your sales rank with other titles. This is good intelligence to have when determining how you are doing relative to your category. Two things you will need to do to use this functionality:
You’ll need to create an account (there is no charge)
You’ll need the Adobe SVG browser plug-in to see the charts
Perhaps someday the arcane calculations within the sales rank algorithm will be exposed. Until then, authors, publishers and curious onlookers will have to examine the nuanced correlations between book sales and sales rank and speculate about the algorithm’s internal machinations. And of course Amazon will continue to adjust the parameters, weightings and equations, resetting the analysis and frustrating the cognoscenti. A corporate powerhouse, using secret code administered by a techno priesthood to strike fear (and sometimes exhilaration) into an anxious population of authors / publishers – the whole thing has the makings of a good techno conspiracy thriller.
- Long Tail Dynamics vs. the 80-20 Rule
- New Rules!
- Impact = Content x Consumption x Engagment
- The Eternal Sunshine of Open Publishing and the Transparent Booksphere
- The New Author Book Pitch
- Publishing Analytics