This is actually an older article that I chickened out of posting at the time of writing. Though some of the Amazon features described are no longer with us, I think there’s still much of relevance, especially given new interest in the subject.—Aaron
To most people, Amazon is a black box, operating by rules and principles diligently veiled from public view. Even those of us who follow Amazon closely must sometimes shrug our shoulders and accept that we don’t know how some things work. But the veil has become at least a bit less opaque, thanks to the Great Amazon Counter Experiment.
The idea developed almost by accident. In April 2009, self publisher Wayne Roseberry contacted me about the sudden disappearance of the image upload function from the posting form for AmazonConnect—the now-defunct Amazon program of author blogs. As a substitute for that function, I then worked out that Amazon allowed placement of a standard image tag in the post’s HTML code, letting you display an image hosted on an outside Web site.
Wayne then suggested that such an image, if inserted in a post assigned to a single book’s detail page—as most of my posts had been—could be used to track how often that page was viewed. All you had to do was check the outside site’s statistics report and see how many times the image had been accessed.
Of course! Within days, I had added image tags to one post each for most of the books I’ve published myself. And so, for the entire month of May, I had “counters”—tiny, invisible graphics hosted on my own site—included on Amazon’s detail pages for those books.
Unfortunately, the experiment was ended by the unceremonious closing of AmazonConnect in late May and the yanking of all past posts from book detail pages in early June. But this was not before I gained some valuable insights into Amazon’s workings.
The most basic insight, of course, was an idea of how many Amazon customers were actually looking at my books. I found that the four most popular had weekly totals ranging from 135 to 340. It’s possible that Amazon was caching my counter images, which would make these figures too low—but they at least stand as minimums.
I had initially hoped to also use the figures to determine sell-though—the proportion of Amazon customers who bought my books against those who just looked. (Amazon’s Ultimately Buy figures cannot be used for this because they include only customers who bought any book at all.) Because of the caching issue and others, I decided this calculation could not be reliable. Still, it gave me figures for upper limits—which, for those same four books, turned out to range from 12% to 21%.
An unexpected and even more interesting benefit of the counters was worked out with the help of my friend and publishing colleague, Morris Rosenthal. This had to do with the customized Web page addresses Amazon uses in its internal links. These addresses include not only what’s needed to call up the linked page but also coded info for Amazon’s record keeping and analysis.
As Morris pointed out, I now had access to the customized addresses in all links clicked to view my books. Since the visited pages were requesting my counter images, those addresses were showing up right in my server logs.
Though I had previously analyzed Amazon addresses to see which parts were actually needed for links from outside, I had never tried to figure out the parts used internally. But now Morris and I became fascinated with the contents of one such part—Amazon’s referrer codes, or “ref codes,” for short.
It turned out that every link from any Amazon page to any other included a ref code to describe the location of that link. And that location included not only the page that the link was on, but also the feature that displayed the link, and the link’s position relative to other links displayed by that feature.
Let’s look at a typical address you might collect from the browser’s address bar while visiting one of Amazon’s book detail pages. This is like one you might see for my book Aiming at Amazon.
Can you spot the ref code? Here it is on its own.
What this code tells you is that you reached this page by performing a search (“sr”) and then, on the first page of results, clicking on the sixth book listed. (Later in the address, a query ID—“qid”—tells Amazon exactly what search produced this positioning for this book, while the ref code is repeated in another format for convenience at the end.)
Amazon’s ref codes, combined with sales history, help Amazon learn how best to design its pages so you are directed to the books you want to see and then buy them. And now, with these same codes coming to me, I too could see which Amazon features were most effective in bringing customers to my books, and from there, fine tune my own marketing.
The first task, shared with Morris, was to map the ref codes—a challenging job that gradually yielded the secrets of the bulk of them. I then analyzed a full week of ref codes for Aiming at Amazon. Here’s what I learned.
• Despite my having two fairly popular Web sites that support this book—and despite the book being featured on many other sites—links from outside Amazon accounted for only about 20% of visitors to the book’s detail page. (This was determined by looking for addresses with no Amazon ref code.)
Conclusion: Marketing through your own Web site can be a valuable addition, but it is not as efficient as proper marketing on Amazon. (This assumes that the book itself, and not just its marketing, is optimized for Amazon.)
• Links in all Amazon Community features together accounted for only 10% of visits to my book’s page from other Amazon pages. This 10% included 1% from Listmania lists, 1% from “So You’d Like to . . . ” Guides, and 3% from linked tags. (And no, these low figures were not due to a lack of my activity in these areas. Though I don’t use Listmania or SYLT heavily, Aiming at Amazon appears in plenty of lists and guides by others.)
Conclusion: The emphasis on Community features by most Amazon “experts” is entirely misguided.
This conclusion is strengthened by the fact that my Amazon Profile—a feature beloved of such “experts”—was viewed by only five customers that week, and my Amazon Blog was viewed by none. (I determined this by looking at some combination of counters that unfortunately was too tricky for me to remember.)
• Amazon’s customized recommendations to customers accounted for 55% of detail page visits—over half! That included 5% from Frequently Bought Together placements, 16% from Also Bought listings, and 3% from Ultimately Buy listings. (The extent of this development was the biggest surprise to me from this experiment.)
Conclusion: The dominant means of book discovery on Amazon is now the one hardest to influence directly!
• Customers searching and then clicking in results accounted for 31% of detail page visits. While impressive enough in itself, this figure may be deceptively low. The ref codes for Aiming at Amazon showed that most of the links clicked after a search were for the first book on the first page of results. This suggests that an even higher percentage of customers may reach their first detail page through search results, even if many of them move on from there through Amazon recommendations.
Conclusion: Optimizing a book and its listing for search is still the most important means you have to make it visible on Amazon.
Hey! I knew that!