Amazon Top 100 Kindle Books – Exploring Reader Engagement

Recently, Amazon released lists of its 100 top selling titles for Kindle ebooks.  In some previous blogs, we looked at how Indies (self or independently published ebooks) compared to Trads (traditionally published or trade published ebooks) in terms of books sold (as measured by ranking within the top 100), reader satisfaction (as measured by Amazon reviews) and imputed revenues/earnings.  We also looked at how well sales could be estimated from reviews, via several lines of evidence.  (For more information regarding the findings from those earlier analyses, see the previous blogs  Amazon Top 100 Kindle Books - Indies versus Trads Part 1, Amazon Top 100 Kindle Books - Indies versus Trads Part 2, Amazon Top 100 Kindle Books - Indies versus Trads Part 3, and Amazon Top 100 Kindle Books - Relationship between Sales Rank and Number of Reviews.)

Note: Since this blog is a bit lengthy, I will put an “executive summary” at the front, repeating the summary remarks made at the conclusion of the blog.  So, to sum up (prematurely):

·         The top best-sellers get fewer reviews than expected, based on sales rank.  Books further down the list got more reviews than one would expect from their sales ranking.

·         Male writers were slightly more likely to be reviewed, relative to sales, but the effect was small.

·         Indies were slightly more likely to be reviewed than Trads, relative to sales, but again the effect was small.

·         Higher priced books were reviewed more often than would be expected from sales rank. The reverse was true for lower priced books.

·         Non-fiction was reviewed more often than would be expected from sales rank, but the numbers of such books were too small to be able to say whether this was meaningful.

·         Romance and Thrillers were not reviewed as much as would be expected from sales rank.  Science Fiction and Fantasy were “over-reviewed” as were the other categories.

Now for the detailed blog

In this fifth blog we continue to looks at sales rank vs number of reviews rank, in order to explore something we might call reader’s public engagement.   That won’t be a measure of sales or of number of reviews, but rather a blend of the two.  Our intent is to discover what factors might correlate with readers’ tendency tell the world about the book they just read.   Note that this measure of engagement could be either positive or negative – it is the willingness to do a review that matters more than the rating given to the book that is important here.

To this end, we will compare a books sales rank to its number of reviews rank, in the Amazon Top 100 list.  So, for example if a book was in the ranked 15^th in sales but 85^th in number of reviews, we would conclude that readers had low levels of engagement with the book – at least in the sense of engaging in a public space such as the Amazon reviews system.  They may have liked it, but they evidently weren’t motivated enough to review it.  Conversely, a book that was 85^th in sales rank, but 15^th number of reviews rank would have high engagement.   People who read it were evidently more motivated than average to write a review.  Lastly, a book whose sales rank and number of reviews rank were the same or nearly so (say ranked 50^th in both) would be considered to have average engagement.  So, given the limitation of our data, what we are measuring is relative engagement – does the book receive more or fewer reviews than would be expected from its sales rank.

Please note that the number of reviews ranking relates to ranking within that list, not the entire set of Amazon books.  But we don’t have access to those numbers, so we work with the data that we have, rather than the data that we wish we had.

So, let’s look at some data.  First, we will look at how engagement varied by sales rank, grouping the data by decile (i.e. into 10 equal sized categories).  To that end, the table below indicates that books with high sales rankings didn’t tend to receive as many reviews as we would expect ,while those with lower sales ranking received more reviews than we would expect.  For example, none of the books in the second decile of sales (ranked 11^th to 20^th) were ranked in the second decile of reviews – they were all ranked somewhere further down the list. On the table, that is indicated by the column headed “3-RevRan<SalesRank”, which shows that all ten books in this decile had review ranks less than their sales ranks.  This is further reinforced by the column “Sales Rank minus Rev Rank”, which shows that books in this decile were rated 20 places higher in sales than in reviews, on average.   Conversely, only one of the books ranked in the 9^th sales decile (81^st to 90^th rank) was in that review decile – most had review rankings higher up the list.  On average, they were rated 16 positions higher in their review rank than their sales rank.

 

Rank2	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
1	2	2	6	3.0
2			10	20.8
3	4	1	5	11.3
4	4		6	10.6
5	5		5	5.2
6	2		8	6.0
7	6		4	-7.5
8	7		3	-16.4
9	9		1	-15.9
10	9		1	-17.4

 

Next, we will look at the writer’s gender.  The table below shows that there doesn’t seem to be a very pronounced gender effect, though males do seem to be somewhat more likely to be reviewed than females, relative to their sales rank.  On average they are 9 ranks higher in reviews than sales, while females are 4 ranks lower in reviews than sales.  Note that the numbers above don’t balance – that’s because there are more women writers than men in the Amazon Top 100 list.

WriterSex	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
Female	33	1	36	3.8
Male	15	2	13	-9.0

Now, let’s look at publisher, Indie vs Trad, always an interesting category.  As the table shows, there really wasn’t any difference between Indies and Trads when it comes to this measure of reader engagement.  Indies were slightly more likely to be reviewed than one might expect from sales rank, but the effect was small – a difference of about 5 ranks.

Pub3	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
Indie	12		12	-4.8
Trad	36	3	37	1.5

And here’s a more detailed look at publishers.  There doesn’t seem to be any notable trend in this table. 

Publisher2	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
Doubleday			1	6.0
Hachette	8		11	6.7
Harlequin	1		1	6.5
Harper Collins	3		1	-18.8
Indie	12		12	-4.8
MacMillan	1			-28.0
Penguin	7	1	12	4.0
Random House	7	2	6	2.9
Simon & Schuster	8		5	-2.1
William Morrow	1			-27.0

Next up is price range, broken out as low (under $4.00), moderate ($4.00-$7.99) and high ($8.00 and over).  There does seem to be an interesting trend here – people appear to be more willing to review higher priced books than lower priced books.  Lower priced books sales ranks tended to be about 8 places higher than their review ranks.  For higher priced books, the opposite was true, and moderately priced books had sales ranks and review ranks that were almost identical, on average.  So, there may be some sort of social status effect here, whereby people are signalling their socio-economic status by reviewing higher priced books disproportionately.  Or perhaps they just feel more “invested” in a higher priced book, and thus more willing to spend a few minutes on a review.

Price2	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
1-Low	11		19	8.2
2-Mod	26	3	23	-1.7
3-High	11		7	-9.1

Now we get into the genre categories, which are usually quite interesting.  First up is fiction vs non-fiction. While it is true that most of the Amazon Top 100 ebooks were fiction, the data does seem to show and interesting effect, whereby non-fiction readers were disproportionately more likely to do reviews.  But the numbers are small, so we can only consider this to be a very provisional result.

Fict_or_NF	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
Fiction	45	2	49	0.8
Non-fiction	3	1		-19.5

Here’s a more detailed look at genre.  The main effect here is that Romance and Thrillers tend to be “under-reviewed” while the other categories are “over-reviewed”. 

Genre1	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
Business		1		0.0
Historical Fiction	2			-24.5
Humour	2			-22.5
LitFic	6	1	4	-6.3
Religion	1			-28.0
Romance	19		28	6.4
Self-help	1			-36.0
SFF	6		1	-28.6
Thriller/Suspense/Crime	11	1	16	4.4

A lot of the categories in the previous table were pretty small, so we will repeat them with the collapsed genre categories below.

Genre2	1-RevRank > SalesRank	2-RevRank = SalesRank	3-RevRank < SalesRank	Sales Rank minus Rev Rank
LitFic	6	1	4	-6.3
Other	6	1		-22.6
Romance	19		28	6.4
SFF	6		1	-28.6
Thriller/Suspense/Crime	11	1	16	4.4

Again, the outstanding feature of the data is how Romance and Thrillers don’t get as many reviews as might be expected from their sales rank. The big winner here seems to be Science Fiction and Fantasy.  Perhaps it is not surprising that they get a lot of reviews, as the readers of these categories are often well educated, literate, and confident in their communication skills.  This was also true of literary fiction, though to a smaller extent.

So, to sum up:

·         Best-sellers get fewer reviews than expected, based on sales rank.  Books further down the list got more reviews than one would expect from their sales ranking.

·         Male writers were slightly more likely to be reviewed, relative to sales, but the effect was small.

·         Indies were slightly more likely to be reviewed than Trads, relative to sales, but again the effect was small.

·         Higher priced books were reviewed more often than would be expected from sales rank. The reverse was true for lower priced books.

·         Non-fiction was reviewed more often than would be expected from sales rank, but the numbers of such books were too small to be able to say whether this was meaningful.

·         Romance and Thrillers were not reviewed as much as would be expected from sales rank.  Science Fiction and Fantasy were “over-reviewed” as were the other categories.