Author Facebook groups can sound a bit like the three-eyed alien toys from the movie Toy Story. Remember the ones in the claw machine who said “The claaaaw!”

Except in author Facebook groups, instead of talking about the claw, they talk about the Amazon algorithm. 

“The algorithm is our master.”

“The algorithm chooses who will go… and who will stay.”

Then the algorithm scoops down, picks a book, and the author squeals, “I have been chosen! My metadata is pure! Farewell, my friends!” 

And I get it. If the Amazon algorithms favor your book, your sales will soar. But when you’re focusing on the algorithms, it’s easy to forget about your target reader, someone I like to call Timothy. 

Like the aliens in Toy Story, we can get so focused on the claw that we fail to notice the human behind the glass who is moving the claw. And as you are about to learn, it is *readers* who program the Amazon algorithms with their actions. Readers control the claw!

So what are the Amazon algorithms? How do they work? And most importantly, how do you get them to put your book in front of likely readers?  

In this episode, you’ll learn how the Amazon algorithms work as well as how similar algorithms on Instagram and YouTube work.

So what is an algorithm anyway? 

How Algorithms and Machine Learning Work

At their core, algorithms are step-by-step instructions or rules that computers follow. Think of an algorithm as a recipe. A human programmer acts like a chef and writes the recipe, and the computer acts like the cook following the recipe to create the meal.  

Machine learning takes this idea a step further by allowing the computer to develop its own “recipes.” In other words, the computer can program itself. Instead of the human chef dictating each step, the human defines the desired outcome, which could be ensuring that the restaurant patron leaves a five-star Yelp review.

Now, let’s imagine that in this fictional restaurant, the cook doesn’t know anything about cooking but is able to cook thousands of dishes at the same time. It just starts trying things. At first, the patrons are not very happy with the food, but over time, the cook starts to find recipes that work better than others. The machine is learning how to cook well by cooking a lot. Eventually, after millions of tries, the machine-learning cook has been able to create entirely new recipes that patrons love and rate with five stars on Yelp. 

If given enough time and data, this cook could start to customize the meals according to the preferences of individual patrons. If the cook knew the last 500 meals you ate, and whether you liked them or not, this cook could create a meal tuned to your specific palette. 

Every time you swipe up on TikTok, YouTube Shorts, or Reels, a machine learning algorithm selects a new meal for your palate. This is why those platforms get more addictive over time. Each time you click “like,” you are telling the machine more about your preferences. 

This may seem like AI, and it is, but it is not new. Machine learning was used in the 1950s and has been in active use for decades. So when AI makes you uncomfortable, remember that computers have always been thinking machines; the thinking is what makes them different from typewriters, and every year they get smarter.  

So when people talk about the “YouTube algorithm” or “Amazon algorithm”, they are talking about a set of machine learning models that “choose” what to show you. It’s actually more complicated than that, so if you want to dive deeper, just Google “neural networks” and enjoy the ride. 

Do algorithms have a will?

This video explains Machine Learning better than I can.

Understanding machine learning helps explain why people sometimes refer to algorithms as having a will. It seems like the Amazon algorithm wants to help you buy a book. But the actual will comes from the humans who created the initial versions of the algorithm and set the criteria for successful outcomes. 

Most algorithms share a common goal: to maximize profit. Once you understand how the company behind an algorithm makes its money, you can determine what the algorithm is optimizing for.

For example, social networks make money from advertising. Every few minutes, you see an ad. The more minutes you spend on the social network, the more ads you watch, and the more the company’s profit is maximized. 

But here is where it gets really interesting. Because our cook is making millions of experimental meals, there is no way for the chef to know what is being cooked at any given time. The human developers at Instagram have no way to know what video you will see next when you swipe up on a reel. Likewise, no one at Amazon knows which books will show up for you when you browse a certain category.  

Because of this unpredictability, it’s easy for algorithmic superstitions to spread among authors.

Common superstitions say, “Amazon looks at completion rate,” or “Amazon reads your book.” The better you understand the why behind Amazon’s algorithms, the less susceptible you will be to time-wasting superstitions and the better your book will sell. 

What is the difference between the A9, A10, and A11 Amazon Algorithms? 

For Amazon, success means keeping customers buying and happy with their purchases, because that’s what drives their bottom line.

For years, Amazon owned a search engine company called A9. The engineers behind that company were very transparent about the changes they were making and how the algorithms were tuned. This was known to Amazon optimizers as the “A9 Era.” Then Amazon shut down a9.com in 2019. 

Around 2020, marketers started to refer to the evolving Amazon algorithm as the “A10 Algorithm.” A10 is a colloquial term to refer to the new era of secrecy at Amazon, when algorithm changes had to be reverse-engineered. Amazon experts like Dave Chesson at Publisher Rocket, Alex Newton at K-lytics, and Joe Solari at Author Nation have dedicated their careers to reverse engineering various parts of the algorithm. 

This is not to mention the many, many marketing gurus in the mainstream Amazon Seller FBA marketplace. Remember, books make up only a tiny slice of Amazon’s profit. Many of the changes that impact authors also impact sellers who make USB cables and a myriad of other widgets. 

I suspect the A10 era is coming to an end. Soon, more of the Amazon algorithms will be fortified with LLM-generated metadata. I propose we start referring to Amazon’s LLM algorithm era as the “A11 Algorithm.” We’ll see if the new era name catches on.   

Different Algorithms for Different Customers

To understand the Amazon algorithms, we have to start with the customer. There are four primary ways readers find books on Amazon:

• Browsing category pages
• Browsing specific book pages
• Browsing bestseller lists
• Searching via Amazon search

Each user has a different purchase intent, and Amazon has specific algorithms for the pages it uses. 

#1 Amazon’s Recommendation Engine

Amazon’s recommendation engine runs on almost every page of Amazon. But the best place to understand how it works is the category page. When I go to the “history” category on Amazon, I will see different books than you do when you visit the exact same page. 

How the Amazon Category Algorithm Works

When I visit the “history” category, I see mostly audiobooks because Amazon knows I mostly purchase audiobooks. But it knows a lot more about me than that. Every time I browse Amazon while logged in, and every time I make a purchase, Amazon gets more metadata about me that it can feed its algorithms. The goal of gathering this data is to make me more likely to purchase something. Amazon doesn’t care so much what I buy as long as I buy something. 

The “history” category shows me a book about Napoleon as the top recommendation. Napoleon just so happens to be the historical figure I’m the most curious about at the moment. The second recommendation is a pocket constitution for $2, which I am very tempted to buy. In short, Amazon’s recommendation engine is very good. 

The technology powering this is called “item-to-item collaborative filtering,” where the algorithm builds “similarity tables” by analyzing user actions like purchases, views, ratings, cart additions, and browsing history. It then combines this with metadata about books. 

This can be as simple as people who buy printers buy ink, as David Gaughran puts it. But it can also be more complicated than that. I suspect the pocket constitution was such a high recommendation for me because I recently purchased a “Don’t Tread on Me” flag. People who buy a Gadston flag also buy pocket constitutions. But perhaps it was my purchase of a Holosun optic or my purchase of a shirt that said, “I love one woman and many guns.” Or perhaps it was all that, plus the knowledge that I live in Texas. 

Amazon combined metadata about me with metadata about the pocket constitution. The data showed that I fit the profile of the kind of person who buys pocket constitutions. And Amazon was not wrong. 

So what does the Amazon recommendation engine look at when it sees your book? 

According to Amazon, the recommendation engine looks at: 

Book Metadata:

• Title (Listen to my episode on book titles and how they affect your rankings on Amazon.)
• Subtitle
• Categories  
• Description
• Reviews (These are growing in importance in the A11 era)

User data:

• Purchases 
• Reviews
• Site Activity 
• User Demographics

Note that the pocket constitution book recommendation was likely influenced by my non-bookstore purchases. This is one reason Amazon is so hard to compete with. They know so much more about their customers than Barnes and Noble does. 

So the algorithm looks at similarity tables and metadata. 

Tips for Optimizing Your Book’s Algorithmic Readability

Tip 1: Pick Deep & Distinct Categories

Start by picking two or three specific categories that fit your story perfectly and that have very little overlap. Go as deep in the category choices as you can reasonably go. Remember that your book is automatically entered in all parent categories. 

So let’s say you wrote a Christian Historical Fiction that takes place in WWI. If you put the book in “Books › Religion & Spirituality› Christian Books & Bibles › Literature & Fiction › Historical,” you are already in one child category and four parent categories. 

If your second category is Books ›  Literature & Fiction › Genre Fiction › Historical Fiction › 20th Century › World War I, you’ve added one more child category and four more parent categories. That’s a total of 10 categories before you’ve picked your third child category. So remember that the more similar your parent categories are, the fewer total categories your book will be listed in. 

Amazon may also add your book to additional categories based on keywords and reader reviews. 

Warning: Choose strategically and avoid non-browsable “ghost categories.” For help researching your category strategy, I recommend getting both K-lytics and Publisher Rocket (Affiliate Links). Authors who have neither of these tools are walking through Amazon blind. Getting one tool is like having one eye open; having both tools is like having both eyes open. 

I also built a Patron Tool called the Amazon Keyword & Category Planner. It reads your book and recommends the categories and keywords that best match it. It’s a great supplement to K-lytics and Publisher Rocket, though it doesn’t provide the same depth of platform analysis those tools offer.

Tip 2: Maintain Category Consistency 

Ensure your titles, descriptions, and backend keywords match the categories you picked. If Amazon doesn’t think your book fits a category, it will yank your book from that category.

All of your metadata must consistently convey what kind of book it is.

Warning: Don’t stuff your book into the wrong category in hopes of getting a bestseller badge. This could poison the recommendation engine associated with your books and lead to poor reviews. Negative reviews can cause you to fall out of your category and lose the badge. If your reviews are complaining that “this is not a real romance,” Amazon may take action. 

My Trope & Genre Finder tool will read your book, give you a list of tropes, and give a genre alignment score based on the tropes from your book. 

Individual Book Pages Are Mini Category Pages

Amazon’s recommendation engine doesn’t just run on the category pages; it runs all over Amazon, on the homepage, deals pages, various checkout pages, and particularly on individual book pages.

Each individual book page shows covers for other books. It displays so many related books that it acts as a mini category page. I looked up my book on Amazon, and I saw 48 book covers for other authors’ books. The sections had headings like “Customers Who Bought This Also Bought…”, “Inspired by Your Browsing History,” and “Books You May Like.”

Now you may think of this as other books promoted on your book page. But you can flip that script. Your book could be recommended on book pages all over Amazon. Or it may be on zero book pages. It’s all up to the Amazon recommendation engine. 

One fundamental element of the recommendation engine is this simple phrase, “Customers who bought this also bought this.” This is what is known among indie authors as the “also boughts.” This concept helps power the entire recommendation engine, not just the “also boughts” section. 

So, yes, “also boughts” matter on category pages. This is one reason why professional authors obsess about the “also boughts” section on their book pages. 

Tip 3: Guard Your Also Boughts

If you attract the wrong kind of customer early in your book’s launch, you can forever ruin the item-to-item collaborative filtering for your book. When your book first launches, Amazon doesn’t know much about it yet. This is called a “cold start,” and during the cold start, initial purchases matter a lot. 

If your first buyers are your grandmother and her knitting club, your book will get linked with knitting books and knitting products in the recommendation engine. This means Amazon will show your book to the wrong people. 

Why is this bad? 

Because knitting enthusiasts won’t buy your book, and if they do, they won’t leave good reviews. Amazon will see that your book is not selling to the people “most likely to buy” (according to their initial purchaser data) and may conclude your book is a stinker. Amazon has 30 million books to promote, and it prefers to promote books that sell.

So how do you let your grandmother and her knitting club buy your book without poisoning the recommendation engine?  

Invite her to buy your book from your Kickstarter campaign rather than from Amazon. She would prefer to pay extra for an early access signed hardback anyway. She can also share the Kickstarter campaign with her knitting club.

Invite her and her friends to your launch party, where you can sell in person.  

Your initial Amazon customers should be genre readers. That’s why don’t want only friends and family to be your first or primary Amazon customers. And for the love of accurate book recommendations, you don’t want other authors to be your first purchasers! This is part of the reason I do not recommend having other authors on your launch team.

So focus your early promotion on the kinds of genre fans who read and review books in your genre. To learn more, listen to my episodes on book launches.

Tip 4: Promote Similar Authors

If you promote a similar author’s books to the subscribers on your email list, you can make affiliate revenue. When readers who’ve purchased your book also purchase a similar book, you create a powerful item-to-item collaborative filtering signal to the algorithm. 

Or in plain English, promoting similar authors creates an “also boughts” connection between your book and theirs. 

And the best part is that you don’t need the other author’s permission or knowledge to promote their book!

Follow these simple guidelines for maximum impact:   

1. Only recommend books your readers would love. Your reputation for having good taste is far more important than some temporary algorithm bump.
2. Only recommend books you have read. Don’t assume a book is good or similar to your book. Read other authors’ books as you would have them read yours. 
3. Review Books You Can’t Recommend. If a book is similar to your book but you didn’t like it, you can review it rather than recommending it. Some of your readers may go on to buy it, and you can still get that item-to-item connection while preserving your reputation for having good taste. So share what you liked and didn’t like about the book, along with a link to buy it. 

I have an episode to help you with this called How to Write Book Reviews Readers Will Want to Read.

I also have a patron tool called Review Drafter. It will draft a full book review based on a bulleted list of what you liked and didn’t like about the book. 

Reviewing similar books fixes the “I don’t know what to write about in my email newsletter” problem. You can even start promoting and reviewing books long before your book launches. 

Many authors can’t use this technique because they are not well-read. They don’t read in their own genre. They don’t know which books are hidden gems, and they couldn’t make good recommendations even if they wanted to. If you’re willing to read in your genre, it will help you improve your taste and voice, which will make you a better writer. And the cross-promotion can make you affiliate money and boost your rankings. It’s a win-win-win. It just requires a little humility and hustle on your part.   

Remember, if you want to become a better author, read:

• Books on craft 
• Genre classics 
• Genre bestsellers  

Tip 5: Host a Genre Podcast

If you host a podcast on your subgenre, each week is an opportunity to promote a similar book. It’s also a great opportunity to connect with other authors in your genre. Hosting a genre podcast is a lot of work, but it will give you an edge over all the other new authors in your genre. 

If promoting other authors makes you feel uncomfortable, remember that your true competitors are TikTok, Netflix, and Candy Crush. Fellow authors are your allies in the great war against mindless mobile games. 

Tip 6: Promote Similar (Non-Book) Products

Remember that pocket constitution Amazon recommended to me because of non-bookstore purchases I made? The recommendation engine runs across the whole Amazon store. Most authors don’t think outside the bookstore, but this is where you can get an edge. 

For example:

• If you write a diet book, recommend supplements on Amazon that are compatible with your diet. 
• If you write fantasy, recommend fantasy board games. 
• If you write cozy mysteries, recommend jigsaw puzzles. 
• If you write romance, recommend bath bombs and candles. 

To maximize this technique, avoid recommending the most popular products. Be specific and focus on items your Timothy doesn’t already know about but would love as soon as he learns about them. 

#2 Amazon Search

Unlike the recommendation engine, which is the best in the world, the Amazon search engine is decent but not great. It still returns a lot of poor results. 

Recently, I did a search for “zoysia grass seed,” and the first results were for various kinds of fescue grass seed. That’s a totally different kind of grass! The first zoysia grass seeds don’t appear until halfway down the page. Amazon search prioritizes what sells over exact matches, which creates opportunities for you. 

If you know what you are doing, like those fescue seed sellers, you can rank for all sorts of search terms related to your book. I cut my teeth with Google SEO, and compared to that, Amazon SEO is simple. 

Success just requires work on your part. Here’s a breakdown of what to do.

Tip 1: Optimize Your Page for Phrases Your Readers Actually Type 

Search engine optimization only helps if you rank for phrases readers are already using. 

To find out what phrases readers are using, open an incognito browser tab and start typing in Amazon’s search bar for your book. What auto-suggestions appear? Those are real phrases shoppers already use. Grab the dozen or so phrases that best fit your book.  

These are the phrases you want to try to rank for.

There are also some tools you can use to help:

• Publisher Rocket (affiliate link)
• Patron Toolbox: Amazon Keyword & Category Planner will read your manuscript and generate keywords for you.  

Now that you know what you want to rank for, let’s start optimizing. 

Tip 2: Use a Keyword-Rich Title & Subtitle 

The most effective place to include keywords is in your book’s title and subtitle. This strategy is so powerful that I devoted an entire episode to it called How to Pick a Strong Book Title. 

This technique requires some finesse. Stuffing keywords into your title can turn off readers if you don’t know what you are doing. So make sure to listen to that episode to learn the nuance required for this technique. 

I also have a Patron Toolbox tool called the Book Title Optimizer. It helps you generate search-optimized titles, subtitles, and series names.

Tip 3: Fill the Keyword Boxes

Unlike Google, which ignores meta keywords, Amazon still uses them to inform its search engine. 

So make sure you fill all seven keyword boxes. You can fill these in your KDP dashboard or by asking your publisher to do it for you (if you are traditionally published). 

Here are some best practices for KDP Keywords:

• no commas 
• no quotes 
• no more than 50 characters per box
• no repeated words

My Amazon Keyword & Category Planner can help you. It reads your book and generates keyword phrases to fill the boxes with no overlapping words.  

Tip 4: Search Optimize Your Book Description

In your description, weave keywords naturally into the first 200 characters. The Amazon Keyword & Category Planner can help you with this as well. It will give you a few keyword super sentences that describe your book in a keyword-rich way. You can work those sentences into your existing description for an SEO boost.    

Warning: Remember that your primary audience for your book description is the human readers trying to decide if your book is worth buying. 

Remember that Sid was operating the claw in Toy Story. 

Readers have countless options for what to read next. They can never get back the hours they spent reading your book. If it takes ten hours to read, then reading your book moves them ten hours closer to death.

Your book description must convince cautious readers that your book is so good it’s worth moving ten hours closer to death. You must persuade them that they can’t risk not reading it because it’s that compelling.

If you can’t do that, the keywords won’t matter. But if you can do that while also using strong keywords, you set your book up for enduring success.

If you want help crafting a convincing book description, check out my episodes on pitching fiction and pitching nonfiction. I also have a bunch of patron tools to help you craft an Amazon description. 

Tip 5: Get Reviews Fast 

Amazon stated in a recent blog post that reviews, ratings, and returns are search ranking factors.

As it has been said of old, “to him who has reviews, more reviews will be given. But to him who has not reviews, even the reviews he thinks he has will be taken away by Amazon.”

It is easy to get so focused on the algorithm that we forget about the reader and specifically about reader psychology. One major psychological factor impacting sales is social proof. Readers want to read what other readers are already reading. Review count is a key social proof indicator. 

This is another reason it’s so important to put your book in front of readers who are likely to love it enough to write a review. The more reviews a book has, the more informed buyers feel, which leads to fewer returns.

I have at least four episodes on how to get more reviews:   

• How to Get More Book Reviews With Joe Walters
• How to Get More Book Reviews With Derek Doepker
• How to Get More Book Reviews With Jim Kukral
• How to Get More 5-Star Reviews With Beta Readers, Editors, and Launch Teams

If you are struggling to get reviews, I strongly recommend you listen to all four episodes. 

Tip 6: Drive Warm Traffic to Your Amazon Book Page

Sending your email list of super fans straight to Amazon will help boost your rankings. 

The Amazon algorithms know if people check out or not. Book pages with high conversion rates can expect better rankings. So make sure to optimize your Amazon book page. I also have an experimental Amazon Page Optimizer Patron Toolbox tool that can sometimes scan your Amazon page and give you optimization tips. 

Another way to increase your conversion rate is to send “warm traffic” to your Amazon page. In other words, convince readers to buy your book before they go to Amazon.  

For example, if you advertise on places like Facebook, consider sending traffic to the book page on your website first. This page either convinces them to buy, in which case they will go to Amazon warm, or they will decide not to buy, and Amazon never sees the reader say no to your book.

Pre-warming your traffic can boost your Amazon conversion rate and your rankings. 

All your marketing (if done well) can help boost conversion rates:

• Author Website
• Podcast Interviews
• Email Marketing
• Public Speaking
• PR
• Advertising

In other words, make sure to subscribe to Novel Marketing to keep getting training on all these things in the future. Like our videos on YouTube to ensure the YouTube algorithm continues to show them to you.  

Tip 7: Launch Team Activation

The next tip is from Dave Chesson, who developed a cool technique for how you can use your launch team to help you with SEO. 

During launch week, instruct your launch team to search Amazon for your target key phrases, find your book, click the listing, and then buy the book. Just a few of those focused purchases can rocket you to the first page of results for that phrase. 

Tip 8: Follow The Rules

Amazon has no problem banishing your book to page 500 of the search results if you break the rules. 

• Don’t stuff keywords
• Don’t buy reviews. 
• Don’t put your book in categories that don’t fit. 
• Don’t violate the Amazon KDP Terms of Service. 

#3 Amazon’s Bestseller Algorithm 

You may be wondering, “If I can’t be on the first page of the bestseller rankings, why should I care about my bestseller status? Can’t I just ignore this portion of the episode?” 

No. 

Your Amazon Best Seller Rank or BSR impacts both the Amazon recommendation engine and the Amazon search engine. All three systems influence each other. The higher your BSR, the more the recommendation engine will recommend your book and the more the search engine will surface it. 

So let me put this as bluntly as I can: Amazon doesn’t want to recommend slop readers won’t like. If your book can’t outsell the slop flooding Amazon, the algorithms will treat your book like slop.

But take heart, you are not racing against the bear. You are racing the other hikers. Or more specifically, you are competing with the other books in your category. 

Joe Solari’s analysis suggests Amazon uses sales data from the last 180 days to calculate the BSR. Amazon itself only says that BSR reflects a mix of recent and historical sales, without publishing the current formula. But I suspect Joe’s analysis is correct. 

The rankings are no longer updated hourly. In 2010, there were a lot of tricks to grab bestseller status for a short time. Those tricks don’t work anymore.

But how can a brand-new book have sustained sales over the last 180 days? Are new books at a disadvantage? 

Warm Start vs. Cold Start

Amazon calls this the “cold start problem.” The algorithm can mitigate cold starts by occasionally surfacing new books to users to see if they get clicks, but with millions of AI books flooding Amazon every year, your book is not guaranteed this kind of help.

During your cold start, the denominator in the ranking algorithm is the number of days your book has been published. This means a surge of sales during your launch week will have a much bigger impact than a spike in sales at any other time. 

Not only that, but the spike in sales will start to feed the algorithm customer data about the kinds of readers who buy your book. This helps the recommendation engine know who to recommend your book to. 

You will still need steady sales. If your book sold 900 copies in the first week, that looks really good to the ranking algorithm because 900 sales in a week is about 129 sales per day. That is good enough to rank in some subcategories. But 900 sales in three months is only ten sales per day, which isn’t enough to rank anywhere that matters.  

Some authors learn about the 180-day weighted rolling average and claim that launches don’t matter. That is not true for several reasons that will make sense after you read the tips below. 

Tip 1: Strive to Earn the “Number #1 New Release Badge” 

For the first 30 days after your book’s release, there is a special bestseller badge you can get called “#1 New Release.” It’s much easier to get than a number-one category bestseller because you are only competing with other brand-new books in your category. 

The #1 New Release badge helps you get more sales and reviews that will increase sales and reviews for the rest of the book’s life. It also acts as social proof, compensating for your relatively low initial review count. 

In my experience, books that get #1 New Release badges tend to sell more lifetime copies than books that don’t.   

Tip 2: Launch Strong to Trigger Ongoing Word of Mouth

It is more fun to read a book while your friends are reading the same book. If you want to maintain the amount of book sales that allows you to rank on the bestseller list, you need ongoing word-of-mouth promotion from your readers. The bandwagon effect is real. Never let an algorithm prevent you from getting your human readers hyped about buying your book as soon as possible. 

If you are ever tempted to keep readers from buying your book, you’re focused too much on the claw and not enough on Sid.

The more people you get reading your book right away, the more word of mouth you can spark—if they love your book. Some authors can persuade readers to buy, but the book isn’t strong enough to hold their attention for 200-plus pages. Those readers don’t finish, and they don’t recommend the book. The authors then blame the algorithm, when the real issue is a boring book.

Generally, readers who don’t finish a book don’t leave reviews. The “DNF” (did not finish) review designation began on Goodreads and is slowly making its way onto Amazon.

What About Pre-Orders? 

Pre-order sales can help offset the cold start problem by feeding the algorithm item-to-item collaborative filtering data ahead of launch day.

For first-time authors, I recommend using the maximum pre-order window to give you time to fix problems, make tweaks, and learn what you don’t know about what you don’t know. So many rookie mistakes are caused by indie authors being rushed at the end of the process. 

For more on this, listen to my episode on 10 Reasons to Delay Your Book Launch.

I have over 30 episodes on book launching, so here’s a link to the Book Launch tag on AuthorMedia.com so you can easily browse the related episodes. 

Tip 3: Put Your Book in KU 

The initial KU borrow (when a reader downloads your book through Kindle Unlimited) counts as a sale for ranking purposes, which contributes to your book’s sales velocity and overall BSR just like a regular purchase. This means that KU (Kindle Unlimited) borrows influence the bestseller rankings in both the Kindle store and the overall books categories on Amazon. 

I have found no evidence that subsequent page reads count as additional sales. From everything I’ve read, page reads are a way of rewarding royalties and are not used as a ranking factor. Only the initial borrow itself influences rankings.

Tip 4: Advertise 

While advertising has no direct impact on the algorithms, it’s still a good idea.

Advertising has a massive indirect impact. Advertising drives many factors we’ve already talked about (sales, velocity, reviews, ratings). All these secondary factors impact the algorithms. 

Remember, the more sales a book gets, the higher it ranks, and nothing drives steady sales like continued advertising. 

So where should you advertise?

You should advertise in the same places where the books you want to be linked with are advertised. You want the same readers who buy those books to buy yours. If you do this well, your book cover will start appearing on book pages all over Amazon. Check out the following episodes on advertising. 

As I said before, I recommend sending traffic to your website first. But, if you prefer to send readers directly to your Amazon page, make sure to use Amazon attribution links to track which ads work and which ones don’t. 

Tip 5: Price Pulse (Yes, it still works.)

Price pulses on platforms like BookBub won’t rocket you to a number-one ranking like they used to, but they can still create a surge of reviews and word-of-mouth buzz that helps generate ongoing sales. 

Certain readers only buy books from BookBub deals, and it is better to get that reader’s money, reviews, and recommendations than not. 

Resist an elitist spirit that doesn’t want to sell to low-budget readers on BookBub.

Tip 6: Write the Next Book

After your book goes 180 days with low sales, it becomes extremely difficult to revive it through the algorithm. So what should you do?

Write the next book.

This gives you a new launch window to work with. Launching a new book typically boosts sales for old books, especially if you know how to optimize your backmatter.  

It doesn’t have to be a sequel. In fact, if the first book sold poorly, a sequel is likely a mistake. Even a new stand-alone book can boost your sales, particularly if it is similar in genre and tone to your previous book. 

You can’t squeeze blood from a stone. Eventually, a book sells all the copies it’s going to sell. Books function like stepping stones in your career: at some point, you need to step onto the next stone and leave the last one behind. 

Amazon Algorithm Superstitions  

Let’s talk about some rumors spreading about the Amazon algorithms.

Superstition 1: Amazon is looking at the read-through rate and completion rate. 

I spent a week researching this question. There is no official evidence from Amazon to support this claim. Also, it doesn’t make sense if you expand your view outside of genre series.

Consider the reader who purchases

• A nonfiction book: The reader reads part of it, finds the answer to her question, and puts the book away. This reader had a positive customer experience and is likely to go on to buy the next nonfiction book from that author.
• The book for a book club: The reader did the bare minimum of reading to keep up with the book club.  
• The book for school: The reader only read the selections assigned by the professor. 
• The book, read to completion: The reader hated the ending and threw the book across the room. This reader completed the book but will never buy from that author again.
• The books about diets: The reader buys a diet book, reads the first chapter, buys another diet book the next month, and another the next month. This reader is a better customer for Amazon than the reader who purchased one diet book, read the whole thing, and stopped buying diet books.  

I could go on and on with examples. 

The reality is, most readers don’t finish most books. Just because a reader finished a book doesn’t mean she had a good customer experience. 

And for books that people do “finish,” they don’t always finish in a technical sense because they didn’t read the third appendix. Did you finish reading Return of the King, or did you *finish* it? Do you know what caused the Kingdom of Arnor to fall and why the Dúnedain survived? If you don’t know, you didn’t finish the book. 

Completion rate is a noisy quality signal. Noisy data for the algorithm leads to a garbage-in-garbage-out situation.  

It is easier and more effective for Amazon to track something far less noisy, like “sequel purchase rate.”

Another less noisy metric would be brand loyalty. How likely is a reader to purchase another book from that author? This kind of metric works on books as well as other products in the store. Remember, books make up a tiny slice of Amazon’s total profit.    

We know for sure Amazon tracks brand loyalty for brands in the brand registry. What we don’t know for sure is whether Amazon uses brand loyalty or sequel purchase rate as factors for authors who are ineligible for the brand registry. If I were on the Amazon machine learning team, I would use brand loyalty and sequel purchase rate as factors long before using the much noisier completion rate. 

Telling all authors to focus on completion rate could really harm nonfiction authors, who often benefit from appendices. But it can even harm novelists who want to be creative with their back matter. 

My advice is to ignore this superstition and focus on thrilling your readers. 

Should you write books that readers want to finish? Of course. But don’t be afraid to include an appendix because you think the algorithm will punish you for having a low completion rate.  

I’m willing to be convinced on this. But unless I see convincing evidence from Amazon itself, I won’t believe completion rate matters. You can call me Doubting Thomas.

Superstition 2: Amazon reads the whole book for the recommendation engine. 

There is no publicly available evidence to support the claim that Amazon is using an LLM to “read the whole book” as part of its recommendation engine. Based on Amazon’s own research pages, the recommendation process relies primarily on behavioral data, metadata, and aggregated user interactions rather than deep content analysis of the books themselves. 

The whole point of that kind of project would be to create metadata, and Amazon already has human-generated metadata for each book.  

The other reason I don’t think Amazon is doing this yet is that, as the creator of the Patron Toolbox, I know how expensive AI tokens are. 

LLMs didn’t have context windows big enough to read an entire Brandon Sanderson book until a couple of months ago. Using bleeding-edge models is insanely expensive. Using them on millions of books is prohibitively expensive.

Now, LLM costs are falling, and as we move into what I’m calling the “A11 Era,” Amazon might start feeding books to LLMs. But for now, it is much cheaper and more effective to rely on the user behavioral data and the book metadata that Amazon already has.

That said, it is possible that Amazon is using an LLM to read the first few pages of each book, particularly to help verify category placement. I think it is safe to assume Amazon lets its AI bot Rufus read the Kindle Instant Preview pages. Running an LLM on the first few pages solves the cost and context window challenges. 

Possible Superstition 3: External traffic counts more.

The final possible myth is that Amazon rewards authors who bring in external traffic. Every expert I could find on the algorithm either holds this view or is silent on the topic. The most notable experts are Dave Chesson and Joe Solari, but neither of them cites a source for the claim.

That said, both Dave and Joe do their own direct testing on the algorithm, so they are both sources themselves, and they agree. It is possible Amazon wants to avoid antitrust trouble and therefore won’t share this factor publicly.  

On the other hand, David Gaughran hasn’t said anything about external traffic one way or the other, which is notable since he literally wrote the book on the Amazon algorithm (Affiliate Link).

It makes intuitive sense that Amazon wants to reward sellers who send it traffic. But we have not heard anything about this from Amazon directly.

We know for sure that overall sales matter since Amazon has a whole page about it on its blog. Even if external sales don’t count for extra, they still count. And we know Amazon doesn’t count sales from you or from a competing bookstore. 

So this is a very nitpicky quibble. Sending external traffic to Amazon helps boost your rankings a little or a lot. 

Remember Your Timothy

As you learn more about the algorithms, never forget your Timothy. Or, to go back to our Toy Story reference, remember Sid behind the glass controlling the claw. The algorithms are tuned to help readers find the book they would most enjoy. Reader actions like buying, reviewing, or returning books train the algorithm. So remember to write the book Timothy would most enjoy. Love your reader as much as you love your book, and you will do well. 

Don’t Forget to Like and Subscribe!

The YouTube machine learning algorithm doesn’t use a sale as a success point for training. It uses likes, subscribes, and most importantly, watch time. So thank you for watching this video to the end. This will encourage YouTube’s item-to-item collaborative filtering system to show you more episodes like this one. 

If you thought it took a long time to watch, just imagine how long it took to prepare. 

If you really want to juice the YouTube algorithm, you can leave a comment, ask a question, respond to someone else’s question, or angrily point out some mistake I made. It all serves the algorithm. The more comments, the more YouTube will promote this video to other authors.  

But I do invite you to fact-check me. In over ten years of podcasting, there is only one episode I spent more time researching than this one. The more I learn about machine learning, the more I learn there is to learn. I’m sure Jonathan will have some fact checks and rebuttals for me in the next episode of Author Update. 

Featured Patron

Kaye Brownstone, author of Whispers of Kith and Kin (affiliate link)                                    

On the run from a cartel, Lily Jordan enters Witness Protection, hoping to disappear in a quiet southern town. But when a despised gossip columnist is murdered, Lily becomes the sheriff’s prime suspect. To clear her name, she must unravel a crime implicating half the town. In this cozy mystery, the whispers of Kith and Kin can be deadly. 

Related Episodes

• Search Engine Optimization for Author Websites
• How to Rank in Amazon Search Results with Dave Chesson
• How to Create Sales Optimized Book Pages for Your Author Website

YouTube Version