View in browser

Sorry about the delayed email! I’ve been trying to finish off Amazon Decoded 2 and doing that usual thing where I reach the last fence in fine fettle and then… stop. Eat some grass for a while. Consider the lilies in the field, and so on. Every. Damn. Time. Stupid brain.

ANYWAY.

One of the chapters that I was chopping and changing and rearranging became particularly relevant last week when a reader (that’s one of you guys!) emailed me asking about the “A10 algorithm” and what these changes meant for authors. Which was a bit like someone drawing a map of our solar system being asked what color they were going to paint Xanadu. (It’s all glitter, all the time, but I digress.)

Having never heard of this supposed “A10 algorithm” I immediately headed to Google and not only did I discover 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 course of an “A8 algorithm” and the curious absence of same gives us the key to understanding all this. Or to grasping this colossal misunderstanding, to be more accurate.

So, here’s what happened. In 2003, Amazon created an independent subsidiary called A9.com – something you may have heard about in the news this summer when the Wall Street Journal ran a big story about it. (Note: I don’t necessarily agree with a lot of the analysis in that story, I’m more interested here in the history of A9 than the present – I think the tensions the Wall Street Journal describes could also be framed as typical tensions within a company between engineers and sales, but whatever. I do agree with some of it, I should also say. MOVING ON.)

Unlike Amazon, A9.com wasn’t headquarted in Seattle, but in the search capital of the world: Palo Alto. This was around the time I joined Google, and Google finally started overtaking Yahoo to become the word’s favorite search engine. Please note I’m not claiming these events are linked. Just saying that I remember this time, and search was the hottest commodity in tech. Everyone wanted a piece of it – including Jeff Bezos, who was one of the earliest investors in Google, people often forget. And as I was working for Google, we were all keeping an eye on the competition there, and watching Amazon’s moves.

Amazon also acquired similar properties around this time like IMDB.com and LibraryThing and  Shelfari, and then Goodreads a little after that. Some of these sites were pretty much left to continue as is, others just seem to languish. In all cases, widespread speculation was that Amazon was more interested in the data and the algorithms under the hood, than any more direct operational value.

In simple terms, Amazon was interested in making it easier for people to find things they liked in its store, and for the store to improve its recommendations to customers – all of which is powered by data and machine learning and algorithms.

Note the plural.

The reason there are no changes you need to worry about deriving from the supposed rollout of an “A10 algorithm” is because there is no “A10 algorithm” just like there was never an “A9 algorithm.” I wrote a post last year about the “Amazon algorithm myth” – which touched on this myth that is usually spread by BS artists who want to sell you The Secret.

Here’s the real secret: there are many algorithms working away under the hood. Different algorithms determine Popularity, others calculate the Sales Rank of each product. More algorithms again will decide which books get sent out as part of the millions of recommendations emailed to readers every day. The idea of One Algorithm To Rule Them All has always been nonsense.

Don’t want to take my word for it? Fine. Here’s a former Amazon Software Engineer responding to a specific question about the A9 Algorithm:

“There is no single “A9 algorithm.”

Seems pretty clear, right? Yet this rumor is remarkably persistent. But if you look at the source of all the articles about the “A9 algorithm,” they are invariably internet marketers, Amazon resellers, often guys flogging some kind of dubious course costing thousands of dollars. I recommend steering clear of these sources completely, in case that isn’t obvious. These guys are also the source for demonstrably false contentions like “reviews affect Sales Rank” and any number of ridiculous ideas that routinely invade the authorial blogosphere.

Anyway, here’s where thinking about One Algorithm To Rule Them All can seriously lead you astray: the factors that influence Sales Rank are very different to those which affect Popularity, and if your marketing plan is only optimized for the first – and takes no account of the second – then you are missing a trick. A huge one if you are in Kindle Unlimited, because Popularity is a huge driver of recommendations to Kindle Unlimited subscribers.

How different are these factors? Let’s do a quick overview:

Sales Rank doesn’t count free downloads (except in a separate chart) but does count borrows. Popularity doesn’t count borrows but does count freebies. However, Popularity has price-weighting, whereas Sales Rank does not. Sales Rank is updated hourly (ish), and sales are heavily weighted for recency. Popularity is a rolling 30-day average of your sales, with no recency weighting.

That’s just for starters. The point here isn’t to tease out all the implications of that, but just to show you that these two things are very different, and a Kindle Unlimited-slaying strategy in particular must take account of these very different quirks baked into Popularity.

And I’ll delve into all that in Amazon Decoded 2 when it releases early next year.

Dave

P.S. Writing music this week is a live version of a slow burner from Led Zeppelin: Goin’ to California.

David Gaughran

Broomfield Business Park, Malahide, Co. Dublin
Ireland

You received this email because you signed up on our website or made a purchase from us.

Unsubscribe