Sunday, March 6, 2016

Discoverability Myths and Keywords in Titles

tldr version: putting keywords in the title increases discoverability, discoverability doesn't appreciably drive sales, and algorithms are weird.

Back in the days before Yelp and the internet, if you needed a plumber you opened the phone book and called the one with the biggest ad or the one listed first.  This is why so many plumbers were called things like "AAAAAAAAAuger Plumbing".

Indie authors do the same thing today with regard to keywords.  You're allowed 7 keywords, but in practice it's really as many as you can fit with 400 characters.  Using these 400 characters to accurately describe elements of a book is generally acceptable.  Using them to stuff as many popular keywords that have nothing to do with your book - "Romance Coming of Age Clean Menage James Patterson Shifter Horror Cozy Mystery Erotica Fantasy" - not so good.

And if you put the keywords in the freaking title and subtitle, there is a special place in hell reserved for you.  The title is supposed to be what the book is called - what is on the actual cover.

"Hey, I read a great book last weekend."
"Really, what's it called?"
"THRILLER: MYSTERY The Sassafras Conundrum (THRILLER MYSTERY THRILLER THRILLER)"

That kind of thing.  It's against Amazon policy, and authors occasionally get an email telling them to change it.  There's a reason that people do that, and I tried to figure out if it works.

The idea is to improve discoverability.  For my example I used the keyword Horror.  There are 163,999 Kindle titles on Amazon related to horror, and just like poor Zany Zach's Plumbing, most of these folks are going to be buried deep deep down at the bottom.


Search ResultsSales RankSalesKeywords in Title
13501112y
253126y
35109913y
47263914y
58593n
682618n
721349917y
818731416y
91061411n
1057227y
1118369015y
12974910n
13538,30619n
147712n
156491n
16no sales20y
1793859n
1838144218y
1921505n
209094n

Search Results: Rank of non-free titles in the order they appear using a keyword search sorted by relevance.
Sales Rank: Overall sales rank at Amazon, the lower the better.
Sales: In order of Sales Rank out of this list
Keywords in Title: Y means there is a keyword that isn't a part of the actual title, like "Horror: The Waffle House Massacre (Horror, Thriller, Romance)".  N means there isn't.  Something like "The Scarening: A Novel of Unrelenting Horror" or "Horror Tales for Tweens" gets an N as well, since it's organically part of the title or subtitle.

First a word about relevance searches.  The algorithm used is secret and probably a little messed up on purpose so folks can't reverse engineer it.  In theory this should show the most relevant, i.e. the most horror-y, results first, then sorted by popularity.  It actually works pretty well, as the shifter erotica doesn't show up for several pages.

Things change if you change the sorting preferences.  One would assume changing to "sort by price" would then by subsorted by relevance - in other words, the order of all the free books would stay the same between searches.  This is not the case - sort by price and all the dreck gets mixed back in.  I suspect it sorts by price, then popularity.

Back to our chart.  The first 20 books I picked happened to have 10 titles with keywords in the title and 10 without.  I noticed looking past 20 books that keywords got fewer and fewer.

Interestingly, only one title in the first 20 search results was in the top 20 paid bestsellers for horror.  This is 15 in our list, which was the 18th horror bestseller.  Another thing to keep in mind - Amazon has been accused of purposely pushing down titles from publishers that they're in a slap fight with.

Titles with keywords did much better in discoverability, taking 8 of the top 10 spots.  For sales, they did worse, taking only 2 of the top 10 spots.  Looking at overall Amazon rank, they did much worse.  Books without keywords had an average rank of 58165.  The book with no sales would rank over 2 million, so using that number, books with keywords had an average rank of 313,573.

Both sets had outliers, so if you take out the worst selling book of each, the difference is even more telling: 4816 without keywords, 126,192 without.

This tells us a few things.  Keywords in titles improve discoverability.  A low selling book gets displayed ahead of better selling books.  I'm going to zoom in on our worst selling book.  As of this writing it has not sold a single copy.  There are 16 keywords in the title, many of them duplicates.  The actual title has three words and "scream" is misspelled.  If you do a search for "horror", this book appears three pages before Stephen King.  Pretty awesome, right?

You missed the part where I said it had zero sales.

Now, an argument could be made that these books would sell even worse without keywords in the titles.  Maybe so, maybe not, though some couldn't sell much worse.  However, I think we can put to rest the myth that discoverability is the key to success.  There are several titles in the top 10 search results that maybe sell a title a month, while Stephen King is doing fine languishing on page 7.

Looking at it from the other direction, none of the top 100 paid horror titles start with a keyword.

Could the keywords in titles do more harm than good?  There's no way to tell from these numbers, but as a reader browsing through titles, I look at "HORROR SUSPENSE THRILLER" the same way I do "RESIDENT" on my mail.

No comments:

Post a Comment