Amazon and the Myth of the A9 Algorithm

There is more nonsense written about Amazon than almost any other topic because mere mention of Amazon will get you more clicks than anything else, and the “A9 Algorithm” might be the most curious example of all.

Amazon also happens to be an incredibly divisive subject generally, one of those where little nuance seems to be permitted — and someone writing articles about Amazon tends to regularly get labeled a “shill” or a “hater,” depending, sometimes off the back of the same piece!

But that doesn’t dissuade anyone, really, because talking about Amazon can also be quite lucrative. The currency of the internet is attention — as I think was once said by Jeff Bezos, although I’m scared to Google it in case I start getting hunted down on every corner of the internet with hyper-personalized ads offering to make me a Kindle Publishing Millionaire or help me build a Drop Shipping EMPIRE.

The A9 Algorithm

Internet marketers are not known for their rigorous application of the scientific method. One intrepid black-hatted pioneer will discover a tasty data-morsel, dress it up in distracting finery, and then parade it about as part of a $2000 course. And then a dozen more will riff off that for their own courses and Patreons and books and masterminds and exclusive online workshops and virtual conferences; it’s like the most expensive game of telephone ever.

Sadly, it’s also quite value-free if you like hard facts.

This kind of environment shows some of the drawbacks of the brave new world ushered in by the internet and Google. If search for the phrase “amazon algorithm,” for example, the very first result is an article titled “Everything You Need To Know About Amazon’s A9 Algorithm,” and my BS alarm immediately goes off — use of the singular “algorithm” is a dead giveaway that the person doesn’t have a clue what they are talking about. Use of “A9” in this manner is another.

Indeed, the author of that article is a content marketer, presumably ranked first because the site he is writing for has solid SEO. Of course, it doesn’t reference a single bit of research to back up its claims.

Algorithms aren’t infallible, people.

Google “A9 algorithm” specifically, and you will fall even further down the rabbit hole — a giant, seething mess of articles dissecting this “A9 Algorithm” and speculating about how to apply it so you can increase sales of your drop shipping empire (or Kindle Publishing business — note these guys never talk about writing books, just publishing them, but I digress).

Authors and the A9 Algorithm

This talk of “the Amazon algorithm” and the “A9 algorithm” sometimes seeps over into our quaint world of publishing books we have actually written. Usually via a few disreputable course sellers in our space, who probably learned the lingo in a previous mastermind they attended and spotted an opportunity to sell this brand of snake oil to authors.

Which is why you get people claiming now, with a straight face, that there is one “Amazon algorithm” or that the “A9 algorithm” is what you need to wrap your head around to have success in the Kindle Store.

This self-referential nonsense reached its apogee when someone heard I was writing the second edition of Amazon Decoded. One of my mailing list subscribers emailed me asking about the “A10 algorithm” and what it meant for authors.

Because I have worked in this sector since the early 2000s, and because I’ve known what A9 actually is since it was first rolled out to the public in 2004 (I’ll return to that) — I immediately smelled a rat.

Nevertheless, I headed over to my trusty search engine again to see if I could garner any information about this mysterious “A10 algorithm” which I somehow missed in all my years studying and researching and writing about this topic.

Not only did I discover a whole group of people pontificating  on the “A9 algorithm” and the “A10 algorithm,” one enterprising soul was even selling the secrets of the “A11 algorithm” — no doubt after reinvesting the profits from his 6-Minute Abs video.

There was no mention of an “A8 algorithm” at all. Which is a pretty giant clue that the Drop Shipping Emperor is wearing no clothes.

I could describe this all as one big misunderstanding, except being utterly wrong about Amazon is as lucrative as it is damaging. The people who are spreading this BS are lining their pockets — often taking the money from hard-up authors who don’t have much to spare, then sending them off in the wrong direction. Authors who will then switch focus and change strategies based on something which is wholly illusory.

Here’s the real story.

The True History of

I was working at Google in 2004 when it was in the process of overtaking Yahoo as the #1 search engine on the planet — which was kind of important as ads beside search results were becoming the biggest money printing machine in history.

Search was the hottest commodity in tech, as a result, and everyone wanted a piece of the action — including Jeff Bezos, who was one of the earliest investors in Google, everyone seems to forget.

Those of us working at Google at the time were closely watching any potential competitors, and even though Amazon wasn’t the most obvious candidate for that title, it had been making acquisitions which pointed towards the possibility that it was moving into Google’s territory: BookPages, Telebook,, Alexa (back when that referred to an internet traffic statistics company),,, Leep Technology Inc., Egghead Software — Amazon even dropped a cool $250m, back when that was more than chump change to Jeff Bezos, on a company called Junglee which it didn’t even launch on the market for another 14 years.

Talk about taking the long-term view!

More seriously, what’s clear now is that Amazon was after something else when buying all these companies, and you can see — with the clarity of hindsight — how many of these acquisitions were stripped down and incorporated into Amazon’s giant AI-powered recommendation engine, that massive system driven by all sorts of algorithms, gobbling down huge amounts of data 24/7, and applying machine learning constantly to successfully pair customers with products.

A9 seemed different though, and definitely got Google’s attention.

Amazon’s new subsidiary seemed properly independent. It was launched publicly in September 2004, after several months in beta and quietly powering Amazon’s search facility, as well as working away on other things, like ads.

A9 was a proper, customer-facing search engine — which people also seem to forget — one aiming to be the next Google, rather than destined to work in the background running Amazon’s search box. And it really did seem like an genuine play for of the overall search business, rather than a data-harvesting effort.

Google was watching it keenly — it had far more respect for Amazon’s leadership team than those running its supposed nemesis at Yahoo. I remember the feeling internally that Yahoo was toast and it was just a matter of time before reality caught up with the inevitable, even if it seemed neck-and-neck to outsiders at the time. And A9 was genuinely innovative in the search space, unlike Yahoo.

A9 was doing StreetView before Google, for example, although A9 called it BlockView. It also had very sophisticated algorithms which remembered your search history and used that to personalize results along with things like your bookmarks and calendar — the first search engine to really do this. And the way A9 was set-up, it was designed to work best if you had its useful toolbar installed. Which basically meant it was tracking all your browsing history, of course.

Hey, maybe it was a data-harvesting operation after all!

Or maybe it was simply that Amazon — like any good engineer — never throws anything out without stripping it for parts first. And Amazon really did seem to try and make A9 work as a public-facing search engine, even launching its own version of Google AdWords, and using that system to deliver ads on also. All of which would be a precursor to Amazon Ads.

A9 Goes Dark

By 2008, Amazon conceded that Google’s trajectory was unassailable and folded up A9 — at least as a public-facing search engine, using it instead to power Amazon’s search, while also diverting some of those A9 resources to help with algorithms generally, parts of its recommendation engine, as well as things like ads (for example, A9 came up with Kindle with Special Offers, which would eventually turn into AMS lockscreen ads and AMG).

A key point: the recommendation engine was already in full swing before A9 was even conceived — something else that is missed by all these hucksters. Amazon had already made its great leap forward in terms of recommender systems, long before A9 came into being.

If these internet marketers spent any time looking beyond their circle-jerk of masterminds, which I believe is the appropriate collective noun, they might have noticed that the engineers and data scientists which built Amazon’s recommendation engine have written multiple papers on how it all works, making a surprising amount of detail public, including aspects of the various algorithms — note the plural — which feed into different parts of it, and explaining how, where, and when exactly Amazon built different aspects, and the specific ways it was innovative, and how it moved recommender systems forward (to the extent that Amazon’s approach is still broadly used by people like Netflix and YouTube today).

And if they looked a little harder, they might also have found various presentations that Amazon routinely makes — even today — at conferences to developers which explain all this stuff quite clearly, if you can wade through pages of engineer-speak, which is not for the faint-hearted, let me tell you.

Here’s the thing. You won’t find a mention of the “A9 Algorithm” anywhere in these papers or talks.

The A9 Algorithm Doesn’t Exist

Which is weird, right? Unless you posit this: there is no such thing as the “A9 Algorithm.” My madcap theory also neatly explains why there has been no “A10 Algorithm” replacing it too, you will note.

Think that’s a crazy idea? Here is an Amazon engineer being pretty explicit:

There is no single A9 algorithm.

Amazon engineer

Not convinced by one source? That’s cool. Here is A9 themselves in a quote from their archived website:

Our ranking algorithms automatically learn to combine multiple relevance features.


Note the plural!

And here’s the most hilarious part, the name A9 literally means algorithms. It’s a geeky in-joke, a numeronym: A + 9 more letters. Algorithms. Plural.

I sometimes wonder what Amazon thinks when looking at this firehose of BS. People spending thousands of dollars on courses teaching them about algorithmic unicorns. The sad thing is, the misunderstandings don’t start or end with thinking that the entirety of Amazon is powered by one single superpowered “A9 Algorithm.”

There is another problem with listening to this crowd of blustering black-hatters, these drop shipping dingbats: they have built misunderstanding on misunderstanding. They think search is exclusively powered by this all-conquering “A9 Algorithm” and think the only thing that matters on Amazon is supposedly tweaking your products and metadata and marketing to gain the favor of our A9 Overlord.

How Amazon Really Works

In reality, appearance in search is determined by a number of algorithms, and a number of different factors, including metadata and Popularity and relevance.

And search is only one tiny component of the overall recommendation engine, all of which is powered by many more different algorithms. And huge chunks of this recommendation engine are much more important to authors than search — for example Sales Rank. Which also has nothing to do with search and predates A9 by years anyway.

Also Boughts are another. Although as I explained elsewhere, it’s not Also Boughts which are important, as much as the connections between products they represent. Either way, that system of mapping connections between products has been informing recommendations to Amazon customers since before A9 ever existed.

And it uses an algorithm or two, as you might imagine.

So be very careful when you hear anyone talk about the “A9 Algorithm” or when anyone uses the word “algorithm” as a singular. Be skeptical if anyone starts talking about “ranking in search” as being critical to an author’s success — particularly an author of fiction — and it’s absolutely crucial not to confuse ranking in search with Sales Rank, which is a very different thing.

There are far more important aspects of the Kindle Store which authors should focus on, things that really do determine whether your book gets recommended to readers.

This “A9 Algorithm” BS is usually a sign that someone doesn’t have a clue about how Amazon works, or what authors need to focus on, for that matter, or what they can safely ignore, and how they can use all this information to sell more books.

Further Resources:

This post on Also Boughts gives you some more background on Amazon’s recommendation engine, and how it was conceived, as well as breaking down how it actually works in practice, and how Also Boughts feed into that — and it’s not what you think! As a bonus: you get to stop stressing about Also Boughts disappearing from your book pages.

FB Also Boughts blog image

And then all of that should whet your appetite for my book Amazon Decoded, which breaks down how all the algorithms powering the Kindle Store and the millions of book recommendations made to readers every single day and how authors can get Amazon to start recommending them.

Decoding Amazon recommendations

David Gaughran

David Gaughran

Born in Ireland, he now lives in a little fishing village in Portugal, although this hasn’t increased the time spent outside. He writes novels under another name, has helped thousands of authors build a readership with his books, blogs, workshops, and courses, and has created marketing campaigns for some of the biggest self-publishers on the planet. Friend to all dogs.