How A Keyword Sells A Book On Amazon

In order for a keyword to create a sale through Amazon search, two events need to occur:
1. The book needs to be present in the search results (or rank) for search queries matching the keywords added to the book.
2. The book needs to convince customers (or convert them) to click on a search result, add the book to their cart, then check out (or, make the purchase).

Each keyword creates an opportunity for sales to be made, as each keyword provides the Amazon search engine with an additional way to surface a book to customers. The more keywords assigned to a book, the more data the search algorithms have to work with. Not every assigned keyword will be matched to customer search queries, but you can think of each keyword as asking the search engine: have you thought about showing my book to customers who search for this query? It's up to the search engine to decide, but the more questions you ask, the more likely it will find more ways to return a book.

Behavioural Influence On Search Rank

Once keywords are ingested by Amazon, they’re subjected to a whole lot of parsing and processing, and testing with customers.

First, let’s remember that the goal of a retailer search engine is to maximise revenue, not help people find information (like on Google). This means that on Amazon, discoverability and sales are tightly linked. The rationale behind this is pretty straight-forward - if you show the books most likely to sell,  more often, you’re more likely to sell more books.

Every time Amazon displays a book to a customer in search, the customers’ action is logged. If the action is positive, such as a click on a link to view the product page, a cart add, a purchase or a read on Kindle, it positively impacts the books search rank. The more positive actions a book accrues, the more visibility it’s rewarded with. 

Conversely, if customers don’t respond in a positive manner to a book in a search result, then the book will eventually be outranked by other books that customers respond more positively to.

For example, if a book is shown in search, and the books above and below it are clicked, and not the book itself, this is regarded as a negative signal and will mean the book-query pair will be weighted downward. The book’s search rank for the query hasn’t encouraged positive behaviour so it is penalized.

A simpler way to think of this is - if book A sells more than book B when shown in search, keep showing book A ahead of book B, because book A is making more money. Let’s also try putting book C ahead of book B, and see if that book makes more money instead.

The ability to maintain and grow search visibility is dependent on the book converting through this funnel of customer actions. Sales are the ultimate goal for publishers and retailers, and are a strong search magnifier, but these sales are dependent on conversion.

All of these scores attributed to customer behaviour decay over time, which means momentum matters. It’s easier to improve the discoverability of a book that is already selling and converting well, using keywords, than it is to improve a book starting from low sales.

Conversion And Social Proof

We've all searched for books on Amazon before - you type in a search query and are presented with a list of books. Perhaps you were looking for an exact title, or perhaps you were searching for books about a topic - in either scenario, you might have been swayed to click on a search result if it looked interesting to you.

Think about the factors that persuaded you to click into the book's product page from search. If you were encountering the book for the first time, the only information you would have at that point was the data visible to you in the search result - cover, title, subtitle, price and importantly: the review count and average rating. These last two elements capture the social proof of the expected reading experience.

Social proof is one of the most powerful on-page converters, as customers often trust each other more than the organization trying to sell to them. Reviews are a book’s public record - they convey more value to a reader than whether book it has sold well, and let them know what other people thought about their experience (good or bad) and why.

Reviews are a permanent, mostly unbiased and irrefutable history of a book by it’s readers. And unlike search rank, the social proof from reviews is cumulative and compounds over time - it doesn’t really decay.

Once you have 500 reviews you’ll always have at least that many reviews. With an established positive star rating across a large base of reviews, it’s also highly likely future reviews and ratings will follow a similar pattern, further strengthening the book’s ability to convert.

In a future post, we'll analyse some data to see how social proof relates to search visibility on Amazon, including examples where Top and Recent customer reviews have been impactful. But in the meantime, thinking about a keyword's place and ability to influence a sale in the search funnel, will help you to focus on which element of metadata to prioritize to maximize sales potential.
 

Do Writers Write What Readers Want To Read?

Have you ever wondered if the genres authors most enjoy writing in, match the genres readers most enjoy reading? Before self-publishing, all new books for sale were filtered by agents and publishers, who acquired and worked on books they thought would sell well. If there was an oversupply of manuscripts by authors in a particular genre, the competition to be chosen and published within the genre, would be higher too. Enter self-publishing: now any writer can publish, without filter, into any genre they desire. Given the influx of new books across genres, does the proportion of books in each genre meet with readers' demand?

(We focussed on fiction for this experiment).

Methodology (or How to Speed Read 3000 books in 3 hours)

To answer our question, we needed a way to read and understand a good sized sample of self-published books, to determine their genre. You might ask why we couldn't simply use the categories or tags authors themselves apply to their books? The reason is accuracy and consistency - most indie authors don't have years of book categorization experience, working across a number of titles. Even traditionally published books are categorized inconsistently from book to book and from publisher to publisher. The inconsistency is not because publishers are poor at the job, but because standardizing the process would require centralizing the categorization effort. (We've worked with data feeds from all major publishers and have experienced this phenomenon first hand). The only way to derive accurate and consistent categorization is to read a large sample of books, understand how each book relates to each category, and assign it, while ensuring consistency across the sample. One of our systems does just this.

We gave our categorization system over 3000 self-published novels to read and understand (these were books offered free by the author). For each book, our system identified all the topics the novels were about, then used this topical knowledge to assign each book to one or more categories and genres. Overall, our system read over 260 million words and figured out all the genres, categories and topics in the data below, in a few hours.

What writers write

Writer’s Genres

The top genres (by count) detected by our system were Romance, Fantasy and Science Fiction.

Romance was the most popular genre, with 24.4% of books tagged. By combining Science Fiction and Fantasy though, to derive a total score of 32.1%, we can deduce that writers enjoy writing in this genre more than any other. Literary and Mystery & Detective both came in around 6%. How does this compare to what readers read?

Reader’s Genres

To understand the genres readers enjoy reading the most, we looked at revenue data. This doesn't incorporate units purchased or read, or ratings, but in aggregate, revenue is a good proxy indicator for reader enjoyment.

Source: Leading book genres worldwide as of January 2014, by revenue (in million U.S. dollars)

The highest selling fiction genres were Romance/Erotica, Crime/Mystery and Science Fiction and Fantasy. Romance was high in both charts, but we can broadly extrapolate that there’s a potentially underserved market for Crime and Mystery & Detective and an oversupply for Science Fiction and Fantasy books (when combining the two genres in our first chart).

The correlation isn’t perfect of course, as our sample size is small, we're not considering units sold vs. price, and the revenue data is based on the less consistent human classification of books. We also assume the novelists in our sample wrote their books for the joy of it, and didn’t select their categories purely for commercial potential. These points aside, for the purposes of this post, the proportional difference in genres across the two charts is interesting.

BISAC Categories

We also wanted to understand the categorical split of each genre, so we dove deeper and analyzed the individual BISAC categories that made up each genre. The chart below is measured by category composition - which analyzes how much of each book belongs to a category. For example, instead of tagging a book as Romance and Fantasy, our systems tell us the book is 30% Romance and 70% Fantasy.

(BISAC is the US publishing industry’s system for categorizing books. You can read all about it here - https://www.bisg.org/tutorial-and-faq)

This chart closely matches our genre chart, but tells us that Romance books typically consist of more granular categories than Science Fiction and Fantasy categories. This is somewhat reflected in the number of different BISAC subcategories for the genres - Romance has almost 50% more sub-categories than Science Fiction and Fantasy combined. It also alludes to a level of variance in the categories - our system was more easily able to split Romance titles into clearly distinct categories, but for Sci Fi and Fantasy, most content was generalized to Fantasy / General or Science Fiction / General.

(We've also classified tens of thousands of freely available Gutenberg books, which you can browse here. Many of these books were published before BISAC was invented.)

Topics

Next, we dove even deeper to look at the topic composition of our sample of books, and analyzed how much each book was made up of each topic. The topics listed below aren’t industry standard, and were created by our team. Topics allow us to quickly and programmatically understand, at a more granular level, what a book is about.

Given the strong bias for Science Fiction and Fantasy, the top few topics aren't particularly surprising. One observation we can make from this data, is that some genres have a higher proportion of genre-specific content than others. For example, a Romance novel will have many romantic scenes and dialogue, and be romance-themed. But the story will often revolve around another topic (western, military, etc.). A Science Fiction or Fantasy novel will usually contain a high proportion of genre-specific content - the whole world of the story will usually relate to the genre. Books in these genres are also likely to encompass elements of other genres too. Therein lies the categorization challenge we discussed earlier - should a novel be FIC027130 (Romance / Science Fiction) or FIC028000 (Science Fiction / General) or both? Are the romance elements strong enough for a book to be categorized as a 'Romance' book? Publishers of course, use knowledge of the book as well as strategic category selection, to influence placement of their books on bookstore bookshelves.

A few notes on the topics above. 'Existence' - covers concepts such as consciousness, the universe, humanity and realms - elements often found in Sci Fi / Fantasy. 'Vampires' have their own topic (instead of being part of 'Creatures & Monsters') which reflects the more prominent showing of vampires compared to other monsters, in recent fiction. 'Erotica' as a topic is smaller in representation for the reasons we discussed above for 'Romance'.

Conclusion

We speculate that writers write more Sci Fi and Fantasy books, as it's simply a lot of fun to create entire worlds with their own rules, creatures and customs. Mystery & Detective or Crime novels, while also fun to write, are often set in our reality, and typically require some technical or specialized knowledge - details which may need to be fact checked and accurate. Many authors in these genres have had prior experience in the field, or have spent significant effort researching their topics. These books will often teach the reader something, which is appealing to readers.

As a writer, should you switch to Crime and Mystery in order to increase your odds of landing an agent or selling more self-published books? We don't think so. Write in the genre that is the best fit for you, as doing so will be reflected in your published work.

We hope you enjoyed this glimpse into what self-published authors are writing. Please let us know how you interpreted our results in the comments below.

If you'd like to see this data for your book, analyze your manuscript at Author Checkpoint.

How Book Categories Have Changed This Year

Amazon continually updates it's browse node categories for books, to cater to the shifting needs of the market. In the last six months, 563 new categories were added, and 122 were removed. There were also a number of category name refinements. Categories serve the purpose of helping readers find similar books. As the number of books allocated to a category fluctuates, the granularity of the categories needs to change too. If categories remained static, they'd become unbalanced, with too few or too many books, which would make browsing and searching for books a challenge.

We've summarized the changes to book categories (browse nodes), that have occurred over the past six months, looking at the impact on various top level categories.

 

Genres with new categories added

categories-with-new-browse-nodes-added.jpg

There were 345 new 'Teen & Young Adult' categories added in the past six months, which is likely a reflection of the huge increase in YA sales over the past year. 

Only one 'New Adult' category was added (Science Fiction & Fantasy/Fantasy/New Adult & College), to take the total to two categories (the other is: Romance/New Adult & College). Rather than expand this newer category, the breadth of 'Teen & Young Adult' has been expanded to accommodate the influx of titles in this area.

Additions to 'Computers & Technology' were dominated by 'Software' (Adobe, Enterprise Applications and Business) and 'Web Development & Design' (Programming and Web Design) sub-categories. 

The largest increase to the 'Religion & Spirituality' genre was in the 'Religious Studies' sub-category, comprising 24 of the 52 additions.

Teen & Young Adult Category Additions

Digging deeper into the 'Teen & Young Adult' genre we see that of the 345 additions made, 152 were in fiction and 193 were in non-fiction. These were broken up as follows:

teen-and-young-adult-category-additions.jpg
categories-browse-nodes-removed.jpg

The 'Computers & Technology' genre had the most categories removed (88), but had an almost equal number of categories added (87). Technology experiences rapid changes, so a commensurate shift in categorization of the subject matter is likely to occur.

The second largest genre with categories removed was 'Crafts, Hobbies & Home', and within that genre, almost half were in the 'Home Improvement & Design' sub-category.

For a full list of the removed categories, click here.

Conclusion

Changes in browse node categories reflect shifts in the type of books available for sale. An increase in categories for a genre is likely driven by a combination of additional supply of books in the genre, and of increased effort to improve the searchability of those books. One may speculate how these two factors reflect increased demand for books in a genre.