Understanding Amazon’s New Algorithms Is As Easy As ABC Publishing

Amazon’s KDP Select introduced a new tranche of self-publishers to the upper reaches of the charts for the first time. For the first couple of months of this year, a new seam had been discovered in this self-publishing “gold-rush.”

It didn’t last too long, however. By the end of March, even those newly minted authors were openly considering leaving KDP Select, despite how successful it had been for them. Self-publishers were noticing that even when they had a stellar free run, garnering thousands and thousands of downloads, it was no longer catapulting them up the charts on their return to the paid side.

Science fiction and fantasy author Ed Robertson penned an excellent hypothesis and gave me permission to re-post. If you don’t understand exactly why successful free runs used to almost always translate into a run at the charts, then read my post on Popularity Lists first for background. That said, here’s Ed:

BACKGROUND

Around March 19, Amazon changed the way they sell books. In a Kindleboards thread devoted to the subject, authors tracking the performance of books during and after a free promotion began reporting strange results. Prior to then, books that gave away several thousand copies during a promo would shoot to the top of the popularity lists some 36-48 hours later. It was like clockwork. Clockwork that paid you several hundred dollars.

Because the popularity lists are a big deal. These are the default book listings you’ll see when you’re browsing around by genre. Here’s the Fantasy list, for instance, with GRRM clogging up the top 10 like the greedy goose he is. If you could ride a free promotion to the top of those lists, your book would be extremely visible to shoppers. Depending on genre and your book’s presentation, topping the pop lists could snag you dozens or hundreds of sales before other books overtook you. Sometimes that visibility was enough to launch a book into the stratosphere, where the stratosphere is also made of money. It was kind of a big deal.

Then, things changed. Except they didn’t change. Not for everyone. Authors began reporting lower sales than expected as well as strange-looking lists. Chaos reigned! Dogs and cats living together, watching couch-bound authors tear out their hair. After a couple weeks, we thought we had it figured out: there was no longer a single popularity list. There were two, and books no longer seemed to be vaulting to the top no matter how many free copies they gave away.

Well, we were wrong. There weren’t two lists. There were three.

Because I am extremely imaginative, I’m going to refer to them from here on out as List A, List B, and List C. I’ll get into the methodology in a bit, but for now, I worked this out through carefully observing my books, reading other Kindleboard authors’ results obsessively, and lobbing theories around with other authors. I would never have figured this out on my own. I know, never say never. Trust me, eventually I would have gotten frustrated and left to play Mario Kart instead. One other author in particular did tremendous heavy lifting. Like the Eye of Sauron, he (or she?!) is far-seeing and awesomely powerful. And much like Sauron, you can’t invoke his or her name without facing terrible wrath. Some of the Eye’s secrets must remain just that.

But the outcome of that info can be revealed. So without further ado, here’s how the three lists work.

THE CHANGES

List A is the same version of the pop lists that existed prior to March 19. It is Select- and freebie-friendly. Here’s roughly how it works:

  • Ranks are heavily weighted to the last few days
  • Free book downloads are weighted equally with paid sales
  • Borrows count as sales

List B appears to be a throwback pop list, one that was running throughout most of last year. Here’s how book ranks are calculated on it:

  • Ranks are determined by the last 30 days of sales, with no extra weight given to the most recent sales
  • Free book downloads are discounted heavily–maybe as little as 10% the value of paid sales
  • Borrows don’t count as sales

List C is a lot like List B, with a couple major differences:

  • Free book downloads aren’t counted at all
  • Recent sales are weighted somewhat more heavily than List B(?)
  • Borrows don’t count as sales

What does that mean in practice? A lot. A lot a lot a lot. Here’s where my book The White Tree ranks on all three lists at this moment in time. Each shot will look a bit different because they’re taken from different browsers–that’s one way to see different lists. The list in question is Fiction > Fantasy > Series, a fairly quiet little fantasy subcategory.

List A:

List B:

List C:

Pictured: Oh shiiiiii–

METHODOLOGY

Most of this was achieved through comparing tons and tons of different books on different browsers, just like the screenshots above. Here’s some stats for the book in question that helped me figure out what was happening here. On March 28-29, The White Tree was downloaded 4700 times (free). On April 17, it was downloaded an additional 1300 times. In April, its paid numbers came in at 210 sales and 46 borrows.

Since March 19, my main browser’s been displaying List B. My big clue to List B came on April 28, when I noticed my book had, over the span of a day or two, dropped from #67 in Epic Fantasy to #165. Rank didn’t slide–it instantly dropped off a cliff. Why? Because it had been 30 days since all those free downloads had come in. I’d noticed the same thing around March 23–I’d done a huge giveaway February 22-23, and once 30 days elapsed, it suddenly plummeted from around #45 to around #255. I didn’t know what it meant then, in fact I don’t think I even knew there were two lists at that point (let alone three), but when it happened again, I had a pattern.

I also had several weeks of observations piled up by then to help me understand new data. For weeks, List B had been showing me very static lists. The books at the stop stayed at the top. There was very little churn. There were very few Select books, i.e. books that were likely to have recently been free, especially within the top ~60 results (first five pages). Most books at the top were traditionally published. List C was even more trad-dominated; generally speaking, an indie title on List B would be ranked 15-25% worse on List C if that title hadn’t been free, and would rank much, much worse if their List B rank was dependent on free downloads (like, hundreds of places).

When I compared the top 240 titles in Epic Fantasy between List B and List C, here’s what I found: on List B, 188 titles weren’t in Select, and 52 were. On List C, 217 titles weren’t in Select, and just 23 were. With no benefit from freebies, and with fewer paperback sales to pad the numbers, most indies get killed in List C.

When it came to figuring out that borrows weren’t counted in List B and C, The Eye of Sauron was particularly helpful. We compared Select books with lots of borrows to non-Select books whose sales were roughly equivalent to the Select books’ total sales+borrows. On List B and C, the non-Select book came out ahead by a good chunk. We compared Select books with lots of borrows relative to sales with Select books with few borrows : sales. (None of these books had recently been free, which acted as a “control” between List A and B.) The ones with a higher ratio of sales : borrows almost always came out better on List B than on List A.

While I wouldn’t lay my life on the line for every one of these observations, I am very confident in the overall conclusions reached. There are three different lists. You can see them for yourself–just compare lists on different browsers, computers, and Kindles. If you’ve gone free recently, you’ll note your popularity rank on List A is much better than B or C.

How do you tell which list you’re looking at? Well, that could take a day or three to figure out, but in short, if you see a bunch of Select titles on the first pages of the pop lists, you’ve probably got List A. If it’s almost all traditionally published books, it’s List B or C. From there, compare your lists on another browser/device; if you’re seeing List C, trad books will generally be even more dominant.

WHAT THIS MEANS

What does all this mean? Hey, maybe you haven’t noticed, but this post is already epically long. The internet is only so big, you know. I’ll save that for a future post. For now, here’s what’s key: there are three different lists. Your book is listed on all three, but any given shopper is only seeing one version of the lists. (In other words, different people see different lists.) If you’re an indie in Select, one of these lists is good. The other two? Well, let’s just hope they’re not here for too much longer.

UPDATE: The same day I posted this, Amazon changed their sales algorithms again. This post will provide a lot of the background to what I talk about in the followup post.

* * *

I want to thank Ed for allowing me to re-post this. His blog is excellent, and particularly strong on things like understanding the inner workings of Amazon.

If you would like to check out some of his books, the fantasy novel The White Tree (pictured up top) is particularly recommended and available on Amazon and Amazon UK, but he has plenty more titles here.

I can also vouch for his data geek bona fides and the methodology employed above. We’re part of the same online writers group and I watched the testing of this theory unfold as it was verified by multiple participants.

While the implications of some of this may seem ominous for self-publishers – especially those depending on the power of Select and the associated free runs – I want to urge everyone not to panic, and to keep speculation from veering into tin-foil hat territory.

Amazon is always testing things, making frequent changes to their algorithms. Some of those changes have favored self-publishers, others haven’t. As I have said on many occasions, Amazon don’t care who has published a book, and have previously made changes that disfavor books from their own imprints.

What they do care about is recommending the book the customer is most likely to purchase. And the basic equation never changes: write good books and lots of them, and let readers know they are there.

Speaking of good books, Ed Robertson’s dystopian novel Breakers is free today.

UK peeps can grab a free copy here.

One final thing (apologies to Ed). I’m blown away by the response to my open letter to the DOJ. I was expected a handful of writers to co-sign, tops. There must be around 150 names now. So, a big thank you from me. I’m going to keep it open for another day or so as names are still pouring in.

Then it gets sent.

UPDATE: Ed’s Breakers is #8 in the free store. Go Ed!