I've got some experience with algorithms, though my experience with search algorithms goes way back to grad school: I did a Master's in computational linguistics and another in library and information science. Organization, search, ambient finadability, and algorithms were my forte. I hoped to work for Google for a while... but I got swept into the statistics and finance fields after grad school. They like us for our ability to find patterns and anomalies in large, weird data sets.
Here are my entirely anecdotal observations and guesses. I can ramble on about this stuff, but I'll try to keep it short. (ETA: I failed. Sorry for the wall of text.)
In the case of Amazon, a rumor on Facebook mentions A10 has likely gone into effect, a change of some sort from A9. (They must threaten death to anyone who makes their algorithms public. It's the best kept secret in the world.) The suspicion has to do with lower book sales in October for many authors. The guess is that A10 has something to do with sales coming through links other than Amazon's own pages or ads. If true, that would encourage, not discourage, buying outside ads instead of AMS ads which is reverse of what was supposed before. How would Amazon benefit from that, I wonder.
They like to keep it secret because all algorithms work by applying number values to a set of circumstances to sort things. In Amazon's case, they want to offer customers the things that A) customers are most likely to want, B) customers are most likely to buy, and C) are going to make Amazon the most money. If people know those exact values, they can manipulate the system (i.e. jump the line ahead of other products that fit A, B, and/or C better), and then the point of the algorithm is destroyed.
As far as the Facebook rumor... it's a rumor. I wouldn't put much thought into it unless several people can provide supporting data.
I can see a benefit to Amazon if they somehow give more credit in the ranking to a book that is bought after clicking on an AMS ad. To make that more to the author's benefit, they could perhaps raise the rank two steps instead of just one. But really, does ranking really sell books before it gets into the top 100?
You're viewing this from an author standpoint, that Amazon wants to bribe you to use their ads. They do, but they want customers on their site more. Someone clicking an Amazon ad is already on the website. If they can encourage you to bring more people from outside (i.e., Facebook), that's a bigger win for them.
As for ranking, not so much: rank comes from sales, not the other way around. YES, it can result in residual sales, but any top seller will tell you that it isn't a perpetual motion machine. If you stop promoting, everything returns to entropy: when you stop advertising consistently (whether that means AMS/FB Ads/email ads/social media/email newsletter/whatever), rank drops. Ranks don't create book sales, book sales create ranks.
If there truly is such a thing a A10 and we were Amazon, what amazing thing would we tell A10 to calculate?
Find the black hats.
I suspect Apple may have some legally acceptable reason it could cite. Whether that's the real reason or not is another question. That's why enforcement is often difficult.
I strongly suspect you're right. I also suspect it may be based on the actual history of borrowers and payment histories, wherein if the gender/sex of the borrower is known, men have historically shown to be better borrowers. In any case where you're recording gender/sex, it's incredibly unlikely that you'll get a dead even average on stats: one will always do better on paper, even if it's incredibly slight, and machine learning bias is born. It's up to Apple to explain this one to any fair lending enforcement that comes knocking.
The thing is, sales go up and down and no one is quite certain why. When a lot of authors report fewer sales at the same time, we think it must be some sort of change Amazon has made, but we will never know for sure. This month, I'm selling more on Apple than Amazon. They usually run about the same. I run AMS ads on Amazon, but of course, I don't on Apple. I would if I could. So, what does that tell me? It tells me it is harder to sell books on Amazon than on Apple right now.
I think it boils down to your (generic "your") peers. Remember what I said earlier about advertising consistently, and how it relates to rank? Part of rank decay also has to do with how strong your ties are to other "peer" books (i.e. your also boughts, your sponsored ads, things that share common search terms, etc.). If your peers are selling, you're getting a vicarious boost on that traffic. If one of your peers runs an email promo, or new release push, or gets a Bookbub, or a search term gets mentioned in a major news story, you get a boost, too.
In my humble theory, this is why it appears that the rich get richer on Amazon. Top sellers have peers who are also top sellers, and those folks have their marketing and advertising nailed down: it's consistent and effective. (They also offer great, quality books, which should be a given, but this is why *these* great books are consistently up in the charts while some others aren't.) They all consistently sell themselves and their peers. If you don't market/advertise/sell consistently, the algorithm knows that based on your history, and your peers look like you except for the occasional push or lazy days. If you want to step into a different peer group, you have to show the algorithm that your behavior has *consistently* changed.
It's also the reason why people emphasize new releases as the time to hit advertising. When you have no history in the algorithm, you get a temporary benefit of the doubt until your history is established. It's also the reason some Amazon Ads seem to do good out of the gate, but then impressions slow or stop: if the ad isn't getting clicked as much as other similar ads, it starts getting passed over to show things more likely to sell. The algorithm isn't evil or out to get anyone, but it does know what sells consistently, and it rewards those products/ads.
Again, just my thoughts. YMMV.
