How I Used Amazon Reviews to Predict Sales of My Book

This research project started with a simple question that has crossed every author’s mind: 

How can I tell which books are successful?

TV shows have Nielsen ratings, music albums have the Billboard Top 100, and movies regularly publicize their box office earnings, but not books. Bestseller lists like The New York Times are notoriously subjective, while more objective ones like Amazon and The Wall Street Journal only provide rankings, not raw numbers.

The most reliable sales figures in the publishing industry come from NPD BookScan, but a subscription to that service costs hundreds of thousands of dollars per year and is only accessible to people in the publishing industry. Academic research on BookScan is explicitly banned!

Now, why is it a problem that we don’t have access to reliable sales figures for books? 

Without that information, authors don’t know who to look to as models. We don’t know if the seemingly meteoric rise of a book or author in the media is just hype, or if those books are actually flying off the shelves. As a result, we can’t learn what the most successful authors are doing right and thereby open the door for more authors to succeed.

That is, until now…

Using official sales figures from BookScan, publicly available data from Amazon, and some basic math, I’ve created a model for determining with a considerable degree of certainty which books are selling, to what extent they are selling, and the odds they are likely to continue selling in the future.

Here’s a summary of my findings:

• The total number of Amazon reviews a book has is an excellent indicator of its actual sales (88% correlation)
• A book rapidly receiving a large number of Amazon reviews upon release is a strong indicator of lifetime sales potential (65% correlation)
• A book’s average Amazon rating is a weaker predictor of sales (30% correlation), meaning that it doesn’t matter so much whether a review is positive or negative
• The average book in my sample received 15 reviews per day over its lifetime, on average 1.5% of book buyers left a review, and each review on average corresponds to $115 in sales
• My book Building a Second Brain is accumulating 11.7 Amazon reviews per day on average – faster than 11 books in this sample, and slower than 6
• As more time passes since a book’s publication, its average rating drops slightly, and the number of reviews it receives per day drops substantially

Read on for more detailed findings.

How can I tell if a book is selling well?

I started by selecting 20 books (including my own for comparison) to study closely. Each of these books had to meet the following criteria:

• Published no more than 10 years ago (5.2 years ago on average)
• Received at least 2,000 Amazon reviews (with an average of 23,000)
• Rated highly on Amazon (with ratings between 4.4 and 4.8 and an average of 4.6)

I only chose books that I personally admired and felt had made a significant cultural impact, so that I’d be confident drawing lessons from them for myself.

Here are my subjects:

• How to Take Smart Notes, by Sonke Ahrens
• The Life-Changing Magic of Tidying Up, by Marie Kondo
• Building a Second Brain, by Tiago Forte
• Ultralearning, Scott Young
• Sprint, by Jake Knapp
• This Is Marketing, Seth Godin
• Four Thousand Weeks, Oliver Burkeman
• The Bullet Journal Method, by Ryder Carroll
• The Coaching Habit, Michael Bungay Stanier
• Big Magic, Elizabeth Gilbert
• Deep Work, Cal Newport
• The Subtle Art of Not Giving a F*ck, by Mark Manson
• Show Your Work!, by Austin Kleon
• The Psychology of Money, by Morgan Housel
• Atlas of the Heart, by Brene Brown
• The Daily Stoic, by Ryan Holiday
• The Body Keeps the Score, by Bessel van der Kolk
• Atomic Habits, by James Clear

I then looked up official BookScan sales figures for these books, via an industry contact. BookScan claims to measure about 80% of book sales, which doesn’t include ebook, audiobook, international, and certain retailer sales. However, since I have access to much more accurate and comprehensive data on sales of my own book, I was able to determine it’s actually far below that. I decided to apply a multiple to the BookScan sales figures to arrive at a more realistic “projected” sales volume for each book, which I then used in the rest of my analysis.

Each of the bestselling books I studied has sold 3.1 million books on average as of January 2023, ranging widely from 53,000 to 11.8 million copies. By cross-referencing these numbers with information from Amazon, I found the following correlations (ranging from -100% for a perfectly negative correlation, to +100% for a perfect positive correlation):

• The total number of reviews positively correlates to total sales: 88%
• The average Amazon rating positively correlates to total sales: 30%
• The time since publication positively correlates to total sales: 29%

In other words, a book’s total number of Amazon reviews is an excellent indicator of a book’s actual sales. This makes sense since Amazon is the dominant retailer of books today and the more people buy and read a book, the more of them will likely leave a rating or review. Thus, a book’s number of Amazon reviews can be used as a publicly available signal of how it’s selling, a useful rule of thumb for any aspiring author looking for models of success.

The average Amazon rating for a book also correlates with more sales, but much less so, which means that it doesn’t matter very much whether a review is positive or negative. Just the act of leaving a review signals that a book is selling well. There’s no such thing as bad publicity apparently. The number of days a book has been on the market also somewhat correlates with its sales since obviously, they have had more time to make sales.

So far, not much of a surprise: books with more Amazon reviews, higher ratings, and more time on the market sell better!

How can I tell if a new book will be successful long term?

My next question was, “Are there any publicly available early signals that a book is going to be a long-term bestseller?”

While it’s useful to know whether a book published years ago has been successful, the lessons I can draw from it are limited. What worked well back then may no longer work well now. On the other hand, strategies that are working well for them now might not work for someone releasing a new book in today’s market. Lifetime sales aren’t enough – I needed to find a way to control for how long a book has been on the market.

To find the answers to these questions, I calculated the “average number of reviews per day” that each book on my list had received over its lifetime. In other words, at what rate did they receive reviews on average? I call this metric “review velocity.”

The review velocity represented a very widespread, from 1.2 reviews per day for Sprint, to 59.4 for Atomic Habits, with an average of 15 reviews received per day across all 20 books.

The reason review velocity is so helpful is that it allows me to compare the “trajectories” of books that were published at completely different times, including recently published books. I have no idea how a book’s reviews are distributed across time (Do a lot of them come in early when it gets released and then decline over time? Or do they start slow and pick up speed over time?), but working with averages allowed me to see the broad strokes.

When I compared “review velocity” with “sales to date” it produced quite a strong positive correlation of 65%. This means that a book’s review velocity isn’t as strong a predictor of sales as the total number of reviews, but a lot stronger than the average rating or time on the market. In other words, a book rapidly receiving a large number of Amazon reviews is a fairly strong indicator of lifetime sales potential.

I also noticed a lot of variation in the percentage of book buyers who left a review. It ranged from 0.3% for The Life-Changing Magic of Tidying Up to 4.8% for my book, with an average across all 20 books of 1.5%. I’m guessing this number starts high as hardcore fans leave reviews and then declines as the book spreads to more mainstream, less dedicated audiences. But it’s a good sign!