Thanks to advancements in Publisher Rocket, we’ve been analyzing data from Amazon for a while. Because of this, and the use of other tools, we picked up on a pattern in how Amazon treats books and their discoverability that I think changes a lot in the industry, and how we authors should operate.
Basically, through analytical proof, we found that when a book has a rise in sales on Amazon through just about any means, Amazon responds by increasing the number of keywords that that book shows up for naturally, and its rankings for those keywords as well. Therefore, because of the increase in a book's popularity, the book would show up more often in the Amazon store.
I’m calling it the Amazon Popularity Effect.
On the one hand, this makes sense. However, this also opens up a lot of questions like how does Amazon do this and what does this mean for authors? Are there things we can do to help ride this popularity effect or take advantage of it? Make better decisions? Plan better launches?
But to answer these questions, we need to understand more about the popularity effect, and how far it really goes. Therefore, in order to help understand this phenomenon and see what else we can learn from Amazon, I decided to conduct a full experiment, record the data and give my thoughts on what is going on here.
- What the Amazon Popularity Effect is
- How this affects authors
- The effect of a single promotion
- The data and science behind an excellent book launch
- Does Amazon prefer newer books?
- What tactics you should employ based on its findings
Table of contents
- How We Got the Data for This Experiment
- Amazon Popularity Graphs Explained
- Experiment 1: Sales vs Number of Keywords
- What Kind of Keywords Does Amazon Add to Popular Books?
- Experiment 2: Single Book Promotion or Sales Spike Effect
- Experiment 3: Book Promotion Launch Effect
- How Authors Can Use This Information
- Recap: What Authors Can Learn from This Experiment
PS: I want to give credit to Ed Robertson – who to the best of my knowledge, popularized the phrase “Amazon Popularity” in his somewhat different but similar context. Also, credit to David Gaughran who is the first I had heard of the phrase in his AMAZING book “Amazon Decoded” (a must in any author's book shelf).
How We Got the Data for This Experiment
We started with the hypothesis that Amazon gives preferential treatment in discoverability to books that show higher conversions and sales over time. In order to prove this, we needed a way to not only track sales of books over time in mass quantity, but also to record the number of keywords a book shows up for and its rankings throughout the entire store. If our hypothesis is correct, as a book makes more sales, we should see Amazon start to show the book more often throughout their site in more search terms.
For the first part of the necessary data collection, Publisher Rocket served beautifully. Over the past year, we’ve added new features, and improved capabilities within Publisher Rocket. But one of our biggest additions that we haven't talked much about is that we are now actively pulling information from Amazon instead of passively, and storing it on our servers so as to provide Publisher Rocket users more longer term data and better information.
This is allowing us to see historical values, trends in the markets, and more. More importantly, we will be releasing new features soon that use this information to help Publisher Rocket owners gain even more knowledge about Amazon, the book markets, and more.
Here are some of the things to come, thanks to our data collection:
If you already own Publisher Rocket, then these will be a free upgrade for you when we release them soon.
The second tool, was the crawler I built for last year's experiment on figuring out the best way to fill in your 7 kindle keywords. This crawler was designed to crawl through Amazon and mark how many times a book was listed or indexed for keywords and their rankings. Just like I mentioned in that experiment, it is quite costly to operate (since Amazon hates crawlers) – which is why it can't be a viable or dependable service for authors (so not worth the cost to break even). Instead, it serves as a great tool for the pursuit of knowledge like this one. (PS: I wanted to give a special thank you to all of those out there who have supported me in being able to do experiments like this – the coffee fund has been a major help)
Now that we have the two tools established, and their methods for collecting data, let's discuss the process.
In order to test the hypothesis, we set both systems under their current capabilities to track well over 500+ books on Amazon. We looked for books that had a certain sales level, and then all of a sudden had a jump in sales – either a quick spike or something more consistent and sustainable. From those, we found 164 books that met our condition. These books did range from new to old, published to self published and covering both fiction and nonfiction. From there, we continued to record their sales over time, and the number of keywords they ranked for.
From this, I was able to take the data and answer three questions, which I label as three specific experiments below:
- Do the number of sales affect the number of keywords Amazon chooses to rank a book for?
- Do sales spikes influence Amazon to naturally show the book more often or do they not play a role?
- What is the effect of a consistent increase of sales?
Based on the answers to these three questions, we will expand on action items we authors should take and use to our benefit.
Limitations to the Experiment
As with any experiment, there are certain limitations to what can be done. However, as you'll see later, I believe there is enough data to show a strong enough conclusion. Here are some of those limitations that should be considered:
- We Only Followed 164 Books: Statistically speaking this is a lower number than I'd like. However, running the scraper and correlating the data from it and Rocket was extremely time consuming and expensive. Furthermore, from the number of books, we did see a strong enough pattern that I believe is significant enough to form conclusions without having to expand testing further.
- Granularity in Genres and Subjects were not Considered: It should be noted that we did not specifically track certain genres or topics. However, I believe that Amazon's response to books isn't predicated off of their genre or topic and so this potential variable could be ignored.
- Unknown promotions or tactics of books: Almost all of the books we tracked, we did not coordinate with. We do not know what those books did or whether or not they did anything. Their jump in sales could have been from emails, promotions site, etc. Or they could have just been lucky with a specific keyword that took off. I think this experiment would be even more concrete if I had coordinated the promotion sets over over 100+ books and recorded, but I am currently not capable of such a momentous task…there aren't enough Focus Aids in my fridge. However, as you'll see in a certain experiment below, I did coordinate with a couple of books to isolate certain things so as to answer some questions.
- Affect of Also Bought and Other Methods: Sadly neither tool was capable of analyzing the books that show up in Also Boughts or other areas where Amazon will choose to show a book as well. This doesn't negate the information collected since, as you'll see, the number of keywords a book was shown for did in fact increase. However, I have a feeling that the number of times a book showed up in another book's Also Bought probably increased as well. So, I'd hypothesize that this too adds to the overall Popularity Effect from sales – it just wasn't taken into consideration during this experiment.
There are a couple of more limitations but these are some of the biggest.
Amazon Popularity Graphs Explained
Before we get into the results of experiments, I wanted to first talk about the graphs you’ll see below.
In order to create the following graphs, we took the total number of graphs built from both Publisher Rocket’s data collection and the scraper and overlaid them on top of each other to create the general graphs that depict the overall trend. The orange represents the number of sales, and the white dash line represents the number of keywords that the book is ranking for.
We removed the y-axis because we didn’t want people to assume that there is an arbitrary number of sales that equals a certain number of keywords. Nor is this indicative of the number of keywords being less than the number of sales. It is only created to show the respective correlation.
So, with that, let's go ahead and jump in.
Experiment 1: Sales vs Number of Keywords
In this first experiment, we wanted to see if books that had a clear increase in sales, would also have an increase in the number of keywords they show up for. As you can see from the graph below, this did in fact happen in a majority of books analyzed. Not only did the number of keywords a book showed up for increase, but so did the rankings for certain keywords as well.
However, it should be noted that the effect of the increase in the number of keywords took on average between 6-8 days before the number became noticeable.
On the whole, this sort of response by Amazon makes sense considering that the increase in a book's popularity might be cause to believe that the book is more popular in general and can convert more. Considering that the overall objective of Amazon's search algorithm(s) is to help increase Amazon's sales by showing the right product to the right customer more often, it can be assumed that Amazon would take such a action, and try to show more proven popular books over less popular.
Basically, if I were the one in charge of the algorithm, I'd make it so that when a product starts to increase in popularity, I would work to show that product more to see if I can ride a trend. So, these results track logically.
So, how did they do this? I will explain my ‘belief' in the next section.
Conclusions From the Results
1.The more sales a book made, no matter where the sales came from, a book would start to index for more phrases on Amazon over time.
2. It took about ~6-8 days of consistent new sales before Amazon would start to show any notice.
What Kind of Keywords Does Amazon Add to Popular Books?
Caution: Please understand that the below information is my subjective opinion developed by looking at the above data and the new keywords that were added to books. I want to make this clear.
If Amazon is choosing to show a book for more keywords because it is selling more, then the next real question is what kind of Keywords is it showing a book for, and what can we do about that.
After analyzing the data, I believe that Amazon started showing books for two things:
- More terms that are like the ones the book was already ranking for
- Broader or tertiary terms that fit the natural character of the original keywords
The first one makes sense in that Amazon knows what keywords do well and how certain keywords are related to others. They track everything from previous books and would know that if a book is doing great for ‘Keyword A,' then historically, it would also perform well for ‘Keyword B.' I saw a lot of keywords added to popular books that were similar to those it was previously ranking for.
As for the second point, I noticed that certain books would also be added to more broad terms than what they were originally listed for (this part is using last years experiment where we did control the keywords people selected and changed and watched what happened – you can see that here). Here’s an example:
In the above example, a book was ranking for the first set of terms on the top. Then, because of the book's performance, Amazon started showing more broad terms in better rankings. These terms aren’t as specific as the original group of keywords, but they do get searched more and the book still ‘fits’ in these.
Then, considering that the book performed relatively well in that group, Amazon then added it to even more broad terms with better rankings. However, in this case, the book then faltered and was not performing as well as Amazon would expect books should for those. So, it was removed from that more broad group and went back to the area of keywords it was performing better than expected.
This shows the importance of the Kindle keywords you select. If you select keywords that your book doesn’t legitimately fit for, then your conversion rates and performance for those keywords will be poor, and your ability to grow in discoverability on Amazon will be reduced. Furthermore, Amazon might have a hard time linking you to tertiary or more broad keywords as well, since it isn't evident where the book should fit.
It is for this reason that authors need to know which are the best keywords for their book. Using Publisher Rocket’s keyword feature and seeing how many people type in a keyword is critical. But it also requires you know which keywords truly fit and describe your book. It is this reason that I recommend you follow the steps to selecting your 7 kindle keywords in this article.
Now that we've seen a direct correlation between sales, and Amazon increasing the number of keywords a book ranks for, as well as discussed my thoughts on how this actually happens, let's look to see if a sales spike has any effect or not. This will help us in understanding book promotions and how we should best treat them.
Experiment 2: Single Book Promotion or Sales Spike Effect
As many authors have seen, when you do a major but specific book promotion, there is usually a spike in sales or downloads. The hope is that when the promotion is over, the momentum of the promotion continues and sales continue beyond the promotion time period.
This was a major tactic that authors used back in the day to beef up their sales. Using a Free Kindle promotion, or some tactic of the sorts, they could keep their books relevant. However, as most authors have noticed, this sort of practice has lost its efficacy. Therefore, if the popularity effect is correct, we should see data to help us understand this better.
To answer this, we did a couple of experiments where we tracked books sales before a promotion, during, and after. We compiled the data to create this general graph of our findings:
While the above graph would show that there is a higher number of sales during the promotion than before, it appears as though there was not enough time for the Amazon Popularity Effect to take effect. The problem with promotion site marketing efforts is that they are generally ephemeral and as we’ve seen under previous studies, spikes in marketing efforts do not gain Amazon’s attention.
We can conclude that a single promotion will have little effect on your book’s popularity on Amazon. Sure, you could have more downloads and sales during that brief period of time. However, this will not be enough for you to breathe new life in your book or grow it…unlike what it used to be 4+ years ago when this would help.
As I’ve said before, it is much better to directly sell 1000 copies evenly over a 30 day period, than to sell 1000 copies on the first day, and none for the next 29 days. As we'll see in the next experiment, consistent sales gain Amazon’s attention.
Conclusions From the Results
1.Spikes in sales do not have a long term effect
2. Book promotion sites are not as effective as they used to be if they are not stacked one after the other
3. Authors should look to build a series of promotion efforts so as to feed off of each other, rather than isolated tactics
Experiment 3: Book Promotion Launch Effect
One thing I’ve always wondered is, does Amazon give preferential treatment to new books? Even more so, what does a well-planned book promotion launch look like, and if done right, how does Amazon respond?
Well, after analyzing the data, here is a general representation of what we saw:
The above is an example of a well planned out book launch. In this, the promotions and marketing tactics were spread out over time and thus minimized giant spikes left with low valleys. Instead, it created a high level of consistent sales.
And, as we’d expect, Amazon rewarded these books with increased keywords and popularity in their store.
But then something happened.
After the a specific point in time (ranged differently between books), the number of keywords would drop even if the sales stayed consistent. Now, the drop wasn’t terrible if the book continued consistent sales…but it was still there and noticeable.
I believe that this is an indication that Amazon does give preferential treatment to books that are newer, rather than older. I like to say that at about the 30-40 day mark, the honeymoon period is over and the climb is a bit harder. I've expanded on this point, and publishing a full article on this honeymoon period because there is even more to dig into on this subject.
Conclusions From the Results
1. Combining promotions together has a great effect on Amazon’s Popularity Effect
2. Maintaining a consistent book launch will increase your overall popularity on Amazon
3. Amazon does seem to favor newer books and gives them preferential treatment
4. You can minimize the drop of the effect by having consistent sales
How Authors Can Use This Information
From the data above, it appears as though choosing the right keywords, and creating consistent sales over a period of time can play a major part to benefiting from the Amazon Popularity Effect. So, what specifically can we do?
Here are steps authors should look at:
Step 2: If you already have keywords set, then perhaps look at updating them using this process
Step 3: If you do a book promotion, be sure to stack your promotions one after another – this will ensure the Amazon Popularity Effect takes effect
Step 4: Using methods that trickle sales over time seems to be incredibly important for long term sales. This includes autoresponders to your email list, advertisements (especially Amazon ads – take my free course), and other tactics to continuously send traffic to your book
Furthermore, you can always supplement your marketing efforts by checking out this list of book sales ideas and see what you can use next to each other so as to stack them over time.
Recap: What Authors Can Learn from This Experiment
Books that make consistent sales show up for more keywords and have better rankings. It appears as though Amazon puts those books in tertiary terms or phrases they know are connected with the phrases the book is doing well in. Therefore, this Popularity Effect grows a book’s discoverability.
Book Promotion Sites are not as effective as they used to be. The Amazon search algorithm does not give short-lived spikes any attention that helps a book stick. There are more sales than if you hadn’t done a book promotion, but it does not solely turn the tide on a book’s discoverability.
When launching a book for the first time, it appears as though there is some kind of new book factor in place. Our data shows that new books get a benefit of the doubt over older books. I like to call this the Honeymoon Period.
Amazon Ads create continuous consistent sales if done correctly. Those sales help to kick in a book’s Amazon Popularity Effect and thus, Amazon keeps the book listed in more keyword phrases than it would have if it hadn’t done the ads.
This data shows that there is extreme importance in spreading out your book marketing efforts over time so that you gain consistent sales during that time period. This has a greater effect on your keyword rankings and thus improves your Amazon Popularity Effect even more.